Prison Labor: The Price of Prisons and the Lasting Effects of Incarceration

Abstract

Institutions of justice, like prisons, can be used to serve economic and other extrajudicial interests, with lasting deleterious effects. We study the effects on incarceration when prisoners are used primarily as a source of labor using evidence from British colonial Nigeria. We digitized sixty-five years of archival records on prisons from 1920 to 1995 and provide new estimates on the value of prison labor and the effects of labor demand shocks on incarceration. We find that prison labor was economically valuable to the colonial regime, making up a significant share of colonial public works expenditure. Positive economic shocks increased incarceration rates over the colonial period. This result is reversed in the postcolonial period, where prison labor is not a notable feature of state public finance. We document a significant reduction in contemporary trust in legal institutions, like police, in areas with high historic exposure to colonial imprisonment. The resulting reduction in trust is specific to legal institutions today.

Full text

Prison Labor: The Price of Prisons and the Lasting
Effects of Incarceration∗
Belinda Archibong†
Barnard College and NBER
Nonso Obikili‡
United Nations
August 16, 2023
Abstract
Institutions of justice, like prisons, can be used to serve economic and other extra-
judicial interests, with lasting deleterious effects. We study the effects on incarceration
when prisoners are primarily used as a source of labor using evidence from British
colonial Nigeria. We digitized 65 years of archival records on prisons from 1920 to 1995
and provide new estimates on the value of colonial prison labor and the effects of labor
demand shocks on incarceration. We find that prison labor was economically valuable
to the colonial regime, making up a significant share of colonial public works expendi-
ture. Positive economic shocks increased incarceration rates over the colonial period.
This result is reversed in the postcolonial period, where prison labor is not a notable
feature of state public finance. We document a significant reduction in present-day
trust in legal institutions, such as the police, in areas with high historical exposure to
colonial imprisonment; the resulting reduction in trust is specific to legal institutions.
JEL classification: H5, J47, O10, O43, N37
Keywords: Prison, Incarceration, Tax, Convict Labor, Colonialism, Trust
∗Thanks to Stelios Michalopoulos, Nathan Nunn, James Fenske, James Robinson, Liz Ananat, Ebonya
Washington, Eduardo Montero, Nancy Qian, David Weiman, Marlous van Waijenburg, Suresh Naidu, Warren
Whatley, Lee Alston, Leonard Wantchekon, Michiel de Haas, Denis Cogneau, Rajiv Sethi, Ewout Frankema,
Florence Bernault, Naomi Lamoreaux, Gerald Jaynes, Tim Guinnane, Robynn Cox, Peter Henry, Marcellus
Andrews, Claudia Goldin, Owen Ozier, Bryce Steinberg, Ellora Derenoncourt, Christopher Muller, Larry
Katz, anonymous referees, and participants at the NBER and BREAD meetings, the Harvard, Yale, Duke,
Brown, UC Berkeley, Columbia, Queens College, Tennessee, Williams, University of Michigan, Stanford, Ox-
ford and PSE seminars, the SSHA, LACEHA, AEHN, ASA, AEA and other conferences for useful comments
and suggestions. Thanks to Yuusuf Caruso, Monique Harmon, Anamaria Lopez, Robrenisha Williams, Chloe
Dennison, and Serena Lewis for excellent research assistance. All errors are our own.
†Corresponding author. Barnard College, Columbia University. 3009 Broadway, New York, NY 10027,
USA, and NBER. ba2207@columbia.edu.
‡me@nonsoobikili.com. The views expressed here are the authors’ and not necessarily those of the UN.
1

1 Introduction
Across Europe’s African colonies, prisons were viewed as a legally sanctioned means of ad-
dressing chronic labor shortages and minimizing the costs associated with hiring wage labor
to construct and maintain public infrastructure like the railroad and roads. Unlike other
parts of colonial administration which were often characterized by indirect rule, where local
officials exerted significance influence over economic extraction activities like tax collection
from local populations, colonial prisons were administered directly by colonial officials who
maintained detailed records on the use of prison labor, exemplified by the below quote from
Inspector Jackson in Nigeria (Archibong, 2019; Lowes and Montero, 2021a).
“The Prison at Port Harcourt has been considerably developed and at the close
of the year there were 829 prisoners in custody and these are employed by the
Eastern Railway. The Engineer in charge at Port Harcourt is highly pleased with
the way the prisoners are worked; they have given no trouble and have been of
great assistance in developing that station. It was my intention to have 1,000
prisoners stationed there before the close of the year, but this was impossible as
two prisons...which should have supplied the drafts to make up the number, had
an outbreak of chicken-pox...”
- E. Jackson, Acting Inspector of Prisons, Lagos, April 23, 1915
These records provide a window into the economics of colonial extraction using formal
legal institutions. The historical evidence allows us to answer important questions on the
economics of colonial institutions, including why and how prison labor was so valuable to
the colonial regime, the incentives for incarceration that arose from colonial governments
viewing African prisoners primarily as a reserve of labor, and the long-term effects of the
colonial prison labor system on populations’ views of state legitimacy and contemporary trust
in legal institutions. They also provide important contrasting evidence to the hypothesis
that European colonizers, focused on extraction, established weak institutions to enable
this extraction, and that strong institutions, centered on laws protecting property rights,
led to positive long-run development outcomes (Acemoglu and Robinson, 2001; Dell and
Olken, 2020). The evidence from colonial prison labor institutions in Africa illustrates that a
complex, formal system of prisons, laws, courts and police was crucial for economic extraction
in Europe’s African colonies, with negative short-run and long-run consequences for local
African populations. The case of colonial prison labor highlights an important insight on the
development effects of colonial institutions- that the implementation of these institutions,
2

not just whether or not they were “inclusive” or “extractive”, was important for determining
their impacts on economic development.
We study the economics of colonial prison labor and its long-run effects on African de-
velopment using evidence from British colonial Nigeria, covering a period between 1920 and
19591, when prison labor was a feature of state public finance and administration; and from
postcolonial Nigeria, covering a period between 1971 and 1995, when prison labor was not a
major feature of state finance. We construct a novel dataset from 65 years of archival records
on prisons from 1920 to 1995, assembling data on prisons, wages, prices, and colonial public
finance from colonial and postcolonial archives, along with geocoded climate information
from high-resolution NASA data to test our hypotheses. Our new dataset represents one of
the most comprehensive records on the economics of colonial prison labor and incarceration
in Africa, spanning both the colonial and postcolonial periods. Colonial Nigeria is an in-
formative region in which to study these issues for a number of reasons. Notably, colonial
Nigeria had relatively high incarceration rates. As of 1940, the British colonial government
in Nigeria was incarcerating more people (0.3%-0.4%) than the United States (0.1%) and
countries in Europe over a similar period (0.06% in 1950) (Muller, 2021; Jacobson, Heard,
and Fair, 2017). To put these figures in context with contemporary data, Figure 1 presents
the top 40 of 222 countries/jurisdictions in the world as of 2018 by incarceration rate. If
we place colonial Nigeria’s incarceration rate in 1940 on the chart, it would have ranked at
number 15 of 222 in 2018, right between the Seychelles and Panama. Nigeria incarcerates a
much lower share of people currently, ranking at around 211 of 222 by World Prison Brief
estimates.
We conduct our analysis in three steps. First, to understand why prison labor was so
valuable to the colonial regime, we calculate the value of unpaid prison labor, and estimate
the share of prison labor in colonial public finance. A key insight from the historical archives
is that prison labor on government public works was a mandated aspect of incarceration, as
part of explicit colonial policy, and prisoners were not paid a wage for their labor.2Unpaid
prison labor was a key input in the construction and maintenance of essential revenue-
generating public works, such as the railroad, which was used to transport agricultural
commodities for export. We estimate the value of unpaid wages to prisoners or unpaid
1Nigeria as an amalgamated entity was a British colony from 1914 to 1960; hence, our dataset covers
almost 40 of the 47 years of the colonial period. The country was under military rule for most of 1960 to
1999, before transitioning to democracy in 1999.
2The 1916 Prison Ordinance explicitly outlined the use of convict labor explicitly (Kingdon, 1923).
3

prison labor and show that it was economically valuable to the colonial regime. The overall
gross value of prison labor was strictly positive over the entire colonial period. Even after
accounting for the most expansive set of prisoner maintenance costs, the net value of prison
labor was nonnegative in 60% of the years from 1920 to 1959 in Nigeria. On average, the gross
value of prison labor was more than double the amount spent by the colonial government on
public works between 1920 and 1959.
Second, we show that, faced with labor shortages and aiming to minimize the costs
of administration, the colonial regime used prison labor to address their economic needs.
Chronic labor shortages were an endemic feature of the labor market in colonial Africa
due to the joint factors of high land to labor ratios, rising demand for labor for colonial
construction and subsequent relatively high prices of labor (van Waijenburg, 2018). Historical
evidence suggests that incarceration rates increased when there was a shortfall of labor for
construction and maintenance of public works; this was particularly true during periods
of positive agricultural productivity shocks, when agricultural wages increased and wage
labor was relatively more costly to attain (Abiodun, 2017; Bernault, 2007). Labor shortages
and tight labor markets, worsened by wage ceilings in the government public works sector,
increased the demand for unpaid prison labor, in line with predictions from theoretical models
of labor coercion (Acemoglu and Wolitzky, 2011). One way colonial authorities addressed
these labor shortages was to increase the number of incarcerated individuals by, for instance,
increasing prosecutions of minor, misdemeanor crimes, and switching the punishment of
these crimes from fines to imprisonment (Killingray, 1999). The use of prison labor continued
through the end of the colonial period in 1960. After independence, prison labor become less
important for the regime. The discovery of oil in 1956 and an oil price boom in the 1970s,
along with increased protests from labor unions, led to a steep decline in the use of prison
labor in postcolonial Nigeria (Abiodun, 2017).
We quantitatively test these historical accounts of the effects of labor demand shocks
on incarceration and the use of colonial prison labor by estimating the effects of shocks to
economic productivity on incarceration rates using a panel regression framework. We con-
struct two measures of shocks to economic productivity. The first measure uses agricultural
commodity export prices and district-level crop suitability, and the second measure exploits
district-level rainfall deviations in a primarily agricultural setting. We show that the in-
carceration rate is procyclical during the colonial period. Positive economic shocks increase
the colonial incarceration rate and the use of prison labor. The positive effect is specific to
the short-term incarceration rate only, with shocks increasing the share of prisoners with
4

short sentences of fewer than six months. There is no effect on incarceration of prisoners
with sentences longer than two years. Using an index of export crop prices, we show that a
10% increase in export prices for a major cash crop in producing regions is associated with
a 5% increase in short-term incarceration relative to the sample mean. In a second spec-
ification, moderate positive rainfall shocks that raised agricultural productivity, increased
the short-term incarceration rate by 16.7 prisoners per 100,000 people, representing a 12%
increase relative to a mean of 134.7. This effect is reversed in the postcolonial period wherein
prison labor is not a main feature of state policy, and negative productivity shocks increase
the incarceration rate. The results provide support to the qualitative historical evidence
on the use of prison labor to address labor shortages in the colonial period, but not in the
postcolonial period.
To provide further evidence on the mechanisms driving the colonial incarceration re-
sults, we test the tight labor market hypothesis by examining the effects of rising wages on
the rate of incarceration and how this varied depending on a prison’s distance to the colonial
railroad. The rationale here is that since a major use of prison labor was for the construction
and maintenance of the railroad, examining the correlation between wages and incarceration
rates at prisons around the railroad can provide insights into how demand for colonial prison
labor changed during periods of higher wages. While short-term sentenced prisoners near the
railroad were generally used as a reserve of unpaid labor for railroad construction and main-
tenance, rising wages intensified the demand for unpaid prison labor. To increase the use
of prison labor, colonial officials would need to increase the numbers of prisoners in prisons
farther away from the railroad as well, since prisons farther away from the railroad generally
had fewer prisoners, on average, than those closer to the railroad. In other words, prisons
that were located farther away from the railroad potentially had more space to respond to
increased demand for prisoners since they generally had fewer prisoners than prisons close
to the railroad. Prisoners, by law, could not be transferred across districts, but could be
transferred from one prison to another within the same district (Foreign and Office, 1947).
Colonial officials could then transport prisoners within the colonial district to conduct work
on the railroad and associated public works as needed during periods of labor shortages
and tight labor markets. The results show that when wages increased, prisons farther away
from the railroad but within the same district, experienced an increase in the number of
prisoners incarcerated for short-term sentences. We also provide suggestive evidence of the
sentence-switching channel, and show that increasing the prison sentences of prisoners that
were already incarcerated may be one way that colonial officials intensified the use of prison
5

labor in response to labor demand shocks.
Finally, in the third step of our analysis, we study the long-run effects of colonial
use of prison labor on present-day trust in legal institutions. The judicial system in colonial
Nigeria was highly centralized, with the colonial courts working in concert with prison officials
and police to “maintain law and order”, protect European property, and meet the revenue
objectives of the colonial government under the prison labor system (Rotimi, 1993; Onoge,
1993). The police were often the first point of contact with the judicial system for local
populations and were especially involved in enforcing tax collection, including tax raids and
arrests associated with violations of colonial crimes. They were also notoriously feared for
their frequent use of violence against local populations. Much of the administrative structure
of the judicial system remained in place through the postcolonial period. Since the origins
of the modern prison and accompanying legal system in Nigeria and other former British
colonies are rooted in the use of state policy around labor coercion, what are the long-
term effects, if any, of exposure to these systems on populations’ trust in these institutions?
We use Afrobarometer data on trust in historical legal institutions in Nigeria (e.g., police,
courts, and tax administration) to examine whether past exposure to coercive, ostensibly
economically influenced, colonial prison systems is associated with current trust in legal
institutions. We document a significant reduction in contemporary trust in legal institutions,
and police, in particular, in areas with high historic exposure to colonial imprisonment. The
resulting reduction in trust is specific to legal institutions, with no evident effect of colonial
imprisonment on interpersonal trust in individuals.
We add to several literatures and debates on the effects of historic, particularly colonial
era, institutions on economic development (Lowes and Montero, 2021a; Dell and Olken, 2020;
Michalopoulos and Papaioannou, 2016, 2014; Acemoglu and Robinson, 2001). Our paper
illustrates that “inclusive”, strong formal institutions and laws can be implemented in ways
that actually reduce economic returns to local populations in the short-run, through their
incarceration and use for unpaid labor, and weaken contemporary institutions in the long-
run by lowering trust in the judicial system. While previous work has examined the long-run
impacts of institutions like the slave trade (Nunn, 2008), colonial labor concessions (Dell,
2010; Lowes and Montero, 2021a; Dell and Olken, 2020) and health (Lowes and Montero,
2021b; Alsan and Wanamaker, 2018) on development outcomes, interpersonal trust (Nunn
and Wantchekon, 2011; Okoye, 2021) and trust in modern medicine (Lowes and Montero,
2021b; Alsan and Wanamaker, 2018), our paper breaks new ground by exploring the incentive
effects of colonial prison labor systems and their long-term impacts on societal trust in legal
6

institutions. This kind of exploration is needed, particularly in light of research linking
environments of low trust in legal institutions and low views of state legitimacy with conflict
(Rohner, Thoenig, and Zilibotti, 2013), low domestic investment and higher transaction costs
from weak contract enforcement (Knack and Keefer, 1997), as well as issues such as effective
policing, crime, and law enforcement (O’Flaherty and Sethi, 2019).
We also add to the literature on the economics of forced labor and coercive labor
contracts (Acemoglu and Wolitzky, 2011; Bobonis and Morrow, 2014; Dell, 2010; Gregory and
Lazarev, 2013; Juif and Frankema, 2018; Lowes and Montero, 2021a; Naidu and Yuchtman,
2013; van Waijenburg, 2018; Saleh, 2019; Dippel, Greif, and Trefler, 2020; Sokoloff and
Engerman, 2000). Research in this area has examined the impacts of economic shocks on
coercive contract enforcement (Naidu and Yuchtman, 2013), and estimated the share of
forced labor in colonial public finance (van Waijenburg, 2018). However, there is very little
evidence on the economics of prison labor. Most research on convict labor is concentrated
on the United States and focused on the institution of convict leasing in the 19th century
when Black-Americans, in particular, were economically exploited by the US government,
in concert with private employers, for their labor (Muller, 2018, 2021; Poyker, 2019; Travis,
Western, and Redburn, 2014; Cox, 2010). A key difference between the US convict leasing
system and the use of prison labor in British colonial Nigeria, is that while, prisoners were
primarily used by private employers in the US, prison labor in British colonial Nigeria was
mainly used by government employers, with different resulting implications for the effects
of labor demand shocks on incarceration rates across the two systems (Muller, 2018, 2021;
Poyker, 2019).3Relatedly, we add to the literature on the economics of incarceration (Becker,
1968; Avio, 1998; Katz, Levitt, and Shustorovich, 2003). While previous work has focused
on the effects of crime and prison conditions on incarceration rates and recidivism (Becker,
1968; Freeman, 1999; Bhuller et al., 2020; Katz, Levitt, and Shustorovich, 2003), we highlight
the role of economic shocks in increasing incarceration under coercive state institutions.
The paper is organized as follows: Section 2 provides historical background on prison
labor in colonial Africa. Section 3 reports quantitative estimates of the value of prison labor
to the colonial regime. Section 4 describes the data on incarceration and economic shocks,
3While similar systems of coercive laws, courts and policing worked to coerce Black labor in the US and
Nigeria, the differences in the uses of prison labor led to varying implications for the effects of labor demand
shocks. White private sector employers were able to both affect incarceration rates and draw on imprisoned
Black populations under the US system, in ways that they could not under the British colonial system,
leading to sometimes opposite effects of labor demand shocks on incarceration and the demand for convict
labor across the two systems (Muller, 2018, 2021).
7

and presents the results on the effects of economic shocks on the incarceration rate and
the use of prison labor. Section 5 discusses the links between colonial imprisonment and
contemporary trust in legal institutions. Section 6 concludes.
2 Prison Labor in Colonial Africa
2.1 A History of Forced Labor
Prison labor was a small aspect of a larger regime of domestic forced labor in colonial Africa.
European colonial governments were tasked with pursuing strategies to maximize revenue
while minimizing the cost of administration in Africa (Gardner, 2012). Attempts to raise
revenue to fund expenditures on key public works projects, such as roads and railroads,
which were necessary for both revenue extraction from cash crop exports and expansion of
colonial control, depended crucially on the colonial government’s ability to raise revenue
through direct or indirect taxation, and cut costs associated with spending on capital and
labor. Labor shortages were an endemic feature of African colonies (Okia, 2012; Ash, 2006).
Shortages were partly driven by an unattractive wage labor market for government projects,
which itself was partly spurred by artificially imposed below market wage compensation, set
both as a cost-cutting measure and to prevent competition with the private sector4(Okia,
2012; Maul, 2007; Ofonagoro, 1982).
European colonialists were particularly concerned with the “Africa labor question,”
wherein, faced with the joint realities of labor shortages and colonial objectives to minimize
labor costs and maximize revenues, colonial administrators questioned how much coercion a
“civilized government” could use to attain labor (Cooper, 1996; Buell, 1965). After numerous
colonial forced labor scandals,5the Forced Labor Convention at the 1930 International Labor
Organization (ILO) conference was passed. The Forced Labor Convention prohibited the
use of forced labor for private industry, defining forced labor as “all work or service which is
extracted from any person under the menace of any penalty and for which the said person has
not offered himself voluntarily” (Cooper, 1996).6Crucially, the Convention made exceptions
for the use of forced labor for public works, “penal and communal labor in the public sector
and compulsory military service” (Kunkel, 2018; Killingray, 1989).
4And also to satisfy the economic and political demands of white settler employers in colonies like Kenya
(Okia, 2012).
5Most infamous of which was the torture and murder of millions of Congolese for the rubber extraction
trade under Belgium’s King Leopold (Lowes and Montero, 2021a).
6ILO 29, Article 2 s 2a, c, e, Articles 4 and 5
8

The answer to European colonial administrators’ “Africa labor question” involved the
institution of various coercive labor regimes, enforced by legislation and through the partic-
ipation of local chiefs or Native Administrators. Among these strategies included the use of
direct taxation, such as hut and poll taxes requiring cash payment, to induce Africans into
the colonial wage labor market; the use of labor tax legislation to force Africans to donate
a certain number of hours of often unpaid labor to private and public sector work; and the
use of precolonial communal labor requirements to compel Africans, under the direction of
the chiefs, to provide unpaid labor for private and public works projects (Okia, 2012; Harris,
1914; Trevor, 1936; van Waijenburg, 2018; Cooper, 1996). The consequences for flouting
such legislation were often fines and imprisonment (Okia, 2012). Another important source
of forced labor for the colonial public sector, sanctioned by the ILO Forced Labor Convention,
was convicts (Hynd, 2015; Bernault, 2007).
2.2 The Prison System in British Colonial Nigeria
Prison was not a main feature of judicial punishment prior to colonial rule across much of
Africa (Bernault, 2007).7In British colonial Nigeria, which lasted formally from 1914 to
19608, labor taxes and labor laws worked in concert with Masters and Servants Ordinances,
vagrancy laws, labor registration, pass laws, and Native Authority Ordinances that mandated
the exploitation of African laborers to work on colonial public works projects (Hynd, 2015).9
There was also military conscription, with Africans conscripted as laborers for wartime
construction and general non-combatant labor (Killingray, 1989).10
As previously noted, the consequences for flouting this legislation often included fines
and imprisonment. The goal of the prison system, codified in colonial law with the 1916
Prisons Ordinance, was twofold. First, prisoners worked as punishment for crimes, as defined
by the colonial government; and second, unpaid prisoners were viewed as a source of cheap
labor, particularly for work on public infrastructure projects in the colonies (Adamson, 1984).
In Nigeria, by law, prisoners were only allowed to work for government agencies in the public
sector and not for private sector employers (Kingdon, 1923; Abiodun, 2017; Foreign and
7The popular punishment for transgressions within communities included ridicule ceremonies, community
sanction and exile in the most severe cases (Bernault, 2007; Onoge, 1993).
8Although the British colonial presence in the region dated back to 1860 (Archibong, 2019).
9These laws were widely used throughout colonial Africa, and were very similar to the Black Codes used
to coerce African-American labor in the US in the 19th century (Adamson, 1984).
10Forced labor for military service was permitted under the ILO Forced Labor Convention and also used
extensively during wartime. A notable instance of the use of this labor was during World War II where
thousands of Africans were conscripted to work on various tasks like porters, builders and cleaners, for the
British colonial government (Killingray, 1986).
9

Office, 1947).11 Additionally, prisoners could not be transported across long distances and
were legally bound to work only within their provincial districts (Foreign and Office, 1947).
Under the 1916 ordinance, Nigeria’s colonial prison system was initially administered
as a dual system, with colonial prisons under the direct supervision of colonial government
officials and the management of the Director of Prisons, and Native Authority prisons di-
rectly overseen by the local chiefs and indirectly supervised by colonial government officials
(Kingdon, 1923).12 Figure 2 presents the distribution of colonial prisons, provinces and
regions in Nigeria along with the distribution of the colonial railroad. The colonial rail-
road was primarily constructed between 1900 and 1930, and was functional only through
the end of the colonial period (Okoye, Pongou, and Yokossi, 2019). The railroad, like the
road networks and other colonial public works infrastructure, was largely used to transport
agricultural commodities and mineral resources to the coast for export. Prisoners made up
a substantial part of the labor on colonial public works like the railroad, and prisons were
frequently located along key public works infrastructure like the railroad, as shown in Fig-
ure 2, to minimize costs of transportation, and adhere to the 1916 Ordinance prohibiting
transport of prisoners across provinces (Ekechi, 1989; Foreign and Office, 1960).13 Under the
1916 Order in Council Act, colonial prisons were strictly classified into three types based on
lengths of prison sentence, including convict prisons, with prisoners serving two or more years
to life sentences; provincial prisons, with prisoners serving greater than six months and less
than two-year sentences; and divisional prisons, with prisoners serving less than or equal to
six-month sentences (Kingdon, 1923; Abiodun, 2017). Most prisoners had shorter sentences
of fewer than two years, with 65%-90% of convicts in provincial or divisional prisons (Hynd,
2015).
2.3 Labor Shortages, Public Works and Prison Labor
To address chronic labor shortages and minimize the costs associated with hiring wage labor
to construct and maintain public works like the railroad and roads, colonial governments
regularly used prisoners for labor. Prisoners in colonial Nigeria worked almost entirely on
11This is unlike in another well-known prison labor system, the US convict leasing system in the 19th
century where prisoners were leased to private companies (Muller, 2018).
12There is little historical information on the functioning of the Native Authority prisons, and we refer
to records on colonial prisons in this study. This means the number of colonial prisoners presented here
represent only a fraction of the total number of individuals imprisoned during this period. We discuss this
further in Section 3 and in Appendix A.2. Prisons were also administered by northern and southern regions,
and further detail on this is also provided in Appendix A.2.
13This strategic placement of prisons and prisoners around key infrastructure like the railway is highlighted
in the opening quote from Inspector Jackson.
10

public works projects (Hynd, 2015; Bernault, 2007). In Nigeria, by law, all able-bodied
prisoners were mandated to work, and only prisoners who had been sentenced could work
(Report, 1925).14 Additionally, prisoners were completely unpaid for their labor. Particularly
during periods of labor shortages, prison departments would hire out unpaid prison labor to
other colonial government departments. These departments would then pay a small fee to
the Prisons department based on the assessed skill needed from the prisoner.
Prisoners were largely engaged in unskilled labor, and prison departments classified
prisoners’ labor into three categories of unskilled hard labor, skilled hard labor (sometimes
referred to as industrial labor), and light or domestic labor.15 Unskilled hard labor included
tasks like quarrying, breaking rocks, felling trees, and other activities for road-making and
railway station upkeep, and was largely assigned to prisoners with short-term sentences of
less than six months. Skilled hard labor included tasks like basket-weaving, brick-making,
tailoring and carpentry, and was usually assigned to prisoners with long-term sentences of
greater than two years.16 Light labor was frequently assigned to sick or old prisoners, or
the 6% of female prisoners within the colonial prison system. In southern Nigeria, between
73% and 91% of prisoners were engaged in either hard or light labor between 1920 and
1938.17 Prisoners engaged in hard labor alone constituted over 70% of convicts over the
same period. The top consumers of this unpaid prison labor were government departments,
including Public Works, Railways and Harbors, Public Health, and Education (Hynd, 2015;
Report, 1925).
Prison labor, sourced primarily from prisoners with short-term sentences who made up
the vast majority of the prison population (76% of sentenced prisoners in Nigeria between
1920 and 1938), was invaluable to the construction of colonial public works and infrastruc-
ture18. Their labor significantly contributed to public works projects such as quarries and
14Between 9% and 26% of prisoners were considered “unfit” for work either due to being non-sentenced
debtors, not yet sentenced individuals in custody awaiting trial, or being too sick to work. Source: British
Blue Books, Nigeria, multiple years.
15Source: Annual Report on Prisons in Nigeria, 1940.
16The rationale provided by colonial prison officials was that these prisoners could be taught a skill given
their long sentences, whereas, prisoners with short-term sentences would not have enough time to learn a
skill. Prisoners in this skilled hard labor category produced items like uniforms that were then sold to other
government departments for profit. Prison labor was reserved exclusively for government use, and colonial
officials were careful to choose sectors for convict labor in order to avoid competing with private industries.
Source: Annual report on the Treatment of Offenders, 1947, British Blue Books and Annual Report on
Prisons, 1920-1959
17Source: British colonial Blue Books, multiple years.
18A regular section in the colonial prison reports highlighted the value of prison labor as shown in Figure
A2 in the Appendix.
11

