Content uploaded by Richard Gilbert
Author content
All content in this area was uploaded by Richard Gilbert on Feb 08, 2019
Content may be subject to copyright.
DE GRUYTER
Richard Gilbert1/ Nora A. Murphy1/ Allison Stepka1/ Mark Barrett2/ Dianne Worku3
Would a Basic Income Guarantee Reduce the
Motivation to Work? An Analysis of Labor
Responses in 16 Trial Programs
Abstract:
Many opponents of BIG programs believe that receiving guaranteed subsistence income would act as a strong
disincentive to work. In contrast, various areas of empirical research in psychology (studies of intrinsic motiva-
tion; non-pecuniary benets of work on social identity and purpose; and reactions to nancial windfalls such
as lottery winnings) suggest that a BIG would not lead to meaningful reductions in work. To test these compet-
ing predictions, a comprehensive review of BIG outcome studies reporting data on adult labor responses was
conducted. The results indicate that 93 % of reported outcomes support the prediction of no meaningful work
reductions when the criterion for support is set at less than a 5 % decrease in either average hours worked per
week or the rate of labor participation. Overall, these results indicate that adult labor responses would show
no substantial impact following a BIG intervention.
Keywords: basic income, basic income guarantee, labor outcomes, evaluation studies, pilot programs
DOI: 10.1515/bis-2018-0011
During the post-millennial period, a revolutionary suite of digital technologies including robotics, 3D printing,
and articial intelligence have dramatically enhanced human productivity (Bryniolfsson & McAfee, 2014). At
the same time, these technologies have decreased job security for skilled and unskilled labor in major sectors of
the economy, including manufacturing, construction, transportation, and retail services (Heath, 2016; Peterson,
2016; Silverberg, 2017; Smith, 2016). In addition, rapid advances in deep-learning systems of articial intelli-
gence have begun to encroach on jobs involving cognitive abilities once believed to be the untouchable province
of human intelligence, such as medical diagnosis, translation services, legal research, and banking and nancial
services (Johnson, 2017; Lohr, 2017; Mukherjee, 2017; Popper, 2016). Cumulatively, these disruptions have led
to the displacement or under-employment of millions of workers with adverse impacts on wages, income and
wealth inequality, and social and political stability (Acemoglu & Restrepo, 2017; Flinders, 2017; Yan, 2016).
Some observers view the mass dislocation of labor as a painful but temporary process as emerging markets
gradually re-absorb workers from obsolete jobs and replenish high wage employment opportunities, similar
to what occurred in the transition from agrarian to industrial production in the nineteenth and early twenti-
eth century (Avent, 2017; Kaplan, 2017). However, others are less sanguine about future prospects for human
workers (Ford, 2015). In their view, machine displacement of labor is a permanent and accelerating feature of
post-industrial society that poses an ongoing challenge to human welfare and social stability. One oft-cited
analysis supporting this perspective indicates that approximately 47 % of current jobs are under threat of dis-
placement by automation in the next twenty years (Frey & Osborne, 2013.) Similarly, a recent analysis conducted
by the global forecasting company, McKinsey, concluded that currently available technologies could “automate
45 percent of the tasks people are paid to perform and that 60 percent of all occupations could see 30 percent
or more of their constituent activities automated, again with technologies available today.”(Chui, Manyika, &
Miremadi, 2016, p. 1.)
Responses to technological unemployment: Familiar and new
A number of familiar macro-economic strategies have been proposed to address issues of technological un-
employment and underemployment. These include calls to increase preparation for quality jobs via enhanced
Richard Gilbert
1
Brought to you by | Loyola Marymount Law School Library
Authenticated
Download Date | 12/10/18 10:20 PM
DE GRUYTER
educational and job-training programs, raise spending on labor-intensive infrastructure projects, and promote
economic growth and job creation via simulative monetary and scal policies. These traditional economic tools
have been helpful in addressing labor downturns during periods of recession, including the Great Recession
of 2008. However, they are unlikely to oset job reductions, wage stagnation, and increased inequality due
to long-term structural changes in the labor market, as highly capable and intelligent machines increasingly
perform production and service tasks at a fraction of the cost of human workers (Javelosa & Houser, 2017). In
this unprecedented context, novel approaches have been proposed to augment traditional economic tools to
eectively cushion the impact of automation on workers.
One new-era approach is already being implemented. Itinvolves promoting economic activities that operate
outside of the traditional employer-employee model, such as participating in bartering and time-bank organi-
zations, selling products on-line, and providing on-demand services for hosting, ride-sharing, and product
deliveries via specialized apps. It is important to note, however, that the large ride-sharing and delivery sec-
tors of the on-demand economy are themselves poised for disruption in the next decade by the large-scale use
of autonomous vehicles (Newman, 2017) and drone delivery systems (Meola, 2017). A second approach, put
forth by the noted futurist, Lanier (2013), involves changing the business model of the Internet. In its present
form, global monopolies or near-monopolies (e. g., Google for search, Facebook for social media, Amazon for
retail, etc.) provide users free access to their platforms and generate enormous revenue by selling advertising
and information about its users to third parties. In contrast, the users who provide free content for the plat-
forms, and thus create part of the value of these companies, generally derive little or no nancial benet from
the information they provide. In response to this perceived inequity, Lanier oers the bold, but unwieldy, idea
of monetizing and distributing the value of the Internet content individual users provide. In addition, Gates
(2017) recently proposed an innovative strategy to address the adverse impact of automation on labor. In his
approach, governments would tax robots (either at the time of their installation or from the prots a company
derives from labor savings) and use the obtained revenue to support displaced workers in a variety of ways
(e. g. retraining, funding health care benets, etc.)