coalfields in southern Nigeria; the Eastern Railway extending from Port-Harcourt in Owerri
province was constructed using large gangs of prison labor, and prisoners engaged in station
upkeep worked across southeastern Nigeria through the 1950s (Hynd, 2015; Ekechi, 1989;
Abiodun, 2017; Foreign and Office, 1960). Prisoner maintenance costs included expenditures
to feed, clothe and house prisoners, the costs of prison staff salaries, and all other expendi-
ture involved in operating the prisons. Tight labor markets worsened labor shortages, and
periods of higher average annual wages were positively correlated with a larger daily number
of people in prisons as shown in Figure 3.19
The recruitment of prisoners to address labor shortages was also explicitly acknowl-
edged by colonial officials. In one infamous account, a British sanitary inspector wrote to
colonial government officials to request increased funds to employ more wage labor. The
officials denied his request, and “The officials asked the prison department to find ways to
either increase the prison population or recruit convicts from outstation prisons to complete
the tasks”.20 In another example, in the 1916 Annual Report on Prisons, the Inspector of
Prisons, W.H. Beverley, lists two main reasons for creating categories of prisons according
to prison sentence:
“(a) to place ‘special prisons’ in townships which are on good lines of commu-
nication and afford the most suitable description of penal labour. (Abeokuta,
Enugu, Lagos, and Port Harcourt, on the eastern and western lines of the Nige-
rian Railway, provide quarrying, industrial work, labour connected with shipping
and transport, etc.)” and (b) “the ensuring, as far as possible, of an automatic
and constant supply of prisoners to each class of prisons. At the end of the
year, the system appeared to be working well; the prison population was evenly
distributed, and nowhere was there shortage of convict labour.”
So significant was the role of prison labor in colonial fiscal accounting, that in 1911,
the Governor of Northern Nigeria remarked that “The value (calculated at 2/3 of the market
rate) of prisoners’ labor in connection with public works, which would otherwise have had
to be paid for in cash was 3,878 pounds. If calculated at the ordinary market rates the value
of the prisoners’ useful labor would have exceeded the entire cost of the Prison Department”
(Salau, 2015). The use of prison labor to address labor shortages in the public works sector
19The correlation between the daily average numbers in prison and the average annual wage to unskilled
laborers is 0.87, p < .001. We discuss these trends in detail in Section 3.
20NAI, CSO 26/2 09591 Vol.1 ‘Lieutenant Governor Southern Province to Resident Calabar Province:
Memorandum on Prison labor’ 23rd April 1923.
12

continued through the end of the colonial period. By the year of independence in 1960, both
incarceration rates and the use of prison labor had diminished, as shown in Figure 4. As of
1938, 0.2% of the local population was incarcerated; that share had fallen to 0.06% of the
population by 1995. While the laws allowing prison labor for government use remained in
place through the postcolonial period,21 the discovery of oil in 1956, followed by an oil price
boom in the 1970s, altered Nigeria’s tax revenue structure, and the majority of government
revenue transitioned from the agricultural commodities of the colonial period to direct taxes
from petroleum.22 . With the transformation to a capital-intensive oil revenue base, the need
for a large base of unskilled labor for public works construction and maintenance with the
goal of agricultural commodity extraction declined in the postcolonial period. This change
in the government revenue base, along with increased protests from labor unions, led to a
decline in the use of prison labor in postcolonial Nigeria (Killingray, 1999; Abiodun, 2017).
2.4 Crime and Punishment
Although there is limited disaggregated data on the types of crimes individuals were con-
victed of during the colonial period, available data from colonial records in Nigeria show
that, on average, 61% of total convictions in colonial courts were from “offences against
revenue laws, municipal, road and other laws relating to social economy of the colony” be-
tween 1920 and 1939, as shown in Figure 5.23 These offenses included defaulting on tax
payments, and violations of vagrancy, labor ordinance, township ordinance (which restricted
the movement of Africans in the provinces), and other similar, minor, “misdemeanor” laws.24
Combined with “miscellaneous minor offences” which included crimes like “drunk and disor-
derly”, “witchcraft,” and other crimes against “public morality”, these minor misdemeanor
offenses accounted for 76% of colonial court convictions between 1920 and 1939. Fines were
most frequently assigned as the main punishment for these misdemeanor crimes, with im-
prisonment sometimes assigned as punishment, or assigned for individuals who were unable
to pay the fines, at the discretion of the colonial court magistrate (Hynd, 2015). An impor-
21Prison labor was never completely abolished as a feature of imprisonment in Nigeria, and prison labor as
mandated by constitutions in Nigeria since independence (e.g., the 1979 and 1999 constitutions), continues
to be a feature in prison (Shajobi-Ibikunle, 2014). More recent constitutional changes have implemented
stricter mandates regarding the work of prisoners, in many cases, limiting work to within the prisons, with
more rhetoric around the rehabilitation and reformation of prisoners under The Nigerian Prisons Service and
Article 9 of the 1972 Prison Act (Tanimu, 2010).
22As shown in Figure A1 in the Appendix.
23Source: British colonial Blue Books, multiple years. There are no disaggregated crime data by the
categories listed in the colonial records between 1940 and 1960.
24Source: “Policing in Lagos and Provinces, 1899-1929”. Reference: 73242C-01; “Judicial and Police,
1899-1960”, British Foreign and Commonwealth Office.
13

tant lever for colonial governments seeking to increase the prison labor base was to increase
prosecutions of the aforementioned minor offenses, and switch punishments from fines to im-
prisonment (Hynd, 2015; Ojomo and Alemika, 1993). In contrast, in the postcolonial period,
when prison labor was no longer used intensively by governments, a major share of crimes
(46%, on average, between 1977 and 1993) prosecuted for incarceration were for property
theft (Figure 5).
2.5 Policing and Enforcement
The judicial system in colonial Nigeria was highly centralized, with the colonial courts work-
ing in concert with prison officials and police to “maintain law and order”, protect European
property, and meet the revenue imperative objectives of the colonial government25 (Rotimi,
1993; Onoge, 1993). The colonial police force was headed by colonial officials, with rank
and file constables, constituting the majority of the base, recruited from local populations
across the country.26 The police were often the first point of contact with the judicial system
for local populations and were especially involved in enforcing tax collection and frequently
involved in tax raids and arrests associated with violations of colonial crimes (Rotimi, 1993;
Onoge, 1993). The police were notoriously feared for their frequent use of violence against
local populations.27 The 1930s witnessed multiple public protests against police brutality.
One infamous protest against the use of police in tax collection, included reports from ob-
servers like a local reverend commenting on being horrified by the sight of women “hunted
down and dragged about on public streets” by police for defaulting on tax payment (Rotimi,
1993). Much of the administrative structure of the police force remained in place through the
postcolonial period. The police in Nigeria, and across much of colonial Africa, where similar
systems were implemented, remain a national body, and there have been multiple critiques
from scholars highlighting “tremendous continuity in the country’s policing traditions and
goals in spite of a series of organizational reforms” (Ojomo and Alemika, 1993; Sanny and
25The colonial police was administered under 2 separate, but jointly legislated forces in northern and
southern Nigeria and amalgamated into a national police force in 1930.
26A popular strategy of colonial officials was to divide police from local populations by hiring rank and file
officers from separate ethnic regions of the country. The strategy was elaborated by colonial official Freeman
in a letter to the Duke of Newcastle in December 1863, wherein he argued that raising a police force for
Lagos (which had a majority Yoruba ethnic population) from mainly Hausas would make it difficult for a
rapport to develop between the police and the people of Lagos against whom they were to enforce repressive
laws (Ojomo and Alemika, 1993).
27Colonial police fear as a feature of local memory is reflected in local songs like the ones from the Urhobo
ethnic group recalling the humiliation of men who fell victim to police ambush during tax raids and were
subsequently locked up in police cells. The songs crystallized colonial policemen as ‘terroristic bogeys’ in
local communities (Onoge, 1993).
14

Logan, 2020). These critiques have also linked continuity in coercive policing practices to
relatively high levels of mistrust in police in these regions today (Sanny and Logan, 2020).
3 Estimating the Value of Prison Labor
The qualitative accounts from colonial records, described in Section 2, highlight the value
of prison labor in Nigeria. To understand why prison labor was so valuable to the colonial
regime, we quantitatively estimate the value of unpaid prison labor. In essence, we ask, “how
much would the colonial state have had to pay if they had to hire non-remunerated prison
workers for a market rate wage?” To assess this value of unpaid prison labor, we digitized
archival records on prison populations, wages, public works expenditure, and revenue from
the British colonial Blue Books and Annual Report on the Administration of the Prisons
Department28 between 1920 and 1959. The Blue Books present statistical returns that
governors of British dependencies were required to submit on an annual basis and report a
complete record of prisons and colonial public finance in Nigeria between 1920 and 1959.29
The Blue Books and the Annual Report also include qualitative descriptions of the activities
undertaken by prison departments, as reported by the Director of Prisons30. These data
sources and the variables we use in our analysis are described in detail in Appendix A.1. As
noted in Section 2.2, the colonial prisons data represent only a fraction of the overall prison
population in Nigeria.31 The results from this analysis provide us with an estimate of the
size of the prison labor system in colonial Nigeria.
3.1 Empirical Strategy
We measure the value of prison labor to the colonial regime, referencing van Waijenburg
(2018), to estimate the value of unpaid wages to prisoners, and its relative share in expen-
diture on new construction of colonial public works.32
28Subsequently referred to as the Annual Report .
29Nigeria was amalgamated from separate regions into a single country in 1914 and although the Blue
Books data extend back to 1914, some information is missing between 1914 and 1920; thus, we start our
analysis in 1920 for completeness. The Blue Books data on prisons and public finance ends in 1938. For prison
data after 1938, we use records from the Annual Report on the Administration of the Prisons Department.
30An example of the archival data is presented in Figure A3 in the Appendix.
31No detailed data on the Native prisons administered by local chiefs was introduced in the colonial
archives prior to 1940. Available data on Native prisons in the annual reports from 1940 show that the
addition of Native prison estimates to the colonial estimates presented in this paper would almost double
the incarceration rate in 1940, from around 224 per 100,000 population to 399 per 100,000 population. This
suggests that the data we present here from 1920 to 1959 is likely an underestimation of the total level of
incarceration, and the total value of prison labor, during this period. We provide further discussion on this
in Appendix A.2.1.
32We use expenditure on new public works construction only here as a comparison as it reflects value-
adding investment in productive public works rather than just upkeep or maintenance. New expenditure
15

We calculate the overall value of unpaid wages of prison laborers in each year tas
follows:
Value of prison labort=Annual wagest×1
N
N
X
n=1
Prisonersnt (1)
This elicits an overall, gross value of benefits accruing to government consumers of
prison labor. As a measure of wages, we use the average annual market wages paid to
unskilled laborers. The wage measure comes from the colonial Blue Books which reports
annual wages paid to people classified as “Labourers and Carriers” and other “Unskilled
Labour”. This captures the wages for some of the types of work that prisoners were required
to perform, including felling trees and breaking rocks to clear areas for road and railroad
construction, as discussed in Section 2. P risoner snt is the daily average number of people
in prisons over ndays in the year from the archival records.33 This measure captures the
amount of convict labor that was available on any given day. As discussed in Section 2,
prison officials measured the value of prison labor based on the funds generated from hiring
out prisoners to other government departments for a small per diem fee.34 We compare our
estimated total value of prison labor to the colonial reported value of prison labor in the
results in Section 3.2.
The specification in Equation 1 does not factor in the costs of prisoner maintenance,
including food, clothing, housing, and prison staff salaries. The archival data report two
sets of costs for prisoner maintenance, including (i) food, which was reported as the main
represents about 40% of total, new and maintenance, public works expenditure between 1920 and 1959.
In Appendix A.3.3, we compare the value of prison labor figures to total public works spending, including
recurrent expenditure on regular maintenance of public works reported in the archives.
33Nis the total amount of prison days in the year recorded in the prison data. While the exact value
of Nis not explicitly listed in the archival data, we use the explicitly recorded ‘daily average number of
prisoners’ category in the colonial archives. A snapshot of the description of this category from the 1925
Annual Report on the Prisons Department, Southern Provinces, is shown in Figure A9 in the Appendix.
34For example, the Directors of Prisons, W.H. Beverly, E. Jackson, or W. Reeder in the southern provinces
from 1915 to 1921, recorded per diem estimates of the value of labor between 1916 and 1921 in the Lagos
colony and southern provinces for Nigeria. Using the classification of labor into skilled hard labor, unskilled
hard labor and light labor, described in Section 2, hard labor, both unskilled and skilled, was given a value
of five pence per day, with light labor given a value of three pence per day in 1916. Starting in 1917, skilled
hard labor was given a value of one shilling and six pence or 18 pence, unskilled hard labor was assigned a
value of five pence and light labor was assigned a value of three pence. The rates for unskilled hard labor
stay the same from 1918 through 1921, with no reporting on the exact value assigned to skilled hard labor
or light labor over this time. After 1921, the reports no longer included information on the per diem value
assigned to the different classes of labor.
16

cost of prisoner upkeep, and (ii) total prisoner maintenance costs, a measure that includes
all expenses involved in operating the prisons (i.e., everything from food costs, to staff
salaries, costs of transporting prisoners, equipment purchases, uniforms for staff, and any
other spending on prisons).35 Food cost represents an average of 35% of the total prisoner
cost for 1920 to 1959, ranging from 27% to 51% of the total prisoner costs over the study
period. Food cost and staff salaries accounted for over 50% of the total prisoner costs from
1920 to 1959. The total prisoner maintenance cost is the most expansive measure of the
prison upkeep cost. The net value of prison labor is the difference between the total value
of prison labor in Equation 1 and total prisoner maintenance costs. To estimate the relative
value of prison labor, we divide the results from Equation 1 by public works expenditures,
prison expenditures and overall expenditure figures from the Blue Books. We present the
results on the net and relative values of prison labor in Section 3.3.
Figure 3a shows the trends in the reported average annual wage and prisoner food and
overall maintenance costs. The total reported prisoner upkeep cost closely tracks the wage,
reflecting increases in staff salaries over time, with a steep increase after 1940. Prisoner food
cost follows a similar pattern, although the post-1940 increase in cost is less steep than the
wage and total prisoner cost. Figure 3b shows the daily average number of prisoners over
the study period. Wages remained above prisoner food costs in all years, and above total
prisoner costs in over 51% of the years between 1920 and 1959. The daily average number in
prison fluctuated between 1920 and 1940, increasing through 1930, then decreasing between
1930 and 1940, before sharply increasing after 1943. Interestingly, the daily average number
of prisoners also appears to track the average annual wage in Figure 3a. There is a positive
correlation (0.87, p < .001) between the daily average numbers in prison and the average
annual wage to unskilled laborers.
We estimate various versions of Equation 1 in alternate specifications, including esti-
mates using alternate wage measures, adjusting for inflation, and addressing any potential
bias in prisoner estimates by computing a weighted average measure of people committed
to prison for penal imprisonment in each year. The trends in the results remain unchanged
and are detailed in Appendix A.3.
35An example of one breakdown of these costs over 1919 to 1921 from the colonial archives for prisons in
the southern provinces is shown in Figure A11 in Appendix A.3.
17

3.2 Value of Prison Labor Results
Figure 6 presents our estimates of the total gross value of prison labor based on Equation 1,
along with a comparison to the colonial prison officials’ own reports of the value of prison
labor based on fees remitted to the Prison department for prisoners’ labor.36 While our
estimates of the value of prison labor are consistently higher than the colonial governments’
own reports, both measures follow similar trends, and the values are close to each other prior
to 1945. There is a positive correlation (0.7, p < 0.001) between our estimates and the colonial
reported values of prison labor. The estimated total gross value of prison labor starts out
around 178,498 pounds in 1920 and fluctuates- first decreasing, and then increasing through
1927, before mostly declining through 1943, then increasing sharply afterward, peaking at
1,532,634 pounds in 1959.37 The average estimated gross value of prison labor is 313,742
pounds over the colonial period. Although prisoners were not paid, the exact amount of
the payment remitted to the Prisons department from other government agencies for their
labor was recorded for the southern provinces in a few years between 1919 and 1925. These
payments were the per diem prices set by the Prisons department for a prisoner’s labor based
on the level of skilled labor required, as discussed in Section 2.3. We compile these estimates,
and compare these prisoner prices with the daily market wage rate for similarly skilled
workers in the southern provinces.38 Prisoners performing unskilled hard labor, who made
up the majority of the prison population, as discussed in Section 2, were assigned a value
between 60%-80% below the market wage rate. Colonial prison officials were consistently
undervaluing prisoners’ labor to keep administration costs for peer government departments
low, while attempting to balance budgets.39
3.3 Prisoner Costs and Relative Value of Prison Labor
We calculate the difference between the total value of prison labor and the total prisoner
maintenance costs, or the net value of prison labor, and compare these estimates to the
colonial government’s spending on public works, prisons and overall colonial expenditure.40
36We provide more detail and numbers in Table A2 of Appendix A.3.
37Given the debates around the choice of the price index for colonial Africa, we present the figures in
nominal terms here (Frankema and Van Waijenburg, 2012). We present the real estimates in Appendix A.3,
and the trends remain unchanged.
38The estimates are shown in Figure A10 of Appendix A.3.
39This is confirmed in the report written by Prison Inspector Beverley in the 1915 Annual Report on
Prisons, in which he states that values assigned to prisoners’ labor are below “wages demanded by workmen
in civil life”. He recommends a doubling of values to balance prison expenditure amounts, illustrating the
balance sheet calculus that appeared to drive the setting of prison labor values.
40There is still a positive correlation (0.5, p < 0.01) between our net value estimates and the colonial
reported value of prison labor described in Section 3.2 as shown in Figure A12 in the Appendix.
18

Table 1 reports the estimates as decadal averages from 1920 to 1959.41 Even after sub-
tracting out the extensive measures of prisoner maintenance costs described in Section 3.1,
the net value of prison labor is nonnegative and strictly positive in 60% and 57% of years,
respectively, in colonial Nigeria.
Over the four decades of colonial data available, the net value of prison labor, less
prisoner food costs, remains strictly positive, with an average of 195,260 pounds. When we
estimate the net value of prison labor using all prisoner maintenance costs reported, the
mean falls to 31,674 pounds. The only decade in which the net value of prison labor, less
total prison costs, becomes negative (at -16,891 pounds) is during the World War II period
between 1941 and 1950, when, as discussed in Section 2.2, there was a relative increase in the
use of military conscript labor. The net value figure increases notably (to 160,805 pounds) in
the post-war period from 1951 to 1959, following a flurry of public works construction activity
right before Nigeria’s independence in 1960. On average, the gross value of prison labor or
unpaid wages to prison laborers was more than double the amount spent by the colonial
government on public works between 1920 and 1959. After adjusting for extensive measures
of prisoner maintenance costs, the share of the net value of prison labor in colonial public
works expenditure remains economically significant, with a mean of 5% and a maximum of
up to 42% during this period.
We show similar trends for the share of prison labor in total prison expenditures and
overall colonial expenditures over this period in Table 1. Given the relatively small share
of new public works expenditure in overall colonial spending,42 the prison labor share in
the overall colonial expenditure was low, constituting an average of 2% and 0.1% of total
expenditure, using the gross and net values of prison labor (including total prisoner main-
tenance costs), respectively. The quantitative results support the qualitative accounts from
the historical records that prison labor, measured as the value of unpaid wages to prison
laborers, was economically valuable to the colonial regime.43
41We provide more detail and numbers in Table A2 of Appendix A.3.
42An average of 2.2% between 1920 and 1959.
43Furthermore, these estimates of the value of prison labor may be further underestimated if there are
potential spillovers to private sector wages, and market wages are depressed by the supply of unpaid prison
laborers.
19

4 The Effects of Economic Shocks on Incarceration Rates and the
Use of Prison Labor
4.1 Conceptual Framework
The historical accounts and quantitative results in Section 3 show that prison labor was an
important resource used by colonial governments to address labor shortages in the public
works sector. Prisoners’ labor was valued for work on infrastructure projects like roads and
the railroad, which were needed to extract agricultural commodities from the interior of the
colony to the coast for export. The next step in our analysis is to estimate the effects of
labor demand shocks on the use of prison labor. To fix ideas regarding the links between
labor shortages and the use of prison labor, we adapt insights from a recent theoretical
literature on the effects of economic shocks on labor demand and coercion under forced
labor institutions (Acemoglu and Wolitzky, 2011; Naidu and Yuchtman, 2013), and outline
a simple conceptual framework as follows.
We highlight two main predictions on the effects of economic shocks on incarceration
rates under forced labor institutions during the colonial period versus non-forced labor in-
stitutions over the postcolonial period. The economic setting in both periods is primarily
agricultural, with the majority of the local population employed in the agricultural sector.44
In this setting, there are two types of shocks that directly affect agricultural revenue or
surplus, including shocks to quantity and shocks to price. Positive (negative) shocks that
directly increase (decrease) the quantity of agricultural output or crops, increase (decrease)
the demand for labor in the agricultural sector, and can lead to an increase (decrease) in
wages in this sector. Similarly, exogenous shocks that raise the prices of agricultural output
can also increase wages in the agricultural sector by creating demand for the output and
hence associated labor demand. These labor demand shocks can increase or decrease wages
in the agricultural sector.
A simple principal-agent framework (with the African worker as the agent and the
colonial government as the principal), wherein the assumption that the principal must pay
higher wages to induce more effort on the part of the agent does not hold when the principal
can coerce the agent to work. We assume there is excess demand in the labor market and
the principal is a cost minimizer with a preference for lower wages, following the historical
44This is the case in Nigeria, where a major share of workers is employed in agriculture. Estimates range
between 37% and 70% as of 2016 by World Bank and Food and Agriculture Organization (FAO) statistics,
respectively.
20

account in Section 2. We also assume that the colonial government, P, can coerce the African
worker, A.Achooses whether or not to work for Prelative to the reservation wage in the
agricultural sector, w. The wage contract offered by P,w∗, must be greater than wto attract
A’s labor. If w∗< w, the agent chooses not to work for the principal. In this scenario, P
can choose to coerce Ato work by, for example, increasing the incarceration rate and the
use of prison labor. Following the historical account in Section 2, the ways the colonial
government could and did this were manifold and included increased prosecution and arrests
for so-called “crimes against the social economy of the colony” or minor, misdemeanor crimes
like vagrancy, labor ordinance and breach of peace violations and switching punishments for
these minor crimes from fines to incarceration. Pchooses coercion if the benefits of coercion
(in the form of increasing output revenue and closing the excess demand gap) outweigh the
costs of coercion (e.g., enforcement, risk of riots, and conflict with local populations) or the
net benefits are positive. A key insight here is that the colonial government was able to reduce
the costs of coercion through the use of incarceration and prison labor with an expansive
definition of what constituted a criminal act, and a centralized system of enforcement and
punishment for these crimes as discussed in Section 2.
The first prediction from this framework is that positive economic or labor demand
shocks, in the form of rising agricultural commodity export prices, or higher rainfall that
increases agricultural output and associated wages of agricultural workers, will increase in-
carceration rates and the use of prison labor under forced labor regimes such as that of the
colonial period. Symmetrically, negative shocks, will, all else being equal, have the opposite
effect and reduce the demand for prison labor under these regimes.
Under non-forced labor institutions, positive economic shocks that increase agricultural
wages or workers’ reservation wage, also increase the opportunity cost of participating in
economic crimes, like property theft or related assault following previous models from the
economics of crime literature (Freeman, 1999; Becker, 1968). Conversely, negative economic
shocks that reduce wages, increase the likelihood of participating in economic crimes. This
results in the second prediction, that negative shocks will increase incarceration rates under
non-forced/non prison labor regimes, as in the postcolonial period in Nigeria. We test the
predictions of this framework using the data on incarceration rates and economic shocks
outlined in Section 4.2.
21

4.2 Description of Data
4.2.1 Incarceration Rates
To assess the effects of economic shocks on incarceration and the use of prison labor, we
digitized 65 years of archival data on prisons from 1920 to 1995. Available disaggregated
data on incarceration rates at the subnational level spans the colonial period (1920-1938) and
the postcolonial period (1971-1995). The Blue Books report incarceration data at the prison
level, and we aggregate up to the district level, where the district is the colonial province
between 1920 and 1938. We calculate the incarceration rate as the number of newly admitted
prisoners per 100,000 population for each province in each year.45 The incarceration data are
broken down by length of prison sentence, which is classified as prisoners with short-term (less
than six months), medium-term (between six months and two years) and long-term (greater
than two years) sentences. We also assemble available data on postcolonial incarceration
rates at the current administrative state level between 1971 and 1995 from Nigeria’s Annual
Abstract of Statistics.46
Table 2 presents the summary statistics. The average incarceration rate falls by almost
a third between the colonial and postcolonial periods, from around 241 prisoners per 100,000
people to 92, as shown in Figure 4. The spatial distribution of incarceration between the
colonial and postcolonial period also significantly changed, with prisoners being clustered in
the southern provinces over the colonial period, and considerably more spatial dispersion in
the postcolonial period, as shown in Figure 7. Short-term prisoners made up the majority
of the colonial prison population, at 53% of all newly committed prisoners and 76% of
penal imprisonment, on average, between 1920 and 1938, as shown in Table 247. The share
of long-term prisoners in penal imprisonment was comparatively smaller, at 11% over the
same period. The share of prisoners with previous convictions was similarly low, with 11% of
prisoners having one previous conviction and only 2% of prisoners with two or three previous
convictions.
45The population data also comes from the Blue Books, and we use population of provinces in 1939 to
calculate incarceration rates from 1920 to 1938. We discuss the population estimates in further detail in
Appendix A.5.3.
46The postcolonial data do not include a similar breakdown by sentence.
47The incarceration rates by sentence in colonial Nigeria are shown in Figure A5 in the Appendix.
22

4.2.2 Economic Shocks
Cash Crop Export Prices
We use two different measures of economic shock; agricultural commodity export prices and
rainfall to capture economic shocks in a primarily agricultural setting.48 The first measure of
productivity shocks we use is agricultural commodity export prices. The measure uses data
on the major cash crop exports in colonial Nigeria, which include cocoa, palm oil and ground-
nuts; the data are global export prices from the Wageningen University African Commodity
Trade Database (Frankema, Williamson, and Woltjer, 2018). Altogether, exports of palm
products, cocoa and groundnuts accounted for 93% of the volume of agricultural commodity
exports and 78% of total exports in Nigeria over the colonial period. We combine the price
data with land suitability and crop production data from the Global Agro-Ecological Zones
and Blue Books databases, respectively, to identify which prices would have theoretically
affected which districts. Figure 7 presents the spatial distribution of cash crop production,
along with a time series of export prices over the colonial study period. Palm oil and cocoa
are produced in the southern provinces, while groundnuts are the major cash crop export
produced in the northern provinces. Prices for cash crops in the southern provinces, namely
cocoa and palm oil, were two times and one-and-a-half times higher, respectively, than prices
for groundnuts produced in the northern provinces from 1920 to 1938. The most productive
cash crops over the colonial period, by price, were palm oil and cocoa. Palm oil was partic-
ularly valuable, given the relatively high share of provinces (29%) involved in its production
(Table 2). It also had the highest volume of trade of the three cash crops over the colonial
period.49
Rainfall
The second measure of productivity shocks we use is rainfall. A major share of workers in
Nigeria are employed in agriculture, which has remained the case for the past few decades.50
Agriculture in Nigeria is primarily rain-fed, with irrigated agriculture accounting for only
1% of cultivated area in the country, and government investment in agriculture has re-
mained relatively stagnant, at 1% of total government expenditure since 1920 (Xie, You,
48The share of agriculture in Nigeria’s GDP has ranged between 40% and 60% between 1960 and 2012 by
some estimates (Ahungwa, Haruna, and Abdusalam, 2014).
49Between 1863 and 1947, 25% of the value of agricultural commodity exports came from palm oil, and
the figure rises to 61% when palm kernels, a byproduct of palm oil production, are included (Frankema,
Williamson, and Woltjer, 2018).
50Estimates range between 37% and 70% as of 2016 by World Bank and Food and Agriculture Organization
(FAO) statistics respectively.
23

and Takeshima, 2017).51 This combination of facts suggests that the economic conditions
of domestic populations are sensitive to sudden, unexpected changes or deviations in rain-
fall that may reduce crop yields and respective agricultural incomes (e.g., through droughts
or floods). For the colonial period, we use rainfall data from 69 weather stations recorded
in the Blue Books to construct measures of rainfall deviations, or z-scores, as deviations
from the district or colonial province long-term mean.52 For the postcolonial period, we use
precipitation data from the NASA MERRA-2 database, and calculate rainfall deviations as
deviations from the district or postcolonial administrative state long-term mean.53
4.3 Empirical Strategy and Results
To test the predictions of the conceptual framework in Section 4.1, we use three main esti-
mating equations: (1) a specification that identifies the effects of productivity shocks with an
interaction term for agricultural export commodity prices; (2a) a nonlinear, quadratic speci-
fication that allows the effect of rainfall shocks on incarceration to vary more flexibly with the
level of district-level rainfall deviation, and estimates the effects of positive economic shocks
on incarceration rates; and (2b) a specification that identifies the effects of moderate posi-
tive, productivity enhancing, rainfall shocks, on incarceration. We include district (province
or current state for colonial or postcolonial data respectively) and year fixed effects in all
specifications, along with clustered standard errors at the district level. Following Cameron,
Gelbach, and Miller (2008), we apply wild bootstrap-based tests to our estimates to account
for potentially low numbers of clusters in estimating our standard errors, and include wild
cluster bootstrap p-values in our results. The rationale behind each empirical strategy is
discussed in further detail in Section 4.3.1 and Section 4.3.2. Our main specification will
be model (1) that estimates the effects of directly, easily observable (to colonial officials)
agricultural export prices on colonial incarceration rates, although we interpret the results
from all three models here. The rainfall specification allows us to: (a) test the effects of
positive versus negative economic shocks on incarceration rates; and (b) test the effects of
rainfall shocks on incarceration in the postcolonial period as well.
51As shown in Figure A16 in Appendix A.4.
52In alternate specifications, we test results with interpolated data from the University of Delaware
database, and confirm that while there is a significant positive correlation between the rainfall values, the
correlation is low and does not translate to the z-scores which are the main explanatory variable used here.
Given that the Delaware values from 1920 offer fewer fine interpolations than the weather station data, we
use the weather station data here for our main results.
53The NASA MERRA-2 data is not available prior to 1980. The dataset is viewed as the gold standard
for climate/weather analysis among climate researchers (Gelaro et al., 2017).
24

4.3.1 Cash Crop Export Prices and Incarceration Rates
Do higher agricultural commodity export prices that increase agricultural output and asso-
ciated agricultural wages of workers, also increase incarceration rates and the use of prison
labor under the colonial, prison labor regime? To answer this question, following previous
specifications in the literature (Dube and Vargas, 2013; Naidu and Yuchtman, 2013), we
estimate equations of the following form:
Prisonersit =
3
X
c=1
γcCash Cropci ×Cash Crop Pricect +µi+δt+it (2)
where P risonersit is the incarceration rate or number of newly committed prisoners per
100,000 population54 in the colonial district or province iat year t; Cash Cropci is an
indicator that equals 1 if province iproduces one of the three major export cash crops
c∈(palmoil,cocoa,groundnut)over the colonial period, and Cash Crop Pricect is the nat-
ural log of the export price of cin year t; and µiand δtare district and year fixed effects
respectively. Errors are clustered at the district level to allow for arbitrary correlations.55
The coefficient of interest is the interaction term γcwhich measures the effect of increases
in cash crop prices in producing provinces on the incarceration rate.
Cash Crop Export Prices and Incarceration Rates Results
Figure 8 shows the coefficients from individual regressions of short-term incarceration on colo-
nial province and year fixed effects and the interaction between an agricultural commodity
presence variable and year fixed effects. The figure shows the positive relationship between
the prices of the most productive colonial cash crops, palm oil and cocoa, and short-term
colonial incarceration; the positive correlation is particularly visible for palm oil, which had
the highest volume of trade of the three cash crops over the colonial period. Table 3 presents
the results from Equation 2 on the effects of cash crop export prices on colonial incarceration
rates. The results show that the effect of plausibly exogenous positive agricultural export
price shocks signaling increases in agricultural productivity on colonial incarceration rates
and the use of prison labor is concentrated in relatively higher value cash crops, like palm
oil as discussed in Section 4.2.2.56
54The results remain unchanged if we standardize by the adult population only.
55We estimate all models with standard errors clustered at the district level and Conley standard errors
with a cut-off window of 100 km to account for spatial auto-correlation (Conley, 1999). The results are
robust to both specifications, and we present the district level clustering results here.
56There is no correlation between domestic rainfall shocks and agricultural export prices, as discussed in
Appendix A.4.
25

We interpret the coefficients from the full, robust specification of the model in column
(1) of Table 3, with short-term incarceration rates as the outcome of interest. A 10% increase
in palm oil prices is associated with an increase in the short-term incarceration rate by around
7 per 100,000 people, or a 5% increase in short-term incarceration relative to the sample
mean in palm oil producing regions. Short-term colonial incarceration rates increased in
response to higher palm oil prices that signaled increases in agricultural productivity. There
is no effect of palm oil prices on long-term incarceration rates in palm oil producing areas
in column (5). The results for short-term incarceration are similar for cocoa, another high
value crop, in column (1), although the effects are larger and more robust for palm oil, the
most valuable colonial cash crop over this period.
4.3.2 Rainfall Shocks and Incarceration Rates
Nonlinear Effects of Economic Shocks on Incarceration Rates
Following the historical accounts in Section 2, the conceptual framework in Section 4.1,
predicts that positive economic or labor demand shocks, in the form of rising agricultural
commodity export prices, or higher rainfall that increased agricultural output and associated
wages of agricultural workers, will increase incarceration rates and the use of prison labor
under forced labor regimes, as in the colonial period. Conversely, negative shocks will increase
incarceration rates under non-forced/non prison labor regimes, as in the postcolonial period
in Nigeria. Following the framework, one hypothesis is that the main functional form of the
relationship between rainfall deviation and incarceration rates in the colonial period is an
inverted-U. The demand for prison labor peaks during periods of moderate positive rainfall
shocks which increase agricultural productivity. In contrast, extremes in rainfall deviations,
like droughts and floods that lower agricultural productivity, lower the demand for prison
labor. As a falsification test, these effects should only hold for short-term incarceration,
which was more elastic and should be more responsive to short-term economic shocks than
long-term imprisonment.
A further, testable implication of the framework is that, as a falsification test, the
effect of rainfall shocks on incarceration rates should be U-shaped if a major motive for state
incarceration is not prison labor. Under a non-convict labor motivated prison system, such as
postcolonial Nigeria, droughts and floods that lower agricultural productivity should increase
incarceration rates through a rise in economic crimes, like theft, as outlined in Section 4.1.
We can then estimate the causal effect of rainfall shocks on incarceration rates by
assessing panel regressions of the following nonlinear, quadratic form:
26

Prisonersit =β1RainfallDevit +β2RainfallDev2
it +µi+δt+it (3)
where RainfallDevit is the rainfall deviation or z-score for each district in each year relative
to the district’s long-term expectation57 . Our key parameter of interest is β2which should
be significantly negative if the inverted-U hypothesis holds and positive if the U-shaped
hypothesis holds.
Identifying the Effects of Positive Productivity Shocks on Incarceration Rates
While Equation 3 allows us to identify the effects of rainfall shocks on incarceration rates
and the use of colonial prison labor more flexibly, it does not allow us to distinguish between
positive and negative productivity shocks. Specifically, Equation 3 does not allow us to
distinguish between moderate positive rainfall shocks that signal increases in agricultural
productivity, and extreme positive and negative shocks that respectively signal floods and
droughts that can reduce productivity.
Since data on agricultural output from the colonial period are not available, we adapt
definitions of rainfall shocks in Africa from the literature (Dillon, McGee, and Oseni, 2015;
Amare et al., 2018; Jensen, 2000) and estimate transition points in Equation 3 from non-
parametric loess models linking rainfall deviations to colonial incarceration rates. From the
transition points, we distinguish between moderate positive shocks, extreme positive shocks,
and extreme negative shocks as follows: (a) Positive shock (M), where “M” is moderate,
is an indicator equal to 1 if 0 < Rainf allDevit <0.75 and a proxy for increases in agricul-
tural productivity; (b) Positive shock (E), where ‘E” is extreme, is an indicator equal to 1
if RainfallDevit >0.75, and signifies floods that reduce agricultural productivity and (c)
Negative shock (E), is an indicator equal to 1 if RainfallDevit <−0.5, and signifies droughts
that also reduce agricultural productivity.
We can then directly estimate the causal effect of moderate positive rainfall shocks on
incarceration rates by estimating the following specification:
Prisonersit =αPositive shock (M)it +E0
itγ+µi+δt+it (4)
where Positive shock (M)it is the moderate positive rainfall shock. The main parameter
of interest in Equation 4 is α, defined as the effect of moderate positive shocks that in-
57We find no effects when we test the specification using lagged rainfall deviations instead following results
in previous literature (Amare et al., 2018). The results are discussed in Appendix A8.
27

crease agricultural productivity on the incarceration rate. We include vectors of the extreme
positive and negative rainfall shock variables, E0
it, to check the robustness of our results.
A Note on Rainfall and Crop Yields
A key assumption motivating the empirical strategy in Equations 3 and 4 is that rainfall de-
viations have a causal effect on crop yields, and that this effect is nonlinear with extremes in
rainfall, like droughts or floods, resulting in a decrease in crop yields or agricultural output.
The change in crop yield changes the demand for labor and corresponding agricultural wages
as outlined in the conceptual framework. This labor demand shock is what affects incarcera-
tion rates and the demand for prison labor over the colonial period. There is a robust litera-
ture on the nonlinear relationship between rainfall and agricultural output (Dell, Jones, and
Olken, 2014; Lesk, Rowhani, and Ramankutty, 2016; Sarsons, 2015; Kaur, 2019; Jayachan-
dran, 2006; Fishman, 2016; Lesk, Coffel, and Horton, 2020). Most models linking weather
and crop yields, particularly in hotter climates, generally find inverted-U trends between
rainfall and crop yields, where more rain increases yields up to a certain optimal point, but
extremes in rainfall, either too much or too little, relative to some setting-dependent thresh-
old, reduce yields (Fishman, 2016; Lesk, Coffel, and Horton, 2020). In colonial Nigeria, while
detailed data on crop yields are unavailable, numerous reports from the Agricultural Depart-
ment from 1921 to 1952 highlight the sensitivity of crop yields to extremes in rainfall.58 One
example is from a 1923 report on cotton yields, stating:
“The annual reports of the Southern Agricultural Department record remarkable
variations in the crops grown from year to year. . . The bad crops were from time
to time ascribed to one or other the following causes: (a) To the direct effect of
climatic conditions on the crop- too much rain in November or too sudden
drought in December...”
The climate in Nigeria has remained largely stable between the colonial and postcolonial
period (Xie, You, and Takeshima, 2017). Additionally, the practice of, and investment in,
agriculture has also remained largely stable in Nigeria since 1920; hence, we can use data
on crop yields from the postcolonial period to infer the relationship between rainfall shocks
and crop yields in the colonial period as well. In the postcolonial period, although there
is relatively little disaggregated data on crop yields, we digitized four years of available
data from 1992 to 1995 from the Annual Abstract of Statistics, with details provided in
Appendix A.4. The data include seven major crops representing almost one-fifth of domestic
58We highlight more evidence from the Agricultural Department reports in Appendix A.4.
28

production by FAO estimates.59 Crop yield is calculated as the average of volume of crop
produced/area cropped, following previous literature (Jayachandran, 2006). We estimate
Equations 3 and 4 using crop yields as the outcome. The results in Table A4 in Appendix
A.4 confirm the inverted-U relationship between rainfall deviations and crop yields (column
(1) of Table A4). Extreme negative rainfall shocks, like droughts, and extreme positive
rainfall shocks, like floods, decrease crop yields (column (2) of Table A4). We discuss the
crop yield data and results further in Appendix A.4.
Rainfall Shocks and Incarceration Rates Results
Table 4 presents the results from Equation 3 on the effects of rainfall shocks on incarceration
rates following the quadratic specification. While the quadratic term is negative but not
significant when we examine all penal imprisonment over the colonial period in column
(1), the effect is significant and negative for short-term incarceration rates. The negative
quadratic coefficient for short-term incarceration is consistent with an inverted-U relationship
between rainfall deviation and short-term imprisonment or the use of prison labor. β2, the
squared rainfall deviation term is not significant for medium or long-term incarceration rates,
congruent with the predictions in Section 4.3.2.
The results of the falsification test for postcolonial imprisonment are shown in column
(5) of Table 4. β2from Equation 3 is positive and significant for postcolonial incarceration
rates. The positive significant estimate for postcolonial incarceration is consistent with the
hypothesis that the effects of rainfall shocks on incarceration rates should be U-shaped
under non-prison labor regimes; instead, imprisonment primarily increased as a response to
increases in economic crimes, like theft, in the aftermath of negative productivity shocks
(e.g., drought or floods).
Table 5 reports the results from the specification in Equation 4, which identifies the
effects of moderate positive rainfall shocks that raise agricultural productivity, versus extreme
positive or negative rainfall shocks (respectively signifying floods or droughts that reduce
productivity) on incarceration rates. The results from our main specification in column (1)
show that moderate positive rainfall shocks had a significant positive effect on short-term
imprisonment over the colonial period. A moderate positive rainfall shock increased the
short-term incarceration rate by 16.7 per 100,000 population, or around 12%, relative to
the sample mean of 135 per 100,000 people. The effect remains significant, increasing the
short-term incarceration rate by about 9% when we add controls for extreme negative and
59The crops include cowpeas, mangoes, palm oil, pepper, soybeans, tomatoes, and leafy vegetables.
29

positive rainfall shocks in column (3) of Table 5.
In line with the inverted U-shape prediction, columns (2) and (3) of Table 5 show the
opposite result for extreme negative rainfall shocks, which reduced short-term colonial im-
prisonment. Extreme negative rainfall shocks, like droughts, signal a decrease in agricultural
productivity and lowered demand for unpaid prison labor under the colonial prison labor
system; this is reflected in the lowered incarceration rates, with extreme negative rainfall
shocks associated with a 13%-15% decline in short-term incarceration relative to the sample
mean. There are no effects of rainfall shocks on long-term incarceration, as shown in columns
(4)-(6) of Table 5.
In contrast, the postcolonial results show that, while moderate positive rainfall shocks
had no significant effect on postcolonial incarceration rates (column (7) and column (9)),
extreme negative (column (8)) and extreme positive (column (9)) rainfall shocks increased
the postcolonial imprisonment rate. From column (9), the magnitude of the increase in post-
colonial imprisonment from droughts/extreme negative rainfall shocks and floods/extreme
positive rainfall shocks is a 21% and 19% increase, respectively, in incarceration rates relative
to a sample mean of 105 per 100,000 people. The results from Equation 4 are consistent
with the results from the quadratic specification in Equation 3, showing an inverted U-shape
relationship between rainfall deviation and incarceration rates in the colonial era, with an
opposite/U-shaped relationship in the postcolonial period.
4.3.3 Robustness
We conduct numerous robustness checks on our results, with a subset of checks presented in
Appendix A.5. We show that our cash crop export price results are robust to the inclusion
of rainfall controls (Table A5) and using raw prices instead of logs (Table A6); and that
contemporaneous, not lagged, rainfall shocks affect incarceration rates in Table A8 and
Table A9. The results are also robust to trimming provinces to account for potential concerns
around district population estimates (Table A11).
4.4 Further Evidence on Mechanisms
Thus far, our analysis confirms that positive productivity shocks increased incarceration rates
and the use of prison labor during the colonial period. The historical account in Section 2
suggests that colonial governments used multiple methods to intensify the use of prison
labor during periods of labor shortages to work on key public works like the railroad needed
to transport agricultural commodity exports for revenue. Among these methods were: (1)
30

increasing short-term incarceration around prisons close to public works like the railroad, (2)
increasing prosecutions of minor, misdemeanor crimes and sentence-switching, or changing
the punishment for these crimes from fines to imprisonment, and (3) potentially increasing
prison sentences as punishment for already incarcerated populations. We evaluate each of
these three hypotheses around colonial government methods below.
4.4.1 Wages and the Railroad
As previously established, a major use of prison labor was for public works and construction
and maintenance of the railroad, which was essential for the transport of cash crops for
export. Railroad construction began in 1898, and had expanded to its full extent across
the country by the 1950s (Figure 2b). One test of the labor market tightness hypothesis
described in this historical account and the conceptual framework is that when market
wages are higher, demand for coerced prison labor should also increase. Although there is
no available disaggregated data, by district, on wages, one way to test this hypothesis, is to
examine the correlation between wages and incarceration rates at prisons near the railroad,
given the intensive use of prison labor for railroad work.60 Table 6 reports the estimates
for the reduced-form relationship between wages and distance from prisons to the railroad
and colonial incarceration rates. The regressions are at the prison level and show that while
prisons closer to the railroad generally had higher short-term incarceration rates (column
(1)), during periods of higher wages, short-term incarceration rates also increased in prisons
farther away from the railroad but within the same colonial province (column (2)).
The interpretation of this result is intuitive. While short-term sentenced prisoners near
the railroad were generally used as a reserve of unpaid labor for railroad construction and
maintenance, increasing wages intensified the demand for unpaid prison labor and worsened
labor shortages and labor market tightness. To increase the share of prison labor, colonial
officials would need to increase the share of prisoners in prisons farther away from the railroad
as well. Prisoners, by law, could not be transferred across provinces, as discussed in Section
2. Colonial officials could then transport prisoners within the province to conduct work on
the railroad and associated public works as needed (Foreign and Office, 1960). The effects are
specific to short-term sentenced prisoners, with no effects for long-term sentenced prisoners
(columns (3) and (4) of Table 6).
60Market wages are endogenous in this context, and we interpret the results as suggestive correlations
only.
31

4.4.2 Increasing Prosecutions of Minor Offenses and Sentence-Switching
The qualitative accounts of officials in the colonial archives provide suggestive evidence of
the ‘increasing prosecutions of minor offenses/sentence-switching’ channel.61 For example,
in 1926, C.W. Duncan, the Inspector-General of Police of the southern provinces, noted the
uptick in cases and convictions in that year in his report. He then highlighted that while ‘
“offences against property show a decrease of 198 cases compared with those of the previous
year”, and there has been a decrease in “offences against persons”, prosecutions of minor
offenses have increased that year, accounting for the increase in cases and convictions’.62
Although there is no available disaggregated colonial data, by district, on crime and
punishment, to test the “sentence-switching” hypothesis, we estimate Equation 2 using the
difference between custody/awaiting trial and short-term incarceration figures as an out-
come. The rationale here is that, as only sentenced prisoners could legally be used for prison
labor, if there was more sentence switching from “awaiting trial” to short-term imprisonment
in response to positive economic shocks, then positive shocks will decrease the difference be-
tween the custody/awaiting trial and short-term incarceration rates. Table A14 in Appendix
A.5.5 provides suggestive evidence of ‘sentence-switching’ as a strategy to increase the share
of short-term prisoners for prison labor in response to positive productivity shocks. Col-
umn (3) of Table A14 shows that increases in the prices of the most productive cash crops,
like palm oil, are associated with decreases in the difference between the custody/awaiting
trial and short-term incarceration rates. Table A15, estimating the effects using Equation 4,
shows similar results between moderate positive rainfall shocks and the difference outcome.
Given that the coefficients on both custody and short term incarceration rates are positive,
the only way for their difference to be negative is if short-term incarceration is rising faster
than the custody category in response to positive productivity shocks. One interpretation
is that prisoners may have been transferred at a faster rate from custody/awaiting trial to
short-term sentences so that the state could take advantage of unpaid prison labor when
positive economic shocks increased labor demand and worsened labor shortages. The results
are not robust and should be interpreted with caution, but provide suggestive evidence of
61Accounts from the “Policing in Lagos and Provinces, 1899-1929”. Reference: 73242C-01 document from
the British Foreign and Commonwealth Office.”
62Duncan highlights one notable case of tax default in the southern provinces where “The inhabitants
of these villages had fallen into arrears in the payment of their taxes and the Assistant District officers,
having failed to collect these arrears in March, warned the people that their property would be seized if
they persisted in their obstinate attitude”. Source: “Policing in Lagos and Provinces, 1899-1929”. Reference:
73242C-01.
32

the sentence-switching hypothesis.
4.4.3 Punishment
The previous two methods focused on the ways colonial governments increased incarceration
rates in response to labor demand shocks. To test the hypothesis that increasing the prison
sentences of prisoners that were already incarcerated may be one, albeit more minor, way
that colonial officials intensified the use of prison labor during these shocks, we digitized data
from the colonial Annual Report on Prisons on the punishments assigned by colonial officials
to prisoners for infractions while in prison. In line with the cost-cutting objectives of colonial
officials, the most popular punishment was reduced diet, accounting for 53% of punishment
to prisoners between 1920 and 1938, on average.63 After reduced diet, the top categories for
prisoners’ punishment were flogging (21%), solitary confinement (8%), forfeiture of “marks”
or credits for good behavior which could be used to reduce a prison sentence (6%), and extra
prison time (4%). To examine the effects of directly observable, to colonial officials, cash
crop price shocks on punishment of prisoners, we estimate Equation 2 with shares of each
punishment in total punishment assigned to prisoners as the outcome. The results in Table 7
provide suggestive evidence that prison officials may have also employed a “carrot and stick”
approach to motivate prisoners to work (carrot) and punish detractors with more prison time
(stick) during periods of increased prices or labor shortages. Increases in palm oil prices in
palm oil producing areas were associated with more prison time assigned as punishment to
prisoners (column (1), “stick”), but less forfeiture of marks assigned as punishment (column
(5), “carrot”).
5 Colonial Imprisonment and Contemporary Trust in Legal Insti-
tutions
In the historical account in Section 2.5, we discussed the role of, often violent, policing in
enforcing the colonial prison labor system, and its reported long-term effects on contemporary
mistrust in police. To explore the implications of the colonial prison labor system for present-
day views of police and contemporary trust in legal institutions more broadly, we present a
brief discussion and suggestive evidence of the long-term effects of colonial imprisonment.
Given that the origins of the modern prison and accompanying legal system in Nigeria and
other former British colonies are rooted in the use of state policy around labor coercion,
what are the long-term effects, if any, of exposure to these systems on populations’ trust in
63The distribution of punishment over time is shown in Figure A21 in the Appendix. Food costs were
often the major cost of maintaining a prisoner as discussed in Section 3.
33

these institutions today? We use Afrobarometer data from Nigeria from surveys over 2003 to
2014 recording respondents’ stated trust in historical legal institutions (e.g., police, courts,
and tax administration) to test whether past exposure to coercive, ostensibly economically
influenced, colonial prison systems affects trust in legal institutions today. To assess if these
effects, if any, are about legal institutions and not broader interpersonal trust, we also assess
trust in individuals (e.g., neighbors, relatives, and elected local governing council members)
as an outcome. Previous research has shown that interpersonal trust is linked to longer term
historical events like the slave trade (Nunn and Wantchekon, 2011).
To test these hypotheses, we estimate equations of the following form:
Trustaig st =βColonial Imprisonmenti+X0
aigstθ+X0
gsφ+µs+δt+aigst (5)
where Trustasit is the contemporary trust outcome of interest for individual aresiding in
historical colonial province i, in current sub-district or local government area (LGA) g, in
one of Nigeria’s six geopolitical zone regions sfor the Afrobarometer survey administered in
year t. Nigeria’s ethnic distribution is proxied by six geopolitical zones delineating ethnic
homelands of populations64 , and the region fixed effects are included to capture culturally
specific factors, like values around social status or age-based hierarchy, that may affect trust
(Archibong, 2018; Nunn and Wantchekon, 2011; Lowes et al., 2017).
Following the historical account in Section 2 and the analysis in Section 4 linking pri-
marily short-term sentenced prisoners to the use of prison labor for public works in response
to labor demand shocks, we measure colonial imprisonment, Prisonersi, as the long-run av-
erage share of short-term sentenced prisoners in overall penal imprisonment from 1920 to
1938 in each colonial province i. The value captures the intensity of the use of prisoners
as convict labor to satisfy economic incentives over the colonial period, with higher values
indicating that more incarcerated people were being used for prison labor on public works
in a province. The share of short-term sentenced prisoners may more strongly reflect the
level of coercive policing and legal practices in colonial provinces over time, as this cate-
gory of prisoners was most intensely exploited for prison labor. Using available data on the
64Broadly, three ethnic groups- the Hausa, Yoruba and Igbo dominate three zones- the Northwest,
Southwest and Southeast respectively. The Kanuri are the majority group in the Northeast, the
Ijaw/Edo/Bini/Ibibio weakly dominate the Southsouth zone, and the Northcentral is home to the Tiv,
Nupe and other smaller ethnic populations (Archibong, 2018).
34

distribution of rank and file police across colonial provinces from 1920 to 1938, we show a
strong, positive correlation between the share of rank and file, “boots on the ground” police
in the total police force and the share of short-term colonial imprisonment (Table A18).
There is no correlation between the share of rank and file police and the share of long-term
colonial imprisonment (column (2)), and the correlation is weaker with overall year to year
short-term incarceration rates (column (3)). As a falsification test, we estimate Equation 5
using the share of long-term sentenced prisoners as well.
To check that the associations in Equation 5 are not being driven by differences in
crime between high and low colonial imprisonment areas, we also test the following “crime
propensity” outcomes from the Afrobarometer: whether the respondent has feared being the
victim of a crime in their home, and how often the respondent had to bribe a government of-
ficial to obtain a document or permit in the past year. We include vectors of individual level
covariates, X0
aigst, including a respondent’s age, age squared, a gender indicator variable, an
indicator that equals one if the respondent lives in an urban location, and educational at-
tainment fixed effects. The sub-district level covariates, X0
gs, include controls for geography,
disease suitability and precolonial and colonial institutional features.65 Geography controls
include land suitability for agriculture, ruggedness, elevation, and indicators for the presence
of petroleum and access to a seacoast. Controls for disease suitability include the mean
malaria ecology index and tsetse fly suitability. Precolonial and colonial institutional con-
trols include the level of precolonial centralization and the total number of slaves exported
from each ethnic region during the Atlantic slave trade. All regressions include region and
survey-year fixed effects, µsand δt, respectively. Standard errors are clustered at the dis-
trict (colonial province) level and wild cluster bootstrap p-values are included to account for
potentially low numbers of clusters as before.
Figure 9 shows the visual relationship between colonial imprisonment and trust in
legal institutions. The simple binscatter in the top panel, using the share of short-term
sentenced prisoners colonial imprisonment measure suggests a strong negative relationship
between short-term colonial imprisonment and trust in legal institutions. The picture is
largely flipped using the share of long-term sentenced imprisonment measure in the bottom
panel. We present OLS estimates for the effects of colonial imprisonment, using our main
short-term sentenced measure, on trust outcomes in Panel A of Table 8. Columns (1)-(3)
of Panel A show a negative association between colonial imprisonment and contemporary
65Data is described in detail in Appendix A.6.
35

trust in legal institutions, with effects particularly robust for trust in police (column (1)).
Increasing the share of short-term sentenced colonial imprisonment in a province from none
to all decreases the reported trust in police by present-day residents of the region by 0.4
points, or a 57% reduction in reported trust in police relative to the sample mean. There
is no significant association between colonial imprisonment and contemporary interpersonal
trust (columns (4)-(6) of Panel A). Panel B shows no significant association between the
long-term colonial imprisonment measure and contemporary trust outcomes.
To check that the result on the negative association between colonial imprisonment and
trust in legal institutions is not being driven by underlying differences in crime rates between
regions of high versus low levels of colonial imprisonment, we present the results on crime in
Table 9. There is no significant association between colonial imprisonment and propensity
for reported criminal/bribery behavior (columns (1) and (2)). If anything, residents from
areas with high levels of (short-term) colonial imprisonment are less likely to report fear
of being victims of a crime in their homes (column (3)). There is no association between
long-term colonial imprisonment and reported crime (columns (4) to (6)).
The results presented in Table 8 suggest that there is a negative correlation between
colonial imprisonment and contemporary trust in legal institutions, but do not identify the
causal effect of colonial imprisonment on trust. The results provide an initial exploration
of the potentially detrimental long-term effects of coercive colonial prison labor systems on
present-day trust in legal institutions, such as police. The qualitative history in Section 2.5,
and the historically high share of rank and file colonial police in these areas suggest that
persistence in coercive policing may be one channel through which these effects persist.66
6 Conclusion
What are the effects on incarceration when prisoners are viewed and used primarily as a
source of labor to serve economic interests? And what are the potential implications for
citizens’ views of state legitimacy, when an institution of state justice, like prison, is used to
serve economic interests? To answer these questions, we digitized annual data from archival
sources for British colonial Nigeria, and show that prisons were economically valuable to the
colonial regime. We present the first quantitative estimates on the value of prison labor in
British colonial Africa, and find that the value of prison labor was strictly positive over the
colonial period. Even after accounting for an extensive set of prisoner maintenance costs,
the net value of prison labor was strictly positive in the majority of years in colonial Nigeria.
66We discuss other possible channels and avenues for estimating more causal effects in Appendix A.6.
36

Prison labor constituted a significant share of public works expenditures, up to 249% and
42%, using our gross and net values of prison labor respectively.
We examine the effects of shocks to economic productivity on incarceration and the use
of prison labor. We find that incarceration rates during the colonial period are procyclical.
Moderate positive rainfall shocks and positive export price shocks that proxy increased
agricultural productivity increased incarceration rates and the use of prison labor in the
colonial period. We provide quantitative and qualitative evidence demonstrating that a
primary reason for the procyclical behavior of incarceration rates during the colonial period
was increased labor demand for construction and maintenance of public works, like the
roads and railroad, that were needed to intensify exports of agricultural commodities during
periods of positive productivity shocks. Labor shortages and tight labor markets increased
the demand for unpaid prison labor, which was reflected in the rise in incarceration rates.
The effect is reversed in the postcolonial period, when prison labor is not a major feature of
state policy and public finance, and thus negative shocks increased incarceration rates.
We explore the implications of exposure to prison labor systems for present-day views of
state judicial legitimacy and provide suggestive evidence of the negative long-term effects of
colonial incarceration on contemporary trust in legal institutions. We document a significant
reduction in contemporary trust in legal institutions like police in areas with high historical
levels of colonial imprisonment. The reduction in contemporary trust is specific to legal
institutions, with no effect on interpersonal trust. The results regarding trust open up
avenues for future work to explore channels through which these effects on reduced trust
may persist over time. Given the renewed global debates on the use of prison labor and
the judicial system globally, our paper provides new quantitative evidence on the effects
on incarceration when prisoners are primarily used as a store of labor, and its potentially
detrimental effects on citizens’ trust in legal institutions.
37

Figure 1: Top 40 countries/territories for incarceration rates, 2018 with Nigeria incarceration
rates in red (year 1940) and blue (year 2018). Source: World Prison Brief
38

Figure 2: Colonial Nigeria with provinces outlined in 1937, prison locations, regions and
railroad network shown
39

0
30
60
90
1920 1930 1940 1950 1960
Year
Average annual wage and prisoner costs (pounds)
category Prison costs Prisoner costs: food Urban unskilled wage
(a) Average annual wage and prisoner costs, 1920−1959
6000
8000
10000
12000
14000
1920 1930 1940 1950 1960
Year
Daily average number in prison
(b) Daily average number in prison
Figure 3: Wages, prisoner costs (a) and daily average number in prisons (b) in colonial
Nigeria, 1920-1959
100
200
300
1920 1930 1940 1950 1960 1970 1980 1990
Year
Prisoners ( 105)
Mean Nos. of Prisoners (per 10^5 pop.), 1920−1995, Nigeria
Figure 4: Mean number of prisoners per 100,000 population in colonial and postcolonial
Nigeria with independence year highlighted, 1920-1995
40

0.0
0.2
0.4
0.6
miscellaneous
minor
offences
offences
against
property
offences
against
revenue,
social
economy
laws
offences
against
the person
Crime
Share of total convictions
period
colonial
postcolonial
Figure 5: Share of total convictions by crime over the colonial (1920-1939) and postcolonial
(1977-1993) periods in Nigeria
0
500000
1000000
1500000
1920 1930 1940 1950 1960
Year
Value of prison labor (pounds)
category Total value, estimate Total value, reported
(a) Value of prison labor
11
12
13
14
1920 1930 1940 1950 1960
Year
Log value of prison labor
category Total value, estimate Total value, reported
(b) Value of prison labor (Logs)
Figure 6: Total value of prison labor estimates versus value of prison labor reported by
colonial government in pounds (a) and in log values (b), 1920-1959. Figure shows values in
pounds (a) and log values (b)
41

Figure 7: Prison populations in colonial (1920) (a) and postcolonial (1980) (b) Nigeria, and
agricultural commodity production areas (c) and prices (d) for the three major cash crops
(palm oil, cocoa, and groundnut) in colonial Nigeria
42

−50
0
50
100
150
1920 1922 1924 1926 1928 1930 1932 1934 1936 1938
Year
Short−Term Incarceration Coefficient
Short−Term Colonial Incarceration in Palm Oil Areas
0.0
0.5
1.0
1.5
1920 1922 1924 1926 1928 1930 1932 1934 1936 1938
Year
Log Cash Crop Price, Palm oil
Agricultural Commodity (Palm Oil) Prices
−60
−40
−20
0
20
1920 1922 1924 1926 1928 1930 1932 1934 1936 1938
Year
Short−Term Incarceration Coefficient
Short−Term Colonial Incarceration in Cocoa Areas
0.5
1.0
1.5
2.0
1920 1922 1924 1926 1928 1930 1932 1934 1936 1938
Year
Log Cash Crop Price, Cocoa
Agricultural Commodity (Cocoa) Prices
−50
0
50
100
1920 1922 1924 1926 1928 1930 1932 1934 1936 1938
Year
Short−Term Incarceration Coefficient
Short−Term Colonial Incarceration in Groundnut Areas
0.0
0.5
1920 1922 1924 1926 1928 1930 1932 1934 1936 1938
Year
Log Cash Crop Price, Groundnut
Agricultural Commodity (Groundnut) Prices
Figure 8: Agricultural commodity export prices and short-term incarceration rates over
the colonial period (1920-1938). Coefficients are from individual regressions of short-term
incarceration on colonial province and year fixed effects and the interaction between an
agricultural commodity presence variable and year fixed effects; the interaction coefficients
are plotted above.
0.25
0.50
0.75
1.00
0.5 0.6 0.7 0.8 0.9
Colonial Imprisonment (ST)
Trust in Police
0.25
0.50
0.75
1.00
0.0 0.1 0.2 0.3
Colonial Imprisonment (LT)
Trust in Police
0.8
1.0
1.2
1.4
0.5 0.6 0.7 0.8 0.9
Colonial Imprisonment (ST)
Trust in Courts
0.8
1.0
1.2
1.4
0.0 0.1 0.2 0.3
Colonial Imprisonment (LT)
Trust in Courts
0.6
0.9
1.2
1.5
0.5 0.6 0.7 0.8 0.9
Colonial Imprisonment (ST)
Trust in Tax Administration
0.6
0.9
1.2
1.5
0.0 0.1 0.2 0.3
Colonial Imprisonment (LT)
Trust in Tax Administration
Figure 9: Colonial imprisonment and contemporary trust in legal institutions. Top panel
uses the main measure of colonial imprisonment, the share of short-term prisoners in penal
imprisonment. Bottom panel uses the share of long-term colonial imprisonment
43

Table 1: Decadal averages of relative values of prison labor in colonial Nigeria with value of prison labor, and value of
prison labor as a share of public works expenditure (top panel), and value of prison labor as a share of expenditure on
prisons and total colonial expenditure (bottom panel)
Period Gross value of prison
labor (PL), pounds
Net value of PL- less
food costs
Net value of PL- less
total prison costs
Share of gross PL
value in public works
exp.
Share of net PL
value (food) in pub-
lic works exp.
Share of net PL
value (total) in pub-
lic works exp.
1920 −1930 163,748 87,858 19,776 0.96 0.51 0.13
1931 −1940 111,760 69,327 6,135 0.66 0.42 0.05
1941 −1950 222,183 149,702 −16,891 0.88 0.59 −0.08
1951 −1959 875,463 652,486 160,805 1.77 1.24 0.14
Period Share of gross PL
value in prison exp.
Share of net value of
PL (food) in prison
exp.
Share of net value of
PL (total) in prison
exp.
Share of gross PL
value in all colonial
exp.
Share of net PL value
(food) in all colonial
exp.
Share of net PL value
(total) in all colonial
exp.
1920 −1930 1.09 0.59 0.14 0.02 0.01 0.003
1931 −1940 1.03 0.65 0.07 0.02 0.01 0.001
1941 −1950 0.91 0.60 −0.08 0.01 0.01 −0.001
1951 −1959 1.20 0.85 0.18 0.01 0.01 0.002
44

Table 2: Summary Statistics: Economic shocks and incarceration rates
Statistic N Mean St. Dev. Min Max
Prisoners, 1920-1938
All Prisoners Total 324 1,811.76 2,286.76 3.00 10,231.00
Penal Imprisonment Total 324 1,251.83 1,626.78 2.00 7, 010.00
Custody Total 324 509.59 635.57 0.00 3, 039.00
Short-Term (<=6 Months) Total 324 1,051.05 1,409.20 2.00 6, 377.00
Medium-Term (6Mo-2Y) Total 324 127.15 171.34 0.00 882.00
Long-Term (>=2yr) Total 324 68.93 84.10 0.00 417.00
All Prisoners /100,000 324 240.73 254.56 0.26 1, 123.30
Penal Imprisonment /100,000 324 162.03 169.55 0.26 759.99
Custody /100, 000 324 71.73 83.47 0.00 333.66
Short-Term /100,000 324 134.66 144.95 0.16 649.43
Medium-Term /100,000 324 16.56 18.26 0.00 80.45
Long-Term /100,000 324 10.18 12.88 0.00 83.45
Share w/ 1 Previous Conviction 324 0.11 0.15 0.00 0.90
Share w/ 2 Previous Convictions 324 0.02 0.03 0.00 0.32
Share w/ 3 Previous Convictions 324 0.02 0.03 0.00 0.18
Agricultural Commodities and Rainfall Deviation, 1920-1938
Cocoa Producing 393 0.15 0.35 0.00 1.00
Groundnut Producing 393 0.18 0.39 0.00 1.00
Palm Oil Producing 393 0.29 0.45 0.00 1.00
Log Cocoa Price 393 1.04 0.40 0.47 1.96
Log Groundnut Price 393 0.35 0.36 −0.36 0.88
Log Palm Oil Price 393 0.72 0.53 −0.22 1.69
Rainfall Dev. 393 −0.00 0.97 −2.21 4.08
Rainfall Dev. Sq. 393 0.95 1.83 0.00 16.67
Positive Rainfall Shock (M) 393 0.17 0.38 0.00 1.00
Negative Rainfall Shock (E) 393 0.30 0.46 0.00 1.00
Positive Rainfall Shock (E) 393 0.21 0.41 0.00 1.00
Prisoners and Rainfall Deviation, 1971-1995
All Prisoners Total 871 2,005.81 1,210.56 104.00 7, 092.00
All Prisoners /100,000 871 92.48 60.43 9.91 361.99
Share w/ 1 Previous Conviction* 6 0.21 0.02 0.18 0.23
Share w/ 2 Previous Convictions* 6 0.12 0.02 0.10 0.16
Share w/ 3 Previous Convictions* 6 0.13 0.04 0.05 0.18
Rainfall Dev. 560 0.01 0.30 −0.62 1.06
Rainfall Dev. Sq. 560 0.09 0.12 0.00 1.11
Positive Rainfall Shock (M) 560 0.49 0.50 0.00 1.00
Negative Rainfall Shock (E) 560 0.04 0.19 0.00 1.00
Positive Rainfall Shock (E) 560 0.01 0.11 0.00 1.00
Notes: See text and online appendix for details. *denotes that data is based on available time series information from
1975-1980.
45

Table 3: Agricultural commodity export prices and colonial incarceration rates (prisoners per 100,000 pop.)
Outcome:Short-Term Long-Term
(1) (2) (3) (4) (5) (6) (7) (8)
Palm oil x Palm oil price 66.681∗∗ 56.546∗∗ 2.738 5.481
(27.920) (22.867) (5.448) (3.490)
[0.048] [0.045] [0.745] [0.166]
Cocoa x Cocoa price 41.965∗4.146 −6.000 −6.535∗∗∗
(23.638) (16.434) (5.952) (2.491)
[0.185] [0.830] [0.521] [0.013]
Groundnut x Groundnut price 2.809 −49.111∗∗ −8.532 −9.130∗∗∗
(29.852) (24.763) (6.905) (3.208)
[0.956] [0.092] [0.416] [0.015]
Mean of outcome 134.659 134.659 134.659 134.659 10.175 10.175 10.175 10.175
Observations 324 324 324 324 324 324 324 324
Clusters 21 21 21 21 21 21 21 21
District FE Yes Yes Yes Yes Yes Yes Yes Yes
Year FE Yes Yes Yes Yes Yes Yes Yes Yes
Notes: Regressions estimated by OLS. Robust standard errors in parentheses clustered by district, where district is colonial province for colonial data. Wild cluster bootstrap (by
district) p-values are in brackets. Observations are provinces. Dependent variables are prisoners per 100,000 population (1939 pop.) by province in Nigeria broken down by short-term
(less than 6 months) sentence in columns (1) to (4) and long-term (greater than 2 years) sentence in columns (5) to (8) over 1920-1938. Prices are in logs. District FE are colonial
province fixed effects. ∗∗∗ Significant at the 1 percent level, ∗∗Significant at the 5 percent level, ∗Significant at the 10 percent level based on clustered standard errors in parentheses.
46

Table 4: Rainfall shocks and colonial (1920-1938) and postcolonial (1971-1995) incarceration rates (prisoners per 100,000
pop.), quadratic specification
Period:Colonial Postcolonial
Outcome: All Penal Short-Term Medium-Term Long-Term All 1971-1995
(1) (2) (3) (4) (5)
Rainfall Dev 14.147∗∗ 11.995∗1.796 0.759 −6.237
(6.041) (6.433) (1.276) (1.227) (8.570)
[0.038] [0.065] [0.212] [0.655] [0.454]
Rainfall Dev Sq −3.569 −4.884∗0.205 0.752 34.275∗∗∗
(2.479) (2.816) (0.387) (0.739) (9.692)
[0.246] [0.068] [0.629] [0.494] [<.001]
Mean of outcome 162.032 134.659 16.556 10.175 104.802
Observations 324 324 324 324 556
Clusters 21 21 21 21 36
District FE Yes Yes Yes Yes Yes
Year FE Yes Yes Yes Yes Yes
Notes: Regressions estimated by OLS. Robust standard errors in parentheses clustered by district, where district is colonial
province for colonial data, and postcolonial state for postcolonial data. Wild cluster bootstrap (by district) p-values are in
brackets. Observations are provinces. Dependent variables in column (1)-(4) are prisoners per 100,000 population (1939
pop.) by province in Nigeria broken down by all prisoners, penal imprisonment, custody/awaiting trial, short-term (less
than 6 months) sentence and medium-term (between 6 months and 2 years) sentence and long-term (greater than 2 years)
sentence over 1920-1938. Dependent variable in (5) is prisoners per 100,000 population (1990 pop.) by state in Nigeria over
1971-1995. Results remain unchanged when we replace the denominator for the incarceration rates with the adult population
of the province only. Rainfall deviation, and rainfall deviation squared (Rainfall Dev and Rainfall Dev Sq) as defined in
text. District FE are colonial province fixed effects in (1)-(4), and postcolonial state fixed effects in (5). ∗∗∗Significant at
the 1 percent level, ∗∗Significant at the 5 percent level, ∗Significant at the 10 percent level based on clustered standard
errors in parentheses.
47

Table 5: Rainfall shocks and colonial (1920-1938) and postcolonial (1971-1995) incarceration rates (prisoners per 100,000
pop.)
Period:Colonial Postcolonial
Outcome: Short-Term Long-Term All 1971-1995
(1) (2) (3) (4) (5) (6) (7) (8) (9)
Positive rainfall shock (M) 16.727∗∗∗ 12.142∗−1.638 −0.695 −4.387 −2.320
(5.456) (6.964) (1.319) (1.437) (4.132) (4.564)
[0.016] [0.093] [0.336] [0.683] [0.320] [0.620]
Negative rainfall shock (E) −20.290∗∗ −17.225∗−1.060 −0.429 22.722∗∗∗ 22.545∗∗∗
(9.484) (10.259) (2.894) (3.530) (7.814) (7.807)
[0.057] [0.139] [0.762] [0.886] [0.016] [0.012]
Positive rainfall shock (E) −0.404 3.358 20.423∗∗
(13.973) (2.654) (8.268)
[0.977] [0.293] [0.046]
Mean of outcome 134.659 134.659 134.659 10.175 10.175 10.175 104.802 104.802 104.802
Observations 324 324 324 324 324 324 556 556 556
Clusters 21 21 21 21 21 21 36 36 36
District FE Yes Yes Yes Yes Yes Yes Yes Yes Yes
Year FE Yes Yes Yes Yes Yes Yes Yes Yes Yes
Notes: Regressions estimated by OLS. Robust standard errors in parentheses clustered by district, where district is colonial province for colonial data, and postcolonial state for postcolonial
data. Wild cluster bootstrap (by district) p-values are in brackets. Observations are districts. Dependent variables in columns (1)-(6) are prisoners per 100,000 population (1939 pop.) by
province in Nigeria broken down by short-term (less than 6 months) sentence ((1)-(3)) and long-term (greater than 2 years) sentence ((4)-(6)) over 1920-1938. Dep endent variable in columns
(7)-(9) is prisoners per 100,000 population (1990 pop.) by state in Nigeria over 1971-1995. Positive rainfall shock (M) where (M) is moderate, and (E) is extreme as defined in text. District
FE are colonial province fixed effects in (1)-(6), and postcolonial state fixed effects in (7)-(9). ∗∗∗ Significant at the 1 percent level, ∗∗ Significant at the 5 percent level, ∗Significant at the 10
percent level based on clustered standard errors in parentheses.
48

Table 6: Reduced-form estimates of the relationship between wages and distance to railroad
and colonial incarceration rates (prisoners per 100,000 pop.)
Outcome:Short-Term Long-Term
(1) (2) (3) (4)
Distance to railroad −0.301∗−1.479∗∗ −0.018 −0.029
(0.157) (0.681) (0.023) (0.099)
[0.144] [0.074] [0.941] [0.778]
Distance x Log wages 0.401∗∗ 0.004
(0.191) (0.033)
[0.078] [0.917]
Mean of outcome 46.198 46.198 3.990 3.990
Observations 938 938 822 822
Clusters 21 21 21 21
District FE Yes Yes Yes Yes
Year FE Yes Yes Yes Yes
Notes: Regressions estimated by OLS. Robust standard errors in parentheses clustered by district,
where district is colonial province for colonial data. Wild cluster bootstrap (by district) p-values
are in brackets. Observations are individual prisons. Dependent variables in (1)-(4) are prisoners in
each prison per 100,000 population of the province broken down by short-term (less than 6 months)
sentence and long-term (greater than 2 years) sentence over 1920-1938. Covariates are distance
to railroad in km and log urban unskilled wages. District FE are colonial province fixed effects.
∗∗∗Significant at the 1 percent level, ∗∗ Significant at the 5 percent level, ∗Significant at the 10
percent level based on clustered standard errors in parentheses.
49

Table 7: Agricultural commodity export prices and punishment of prisoners
Outcome:Extra Imprisonment Reduced Diet Flogging Solitary Confinement Forfeit Marks
(1) (2) (3) (4) (5)
Palm oil x Palm oil price 0.061∗∗ 0.040 0.028 −0.047 −0.114∗∗
(0.031) (0.179) (0.055) (0.038) (0.049)
[0.075] [0.846] [0.657] [0.260] [0.115]
Cocoa x Cocoa price 0.133 0.542 −0.036 0.244 −0.089∗
(0.100) (0.652) (0.094) (0.160) (0.048)
[0.494] [0.580] [0.738] [0.226] [0.240]
Groundnut x Groundnut price 0.066 0.135 0.084 −0.065 −0.129
(0.040) (0.339) (0.142) (0.049) (0.168)
[0.207] [0.742] [0.635] [0.179] [0.624]
Mean of outcome 0.050 0.556 0.166 0.087 0.071
Observations 228 228 228 228 228
Clusters 21 21 21 21 21
District FE Yes Yes Yes Yes Yes
Year FE Yes Yes Yes Yes Yes
Notes: Regressions estimated by OLS. Robust standard errors in parentheses clustered by district, where district is colonial province for colonial data.
Wild cluster bootstrap (by district) p-values are in brackets. Observations are provinces. Dependent variables are shares of total punishment assigned to
prisoners from extra prison time (1), reduced diet (2), flogging (3), solitary confinement (4), and forfeiture of marks (5), as described in the text. Prices
are in logs. District FE are colonial province fixed effects. ∗∗∗ Significant at the 1 percent level, ∗∗Significant at the 5 percent level, ∗Significant at the 10
percent level based on clustered standard errors in parentheses.
50

Table 8: OLS estimates of the relationship between colonial imprisonment and present-day trust in historical legal Insti-
tutions versus interpersonal trust
Panel A: Colonial Imprisonment (Short-Term) and Contemporary Trust Outcomes
Outcome:Trust in Historical Legal Institutions Interpersonal Trust
Police Courts Tax Neighbors Relatives Local Gov
(1) (2) (3) (4) (5) (6)
Colonial imprisonment (ST) −0.401∗∗∗ −0.541∗−0.750∗∗ −0.382 0.878 −0.255
(0.143) (0.279) (0.383) (0.555) (0.675) (0.220)
[0.002] [0.187] [0.136] [0.544] [0.376] [0.354]
Mean of outcome 0.709 1.274 0.976 1.334 1.913 0.948
Panel B: Colonial Imprisonment (Long-Term) and Contemporary Trust Outcomes
Outcome:Trust in Historical Legal Institutions Interpersonal Trust
Police Courts Tax Neighbors Relatives Local Gov
(1) (2) (3) (4) (5) (6)
Colonial imprisonment (LT) 0.285 0.401 0.304 0.635 −0.563 −0.061
(0.291) (0.386) (0.523) (0.619) (0.908) (0.375)
[0.510] [0.527] [0.649] [0.423] [0.658] [0.887]
Mean of outcome 0.709 1.274 0.976 1.334 1.913 0.948
Observations 6,642 6,590 3,126 3,439 3,317 4,899
Clusters 21 21 21 21 21 21
Individual Controls Yes Yes Yes Yes Yes Yes
Geographic Controls Yes Yes Yes Yes Yes Yes
Disease Controls Yes Yes Yes Yes Yes Yes
Precolonial and Colonial Controls Yes Yes Yes Yes Yes Yes
Region FE Yes Yes Yes Yes Yes Yes
Year FE Yes Yes Yes Yes Yes Yes
Notes: Regressions estimated by OLS. Robust standard errors in parentheses clustered by colonial province. Wild cluster bootstrap (by district) p-values are in brackets.
The unit of observation is an individual. Colonial imprisonment (ST or LT) is the average share of short-term (ST) or long-term (LT) incarcerated populations in each
colonial province over 1920 to 1938 as defined in the text. Trust variables are from the Afrobarometer samples over 2003 to 2014 and as defined in the main text. Trust
outcomes are reported trust levels on a scale of 0-3, where “Not at all”= “0”, “Just a little”=“1”, “Somewhat”=“2”, “A lot”=“3”. All regressions use region fixed effects at
the geopolitical zone level in Nigeria (for 6 geopolitical zones), year fixed effects and educational attainment fixed effects. Individual controls include age, age squared and
gender. Geographic controls include an indicator for whether the respondent lives in an urban location, and, at the sub-district or local government area level, include,
ruggedness, indicators for petroleum, seacoast and mean land suitability for agriculture and mean elevation in alternate specifications. Disease controls at the sub-district
level include malaria suitability and tsetse fly suitability in alternate specifications with results unchanged. Precolonial and colonial controls at the ethnicity-level include
the level of precolonial centralization and total exports of slaves from the region during the Atlantic slave trade. ∗∗∗ Significant at the 1 percent level, ∗∗Significant at the
5 percent level, ∗Significant at the 10 percent level.
51

Table 9: OLS estimates of the relationship between colonial imprisonment and present-day crime outcomes
Outcome:Bribe Doc Bribe HHS Fear Crime Bribe Doc Bribe HHS Fear Crime
Covariate: Colonial Imprisonment (ST) Colonial Imprisonment (LT)
(1) (2) (3) (4) (5) (6)
Colonial imprisonment 0.026 −0.151 −0.467∗∗ −0.263 0.108 0.256
(0.139) (0.175) (0.231) (0.245) (0.246) (0.404)
[0.890] [0.544] [0.117] [0.426] [0.737] [0.669]
Mean of outcome 0.225 0.229 0.571 0.225 0.229 0.571
Observations 4,279 4,343 6,700 4,279 4,343 6,700
Clusters 21 21 21 21 21 21
Individual Controls Yes Yes Yes Yes Yes Yes
Geographic Controls Yes Yes Yes Yes Yes Yes
Disease Controls Yes Yes Yes Yes Yes Yes
Precolonial and Colonial Controls Yes Yes Yes Yes Yes Yes
Region FE Yes Yes Yes Yes Yes Yes
Year FE Yes Yes Yes Yes Yes Yes
Notes: Regressions estimated by OLS. Robust standard errors in parentheses clustered by colonial province. Wild cluster bootstrap (by district) p-values
are in brackets. The unit of observation is an individual. Colonial imprisonment (ST or LT) is the average share of short-term (ST), in columns (1)-(3),
or long-term (LT), in columns (4)-(6), incarcerated populations in each colonial province over 1920 to 1938 as defined in the text. Outcome variables are
from the Afrobarometer samples over 2003 to 2014 and as defined in the main text. Bribe Doc and Bribe HHS is reported frequency of respondent bribery
of government official for document and household services respectively where “Never”=“0”, “Once or Twice”=“1”, “A Few Times ”=“2”, “Often”=“3”.
Fear Crime is how often respondent or family has feared crime in their home where “Never”=“0”, “Just once or twice”=“1”, “Several times”=“2”, “Many
times”=“3”, “Always”=“4”. All regressions use region fixed effects at the geopolitical zone level in Nigeria (for 6 geopolitical zones), year fixed effects and
educational attainment fixed effects. Individual controls include age, age squared and gender. Geographic controls include an indicator for whether the
respondent lives in an urban location, and, at the sub-district or local government area level, include, ruggedness, indicators for petroleum, seacoast and
mean land suitability for agriculture and mean elevation in alternate specifications. Disease controls at the sub-district level include malaria suitability
and tsetse fly suitability in alternate specifications with results unchanged. Precolonial and colonial controls at the ethnicity-level include the level of
precolonial centralization and total exports of slaves from the region during the Atlantic slave trade. ∗∗∗ Significant at the 1 percent level, ∗∗Significant at
the 5 percent level, ∗Significant at the 10 percent level.
52

References
Abiodun, Tosin Funmi. 2017. A historical study on penal confinement and institutional life
in southern Nigeria, 1860-1956 PhD thesis University of Texas at Austin.
Acemoglu, Daron, and A Robinson. 2001. “The Colonial Origins of Comparative Develop-
ment: An Empirical Investigation.” The American Economic Review 91 (5): 1369–1401.
Acemoglu, Daron, and Alexander Wolitzky. 2011. “The economics of labor coercion.” Econo-
metrica 79 (2): 555–600.
Adamson, Christopher. 1984. “Toward a Marxian penology: Captive criminal populations
as economic threats and resources.” Social Problems 31 (4): 435–458.
Adjuik, M, M Bagayoko, F Binka, M Coetzee, J Cox, M Craig, U Deichman, F Don deSav-
igny, C Fraser, E Gouws et al. 1998. “Towards an atlas of malaria risk in Africa.” First
technical report of the Mapping Malaria Risk in Africa/Atlas du Risque de la Malaria en
Afrique (MARA/ARMA) collaboration. Durban, MARA/ARMA .
Ahungwa, Gabriel T, Ueda Haruna, and Rakiya Y Abdusalam. 2014. “Trend analysis of the
contribution of agriculture to the gross domestic product of Nigeria (1960-2012).” Journal
of Agriculture and Veterinary Science 7 (1): 50–55.
Akurang-Parry, Kwabena Opare. 2000. “Colonial forced labor policies for road-building in
southern Ghana and international anti-forced labor pressures, 1900-1940.” African Eco-
nomic History (28): 1–25.
Alsan, Marcella. 2015. “The effect of the tsetse fly on African development.” American
Economic Review 105 (1): 382–410.
Alsan, Marcella, and Marianne Wanamaker. 2018. “Tuskegee and the health of black men.”
The Quarterly Journal of Economics 133 (1): 407–455.
Amare, Mulubrhan, Nathaniel D Jensen, Bekele Shiferaw, and Jennifer Denno Ciss´e. 2018.
“Rainfall shocks and agricultural productivity: Implication for rural household consump-
tion.” Agricultural systems 166: 79–89.
Archibong, Belinda. 2018. “Historical origins of persistent inequality in Nigeria.” Oxford
Development Studies 46 (3): 325–347.
53

Archibong, Belinda. 2019. “Explaining divergence in the long-term effects of precolonial
centralization on access to public infrastructure services in Nigeria.” World Development
121: 123–140.
Ash, Catherine B. 2006. “Forced labor in colonial West Africa.” History Compass 4 (3):
402–406.
Avio, Kenneth. 1998. “The economics of prisons.” European Journal of Law and Economics
6 (2): 143–175.
Becker, Gary S. 1968. “Crime and Punishment: An Economic Approach.” Journal of Political
Economy 76: 169–217.
Bernault, Florence. 2007. “The shadow of rule: Colonial power and modern punishment
in Africa.” Cultures of Confinement: A History of the Prison in Africa, Asia and Latin
America pp. 55–94.
Bhuller, Manudeep, Gordon B Dahl, Katrine V Løken, and Magne Mogstad. 2020. “Incar-
ceration, recidivism, and employment.” Journal of Political Economy 128 (4): 000–000.
Bobonis, Gustavo J, and Peter M Morrow. 2014. “Labor coercion and the accumulation of
human capital.” Journal of Development Economics 108: 32–53.
Buell, Raymond Leslie. 1965. The native problem in Africa. Vol. 2 Archon Books.
Cameron, A Colin, Jonah B Gelbach, and Douglas L Miller. 2008. “Bootstrap-based im-
provements for inference with clustered errors.” The Review of Economics and Statistics
90 (3): 414–427.
Conley, Timothy G. 1999. “GMM estimation with cross sectional dependence.” Journal of
econometrics 92 (1): 1–45.
Cooper, Frederick. 1996. Decolonization and African society: The labor question in French
and British Africa. Vol. 89 Cambridge University Press.
Cox, Robynn. 2010. “Crime, incarceration, and employment in light of the great recession.”
The Review of Black Political Economy 37 (3-4): 283–294.
Dell, Melissa. 2010. “The persistent effects of Peru’s mining mita.” Econometrica 78 (6):
1863–1903.
54

Dell, Melissa, and Benjamin A Olken. 2020. “The development effects of the extractive
colonial economy: The dutch cultivation system in java.” The Review of Economic Studies
87 (1): 164–203.
Dell, Melissa, Benjamin F Jones, and Benjamin A Olken. 2014. “What do we learn from the
weather? The new climate-economy literature.” Journal of Economic Literature 52 (3):
740–98.
Dillon, Andrew, Kevin McGee, and Gbemisola Oseni. 2015. “Agricultural production, dietary
diversity and climate variability.” The Journal of Development Studies 51 (8): 976–995.
Dippel, Christian, Avner Greif, and Dan Trefler. 2020. “Outside options, coercion, and
wages: Removing the sugar coating.” The Economic Journal .
Dube, Oeindrila, and Juan F Vargas. 2013. “Commodity price shocks and civil conflict:
Evidence from Colombia.” The Review of Economic Studies 80 (4): 1384–1421.
Ekechi, Felix K. 1989. Tradition and transformation in Eastern Nigeria: a sociopolitical
history of Owerri and its hinterland, 1902-1947. Kent State Univ Pr.
Feinstein, Charles Hilliard. 1972. National income, expenditure and output of the United
Kingdom, 1855-1965. Vol. 6 Cambridge University Press Cambridge.
Fishman, Ram. 2016. “More uneven distributions overturn benefits of higher precipitation
for crop yields.” Environmental Research Letters 11 (2): 024004.
Foreign, British, and Commonwealth Office. 1947. “Nigeria Annual Report on the Treatment
of Offenders, 1947.”.
Foreign, British, and Commonwealth Office. 1960. “Nigeria Annual Report on the Prisons
Department, North and South Provinces 1914-1960.”.
Frankema, Ewout. 2011. “Colonial taxation and government spending in British Africa,
1880–1940: Maximizing revenue or minimizing effort?” Explorations in Economic History
48 (1): 136–149.
Frankema, Ewout, Jeffrey Williamson, and Pieter Woltjer. 2018. “An economic rationale for
the West African scramble? The commercial transition and the commodity price boom of
1835–1885.” The Journal of Economic History 78 (1): 231–267.
55

Frankema, Ewout, and Marlous Van Waijenburg. 2012. “Structural impediments to African
growth? New evidence from real wages in British Africa, 1880–1965.” The Journal of
Economic History 72 (4): 895–926.
Frankema, Ewout, and Morten Jerven. 2014. “Writing history backwards or sideways: to-
wards a consensus on African population, 1850–2010.” The Economic History Review 67
(4): 907–931.
Freeman, Richard B. 1999. “The economics of crime.” Handbook of labor economics 3: 3529–
3571.
Gallup, John Luke, and Jeffrey D Sachs. 2001. “The economic burden of malaria.” The
American journal of tropical medicine and hygiene 64 (1 suppl): 85–96.
Gardner, Leigh. 2012. Taxing colonial Africa: the political economy of British imperialism.
Oxford University Press.
Gelaro, Ronald, Will McCarty, Max J Su´arez, Ricardo Todling, Andrea Molod, Lawrence
Takacs, Cynthia A Randles, Anton Darmenov, Michael G Bosilovich, Rolf Reichle et al.
2017. “The modern-era retrospective analysis for research and applications, version 2
(MERRA-2).” Journal of climate 30 (14): 5419–5454.
Gregory, Paul R, and Valery Lazarev. 2013. The economics of forced labor: The Soviet Gulag.
Vol. 518 Hoover Institution Press.
Harris, John H. 1914. “Making the Lazy Nigger Work.” Contemporary Review 105: 819–25.
Hynd, Stacey. 2015. ““...a Weapon of Immense Value?” Convict Labour in British Colonial
Africa, c. 1850-1950s, in Christian G. de Vito, Alex Lichtenstein (eds), Global Convict
Labour, Leiden, Brill, Studies in Global Social History, vol. 19, 2015, p. 222-248.”.
Jacobson, Jessica, Catherine Heard, and Helen Fair. 2017. “Prison: Evidence of its use and
over-use from around the world.”.
Jayachandran, Seema. 2006. “Selling labor low: Wage responses to productivity shocks in
developing countries.” Journal of political Economy 114 (3): 538–575.
Jensen, Robert. 2000. “Agricultural volatility and investments in children.” American Eco-
nomic Review 90 (2): 399–404.
56

Juif, D´acil, and Ewout Frankema. 2018. “From coercion to compensation: institutional
responses to labour scarcity in the Central African Copperbelt.” Journal of Institutional
Economics 14 (2): 313–343.
Katz, Lawrence, Steven D Levitt, and Ellen Shustorovich. 2003. “Prison conditions, capital
punishment, and deterrence.” American Law and Economics Review 5 (2): 318–343.
Kaur, Supreet. 2019. “Nominal wage rigidity in village labor markets.” American Economic
Review 109 (10): 3585–3616.
Killingray, David. 1986. “Labour Mobilisation in British Colonial Africa for the War Effort,
1939–46.” In Africa and the Second World War. Springer pp. 68–96.
Killingray, David. 1989. “Labour Exploitation for Military Campaigns in British Colonial
Africa 1870-1945.” Journal of Contemporary History 24 (3): 483–501.
Killingray, David. 1999. “8. Punishment to Fit the Crime? Penal Policy and Practice
in British Colonial Africa.” In Enfermement, prison et chˆatiments en Afrique. Editions
Karthala pp. 181–203.
Kingdon, Donald. 1923. The laws of Nigeria, containing the ordinances of Nigeria, in force
on the 1st day of January, 1923 : and the orders, proclamations, rules, regulations and
bye-laws made thereunder, in force on the 1st day of May, 1923, and the principal imperial
statutes, orders in Council, letters patent and royal instructions relating to Nigeria. Rev.
ed. / prepared under the authority of the revised edition of the laws ordinance, 1923 ; by
donald kingdon. ed. Govt. Printer Lagos.
Knack, Stephen, and Philip Keefer. 1997. “Does social capital have an economic payoff? A
cross-country investigation.” The Quarterly Journal of Economics 112 (4): 1251–1288.
Kunkel, Sarah. 2018. “Forced Labour, Roads, and Chiefs: The Implementation of the ILO
Forced Labour Convention in the Gold Coast.” International Review of Social History
pp. 1–28.
Lesk, Corey, Ethan Coffel, and Radley Horton. 2020. “Net benefits to US soy and maize
yields from intensifying hourly rainfall.” Nature Climate Change 10 (9): 819–822.
Lesk, Corey, Pedram Rowhani, and Navin Ramankutty. 2016. “Influence of extreme weather
disasters on global crop production.” Nature 529 (7584): 84.
57

Lowes, Sara, and Eduardo Montero. 2021a. “Concessions, Violence, and Indirect Rule:
Evidence from the Congo Free State.” The Quarterly Journal of Economics 136 (4): 2047–
2091.
Lowes, Sara, and Eduardo Montero. 2021b. “The legacy of colonial medicine in Central
Africa.” American Economic Review 111 (4): 1284–1314.
Lowes, Sara, Nathan Nunn, James A Robinson, and Jonathan L Weigel. 2017. “The evolution
of culture and institutions: Evidence from the Kuba kingdom.” Econometrica 85 (4): 1065–
1091.
Maul, Daniel Roger. 2007. “The International Labour Organization and the struggle against
forced labour from 1919 to the present.” Labor History 48 (4): 477–500.
Michalopoulos, Stelios, and Elias Papaioannou. 2014. “National institutions and subnational
development in Africa.” The Quarterly Journal of Economics 129 (1): 151–213.
Michalopoulos, Stelios, and Elias Papaioannou. 2016. “The long-run effects of the scramble
for Africa.” American Economic Review 106 (7): 1802–48.
Muller, Christopher. 2018. “Freedom and convict leasing in the postbellum south.” American
Journal of Sociology 124 (2): 367–405.
Muller, Christopher. 2021. “Exclusion and exploitation: The incarceration of Black Ameri-
cans from slavery to the present.” Science 374 (6565): 282–286.
Naidu, Suresh, and Noam Yuchtman. 2013. “Coercive Contract Enforcement: Law and the
Labor Market in Nineteenth Century Industrial Britain.” The American Economic Review
103 (1): 107–144.
Nunn, Nathan. 2008. “The Long-term Effects of Africa’s Slave Trades.” The Quarterly
Journal of Economics 123 (1): 139–176.
Nunn, Nathan, and Diego Puga. 2012. “Ruggedness: The blessing of bad geography in
Africa.” Review of Economics and Statistics 94 (1): 20–36.
Nunn, Nathan, and Leonard Wantchekon. 2011. “The Slave Trade and the Origins of Mistrust
in Africa.” The American Economic Review 101 (7): 3221–3252.
58

O’Flaherty, Brendan, and Rajiv Sethi. 2019. Shadows of Doubt: Stereotypes, Crime, and the
Pursuit of Justice. Harvard University Press.
Ofonagoro, Walter I. 1982. “An aspect of British colonial policy in southern Nigeria: The
problems of forced labour and slavery, 1895–1928.” Studies in southern Nigerian history
pp. 219–44.
Ojomo, A.J., and Etannibi Alemika. 1993. “Information Management and Policing in Nige-
ria.” Edited volume: Policing Nigeria: Past, Present and Future pp. 609–631.
Okia, Opolot. 2012. Communal Labor in Colonial Kenya: The Legitimization of Coercion,
1912–1930. Springer.
Okoye, Dozie. 2021. “Things fall apart? missions, institutions, and interpersonal trust.”
Journal of Development Economics 148: 102568.
Okoye, Dozie, Roland Pongou, and Tite Yokossi. 2019. “New technology, better economy?
The heterogeneous impact of colonial railroads in Nigeria.” Journal of Development Eco-
nomics 140: 320–354.
Onoge, Omafume. 1993. “Social Conflicts and Crime Control in Colonial Nigeria.” Edited
volume: Policing Nigeria: Past, Present and Future pp. 151–186.
Poyker, Michael. 2019. “Economic Consequences of the US Convict Labor System.” Institute
for New Economic Thinking Working Paper Series 1 (91).
Report, Nigeria Annual. 1925. “Nigeria Annual Report on the Prisons Department 1925-
1935.”.
Rohner, Dominic, Mathias Thoenig, and Fabrizio Zilibotti. 2013. “War signals: A theory of
trade, trust, and conflict.” Review of Economic Studies 80 (3): 1114–1147.
Rotimi, Kemi. 1993. “Local Police in Western Nigeria: End of an Era.” Edited volume:
Policing Nigeria: Past, Present and Future pp. 187–238.
Salau, Mohammed Bashir. 2015. “10. Convict labour in early colonial Northern Nigeria:
a preliminary study.” From Dust to Digital: Ten Years of the Endangered Archives Pro-
gramme p. 293.
59

Saleh, Mohamed. 2019. “Export Booms and Labor Coercion: Evidence from the Lancashire
Cotton Famine.”.
Sanny, Josephine Appiah-Nyamekye, and Carolyn Logan. 2020. “Citizens’ Negative Percep-
tions of Police extend well beyond Nigeria’s# EndSARS.” Afrobarometer working paper
.
Sarsons, Heather. 2015. “Rainfall and conflict: A cautionary tale.” Journal of Development
Economics 115: 62–72.
Shajobi-Ibikunle, Gloria D. 2014. “Challenges of imprisonment in the Nigerian penal system:
The way forward.” American Journal of Humanities and Social Sciences 2 (2): 94–104.
Sokoloff, Kenneth L, and Stanley L Engerman. 2000. “Institutions, factor endowments, and
paths of development in the new world.” Journal of Economic perspectives 14 (3): 217–232.
Tanimu, Bashir. 2010. “Nigeria convicts and prison rehabilitation ideals.” Journal of sus-
tainable development in Africa 12 (3): 140–152.
Travis, Jeremy, Bruce Western, and F Stevens Redburn. 2014. The growth of incarceration
in the United States: Exploring causes and consequences. The National Academies Press.
Trevor, Daphne. 1936. “South African Native Taxation.” The Review of Economic Studies 3
(3): 217–225.
van Waijenburg, Marlous. 2018. “Financing the African Colonial State: The Revenue Im-
perative and Forced Labor.” The Journal of Economic History 78 (1): 40–80.
Xie, Hua, Liangzhi You, and Hiroyuki Takeshima. 2017. “Invest in small-scale irrigated
agriculture: A national assessment on potential to expand small-scale irrigation in Nigeria.”
Agricultural Water Management 193: 251–264.
60

A Appendix (For Online Publication)
Contents
1 Introduction 2
2 Prison Labor in Colonial Africa 8
2.1 AHistoryofForcedLabor ............................ 8
2.2 The Prison System in British Colonial Nigeria . . . . . . . . . . . . . . . . . 9
2.3 Labor Shortages, Public Works and Prison Labor . . . . . . . . . . . . . . . 10
2.4 CrimeandPunishment.............................. 13
2.5 Policing and Enforcement . . . . . . . . . . . . . . . . . . . . . . . . . . . . 14
3 Estimating the Value of Prison Labor 15
3.1 EmpiricalStrategy ................................ 15
3.2 Value of Prison Labor Results . . . . . . . . . . . . . . . . . . . . . . . . . . 18
3.3 Prisoner Costs and Relative Value of Prison Labor . . . . . . . . . . . . . . . 18
4 The Effects of Economic Shocks on Incarceration Rates and the Use of
Prison Labor 20
4.1 ConceptualFramework.............................. 20
4.2 DescriptionofData................................ 22
4.2.1 IncarcerationRates............................ 22
4.2.2 EconomicShocks ............................. 23
4.3 Empirical Strategy and Results . . . . . . . . . . . . . . . . . . . . . . . . . 24
4.3.1 Cash Crop Export Prices and Incarceration Rates . . . . . . . . . . . 25
4.3.2 Rainfall Shocks and Incarceration Rates . . . . . . . . . . . . . . . . 26
4.3.3 Robustness ................................ 30
4.4 Further Evidence on Mechanisms . . . . . . . . . . . . . . . . . . . . . . . . 30
4.4.1 Wages and the Railroad . . . . . . . . . . . . . . . . . . . . . . . . . 31
4.4.2 Increasing Prosecutions of Minor Offenses and Sentence-Switching . . 32
4.4.3 Punishment ................................ 33
5 Colonial Imprisonment and Contemporary Trust in Legal Institutions 33
6 Conclusion 36
61

A Appendix (For Online Publication) 61
A.1 Data and Archival Materials . . . . . . . . . . . . . . . . . . . . . . . . . . . 66
A.2 A Further History of Prison Labor in Colonial Africa . . . . . . . . . . . . . 67
A.2.1 North-South Differences in the Distribution of Colonial versus Native
Prisons................................... 69
A.3 Value of Prison Labor Specification Checks . . . . . . . . . . . . . . . . . . . 75
A.3.1 Value of Prison Labor: Adjusting for Inflation . . . . . . . . . . . . . 75
A.3.2 Value of Prison Labor: Measuring Bias in Estimates . . . . . . . . . . 79
A.3.3 Relative Value of Prison Labor: Comparison to Recurrent Maintenance
Public Works Expenditure . . . . . . . . . . . . . . . . . . . . . . . . 79
A.4 Rainfall Shocks and Crop Yields . . . . . . . . . . . . . . . . . . . . . . . . . 83
A.5 Effects of Economic Shocks on Incarceration Rates, Robustness Tables . . . 87
A.5.1 Cash Crop Export Price Shock Results Using Raw Prices, Distance to
RailroadandPrices............................ 87
A.5.2 Accounting for Lags, Leads in Rainfall, Rainfall Controls . . . . . . . 91
A.5.3 Robustness to Population Estimates- Trimming Provinces . . . . . . 91
A.5.4 Effects By Previous Incarceration Status, Alternate Incarceration Mea-
sure .................................... 97
A.5.5 Suggestive Evidence of Sentence-Switching in Response to Short-Term
Economic Shocks, Punishment . . . . . . . . . . . . . . . . . . . . . . 99
A.5.6 Rainfall Shocks and Colonial Incarceration Rates by Region . . . . . 101
A.5.7 Gender................................... 102
A.6 Colonial Imprisonment and Contemporary Trust in Legal Institutions, Ro-
bustness ...................................... 104
A.6.1 Afrobarometer Summary Statistics and Colonial Imprisonment . . . . 104
A.6.2 PossibleChannels............................. 107
A.6.3 Instrumental Variable Strategy and Results . . . . . . . . . . . . . . 110
List of Figures
1 Top 40 countries/territories for incarceration rates, 2018 with Nigeria incar-
ceration rates in red (year 1940) and blue (year 2018). Source: World Prison
Brief ........................................ 38
2 Colonial Nigeria with provinces outlined in 1937, prison locations, regions and
railroadnetworkshown.............................. 39
62

3 Wages, prisoner costs (a) and daily average number in prisons (b) in colonial
Nigeria,1920-1959 ................................ 40
4 Mean number of prisoners per 100,000 population in colonial and postcolonial
Nigeria with independence year highlighted, 1920-1995 . . . . . . . . . . . . 40
5 Share of total convictions by crime over the colonial (1920-1939) and post-
colonial (1977-1993) periods in Nigeria . . . . . . . . . . . . . . . . . . . . . 41
6 Total value of prison labor estimates versus value of prison labor reported by
colonial government in pounds (a) and in log values (b), 1920-1959. Figure
shows values in pounds (a) and log values (b) . . . . . . . . . . . . . . . . . 41
7 Prison populations in colonial (1920) (a) and postcolonial (1980) (b) Nigeria,
and agricultural commodity production areas (c) and prices (d) for the three
major cash crops (palm oil, cocoa, and groundnut) in colonial Nigeria . . . . 42
8 Agricultural commodity export prices and short-term incarceration rates over
the colonial period (1920-1938). Coefficients are from individual regressions
of short-term incarceration on colonial province and year fixed effects and
the interaction between an agricultural commodity presence variable and year
fixed effects; the interaction coefficients are plotted above. . . . . . . . . . . 43
9 Colonial imprisonment and contemporary trust in legal institutions. Top panel
uses the main measure of colonial imprisonment, the share of short-term pris-
oners in penal imprisonment. Bottom panel uses the share of long-term colo-
nialimprisonment................................. 43
A1 Composition of tax revenue in colonial and postcolonial Nigeria, 1930-1980.
Top figure shows the share of direct, petroleum, and indirect (custom/excise)
taxes in total tax revenue in Nigeria. Bottom figure shows the share of direct
and indirect taxes in total government revenue . . . . . . . . . . . . . . . . . 69
A2 Excerpt from the colonial archives highlighting the value of prison labor for
public works (Source: Annual Report on the Prisons Department, Colony and
SouthernProvinces,1920) ............................ 70
A3 Example of colonial archival data on prisons and wages from the British Blue
Books(1922) ................................... 70
A4 Breakdown of estimated public works expenditure, Northern (NP) and South-
ern (SP) Provinces, 1920 and 1935 . . . . . . . . . . . . . . . . . . . . . . . 71
A5 Incarceration rates by sentence in colonial Nigeria . . . . . . . . . . . . . . . 71
A6 Native administration prisons, 1940 . . . . . . . . . . . . . . . . . . . . . . . 73
63

A7 Native prison incarceration rates, 1940 and 1945 . . . . . . . . . . . . . . . . 74
A8 Excerpt from the colonial archives highlighting the value of prison labor for
public works in the northern provinces (Source: Annual Report on the Prisons
Department, Northern Provinces, 1925) . . . . . . . . . . . . . . . . . . . . . 74
A9 Excerpt from the 1925 Annual Report on the Prisons Department, Southern
Provinces on the daily average number of prisoners . . . . . . . . . . . . . . 75
A10 Value of wages for different skill categories in prison and market sectors, 1919-
1925 ........................................ 75
A11 Prison expenditures in colonial Nigeria, Southern provinces, 1919-1921 (Source:
Annual Report on the Prisons Department, Colony and Southern Provinces,
Nigeria,1921)................................... 76
A12 Comparing total values of prison labor estimates with value of prison labor
reported by colonial government (a), and net value of prison labor estimates
with reported colonial value of prison labor (b), plotted against the 45 degree
line of equality, 1920-1959 . . . . . . . . . . . . . . . . . . . . . . . . . . . . 76
A13 Value of prison labor, real vs nominal estimates . . . . . . . . . . . . . . . . 77
A14 Alternate prison and value of labor coercion measures, 1920-1938 . . . . . . 81
A15 Relative value of prison labor, 1920-1959 . . . . . . . . . . . . . . . . . . . . 82
A16 Share of agriculture spending in total government expenditure, 1920-2017 . . 83
A17 Excerpt from archival material: cowpea yields from the Nigeria AAS and
Federal Ministry of Agriculture . . . . . . . . . . . . . . . . . . . . . . . . . 85
A18 No correlation between agricultural commodity export prices and share of
colonial provinces with rainfall shocks, 1920-1938 . . . . . . . . . . . . . . . 87
A19 Agricultural commodity export prices over the colonial period (1920-1938) . 88
A20 Population estimates note from Frankema and Jerven (2014) . . . . . . . . . 95
A21 Share of various punishments in total punishments of prisoners for infractions
while in prison in colonial Nigeria . . . . . . . . . . . . . . . . . . . . . . . . 100
A22 Incarceration rates by gender and sentence, for short-term (ST) and long-term
(LT)sentence,1920-1938............................. 102
A23 Incarceration rates for male and female prisoners, by short-term (ST) and
long-term (LT) sentence, 1920-1938 . . . . . . . . . . . . . . . . . . . . . . . 102
64

List of Tables
1 Decadal averages of relative values of prison labor in colonial Nigeria with
value of prison labor, and value of prison labor as a share of public works
expenditure (top panel), and value of prison labor as a share of expenditure
on prisons and total colonial expenditure (bottom panel) . . . . . . . . . . . 44
2 Summary Statistics: Economic shocks and incarceration rates . . . . . . . . 45
3 Agricultural commodity export prices and colonial incarceration rates (pris-
onersper100,000pop.).............................. 46
4 Rainfall shocks and colonial (1920-1938) and postcolonial (1971-1995) incar-
ceration rates (prisoners per 100,000 pop.), quadratic specification . . . . . . 47
5 Rainfall shocks and colonial (1920-1938) and postcolonial (1971-1995) incar-
ceration rates (prisoners per 100,000 pop.) . . . . . . . . . . . . . . . . . . . 48
6 Reduced-form estimates of the relationship between wages and distance to
railroad and colonial incarceration rates (prisoners per 100,000 pop.) . . . . . 49
7 Agricultural commodity export prices and punishment of prisoners . . . . . . 50
8 OLS estimates of the relationship between colonial imprisonment and present-
day trust in historical legal Institutions versus interpersonal trust . . . . . . 51
9 OLS estimates of the relationship between colonial imprisonment and present-
daycrimeoutcomes................................ 52
A1 Relationship between precolonial centralization and number of colonial vs na-
tiveprisons .................................... 74
A2 Value of prison labor, 1920-1959 . . . . . . . . . . . . . . . . . . . . . . . . . 78
A3 Value of prison labor, real estimates . . . . . . . . . . . . . . . . . . . . . . . 80
A4 Rainfall shocks and crop yields, 1992-1995 . . . . . . . . . . . . . . . . . . . 86
A5 Agricultural commodity export prices and colonial incarceration rates (w/
rainfallcontrols).................................. 89
A6 Agricultural commodity export prices and colonial incarceration rates (raw
prices) ....................................... 90
A7 Reduced-form estimates of the relationship between agricultural commodity
export prices and distance to railroad and colonial incarceration rates . . . . 91
A8 Rainfall shocks and colonial (1920-1938) and postcolonial (1971-1995) incar-
cerationrates(lags)................................ 92
A9 Rainfall shocks (leads) and colonial (1920-1938) incarceration rates . . . . . 93
65

A10 Rainfall shocks in neighboring provinces and colonial (1920-1938) incarcera-
tionrates ..................................... 94
A11 Rainfall shocks and colonial incarceration rates, robustness . . . . . . . . . . 96
A12 Rainfall shocks, agricultural commodity export prices and colonial incarcera-
tion rates by previous incarceration status . . . . . . . . . . . . . . . . . . . 97
A13 Rainfall shocks, agricultural commodity export prices and colonial incarcera-
tion rates, alternate incarceration measure . . . . . . . . . . . . . . . . . . . 98
A14 Agricultural commodity export prices and colonial incarceration rates by cus-
tody/awaiting trial category . . . . . . . . . . . . . . . . . . . . . . . . . . . 99
A15 Rainfall shocks and colonial incarceration rates by custody/awaiting trial cat-
egory........................................ 100
A16 Rainfall shocks and colonial incarceration rates by region . . . . . . . . . . . 101
A17 Rainfall shocks, agricultural commodity prices and colonial incarceration rates
bygender ..................................... 103
A18 Relationship between share of rank and file police in total police force and
colonialimprisonment .............................. 105
A19 Summary Statistics: Afrobarometer Results . . . . . . . . . . . . . . . . . . 106
A20 OLS Estimates: Relationship between colonial imprisonment and trust in
historical legal Institutions versus interpersonal trust by southern ethnicity
status ....................................... 109
A21 OLS Estimates: Soil suitability for palm oil interacted with share of moder-
ate positive rainfall shock years in colonial province instrument and colonial
imprisonment ................................... 111
A22 IV Estimates: Effect of colonial imprisonment on present-day trust in histor-
ical legal Institutions versus interpersonal trust . . . . . . . . . . . . . . . . 112
A.1 Data and Archival Materials
•Primary data from British Online Archives, Nigeria [Colony and Protectorate] Blue
Books, 1914-1940. British Foreign and Commonwealth Office
•Nigeria, Annual Report on the Prisons Department, Northern and Southern Provinces,
1914-1960
•NAI, CSO 26/2 09591 Vol.1 ‘Lieutenant Governor Southern Province to Resident Cal-
abar Province: Memorandum on Prison labor’ 23rd April 1923
•Annual Report on the Treatment of Offenders, 1947, Nigeria
66

•Nigeria Annual Abstract of Statistics, 1975-1997
•Policing in Lagos and Provinces, 1899-1929. Reference: 73242C-01
•Law and Judicial System, 1906-1958 archives. Reference: 73242C-06
•Judicial and Police, 1899-1960
•Nigeria, Agricultural Dept, 1921-1952. Reference: 73242D-04
A.2 A Further History of Prison Labor in Colonial Africa
Between the 1900s and 1950s, millions of Africans were incarcerated in Europe’s African
colonies.67 The unpaid labor of these African prisoners was used extensively for the con-
struction of colonial infrastructure, including the roads and railroads that were needed to
extract, transport and export agricultural and mineral resources, with the aim of raising rev-
enues from Europe’s colonial empire (Bernault, 2007; Hynd, 2015). African prison labor was
recognized as so essential to the functioning of Europe’s colonies, that in 1937, one British
colonial official argued, in his critique of the colonial justice system, that “prison labor is
a weapon of immense value...” (Hynd, 2015).68 In colonial Nigeria, forced labor regulation
included the Native House Rule Ordinance of 1901 and the Roads and Creek Proclamation
of 1903, both of which mandated labor for ‘public purposes’ for all men between 15 and 50
years old and all women between 15 and 45 years old (Ofonagoro, 1982). The Masters and
Servants Proclamations of 1901 and 1903 also instituted forced labor in colonial Nigeria,
granting Native Administrators or chiefs the authority to coerce local laborers for up to 24
working days in a year or 1 out of 12 months. Laborers were frequently employed on public
works projects and physically intensive manual tasks like porterage, carrying pounds of bag-
gage for British officials through often dangerous environments like military expeditions for
“miserable” below market-wage pay (Ofonagoro, 1982; Okia, 2012). This is exemplified by
one account, recorded in Ofonagoro (1982), where in 1925, to defend forced labor recruitment
practices under labor taxes and the use of precolonial communal labor requirements for the
construction of the railroad in the northern provinces, a colonial official stated:
“Were the Government to rely on such labour as can be recruited individually at
current labour rate, it would be impossible to build railways or to undertake any
other public work of any magnitude.” (Ofonagoro, 1982), p. 230.
67In 10 colonies in British colonial Africa alone between 1920 and 1940, the figure was approximately 2
million Africans. Source: Author calculations from archival material described in the Appendix.
68The statement was made by Charles Clifton Roberts, a former magistrate and attorney-general in Nyasa-
land and Uganda (Hynd, 2015).
67

Forced labor was recognized by the colonial regime as so essential to the functioning of
the state, that, in one instance, when the colonial office in Nigeria surveyed commissioners
in 1911 on their preferences for terminating the House Rule Ordinance, which bolstered the
authority of chiefs to coerce labor for the government, the minutes from the meeting report
that “Perhaps most interesting evidence of all is that of the Commissioners who with one
lament ask how is the administration to be carried out if we cannot go to the Head of a
House and demand carriers and paddlers? How is the work of sanitation, road making and
clearing to be carried on if we cannot hold the Head of the House responsible for finishing
the necessary labour? They are all of the opinion that the necessary labour cannot be got,
even at a ruinous price, and that thus the progress and development of the country would
be retarded.” (Ofonagoro, 1982).69
Although prisoners were most often employed on public works, public works expendi-
ture was a small fraction of overall colonial expenditures between 1920 and 1940, composing
an average of 2.8% of colonial expenditures over the period70 . As of 1920, 30% of expen-
diture was on railways, 12% on servicing public debt, and 19% of expenditure was devoted
to defense spending on ‘marine, political and West African Frontier Force’. The majority
of revenues in 1920 were from customs (46%) and railways (23%). By 1936, the share of
expenditure on railways had dropped to 8% of overall expenditure, with public debt, and
pensions and gratuities remaining as the top spending categories for the colonial regime.
Public works expenditure in both years remained low at around 2%. While revenue from the
railway could be used to service railroad expenditure, only 2.8% of colonial expenditures,
on average, was allocated for less costly public works projects, like spending on civil roads,
canals, bridges and “buildings not of a military nature” (e.g. court houses and hospitals).
A breakdown of the top ten, where available, categories for estimated public works
expenditure in 1920 and 1935 for the Northern and Southern provinces is shown in Figure
A471. In the Northern provinces in 1920, roads, public offices, hospitals and court houses
accounted for 80% of overall public works expenditure, while government quarters, industrial
plants and roads accounted for 68% of overall public works expenditure in Southern provinces
in the same year. By 1935, the major public works expenditure categories in both the
Northern and Southern provinces were waterworks, electricity infrastructure projects and
69See also CO/520/107, ‘Native House Rule Ordinance’, minutes by Sir Percy Anderson, 18/12/1911.
70Author’s estimates from Annual Report on Prisons Data over 1920 to 1940.
71We use estimated rather than actual expenditure in a given year to reflect colonial government expecta-
tion around expenditure and to account for unfinished projects and multiple missing entries in the ’spending
to date’ values provided in the Blue Books records.
68

government offices with 100% and 95% of overall public works expenditure in Northern and
Southern Provinces respectively. Convict labor, by colonial officials’ own admissions, was
an essential part of funding these public works projects (Foreign and Office, 1960). The
use of prison labor for colonial public works projects continued through the 1950s in British
colonial Africa with an estimated between 1 in 300 and 1 in 500 Africans imprisoned over
1930 through the 1950s, in contrast with 1 in 2000 British natives in Britain (Hynd, 2015).
0.00
0.25
0.50
0.75
1932 1934 1936 1938 1940 1942 1944 1946 1948 1950 1952 1954 1956 1958 1960 1962 1964 1966 1968 1970 1972 1974 1976 1978 1980 1982
Year
Share of Tax Revenue
variable
custom/tax
direct/tax
petroleum/direct
Share of Direct and Indirect Taxes in Tax Revenue in Nigeria, 1933−1980
0.25
0.50
0.75
1932 1934 1936 1938 1940 1942 1944 1946 1948 1950 1952 1954 1956 1958 1960 1962 1964 1966 1968 1970 1972 1974 1976 1978 1980 1982
Year
Share of Total Govt. Revenue
variable
custom/total
direct/total
Share of Direct and Indirect Taxes in Total Govt. Revenue in Nigeria, 1933−1980
Figure A1: Composition of tax revenue in colonial and postcolonial Nigeria, 1930-1980. Top
figure shows the share of direct, petroleum, and indirect (custom/excise) taxes in total tax
revenue in Nigeria. Bottom figure shows the share of direct and indirect taxes in total
government revenue
A.2.1 North-South Differences in the Distribution of Colonial versus Native
Prisons
There was a dual system of prison administration in Nigeria, under the Native Adminis-
tration, overseen by local chiefs under indirect rule. Under indirect rule, areas with more
centralized precolonial institutions were granted more autonomy to oversee local adminis-
tration, including on the creation and administering of Native Authority prisons. Results
from Table A1 confirm a significant positive correlation between the level of precolonial cen-
tralization and the numbers of native prisons (Archibong, 2019). Although we don’t have
detailed Native Administration prisons data over the 1920 to 1938 period, Figure A6 shows
69

Figure A2: Excerpt from the colonial archives highlighting the value of prison labor for public
works (Source: Annual Report on the Prisons Department, Colony and Southern Provinces,
1920)
Figure A3: Example of colonial archival data on prisons and wages from the British Blue
Books (1922)
70

Offices and Court Room, Kano
Four European Quarters (Type VII) −Kaduna
Office for Station Magistrate and Court House, Kaduna
Native Hospital, Minna
1 European Quarters (Type II), Jos
European Hospital and Fittings, Kaduna
Jos−Bauchi Road
Joinery and Machine Shops, Kaduna
Public Offices, Kaduna
Zaria Sokoto Road
0.0 0.1 0.2 0.3
Share of total estimated expenditure
Category
Public works estimated expenditure breakdown, NP 1920
Butchers’ Stalls − Zaria Township
Landing Ground − Maiduguri
Electricity Scheme − Kaduna
Landing Ground, Kaduna
Landing Ground − Minna
Quaters and Offices, New Protectorate Court
Landing Ground − Kano
Waterworks Scheme − Okene
0.0 0.2 0.4 0.6
Share of total estimated expenditure
Category
Public works estimated expenditure breakdown, NP 1935
Akure Ondo Road
Completion of Agege−Lafenwa Road
Igbara−Oke−Igbara−Odo−Ikere Road
Enugu−Abakaliki Road
Produce Wharves, Lagos
Customs Shed "I", Lagos
Four Quarters, Type IV, Port Harcour t
Four Quarters, Type IV, Benin
Brick and Tile Plant, Ishiago
Ten Quarters, Type IV of 1920, Lagos
0.00 0.05 0.10 0.15
Share of total estimated expenditure
Category
Public works estimated expenditure breakdown, SP 1920
Electricity Supply Works − Enugu
Workshop and Lecture Hall, Higher College, Yaba
Mamfe − Bamenda Road
Waterworks Scheme − Calabar
Waterworks Schemes − Ife
Waterworks Scheme − Abeokuta
Waterworks Scheme − Benin City
Electricty Scheme − Abeokuta
Electricity Supply Works − Lagos
Tank & Non−tank Latrines, Lagos
0.00 0.05 0.10 0.15 0.20
Share of total estimated expenditure
Category
Public works estimated expenditure breakdown, SP 1935
Figure A4: Breakdown of estimated public works expenditure, Northern (NP) and Southern
(SP) Provinces, 1920 and 1935
0
100
200
300
1920 1922 1924 1926 1928 1930 1932 1934 1936 1938
Year
Prisoners ( 105)
prisoners
all
long−term
short−term
Mean Nos. of Prisoners (per 10^5 pop.), 1920−1938, Nigeria
Figure A5: Incarceration rates by sentence in colonial Nigeria
71

the distribution of Native Administration prisons in 1940, for the first year of available data
in the colonial archives.
Native Authority or Administration prisons were more heavily concentrated in the
Northern provinces, which had a more extensive history of organized precolonial institutions
around courts than their southern counterparts (Killingray, 1999). Precolonial political in-
stitutions are proxied using Murdock’s (1967) “Jurisdictional Hierarchy Beyond the Local
Community Level” called the precolonial centralization index here. The precolonial central-
ization index or “Jurisdictional Hierarchy Beyond the Local Community Level” variable is an
index of “political complexity” that assigns a score between 0 to 4 to each ethnic region unit
and describes the number of political jurisdictional hierarchies above the local community
level for each unit. The score is defined as follows: 0 represents so-called “stateless soci-
eties”,“lacking any form of political organization”, 1 and 2 are petty and larger paramount
chiefdoms, 3 and 4 are large, more organized states. Table A1 provides suggestive evidence
of the positive correlation between precolonial centralization and the number of native pris-
ons in a colonial province. While prison labor was a feature of all colonial era prisons, both
Native Administration and colonial government prisons, since Native Authority prisons were
more numerous than colonial prisons72, Native Authority prisons processed more prisoners
than colonial prisons in the north, with the share of prison labor coming primarily from
Native Authority prisons in the Northern provinces.
72On average there were 18 colonial prisons over 1920 to 1938 in the Northern provinces vs 56 Native
Authority prisons in 1940. The ratio for Southern provinces over those periods was 54 to 9. Source: colonial
archives.
72

Figure A6: Native administration prisons, 1940
73

Table A1: Relationship between precolonial centralization and number of colonial vs native
prisons
Native prisons Colonial prisons
(1) (2)
Precolonial centralization 0.599∗0.515
(0.316) (0.339)
Constant 1.447∗∗∗ 2.112∗∗
(0.265) (0.969)
Observations 22 19
R20.124 0.026
Notes: Regressions estimated by OLS. Robust standard errors in parentheses.
Unit of observation is Murdock ethnic region. Precolonial centralization is
Murdock centralization index as defined in text.
∗∗∗Significant at the 1 percent level, ∗∗ Significant at the 5 percent level,
∗Significant at the 10 percent level.
Figure A7: Native prison incarceration rates, 1940 and 1945
Figure A8: Excerpt from the colonial archives highlighting the value of prison labor for
public works in the northern provinces (Source: Annual Report on the Prisons Department,
Northern Provinces, 1925)
74

A.3 Value of Prison Labor Specification Checks
Figure A9: Excerpt from the 1925 Annual Report on the Prisons Department, Southern
Provinces on the daily average number of prisoners
0
20
40
60
1919 1920 1921 1922 1923 1924 1925
Year
Value of wages (pence)
sector
market
prison
Value of wages for bricklayers
0
20
40
1919 1920 1921 1922 1923 1924 1925
Year
Value of wages (pence)
sector
market
prison
Value of wages for carpenters
0
5
10
15
20
25
1919 1920 1921 1922 1923 1924 1925
Year
Value of wages (pence)
sector
market
prison
Value of wages for laborers
0.60
0.65
0.70
0.75
0.80
1919 1920 1921 1922 1923 1924 1925
Year
Fraction below market rate
professions
bricklayer
carpenter
laborer
Percentage of prison wages below market rate
Figure A10: Value of wages for different skill categories in prison and market sectors, 1919-
1925
A.3.1 Value of Prison Labor: Adjusting for Inflation
The measures of values of prison labor used so far have been calculated using nominal values
as shown in Figure A13(a) and Table A2. One potential side effect of using nominal values
when observing trends over time is that is it difficult disentangle the difference between
changes in the observed variable and changes in the price level. To ensure that the trends in
our measure of prison labor are not driven by changes in the price level, we convert the values
into real values using 1920 as the base year, following the technique outlined in Frankema
75

Figure A11: Prison expenditures in colonial Nigeria, Southern provinces, 1919-1921 (Source:
Annual Report on the Prisons Department, Colony and Southern Provinces, Nigeria, 1921)
0
500000
1000000
1500000
0 500000 1000000 1500000
Reported value of prison labor (pounds)
Estimated gross value of prison labor (pounds)
(a) Reported value vs Estimated gross value of prison labor, 1920−1959
0e+00
2e+05
4e+05
0e+00 2e+05 4e+05
Reported value of prison labor (pounds)
Estimated net value of prison labor (pounds)
(b) Reported value vs Estimated net value of prison labor, 1920−1959
Figure A12: Comparing total values of prison labor estimates with value of prison labor
reported by colonial government (a), and net value of prison labor estimates with reported
colonial value of prison labor (b), plotted against the 45 degree line of equality, 1920-1959
76

(2011)73. Figure A13(b) and Table A3 show trends in the value of prison labor, adjusted for
inflation. The trends remain unchanged using real versus nominal estimates of prison labor
and the value of prison labor is not driven by changes in the price level.
Figure A13: Value of prison labor, real vs nominal estimates
73Using Feinstein (1972)’s British price index data.
77

Table A2: Value of prison labor, 1920-1959
Year Total value of
prison labor (PL),
estimate
Net value of PL-
less food costs
Net value of PL-
less prison costs
Total value of PL,
reported
Share of total PL
value in public
works exp.
Share of net PL
value (food) in pub-
lic works exp.
Share of net PL
value (prison) in
public works exp.
1920 178,498.10 55,889.37 1.33 0.42
1921 176,260.50 80, 740.86 27, 912.67 53,661 1.12 0.51 0.18
1922 170,936.80 79, 406.14 19, 618.41 57,312
1923 145,679.00 66, 501.46 -11905.93 64, 244 0.93 0.43 -0.08
1924 176,716.20 112, 860.10 42,908.14 62,222 1.13 0.72 0.27
1925 185,745.60 120, 236.40 47,427.82 60,492 1.17 0.76 0.30
1926 184,522.30 108, 556.80 29,269.52 66,052 1.05 0.62 0.17
1927 188,665.80 110, 374.10 32,701.03 67,859 1.02 0.59 0.18
1928 142,465.90 69, 713.27 -14, 449.62 62, 358 0.71 0.35 -0.07
1929 134,080.40 73, 090.61 8, 683.13 60,851 0.61 0.33 0.04
1930 117,659.00 57, 097.79 -20, 521.35 62, 408 0.48 0.23 -0.08
1931 113,460.70 55, 957.54 -12, 285.62 59, 090 0.44 0.22 -0.05
1932 102,978.70 54, 870.35 -14, 204.48 54, 415 0.41 0.22 -0.06
1933 97,714.65 55,956.14 -2, 798.60 52,434 0.53 0.31 -0.02
1934 102,992.10 59, 841.23 133.75 53,956 0.69 0.40 0.001
1935 94,803.18 62,325.81 -343.81 50,216 0.69 0.45 -0.002
1936 124,892.90 89, 130.29 26, 931.63 44,767 0.98 0.70 0.21
1937 115,976.10 79, 873.06 19, 874.01 44,393 0.83 0.57 0.14
1938 121,687.10 80, 217.16 18, 640.54 49,536 0.72 0.48 0.11
1939 135,812.80 93, 269.02 29, 920.89 54,167 0.75 0.52 0.17
1940 107,276.90 61, 833.98 -4, 521.68 51, 517 0.58 0.34 -0.02
1941 101,133.10 59, 647.90 -11, 764.46 50, 495 0.53 0.32 -0.06
1942 100,486.60 60, 091.00 -30, 949.88 51, 780 0.43 0.26 -0.13
1943 103,498.80 61, 346.58 -34, 436.89 50, 397 0.40 0.24 -0.13
1944 50,640
1945 176,359.10 116, 201.00 0 50, 744 0.60 0.39 0
1946 242,852.30 169, 618.00 28,666.32 56,525 0.73 0.51 0.09
1947 285,395.90 210, 935.60 52,581 0.59 0.43
1948 285,624.40 208, 625.30 -1,372.28 53, 208 1.43 1.04 -0.01
1949 302,473.20 176, 454.90 -127, 471.50 70, 781 1.44 0.84 -0.61
1950 401,825.60 284, 397.10 42,200.86 100, 942 1.77 1.25 0.19
1951
1952 431,855.70 288, 159.40 -15,199.55 118, 364 2.49 1.66 -0.09
1953 518,616.60 352, 824.50 21,240.32 130, 981 2.49 1.69 0.10
1954 631,327.40 2.20
1955 740,092.80 513,126.50 100,460.50 146,406 1.71 1.18 0.23
1956 992,023.60 1.24
1957 1,023, 998.00 745, 241.50 234, 187.20 179, 610 1.09 0.79 0.25
1958 1,133,155.00 818,992.30 177, 577.90 83, 461 1.19 0.86 0.19
1959 1,532,634.00 1,196,574.00 446,565.70 91, 417
78

A.3.2 Value of Prison Labor: Measuring Bias in Estimates
Using the daily average number of prisoners might not properly capture the entire sample
of prisoners whose labor was appropriated by the colonial government. Those who were
charged but sent out on bail for instance would still have to commit their labor but would
not be counted as being in prison.
As an alternative measure to the daily average in prison, we use the number of people
committed to penal imprisonment in each year, that is the number of people who were
arrested and sent to jail for one reason or another and who were expected to serve penal
labor. The number of people committed to prison however does not imply that they spend
the entire year there. Since the Blue Books break down sentences into 3 categories: those
committed for over 2 years, those committed for between 6 months and 2 years, and those
committed for less than 6 months, we weight the number of people committed to prison by
the categories of their duration of stay. Specifically, we assume that those with more than
two-year sentences spend 2 years in prison, those between six-month and two-year sentences
spend 1 year and 3 months in prison, and those with less than six-month sentences spend
3 months in prison. Finally, we assume that imprisonment started at the beginning of the
year hence 1 year in prison would run from January 1st until December 31st.
Figure A14(a) compares the daily average number in prison to our weighted average
measure of people committed to prison for penal imprisonment in each year. The daily
average as measured in the Blue Books tends to be much lower than our weighted average
measure of those committed to prison. This is true especially in the earlier years of our
sample. There however seems to be a convergence in both measures over time.
Recalculating the value of prison labor using our weighted measure of people committed
to prisons shows that using the average number in prison underestimates the value of prison
labor. At its peak, the value of prison labor is more than 60% larger when using the weighted
average of people committed for penal imprisonment compared to using the average number
in prison as shown in Figure A14(b). The trend however remains the same with the value
declining over time.
A.3.3 Relative Value of Prison Labor: Comparison to Recurrent Maintenance
Public Works Expenditure
The relative value of prison labor measures, comparing the value of prison labor to public
works expenditure in the main results used expenditure on new public works construction as
79

Table A3: Value of prison labor, real estimates
Year Real total value of
prison labor (PL),
estimate
Real net value of
PL- less food costs
Real net value of
PL- less prison
costs
Real total value of
PL, reported
1920 178,498.10 55, 889.37
1921 160,933.50 73,719.91 25, 485.49 48, 994.83
1922 134,452.30 62,457.80 15, 431.08 45, 079.40
1923 107,675.80 49,153.25 -8, 800.04 47, 484.70
1924 129,917.90 82,972.27 31, 545.12 45, 744.24
1925 136,556.10 88,395.16 34, 867.89 44, 472.38
1926 134,927.40 79,379.46 21, 402.61 48, 298.89
1927 134,228.60 78,527.04 23, 265.56 48, 279.13
1928 101,359.10 49,598.37 -10,280.36 44, 365.37
1929 94,333.23 51, 423.43 6,109.08 42,812.17
1930 80,454.55 39, 043.15 -14,032.38 42,674.24
1931 74,444.58 36, 715.22 -8,060.92 38, 770.51
1932 65,938.95 35, 134.37 -9,095.36 34, 842.81
1933 61,023.38 34, 944.94 -1,747.74 32, 745.34
1934 64,319.20 37, 371.20 83.53 33, 695.84
1935 59,579.86 39, 169.19 -216.07 31, 558.67
1936 78,983.62 56, 366.98 17,031.86 28, 311.15
1937 76,094.96 52, 406.83 13,039.86 29, 127.42
1938 80,804.11 53, 266.73 12,377.91 32, 893.47
1939 92,868.03 63, 776.84 20,459.74 37, 039.09
1940 85,651.90 49, 369.43 -3,610.20 41, 132.15
1941 89,540.79 52, 810.79 -10, 415.97 44, 707.04
1942 95,323.28 57, 003.32 -29, 359.57 49, 119.37
1943 101,453.40 60,134.20 -33, 756.32 49, 401.01
1944 51,040.32
1945 182,632.70 120, 334.70 0 52,549.12
1946 259,170.50 181, 015.30 30, 592.51 60, 323.12
1947 326,005.60 240, 950.10 60,062.88
1948 351,103.60 256, 452.50 -1, 686.87 65,405.88
1949 382,574.80 223, 184.10 -161,228.80 89, 525.38
1950 524,120.30 370, 952.70 55, 044.60 131, 663.50
1951
1952 670,827.30 447, 615.20 -23,610.36 183, 861.90
1953 830,196.60 564, 798.20 34, 001.31 209, 673.10
1954 1,030, 586.00
1955 1,260, 790.00 874, 140.40 171, 140.20 249,411.00
1956 1,776, 232.00
1957 1,898, 242.00 1, 381,495.00 434, 125.80 332,952.90
1958 2,167, 774.00 1, 566,768.00 339, 714.30 159,664.50
1959 2,944, 111.00 2, 298,557.00 857, 829.70 175,607.40
80

Figure A14: Alternate prison and value of labor coercion measures, 1920-1938
the main category for comparison. The rationale is that new construction represents value-
adding investment in productive public works, as opposed to just upkeep or maintenance.
The archival data also records information on recurrent maintenance public works expendi-
ture, and, in some years between 1920 and 1938 only, an undefined category of public works
expenditure called “extraordinary” expenditure. We estimate the share of prison labor in
total (new and maintenance) public works expenditure and overall (new, maintenance and
the extraordinary category) public works spending. The results are in Figure A15.
Figure A15(c) reports estimates for the share of prison labor in total (new and main-
tenance) public works expenditure from 1920 to 1959. The gross share average is 35% with
the share ranging from 12% to 119%. The net share including the most extensive measures
of prisoner maintenance costs is 3%, with a maximum of up to 24% during this period.
Figure A15(d) reports estimates for the share of prison labor in overall (new, maintenance
and extraordinary) public works expenditure. The gross share average is 25% with the share
ranging from 8% to 119%. The net share including the most extensive measures of prisoner
maintenance costs is 2%, with a maximum of up to 19% during this period.
81

Figure A15: Relative value of prison labor, 1920-1959
82

A.4 Rainfall Shocks and Crop Yields
The share of agriculture spending in total government expenditures in Nigeria has remained
relatively low at around 1% of total expenditures, on average, over the colonial and post-
colonial periods as shown in Figure A16. Farming, which accounts for the major share of
employment, is largely subsistence farming, and irrigated agriculture accounts for only 1%
of cultivated area in the country (Xie, You, and Takeshima, 2017).
Figure A16: Share of agriculture spending in total government expenditure, 1920-2017
The colonial Annual Report on the Agricultural Department documents multiple men-
tions of the links between rainfall shocks and crop yields from 1921 to 1952. Among some
of the excerpts are the following:
•On experimental coffee growth between 1931 and 1932 in Ibadan: “The long dry season
is the limiting factor in the successful cultivation of these better types of coffee, but
the effects of drought can be greatly alleviated by the use of shade trees”
•1942-1943 season: “Unfortunately at a time when maximum production of both food
and export crops was required the whole of Nigeria except the Eastern Provinces ex-
perienced a season in which the rainfall was both very short and badly distributed. In
the North good early rains are followed by a severe drought in June and July and the
total for the year was very much below average in most areas.. . The groundnut and
cotton crops suffered severely as a result of the drought.”
•1950 season, referring to rice cultivation: “Work in the river rain Massagha Swamp
which is being opened by hand, is being extended but considerable damage was done
to a very promising crop by unusually high floods in September. Such losses must be
83

faced in riverain areas in years of exceptional flood and cannot be prevented without
the construction of elaborate levees and sluices, the cost of which is likely to prove
uneconomic.”
•1933 season, referring to yields of cotton and export crops: “The weather in the “export
belt” was unfavorable, partly because the rainfall in July, August, September, when
it is always more than adequate, was exceptionally heavy, but chiefly because only a
fraction of an inch of rain (in some places none at all) fell in October, instead of the
two, three, or four inches in that month which make so much difference to cotton.
These factors caused an exceptionally low yield per acre. . . ”
•1937-38 (1937) season: “Reports from all agricultural stations throughout Nigeria show
that the rainfall in 1937 was below average. In the Southern Provinces, there was a
lack of rain during the early part of the season which resulted in very low yields of
maize, and handicapped the progress of palm planting.”
In the postcolonial period, the Annual Abstract of Statistics provides some disaggre-
gated data on crop yields at the state level in Nigeria between 1992 and 1995 as shown in
Figure A17. The crops include cowpea, mango, palm oil, pepper, soyabeans, tomatoes, and
leafy vegetables, and represent almost one-fifth of domestic production by Food and Agricul-
ture Organization (FAO) estimates. The nonlinear relationship between rainfall shocks and
crop yields is shown in Table A4. The results in Table A4 confirm the inverted-U relation-
ship between rainfall deviations and crop yields (column (1) of Table A4). Extreme negative
rainfall shocks like droughts and extreme positive rainfall shocks like floods decrease crop
yields (column (2) of Table A4).
84

Figure A17: Excerpt from archival material: cowpea yields from the Nigeria AAS and Federal
Ministry of Agriculture
85

Table A4: Rainfall shocks and crop yields, 1992-1995
Outcome: Yield per area
(1) (2) (3) (4) (5)
Rainfall Dev −0.114
(4.254)
[0.985]
Rainfall Dev Sq −17.309∗∗∗
(4.479)
[0.000]
Positive rainfall shock (M) −2.433 −0.952
(1.984) (0.985)
[0.287] [0.353]
Negative rainfall shock (E) −3.195∗−3.523∗∗
(1.718) (1.679)
[0.109] [0.078]
Positive rainfall shock (E) −17.081∗∗∗ −14.587∗∗∗
(2.275) (0.723)
[0.000] [0.000]
Mean of outcome 2.827 2.827 2.827 2.827 2.827
Observations 122 122 122 122 122
Clusters 31 31 31 31 31
District FE Yes Yes Yes Yes Yes
Year FE Yes Yes Yes Yes Yes
Notes: Regressions estimated by OLS. Robust standard errors in parentheses clustered by district or postcolonial state. Wild
cluster bootstrap (by district) p-values are in brackets. Observations are postcolonial states. Dependent variable is average yield
per area, measured in tons per hectare for 7 major crops recorded in the Nigerian Annual Abstract of Statistics: cowpea, mango,
palm oil, pepper, soya beans, tomatoes and leafy vegetables from 1992 to 1995. Rainfall Dev and Rainfall Dev Sq are rainfall
deviation and the squared rainfall deviation term as defined in the text. Positive rainfall shock (M) where (M) is moderate, and
(E) is extreme as defined in text. District FE are postcolonial state fixed effects. ∗∗∗Significant at the 1 percent level, ∗∗Significant
at the 5 percent level, ∗Significant at the 10 percent level based on clustered standard errors in parentheses.
86

A.5 Effects of Economic Shocks on Incarceration Rates, Robustness Tables
A.5.1 Cash Crop Export Price Shock Results Using Raw Prices, Distance to
Railroad and Prices
Figure A18: No correlation between agricultural commodity export prices and share of
colonial provinces with rainfall shocks, 1920-1938
87

Figure A19: Agricultural commodity export prices over the colonial period (1920-1938)
88

Table A5: Agricultural commodity export prices and colonial incarceration rates (w/ rainfall controls)
Outcome:Short-Term Long-Term
(1) (2) (3) (4) (5) (6) (7) (8)
Palm oil x Palm oil price 70.236∗∗∗ 59.364∗∗∗ 2.667 5.192
(26.508) (21.520) (5.358) (3.436)
[0.038] [0.019] [0.732] [0.201]
Cocoa x Cocoa price 43.104∗4.306 −6.023 −6.501∗∗∗
(23.885) (15.868) (5.918) (2.517)
[0.205] [0.809] [0.537] [0.014]
Groundnut x Groundnut price −3.179 −45.769∗−8.412 −8.270∗∗
(31.778) (27.028) (7.111) (4.017)
[0.949] [0.149] [0.465] [0.071]
Mean of outcome 134.659 134.659 134.659 134.659 10.175 10.175 10.175 10.175
Observations 324 324 324 324 324 324 324 324
Clusters 21 21 21 21 21 21 21 21
Rainfall control Yes Yes Yes Yes Yes Yes Yes Yes
District FE Yes Yes Yes Yes Yes Yes Yes Yes
Year FE Yes Yes Yes Yes Yes Yes Yes Yes
Notes: Regressions estimated by OLS. Robust standard errors in parentheses clustered by district, where district is colonial province for colonial data. Wild cluster bootstrap (by
district) p-values are in brackets. Observations are provinces. Dependent variables are prisoners per 100,000 population (1939 pop.) by province in Nigeria broken down by short-term
(less than 6 months) sentence in columns (1) to (4) and long-term (greater than 2 years) sentence in columns (5) to (8) over 1920-1938. Prices are in logs. District FE are colonial
province fixed effects. Rainfall control is total rainfall (in inches) in the district within each year. ∗∗∗ Significant at the 1 percent level, ∗∗Significant at the 5 percent level, ∗Significant
at the 10 percent level based on clustered standard errors in parentheses.
89

Table A6: Agricultural commodity export prices and colonial incarceration rates (raw prices)
Outcome:Short-Term Long-Term
(1) (2) (3) (4) (5) (6) (7) (8)
Palm oil x Palm oil price 20.891∗19.221∗1.028 2.116
(11.955) (9.920) (1.960) (1.393)
[0.131] [0.091] [0.709] [0.201]
Cocoa x Cocoa price 7.949 1.607 −1.537 −1.602∗∗∗
(6.011) (4.026) (1.267) (0.582)
[0.304] [0.729] [0.420] [0.018]
Groundnut x Groundnut price −8.749 −34.540∗−5.926 −6.289∗∗∗
(22.672) (19.243) (4.093) (2.375)
[0.788] [0.122] [0.306] [0.023]
Mean of outcome 134.659 134.659 134.659 134.659 10.175 10.175 10.175 10.175
Observations 324 324 324 324 324 324 324 324
Clusters 21 21 21 21 21 21 21 21
District FE Yes Yes Yes Yes Yes Yes Yes Yes
Year FE Yes Yes Yes Yes Yes Yes Yes Yes
Notes: Regressions estimated by OLS. Robust standard errors in parentheses clustered by district, where district is colonial province for colonial data. Wild cluster bootstrap
(by district) p-values are in brackets. Observations are provinces. Dependent variables are prisoners per 100,000 population (1939 pop.) by province in Nigeria broken down by
short-term (less than 6 months) sentence in columns (1) to (4) and long-term (greater than 2 years) sentence in columns (5) to (8) over 1920-1938. Prices are in pence per kg.
District FE are colonial province fixed effects. ∗∗∗Significant at the 1 percent level, ∗∗Significant at the 5 percent level, ∗Significant at the 10 percent level based on clustered
standard errors in parentheses.
90

Table A7: Reduced-form estimates of the relationship between agricultural commodity ex-
port prices and distance to railroad and colonial incarceration rates
Outcome:Short-Term Long-Term
(1) (2) (3) (4) (5) (6) (7) (8)
Distance to railroad −0.301∗−0.456∗∗ −0.460∗∗ −0.407∗∗ −0.018 −0.022 −0.007 −0.019
(0.157) (0.216) (0.213) (0.194) (0.023) (0.023) (0.024) (0.023)
[0.144] [0.042] [0.028] [0.038] [0.941] [0.403] [0.792] [0.455]
Distance x Palm oil price 0.214∗∗ 0.005
(0.096) (0.019)
[0.059] [0.871]
Distance x Cocoa price 0.151∗∗ −0.010
(0.068) (0.018)
[0.063] [0.789]
Distance x Groundnut price 0.306∗∗ 0.001
(0.129) (0.028)
[0.045] [0.984]
Mean of outcome 46.198 46.198 46.198 46.198 3.990 3.990 3.990 3.990
Observations 938 938 938 938 822 822 822 822
Clusters 21 21 21 21 21 21 21 21
District FE Yes Yes Yes Yes Yes Yes Yes Yes
Year FE Yes Yes Yes Yes Yes Yes Yes Yes
Notes: Regressions estimated by OLS. Robust standard errors in parentheses clustered by district, where district is colonial province for colonial data. Wild cluster
bootstrap (by district) p-values are in brackets. Observations are individual prisons. Dependent variables are prisoners in each prison per 100,000 population of the
province broken down by short-term (less than 6 months) sentence in columns (1)-(4) and long-term (greater than 2 years) sentence in columns (5)-(8) over 1920-1938.
Prices are in logs, and distance to railroad in km. District FE are colonial province fixed effects. ∗∗∗ Significant at the 1 percent level, ∗∗Significant at the 5 percent level,
∗Significant at the 10 percent level based on clustered standard errors in parentheses.
A.5.2 Accounting for Lags, Leads in Rainfall, Rainfall Controls
A.5.3 Robustness to Population Estimates- Trimming Provinces
The colonial incarceration rates presented in the paper are calculated using the population of
the colonial provinces in 1939. One question that may arise is how reliable these population
estimates are, and if any measurement errors in the calculation of these estimates may affect
the results. In their review of historical population estimates in Africa, Frankema and Jerven
(2014) argue that African population estimates in the colonial era are often underestimates
from 1950s period, and that population figures are best guesses as shown in their note in
Figure A20. Any underestimates to the population figures will not substantially affect our
results unless the underestimates vary systematically by province. One way to test this is
to conduct trimming exercises, dropping potentially significant provinces from the sample to
see if the results hold.
One assumption is that any error in population estimates would be in the most populous
provinces (Kano is the most popular province in 1939) and centered around the capitals of
the southern (Lagos/Colony province) and northern (Niger till 1923, and Zaria from 1923-
91

Table A8: Rainfall shocks and colonial (1920-1938) and postcolonial (1971-1995) incarcera-
tion rates (lags)
Panel A: Rainfall Shocks and Incarceration Rates, Quadratic Specification
Period:Colonial Postcolonial
Outcome:ST, t+1 LT, t+1 ST, t+2 LT, t+2 PC, t+1 PC, t+2
(1) (2) (3) (4) (5) (6)
Rainfall Dev 8.125 0.638 3.129 0.0005 −22.836∗∗∗ 6.114
(6.208) (1.361) (7.472) (1.287) (8.453) (9.394)
[0.252] [0.709] [0.704] [1.000] [0.013] [0.539]
Rainfall Dev Sq −3.347 0.454 −0.365 0.327 −0.746 1.633
(2.797) (0.754) (2.117) (0.494) (14.796) (13.400)
[0.353] [0.765] [0.913] [0.628] [0.960] [0.893]
Mean of outcome 134.381 10.432 135.426 10.634 106.348 107.592
Panel B: Rainfall Shocks and Incarceration Rates, Linear Specification
Period:Colonial Postcolonial
Outcome:ST, t+1 LT, t+1 ST, t+2 LT, t+2 PC, t+1 PC, t+2
(1) (2) (3) (4) (5) (6)
Positive rainfall shock (M) −0.112 −1.796∗−3.892 2.201 −6.504 0.183
(8.073) (1.027) (8.632) (2.530) (6.488) (4.804)
[0.990] [0.092] [0.668] [0.558] [0.345] [0.977]
Negative rainfall shock (E) −27.309∗∗ −2.054 −17.044 0.536 20.871∗∗ 14.779∗
(13.399) (3.075) (14.496) (2.226) (9.760) (7.951)
[0.074] [0.608] [0.340] [0.861] [0.042] [0.098]
Positive rainfall shock (E) −9.815 1.405 −10.440 0.565 −2.906 19.127∗∗
(12.121) (2.345) (13.238) (1.695) (11.747) (8.176)
[0.492] [0.584] [0.546] [0.737] [0.826] [0.035]
Mean of outcome 134.381 10.432 135.426 10.634 106.348 107.592
Observations 310 310 296 296 555 554
Clusters 21 21 21 21 36 36
District FE Yes Yes Yes Yes Yes Yes
Year FE Yes Yes Yes Yes Yes Yes
Notes: Regressions estimated by OLS. Robust standard errors in parentheses clustered by district, where district is colonial province for colonial
data, and postcolonial state for postcolonial data. Wild cluster bootstrap (by district) p-values are in brackets. Observations are districts.
Dependent variables in column (1)-(4) are colonial-era prisoners per 100,000 population (1939 pop.) by province in Nigeria broken down by
short-term (ST) (less than 6 months) sentence and long-term (LT) (greater than 2 years) sentence over 1920-1938. Dep endent variable in (5) and
(6) are postcolonial period prisoners per 100,000 population (1990 pop.) by state in Nigeria from 1971-1995. Outcomes are denoted t+1 for
outcomes 1 year later and t+2 for outcomes 2 years later. Rainfall deviation as defined in text. District FE are colonial province fixed effects in
(1)-(4), and postcolonial state fixed effects in (5)-(6). ∗∗∗Significant at the 1 percent level, ∗∗Significant at the 5 percent level, ∗Significant at the
10 percent level based on clustered standard errors in parentheses.
92

Table A9: Rainfall shocks (leads) and colonial (1920-1938) incarceration rates
Period:Colonial
Outcome:Short-Term Long-Term Short-Term Long-Term
(1) (2) (3) (4)
Rainfall Dev, t+1 12.505∗∗ 0.326
(6.029) (1.665)
[0.049] [0.900]
Rainfall Dev Sq, t+1 −1.400 0.456
(2.992) (0.671)
[0.693] [0.788]
Positive rainfall shock (M), t+1 7.824 1.543
(7.410) (1.663)
[0.356] [0.382]
Negative rainfall shock (E), t+1 −16.501 −0.505
(10.748) (2.833)
[0.183] [0.912]
Positive rainfall shock (E) , t+1 10.786 0.744
(17.275) (1.962)
[0.601] [0.713]
Mean of outcome 134.659 10.175 134.659 10.175
Observations 304 304 304 304
Clusters 21 21 21 21
District FE Yes Yes Yes Yes
Year FE Yes Yes Yes Yes
Notes: Regressions estimated by OLS. Robust standard errors in parentheses clustered by district, where district
is colonial province for colonial data. Wild cluster bootstrap (by district) p-values are in brackets. Observations
are districts. Dependent variables in columns (1)-(4) are prisoners per 100,000 population (1939 pop.) by province
in Nigeria broken down by short-term (less than 6 months) sentence and long-term (greater than 2 years) sentence
over 1920-1938. Positive rainfall shock (M) where (M) is moderate, and (E) is extreme as defined in text. One year
in the future is denoted by t+1. District FE are colonial province fixed effects. ∗∗∗Significant at the 1 percent
level, ∗∗Significant at the 5 percent level, ∗Significant at the 10 percent level based on clustered standard errors in
parentheses.
93

Table A10: Rainfall shocks in neighboring provinces and colonial (1920-1938) incarceration
rates
Period:Colonial
Outcome:Short-Term Long-Term Short-Term Long-Term
(1) (2) (3) (4)
Rainfall Dev, j 12.238 −1.803
(8.103) (1.348)
[0.172] [0.266]
Rainfall Dev Sq, j 0.821 0.089
(2.583) (0.411)
[0.765] [0.839]
Positive rainfall shock (M), j 3.831 −2.475
(10.474) (2.759)
[0.774] [0.450]
Negative rainfall shock (E), j −19.266 0.431
(14.171) (2.446)
[0.236] [0.861]
Positive rainfall shock (E), j 19.822 −3.398∗
(15.293) (1.838)
[0.302] [0.167]
Mean of outcome 134.659 10.175 134.659 10.175
Observations 320 320 320 320
Clusters 21 21 21 21
District FE Yes Yes Yes Yes
Year FE Yes Yes Yes Yes
Notes: Regressions estimated by OLS. Robust standard errors in parentheses clustered by district, where district
is colonial province for colonial data. Wild cluster bootstrap (by district) p-values are in brackets. Observations
are districts. Dependent variables in columns (1)-(4) are prisoners per 100,000 population (1939 pop.) by province
in Nigeria broken down by short-term (less than 6 months) sentence and long-term (greater than 2 years) sentence
over 1920-1938. Positive rainfall shock (M) where (M) is moderate, and (E) is extreme as defined in text. Nearest
neighbor districts by prison distance are denoted by j. District FE are colonial province fixed effects. ∗∗∗Significant
at the 1 percent level, ∗∗Significant at the 5 percent level, ∗Significant at the 10 percent level based on clustered
standard errors in parentheses.
94

1966) regions which may attract relatively higher levels of migration, then we can examine
robustness of the results by trimming or dropping these provinces. The results in Table A11
are largely stable and qualitatively similar to the main results in Table 5.
Figure A20: Population estimates note from Frankema and Jerven (2014)
95

Table A11: Rainfall shocks and colonial incarceration rates, robustness
Outcome:Short-Term Long-Term
(1) (2) (3) (4) (5) (6) (7) (8)
Positive rainfall shock (M) 14.075∗12.144∗10.975 11.909 −0.907 −0.692 −1.129 −0.695
(7.912) (6.967) (7.255) (7.521) (1.237) (1.438) (1.526) (1.525)
[0.119] [0.106] [0.171] [0.138] [0.548] [0.655] [0.514] [0.670]
Negative rainfall shock (E) −8.590 −17.141∗−16.771 −15.290 −3.528 −0.403 −0.335 −0.166
(7.350) (10.361) (10.864) (10.891) (2.284) (3.554) (3.607) (3.597)
[0.268] [0.152] [0.197] [0.243] [0.187] [0.897] [0.901] [0.954]
Positive rainfall shock (E) 3.102 −0.469 −3.586 −2.238 1.712 3.337 2.997 2.902
(14.613) (14.146) (14.364) (13.988) (2.211) (2.632) (2.716) (2.692)
[0.835] [0.980] [0.824] [0.882] [0.517] [0.280] [0.311] [0.333]
Mean of outcome 126.542 136.751 140.583 138.096 8.843 10.324 10.661 9.825
Observations 305 319 309 305 305 319 309 305
Clusters 20 20 20 20 20 20 20 20
Province dropped Lagos Niger Kano Zaria Lagos Niger Kano Zaria
District FE Yes Yes Yes Yes Yes Yes Yes Yes
Year FE Yes Yes Yes Yes Yes Yes Yes Yes
Notes: Regressions estimated by OLS. Robust standard errors in parentheses clustered by district, where district is colonial province for colonial data, and
postcolonial state for postcolonial data. Wild cluster bootstrap (by district) p-values are in brackets. Observations are districts.Dependent variables are prisoners
per 100,000 population (1939 pop.) by province in Nigeria broken down by short-term (less than 6 months) sentence in columns (1) to (4) and long-term (greater
than 2 years) sentence in columns (5) to (8) over 1920-1938. Positive rainfall shock (M) where (M) is moderate, and (E) is extreme as defined in text. District FE
are colonial province fixed effects. ∗∗∗ Significant at the 1 percent level, ∗∗Significant at the 5 percent level, ∗Significant at the 10 percent level based on clustered
standard errors in parentheses.
96

A.5.4 Effects By Previous Incarceration Status, Alternate Incarceration Mea-
sure
Table A12: Rainfall shocks, agricultural commodity export prices and colonial incarceration
rates by previous incarceration status
Panel A: Rainfall Shocks and Incarceration
Outcome:One Previous Two Previous Three Previous
(1) (2) (3)
Positive rainfall shock (M) −2.601 1.106 −0.604
(10.213) (1.220) (0.765)
[0.830] [0.477] [0.523]
Negative rainfall shock (E) −19.525∗∗ −0.446 −1.085
(8.446) (1.234) (1.833)
[0.037] [0.792] [0.826]
Positive rainfall shock (E) −1.816 −0.311 −0.968
(6.334) (0.815) (1.035)
[0.781] [0.692] [0.412]
Mean of outcome 38.330 6.413 4.704
Observations 324 324 324
Clusters 21 21 21
Panel B: Agricultural Commodity Prices and Incarceration
Outcome:One Previous Two Previous Three Previous
(1) (2) (3)
Palm oil x Palm oil price 32.964∗∗∗ 0.024 3.335
(7.587) (2.471) (5.487)
[0.003] [0.996] [0.727]
Cocoa x Cocoa price −4.440 3.363∗7.595
(8.805) (1.879) (5.427)
[0.628] [0.246] [0.372]
Groundnut x Groundnut price −20.032 −2.329 3.691
(20.498) (3.170) (7.572)
[0.507] [0.597] [0.798]
Mean of outcome 38.330 6.413 4.704
Observations 324 324 324
Clusters 21 21 21
District FE Yes Yes Yes
Year FE Yes Yes Yes
Notes: Regressions estimated by OLS. Robust standard errors in parentheses clustered by district, where district is colonial province
for colonial data. Wild cluster b ootstrap (by district) p-values are in brackets. Observations are provinces. Dependent variables are
incarceration rates or prisoners per 100,000 population (1939 pop.) by province in Nigeria broken down by number of prisoners with
one previous sentence (1), two previous sentences (2), and 3 or more previous sentences (3). Positive rainfall shock (M) where (M)
is moderate, and (E) is extreme as defined in text. Prices are in logs. District FE are colonial province fixed effects. ∗∗∗Significant
at the 1 percent level, ∗∗Significant at the 5 percent level, ∗Significant at the 10 percent level based on clustered standard errors in
parentheses.
97

Table A13: Rainfall shocks, agricultural commodity export prices and colonial incarceration rates, alternate incarceration
measure
Outcome:Share Short-Term Share Long-Term
(1) (2) (3) (4) (5) (6)
Rainfall Dev −0.003 0.008
(0.016) (0.007)
[0.877] [0.283]
Rainfall Dev Sq −0.005 0.002
(0.004) (0.003)
[0.250] [0.612]
Positive rainfall shock (M) −0.006 0.010
(0.017) (0.014)
[0.748] [0.537]
Negative rainfall shock (E) −0.021 0.028
(0.031) (0.018)
[0.544] [0.202]
Positive rainfall shock (E) −0.032 0.048∗∗∗
(0.022) (0.013)
[0.194] [0.006]
Palm oil x Palm oil price −0.026 −0.004
(0.051) (0.031)
[0.647] [0.901]
Cocoa x Cocoa price 0.032 −0.053
(0.050) (0.035)
[0.567] [0.284]
Groundnut x Groundnut price −0.003 −0.034
(0.076) (0.063)
[0.975] [0.681]
Mean of outcome 0.764 0.764 0.764 0.111 0.111 0.111
Observations 324 324 324 324 324 324
Clusters 21 21 21 21 21 21
District FE Yes Yes Yes Yes Yes Yes
Year FE Yes Yes Yes Yes Yes Yes
Notes: Regressions estimated by OLS. Robust standard errors in parentheses clustered by district, where district is colonial province for colonial
data, and postcolonial state for postcolonial data. Wild cluster bo otstrap (by district) p-values are in brackets. Observations are districts.
Dependent variables are the share of short-term sentenced ((1)-(3)) and share of long-term sentenced ((3)-(6)) prisoners in sentenced prisoners
by province in Nigeria over 1920-1938. Rainfall deviation, and rainfall deviation squared (Rainfall Dev and Rainfall Dev Sq) as defined in text.
Positive rainfall shock (M) where (M) is moderate, and (E) is extreme as defined in text. Prices are in logs. District FE are colonial province fixed
effects. ∗∗∗Significant at the 1 percent level, ∗∗Significant at the 5 percent level, ∗Significant at the 10 percent level based on clustered standard
errors in parentheses.
98

A.5.5 Suggestive Evidence of Sentence-Switching in Response to Short-Term
Economic Shocks, Punishment
Table A14: Agricultural commodity export prices and colonial incarceration rates by cus-
tody/awaiting trial category
Outcome:Custody Short-Term Custody −Short-Term
(1) (2) (3)
Palm oil x Palm oil price 35.976∗66.681∗∗ −30.705∗
(21.791) (27.920) (18.225)
[0.181] [0.048] [0.168]
Cocoa x Cocoa price 15.023 41.965∗−26.943∗
(13.962) (23.638) (13.767)
[0.451] [0.185] [0.131]
Groundnut x Groundnut price 34.607∗2.809 31.798∗∗∗
(19.750) (29.852) (11.956)
[0.156] [0.956] [0.015]
Mean of outcome 71.727 134.659 −62.932
District FE Yes Yes Yes
Year FE Yes Yes Yes
Observations 324 324 324
Clusters 21 21 21
Notes: Regressions estimated by OLS. Robust standard errors in parentheses clustered by district, where district is
colonial province for colonial data, and postcolonial state for postcolonial data. Wild cluster bootstrap (by district)
p-values are in brackets. Observations are provinces. Dependent variable in (1) is prisoners awaiting custody or trial
per 100,000 population (1939 pop.) and in (2) is short-term prisoners with less than 6 months sentences respectively.
Outcome in (3) is the difference between the custody/awaiting trial incarceration rate and the short-term, less than 6
months sentence incarceration rate. District FE are colonial province fixed effects. Prices are in logs. ∗∗∗ Significant
at the 1 percent level, ∗∗Significant at the 5 percent level, ∗Significant at the 10 percent level based on clustered
standard errors in parentheses.
99

Table A15: Rainfall shocks and colonial incarceration rates by custody/awaiting trial cate-
gory
Outcome:Custody Short-Term Custody −Short-Term
(1) (2) (3) (4) (5) (6)
Positive rainfall shock (M) 5.623∗∗ 1.774 16.727∗∗∗ 12.142∗−11.104∗∗ −10.368
(2.201) (2.795) (5.456) (6.964) (4.554) (6.475)
[0.014] [0.558] [0.016] [0.093] [0.040] [0.154]
Negative rainfall shock (E) −6.703 −17.225∗10.523
(6.396) (10.259) (8.004)
[0.371] [0.139] [0.241]
Positive rainfall shock (E) −6.734∗−0.404 −6.331
(4.044) (13.973) (13.161)
[0.093] [0.977] [0.615]
Mean of outcome 71.727 71.727 134.659 134.659 −62.932 −62.932
District FE Yes Yes Yes Yes Yes Yes
Year FE Yes Yes Yes Yes Yes Yes
Observations 324 324 324 324 324 324
Clusters 21 21 21 21 21 21
Notes: Regressions estimated by OLS. Robust standard errors in parentheses clustered by district, where district is colonial province for colonial
data, and postcolonial state for postcolonial data. Wild cluster bootstrap (by district) p-values are in brackets. Observations are provinces.
Dependent variables in (1)-(2) and (3)-(4) are prisoners awaiting custody or trial per 100,000 population (1939 p op.) and short-term prisoners
with less than 6 months sentences respectively. Outcome in (5)-(6) is the difference between the custody/awaiting trial incarceration rate and the
short-term, less than 6 months sentence incarceration rate. Positive rainfall shock (M) where (M) is moderate, and (E) is extreme as defined in
text. District FE are colonial province fixed effects in (1)-(6). ∗∗∗Significant at the 1 percent level, ∗∗Significant at the 5 percent level, ∗Significant
at the 10 percent level based on clustered standard errors in parentheses.
0.0
0.1
0.2
0.3
0.4
0.5
diet
flogging
marks
prison
solitary
Punishment
Share of total punishment
punishment
diet
flogging
marks
prison
solitary
Mean share of total punishment, 1920−1938
Figure A21: Share of various punishments in total punishments of prisoners for infractions
while in prison in colonial Nigeria
100

A.5.6 Rainfall Shocks and Colonial Incarceration Rates by Region
Table A16 reports estimates from the heterogeneity by region analysis. The positive rela-
tionship between moderate positive rainfall shocks and colonial incarceration rates is driven
by short-term incarceration in the southern provinces where the most productive cash crops,
palm oil and cocoa, are located.
Table A16: Rainfall shocks and colonial incarceration rates by region
Panel A: Rainfall Shocks and Colonial Incarceration Rates, Quadratic Specification
Outcome: Short-Term Long-Term
Sample:All South North All South North
(1) (2) (3) (4) (5) (6)
Rainfall Dev 11.995∗∗ 18.884∗1.978 0.759 −0.071 0.236
(5.876) (11.046) (1.234) (1.227) (2.201) (0.338)
[0.065] [0.142] [0.205] [0.655] [0.989] [0.454]
Rainfall Dev Sq −4.884∗−8.686∗∗ 0.860∗∗∗ 0.752 1.381 0.062
(2.572) (4.235) (0.309) (0.739) (1.346) (0.098)
[0.068] [0.046] [<.001] [0.494] [0.541] [0.675]
Mean of outcome 134.659 217.517 18.657 10.175 14.743 3.781
Observations 324 189 135 324 189 135
Clusters 21 10 11 21 10 11
Panel B: Rainfall Shocks and Colonial Incarceration Rates, Linear Specification
Outcome: Short-Term Long-Term
Sample:All South North All South North
(1) (2) (3) (4) (5) (6)
Positive rainfall shock (M) 16.727∗∗∗ 24.826∗∗∗ 0.392 −1.638 −2.609 −0.573
(5.456) (7.795) (1.086) (1.319) (2.127) (0.446)
[0.016] [0.009] [0.729] [0.336] [0.408] [0.174]
Mean of outcome 134.659 217.517 18.657 10.175 14.743 3.781
Observations 324 189 135 324 189 135
Clusters 21 10 11 21 10 11
District FE Yes Yes Yes Yes Yes Yes
Year FE Yes Yes Yes Yes Yes Yes
Notes: Regressions estimated by OLS. Robust standard errors in parentheses clustered by district, where district is colonial province for colonial data. Wild
cluster bootstrap (by district) p-values are in brackets. Observations are provinces. Dependent variables are prisoners per 100,000 population (1939 pop.) by
province in Nigeria broken down by short-term (less than 6 months) sentence( (1)-(3))and long-term (greater than 2 years) sentence((4)-(6)) over 1920-1938.
Rainfall Dev is rainfall deviation from the quadratic specification as defined in the text. Positive rainfall shock (M) where (M) is moderate as defined in the text.
District FE are colonial province fixed effects. ∗∗∗ Significant at the 1 percent level, ∗∗Significant at the 5 percent level, ∗Significant at the 10 percent level based
on clustered standard errors in parentheses.
101

A.5.7 Gender
Women account for just 6% of incarcerated populations between 1920 and 1938, and also had
to work in prisons. They were usually assigned to light labor, cleaning and cooking in prisons.
The effects of economic shocks on incarceration rates are largely driven by male prisoners as
shown in the figures and table below. In ongoing research we explore the effects of women led
protests on the female incarceration rates, particularly for long-term incarceration as well.
Figure A22: Incarceration rates by gender and sentence, for short-term (ST) and long-term
(LT) sentence, 1920-1938
Figure A23: Incarceration rates for male and female prisoners, by short-term (ST) and
long-term (LT) sentence, 1920-1938
102

Table A17: Rainfall shocks, agricultural commodity prices and colonial incarceration rates
by gender
Panel A: Rainfall Shocks and Colonial Incarceration Rates
Outcome:Short-Term Long-Term
Group:All Male Female All Male Female
(1) (2) (3) (4) (5) (6)
Positive rainfall shock (M) 12.142∗6.057 0.156 −0.695 −0.524 −0.117∗∗
(6.964) (4.489) (1.300) (1.437) (1.172) (0.051)
[0.093] [0.212] [0.926] [0.683] [0.678] [0.020]
Negative rainfall shock (E) −17.225∗−19.949∗∗ −5.677∗∗∗ −0.429 −0.498 −0.140∗∗
(10.259) (7.980) (1.539) (3.530) (3.408) (0.071)
[0.139] [0.025] [0.004] [0.886] [0.808] [0.093]
Positive rainfall shock (E) −0.404 5.199 0.680 3.358 1.844 0.100
(13.973) (8.665) (2.460) (2.654) (1.840) (0.106)
[0.977] [0.576] [0.781] [0.293] [0.367] [0.426]
Mean of outcome 134.659 109.005 13.282 10.175 9.164 0.184
Observations 324 316 316 324 316 316
Clusters 21 21 21 21 21 21
Panel B: Agricultural Commodity Prices and Colonial Incarceration Rates
Outcome:Short-Term Long-Term
Group:All Male Female All Male Female
(1) (2) (3) (4) (5) (6)
Palm oil x Palm oil price 66.681∗∗ 56.482∗∗∗ 16.112∗∗∗ 2.738 4.382 0.375∗∗
(27.920) (17.015) (3.680) (5.448) (5.033) (0.164)
[0.048] [0.009] [0.008] [0.745] [0.618] [0.053]
Cocoa x Cocoa price 41.965∗28.612∗∗ 0.874 −6.000 −5.294 −0.036
(23.638) (13.577) (1.238) (5.952) (5.712) (0.042)
[0.185] [0.070] [0.517] [0.521] [0.570] [0.434]
Groundnut x Groundnut price 2.809 0.171 1.181 −8.532 −7.468 −0.087∗∗
(29.852) (23.161) (1.960) (6.905) (7.534) (0.039)
[0.956] [0.998] [0.611] [0.416] [0.530] [0.063]
Mean of outcome 134.659 109.005 13.282 10.175 9.164 0.184
Observations 324 316 316 324 316 316
Clusters 21 21 21 21 21 21
District FE Yes Yes Yes Yes Yes Yes
Year FE Yes Yes Yes Yes Yes Yes
Notes: Regressions estimated by OLS. Robust standard errors in parentheses clustered by district, where district is colonial province for colonial data.
Wild cluster bootstrap (by district) p-values are in brackets. Observations are provinces. Dependent variables are incarceration rates or prisoners per
100,000 population (1939 pop.) by province in Nigeria broken down by short-term (less than 6 months) sentence ((1)-(3))and long-term (greater than 2
years) sentence ((4)-(6)) over 1920-1938; incarceration rates for all, male and female prisoners as specified in the table. Positive rainfall shock (M) where
(M) is moderate, and (E) is extreme as defined in text. Prices are in logs. District FE are colonial province fixed effects. ∗∗∗ Significant at the 1 percent
level, ∗∗Significant at the 5 percent level, ∗Significant at the 10 percent level based on clustered standard errors in parentheses.
103

A.6 Colonial Imprisonment and Contemporary Trust in Legal Institutions, Ro-
bustness
A.6.1 Afrobarometer Summary Statistics and Colonial Imprisonment
Given the rich literature on the long-term impacts of historical institutions, and coercive
labor institutions in particular, on contemporary attitudes and outcomes, to explore the long-
term impacts of exposure to colonial imprisonment driven primarily by economic motives
around prison labor, on views of state legitimacy, we use geocoded data from all rounds
of the Afrobarometer surveys for Nigeria. We use Afrobarometer surveys from all 5 rounds
from 2003, 2005, 2008, 2012 and 2014. Our main outcomes of interest are, following previous
literature (Nunn and Wantchekon, 2011; Lowes and Montero, 2021b), respondent reported
trust in institutions or individuals variables. Trust outcomes are reported trust levels on a
scale of 0-3, where “Not at all”= “0”, “Just a little”=“1”, “Somewhat”=“2”, “A lot”=“3”.
Specifically, we use data on trust in historical legal institutions namely: trust in police,
courts, and trust in tax administration and interpersonal trust: trust in neighbors, trust in
relatives, and trust in the elected local governing council member, to test the hypothesis that
long-term exposure to colonial imprisonment centered around prison labor reduces views of
state legitimacy through lowered trust in legal institutions.
In addition to individual level controls for age and gender and education fixed effects
and an indicator that equals one if the respondent lives in an urban location, to control for
potential covariates that could impact both exposure to long-term colonial imprisonment
and trust in legal institutions, we combine the Afrobarometer data with geographic controls,
disease controls and controls for precolonial and colonial institutions, with descriptions of
the data and summary statistics shown in Table A19. Precolonial political institutions are
proxied using Murdock’s (1967) “Jurisdictional Hierarchy Beyond the Local Community
Level” called the Precolonial centralization index here. The precolonial centralization index
or “Jurisdictional Hierarchy Beyond the Local Community Level” variable is an index of
“political complexity” that assigns a score between 0 to 4 to each ethnic region unit and
describes the number of political jurisdictional hierarchies above the local community level for
each unit. The score is defined as follows: 0 represents so-called “stateless societies”,“lacking
any form of political organization”, 1 and 2 are petty and larger paramount chiefdoms, 3 and
4 are large, more organized states. The colonial institutions include Nunn and Wantchekon
(2011)’s total number of exported slaves in the trans Atlantic and Indian ocean slave trades
from 1400-1900. Disease controls are included for malaria by using climatic suitability for
104

malaria transmission from Adjuik et al. (1998) to address the various hypotheses in the
literature on the negative impacts of malaria on African development outcomes (Gallup and
Sachs, 2001) and tsetse fly suitability following Alsan (2015). Geographic controls include
land suitability for agriculture from FAO, mean elevation in km, ruggedness of the terrain
(following Nunn and Puga (2012)), and indicators for sea coast and petrol, to control for
access to trade routes and mineral wealth on trust outcomes.
Table A18: Relationship between share of rank and file police in total police force and
colonial imprisonment
Outcome:Colonial Imprisonment (ST) Colonial Imprisonment (LT) Short-Term Long-Term
(1) (2) (3) (4)
Share of rank and file police 0.021∗∗∗ −0.001 5.000∗0.433
(0.005) (0.003) (2.743) (0.587)
[0.000] [0.692] [0.078] [0.547]
Mean of outcome 0.764 0.111 134.659 10.175
Observations 234 234 234 234
Clusters 19 19 19 19
District FE Yes Yes Yes Yes
Year FE Yes Yes Yes Yes
Notes: Regressions estimated by OLS. Robust standard errors in parentheses clustered by colonial province. Wild cluster bootstrap (by district) p-values
are in brackets. Observations are provinces. Covariate is the share of rank and file police in the total police force. Outcomes in columns (1) and (2) are
Colonial imprisonment (ST or LT), which is the average share of short-term (ST) or long-term (LT) incarcerated p opulations in each colonial province
over 1920 to 1938 as defined in the text. Outcomes in columns (3) and (4) are prisoners per 100,000 p opulation (1939 pop.) by province in Nigeria broken
down by short-term (less than 6 months) sentence ((3)) and long-term (greater than 2 years) sentence ((4)) over 1920-1938. District FE are colonial
province fixed effects. ∗∗∗ Significant at the 1 percent level, ∗∗Significant at the 5 percent level, ∗Significant at the 10 percent level.
105

Table A19: Summary Statistics: Afrobarometer Results
Statistic N Mean St. Dev. Min Max
Trust and Crime Outcomes
Trust in Courts 11,354 1.21 0.92 0.00 3.00
Trust in Police 11, 486 0.69 0.87 0.00 3.00
Trust in Tax Admin. 4, 480 1.01 0.85 0.00 3.00
Trust Relatives 4, 596 1.97 1.03 0.00 3.00
Trust Neighbors 4,682 1.37 1.00 0.00 3.00
Trust Local Gov. 8,961 0.93 0.87 0.00 3.00
Fear Crime 11, 584 0.59 1.00 0.00 4.00
Bribe (HHS) 8, 082 0.27 0.68 0.00 3.00
Bribe (Doc) 7,987 0.29 0.66 0.00 3.00
Individual Controls and Fixed Effects
Age 11, 603 31.94 12.05 18.00 95.00
Age Squared 11, 603 1, 165.29 987.34 324.00 9, 025.00
Female 11,654 0.50 0.50 0 1
Education 11, 629 3.27 1.92 0.00 7.00
Urban 9, 300 0.46 0.50 0.00 1.00
Geographic and Disease Controls
Agricultural Land Suitability 8,453 4.71 0.76 1.80 6.00
Malaria 9,095 1.00 0.02 0.79 1.00
Ruggedness 9,095 0.26 0.22 0.03 2.28
Mean Elevation 8,332 248.09 234.70 −0.25 1, 284.11
Sea Coast 9,095 0.29 0.45 0.00 1.00
Petrol 9,095 0.34 0.47 0.00 1.00
Tsetse Suitability 7,147 0.91 0.46 −0.78 1.45
Precolonial and Colonial Controls
Precolonial Centralization 9, 095 1.66 0.78 0.00 3.00
Slave Exports 9, 095 150, 841.30 206,271.70 0.00 665,966.00
Colonial Imprisonment and Instrument
Colonial Imprisonment (ST) 11,025 0.75 0.13 0.46 0.92
Colonial Imprisonment (LT) 11,025 0.11 0.08 0.01 0.33
Soil Suitability for Palm Oil
x Share of Positive Shock (M) Years 11, 025 3.09 7.95 0.00 32.34
Notes: See text and online appendix for details.
106

A.6.2 Possible Channels
There are many potential channels through which the use of an, ostensibly, institution of
justice like prisons for primarily prison labor or economic/extrajudicial motives may matter
for populations’ long-term trust in legal institutions. A full exploration of these channels is
beyond the scope of this paper, but we discuss two main ones here. One hypothesis is that
repressive practices like coercive policing and police violence against populations needed to
exert control and imprisonment for prison labor described in Section 2 may have continued
in regions today, even after the prison labor motive for incarceration disappeared in the
postcolonial period.
A second channel is that colonial imprisonment coupled with the existing economically
motivated system of prison labor is highlighted in local memory as unjust and the fear and
injustice retained in residents’ memories has been passed down over generations and is re-
flected in lower trust in legal institution outcomes in current times. Exposure to colonial
imprisonment then reduces residents’ trust in legal institutions with colonial origins such
as modern courts, the police, and systems of tax administration, as a result of repeated
negative experiences and long local memories as described in previous literature (Nunn and
Wantchekon, 2011; Lowes and Montero, 2021a,b). A key assumption here is that there
are relatively low levels of internal migration, with most people residing in their provincial
homelands. Although there are no available data on migration, research has documented sig-
nificant positive correlations (0.7, p < 0.001) between historic (c.1850) ethnic/province-level
residence and contemporary Afrobarometer respondent locations by ethnicity (Archibong,
2019; Nunn and Wantchekon, 2011); this suggests that the low migration assumption is
reasonable here.
One way to potentially assess this hypothesis is test if reported trust in legal institu-
tions is even lower among people who report their ethnicity in the Afrobarometer survey as
being from an ethnic group, historically based in the southern region. The southern region
experienced the most intensive use of prison labor in the country, and the results are largely
driven by the southern region as discussed in Section A.5.6. We match ethnic groups of
respondents in Afrobarometer to their historic ethnic homeland in Murdock’s (1967) Ethno-
graphic Atlas. We then examine the relationship between colonial imprisonment and trust
by southern ethnicity status. The results in Table A20 show no effects of southern ethnicity
status on the trust outcomes. The results suggest that the persistence in policing practices
channel may be the primary channel at work here, although we cannot rule out the local
107

memory channel based on the findings in Table A20.
108

Table A20: OLS Estimates: Relationship between colonial imprisonment and trust in historical legal Institutions versus
interpersonal trust by southern ethnicity status
Panel: Colonial Imprisonment (Short-Term) and Contemporary Trust Outcomes
Outcome:Trust in Historical Legal Institutions Interpersonal Trust
Police Courts Tax Neighbor Relative Local Gov.
(1) (2) (3) (4) (5) (6)
Colonial imprisonment (ST) −0.584∗∗∗ −0.599∗∗∗ −0.766∗∗ −0.326 1.106 −0.568∗∗
(0.161) (0.212) (0.367) (0.545) (0.735) (0.287)
[0.001] [0.064] [0.166] [0.599] [0.358] [0.353]
Southern Ethnicity −0.617 −0.025 0.193 0.207 0.709 −0.632
(0.427) (0.632) (0.864) (0.437) (0.729) (0.601)
[0.327] [0.974] [0.867] [0.719] [0.437] [0.482]
ST x Southern Ethnicity 0.762 0.083 −0.560 −0.417 −1.060 1.060
(0.533) (0.877) (1.182) (0.605) (1.054) (0.783)
[0.312] [0.932] [0.756] [0.581] [0.441] [0.374]
Mean of outcome 0.709 1.274 0.976 1.334 1.913 0.948
Observations 6,163 6,115 2,906 3,192 3,125 4,510
Clusters 21 21 21 21 21 21
Individual Controls Yes Yes Yes Yes Yes Yes
Geographic Controls Yes Yes Yes Yes Yes Yes
Disease Controls Yes Yes Yes Yes Yes Yes
Precolonial and Colonial Controls Yes Yes Yes Yes Yes Yes
Region FE Yes Yes Yes Yes Yes Yes
Year FE Yes Yes Yes Yes Yes Yes
Notes: Regressions estimated by OLS. Robust standard errors in parentheses clustered by colonial province. Wild cluster bootstrap (by district) p-values are in brackets.
The unit of observation is an individual. Colonial imprisonment (ST) is the average share of short-term (ST) incarcerated populations in each colonial province over 1920
to 1938 as defined in the text. Southern Ethnicity is an indicator that equals one if the respondent is from an ethnic group historically located in the former southern
colonial provinces. Trust variables are from the Afrobarometer samples over 2003 to 2014 and as defined in the main text. Trust outcomes are reported trust levels on a
scale of 0-3, where “Not at all”= “0”, “Just a little”=“1”, “Somewhat”=“2”, “A lot”=“3”. All regressions use region fixed effects at the geopolitical zone level in Nigeria
(for 6 geopolitical zones), year fixed effects and educational attainment fixed effects. Individual controls include age, age squared and gender. Geographic controls include
an indicator for whether the respondent lives in an urban location, and, at the sub-district or local government area level, include, ruggedness, indicators for petroleum,
seacoast and mean land suitability for agriculture and mean elevation in alternate specifications. Disease controls at the sub-district level include malaria suitability and
tse tse fly suitability in alternate specifications with results unchanged. Precolonial and colonial controls at the ethnicity-level include the level of precolonial centralization
and total exports of slaves from the region during the Atlantic slave trade. ∗∗∗ Significant at the 1 percent level, ∗∗Significant at the 5 percent level, ∗Significant at the
10 percent level.
109

A.6.3 Instrumental Variable Strategy and Results
While Equation 5 includes a rich set of controls, βdoes not identify the causal effect of colo-
nial imprisonment on trust in legal institutions. It is possible that there exists an omitted
variable, such as lower inherent trust among imprisoned populations, which determines both
(short-term) colonial imprisonment exposure and trust in legal institutions. To address this
issue, we present results using an instrumental variables approach. We construct an instru-
ment for our colonial imprisonment outcome that is the interaction between two variables:
(1) the soil suitability for palm oil and (2) the share of moderate positive rainfall shock
years in the colonial province over 1920 to 1938. The instrument is based on the findings
of the strong predictive power of palm oil production and prices for (short-term) colonial
imprisonment, and the previous results showing that short-term incarceration increased in
response to moderate positive rainfall shocks that increased agricultural productivity. For
instrument validity and for the exclusion restriction to hold, the interacted soil suitability
for palm oil instrument must only affect the trust outcomes through the share of short-term
sentenced colonial imprisonment measure.
Table A21 shows that the interacted instrument strongly predicts the (short-term)
colonial imprisonment measure. Conversely, there is a weaker negative relationship between
the interacted soil suitability for palm oil instrument and the share of long-term sentenced
colonial imprisonment (column (3) and column (4)). Panel A of Table A22 presents the
first-stage estimates for the instrument using the “soil suitability for palm oil x colonial
palm oil production” indicator to predict our colonial imprisonment outcome. The instru-
ment predicts (short-term) colonial imprisonment, with an F-stat greater than 10 across
all specifications. Panel B of Table A22 reports the second-stage estimates for the trust
outcomes. The IV estimates are largely qualitatively similar to the OLS results. While the
estimate is imprecisely measured, the coefficient on trust in police remains negative, with
similar magnitudes as in the OLS results. The coefficients on trust in courts and tax ad-
ministration are also negative and significant, although the large differences in magnitudes
between the OLS and IV estimates suggests measurement error and caution in interpreting
the IV estimates. The estimate on trust in neighbors is now negative and significant, al-
though with similarly inflated estimates. There is no significant effect for trust in relatives
and the elected local governing council.
110

Table A21: OLS Estimates: Soil suitability for palm oil interacted with share of moderate
positive rainfall shock years in colonial province instrument and colonial imprisonment
Outcome: Colonial imprisonment (ST) Colonial imprisonment (LT)
(1) (2) (3) (4)
Soil Suitability for Palm Oil
x Share of Positive Shock (M) Years 0.036∗∗∗ 0.013∗∗∗ −0.019∗∗∗ −0.006∗∗
(0.010) (0.003) (0.006) (0.003)
Mean of outcome 0.750 0.769 0.112 0.104
Observations 11,025 6,745 11,025 6,745
Clusters 21 21 21 21
Individual Controls No Yes No Yes
Geographic Controls No Yes No Yes
Disease Controls No Yes No Yes
Precolonial and Colonial Controls No Yes No Yes
Region FE No Yes No Yes
Year FE No Yes No Yes
Notes: Regressions estimated by OLS. Robust standard errors in parentheses clustered by colonial province. The unit of
observation is an individual. Colonial imprisonment (ST or LT) is the average share of short-term (ST) or long-term (LT)
incarcerated populations in each colonial province over 1920 to 1938 as defined in the text. Where specified, regressions
use region fixed effects at the geopolitical zone level in Nigeria (for 6 geopolitical zones), year fixed effects and educational
attainment fixed effects. Individual controls include age, age squared and gender. Geographic controls include an indicator
for whether the respondent lives in an urban location, and, at the sub-district or local government area level, include,
ruggedness, indicators for petroleum, seacoast and mean land suitability for agriculture and mean elevation in alternate
specifications. Disease controls at the sub-district level include malaria suitability and tse tse fly suitability in alternate
specifications with results unchanged. Precolonial and colonial controls at the ethnicity-level include the level of precolonial
centralization and total exports of slaves from the region during the Atlantic slave trade. ∗∗∗Significant at the 1 percent
level, ∗∗Significant at the 5 percent level, ∗Significant at the 10 percent level.
111

Table A22: IV Estimates: Effect of colonial imprisonment on present-day trust in historical legal Institutions versus
interpersonal trust
Panel A: First-Stage Estimates
Outcome:Colonial Imprisonment (ST)
(1) (2) (3) (4) (5) (6)
Soil Suitability for Palm Oil
x Share of Positive Shock (M) Years 0.013∗∗∗ 0.013∗∗∗ 0.013∗∗∗ 0.013∗∗∗ 0.013∗∗∗ 0.013∗∗∗
(0.003) (0.003) (0.003) (0.003) (0.003) (0.003)
F-Stat of Excluded Instrument 15.31 15.14 15.30 16.72 11.83 16.25
Mean of outcome 0.769 0.769 0.770 0.769 0.768 0.768
Panel B: Second-Stage 2SLS Estimates
Outcome:Trust in Historical Legal Institutions Interpersonal Trust
Police Courts Tax Neighbors Relatives Local Gov
(1) (2) (3) (4) (5) (6)
Colonial imprisonment (ST) −0.531 −4.345∗∗ −4.105∗∗∗ −2.146∗∗ −1.094 −1.357
(0.565) (1.730) (1.525) (1.012) (1.354) (0.978)
Mean of outcome 0.709 1.274 0.976 1.334 1.913 0.948
Observations 6,642 6,590 3,126 3,439 3,317 4,899
Clusters 21 21 21 21 21 21
Individual Controls Yes Yes Yes Yes Yes Yes
Geographic Controls Yes Yes Yes Yes Yes Yes
Disease Controls Yes Yes Yes Yes Yes Yes
Precolonial and Colonial Controls Yes Yes Yes Yes Yes Yes
Region FE Yes Yes Yes Yes Yes Yes
Year FE Yes Yes Yes Yes Yes Yes
Notes: Regressions estimated by OLS. Robust standard errors in parentheses clustered by colonial province. Wild cluster bootstrap (by district) p-values are in brackets.
The unit of observation is an individual. Colonial imprisonment (ST or LT) is the average share of short-term (ST) or long-term (LT) incarcerated populations in each
colonial province over 1920 to 1938 as defined in the text. Trust variables are from the Afrobarometer samples over 2003 to 2014 and as defined in the main text.
Trust outcomes are reported trust levels on a scale of 0-3, where “Not at all”= “0”, “Just a little”=“1”, “Somewhat”=“2”, “A lot”=“3”. All regressions use region fixed
effects at the geopolitical zone level in Nigeria (for 6 geopolitical zones), year fixed effects and educational attainment fixed effects. Individual controls include age, age
squared and gender. Geographic controls include an indicator for whether the respondent lives in an urban location, and, at the sub-district or local government area
level, include, ruggedness, indicators for petroleum, seacoast and mean land suitability for agriculture and mean elevation in alternate specifications. Disease controls
at the sub-district level include malaria suitability and tse tse fly suitability in alternate specifications with results unchanged. Precolonial and colonial controls at the
ethnicity-level include the level of precolonial centralization and total exports of slaves from the region during the Atlantic slave trade. ∗∗∗ Significant at the 1 percent
level, ∗∗Significant at the 5 percent level, ∗Significant at the 10 percent level.
112