Along with the preceding ideas, increased attention has been directed toward another non-traditional re-
sponse to technological displacement of human workers: The implementation of a Basic Income Guarantee or
BIG, which involves a tax-free, subsistence-level income paid directly by a government to individuals or house-
holds with few or no conditions in order to ensure a basic level of economic security (Standing, 2017; Van Parijs
& Vanderborght, 2017). In order to incentivize work, basic income can be supplemented without penalty by
money earned via employment, savings, investments, or public or private pensions, with all income above the
guarantee subject to taxation at prevailing marginal rates. Thus, with the exception of establishing an economic
oor below which no one can fall, all aspects of market-oriented economies are preserved.
The concept of guaranteed basic income is not new. In various forms, it has existed on the fringes of intellec-
tual life as early as the Renaissance (More, 1516/1978) and the post-revolutionary eras in France (Concordant,
1795/1988) and America (Paine, 1796/1974). There have also been periodic expressions of support by a het-
erogeneous mix of twentieth century social theorists, economists, social activists, and politicians (King, 1967;
Lampman, 1965; Moynihan, 1973; Russell, 1918/1966) Now, in the post-industrial era, the concept is gaining
currency among a growing cadre of technologists, social activists, economists, and policy makers as the po-
tential threat posed by the widespread automation of labor comes into sharper focus (Stern & Kravitz, 2016.)
Reecting this expanding interest, new basic income trial programs are in the early implementation or plan-
ning stages in multiple countries including Canada, India, Kenya, the Netherlands, Scotland, Spain, Uganda,
and the United States (Dillow & Rainwater, 2017). In addition several countries have given serious consider-
ation to implementing a basic income guarantee on a national level, including Switzerland (Jacobs, 2016) and
Finland (Sodha, 2016).
Labor responses to a Basic Income Guarantee
Debates over the merits of a BIG invariably center on questions of nance (Is a BIG aordable?), philosophy (Is
subsistence a right or a responsibility?), politics (Is passage of a BIG feasible?) and motivation (Would a BIG
act as a disincentive to work?) With respect to the nal question, both airmative and dissenting opinions have
been expressed in the literature.
Those who believe that a BIG would have a detrimental impact on the motivation to work tend to rely on
logical assertions to support their position, such as Porter’s (2016) statement that:
A universal basic income has many undesirable features, starting with its non-negligible disincentive to
work. Almost a quarter of American households make less than $25,000. It would be hardly surprising
if a $10,000 check each for mom and dad sapped their desire to work. (p. 3)
2Brought to you by | Loyola Marymount Law School Library
Authenticated
Download Date | 12/10/18 10:20 PM
DE GRUYTER
Similarly, Gabe and Falk (1995) argued that leisure is a commodity like any other good or service and, with a rise
in income, people will purchase more leisure by reducing their work eort. In addition, adherents to the work
disincentive position occasionally cite studies of the employment behavior of lottery winners indicating that
12–24 % of recipients exit their jobs post-windfall and another 8–16 % reduce their work hours (Avery, Harpaz,
& Liao, 2004; Faraker & Hedenus, 2009; Kaplan, 1987).
Those who believe that labor disincentives associated with a BIG would be minimal or even nonexistent
also cite studies of lottery winners to support their argument (Avery et al., 2004.) They point out that a large
majority of prize winners who exit their jobs either returned to the same position after a vacation period, moved
to a dierent area of employment after receiving additional education or training, or shifted to self-employed
work. Thus, most individuals who received unconditional income chose not to exit or reduce work at all, and
most changes in labor involved a temporary respite or a transitional period to new forms of employment.
Along with ndings on responses to nancial windfalls, work reduction skeptics reference a number of
psychological theories and areas of research to support the prediction that basic income would have a negligible
impact on the work activity of recipients. Research on intrinsic versus extrinsic motivation (Gagné & Deci,
2005; Thomas, 2009) demonstrates that individuals are motivated to learn and work by internal desires such
as curiosity, knowledge acquisition, and expanding one’s capacities in addition to external rewards such as
social approval or monetary compensation. Similarly, research on “competence motivation”(White, 1959) or
“mastery motivation”(Pike, 2009), suggests that a powerful reinforcement for actions undertaken by human
beings, including their work activities, is an increased sense of competence in their environment, something
that is unlikely to be diminished by a subsistence guarantee. In addition, a body of theory and research has
examined the important role of work in promoting social relationships, social identity, and a sense of purpose
and meaning in addition to providing nancial benets (Kirk & Wall, 2011.) Work motivation theories also
emphasize the psychology behind an individual’s work-related behavior, which go beyond simply monetary
compensation, and include needs, traits, and cognitive factors (Latham & Pinder, 2005; Van den Broeck, Ferris,
Chang, & Rosen, 2016).). Finally, Maslow (1954), a seminal voice within Humanistic Psychology, argued that
there is a hierarchy of human motivations that begins with having the basic necessities for survival (food,
water, clothing, and shelter), extends to social and emotional goals of love, belonging, and self-esteem, and
culminates in a desire for self-actualization, or the full realization of one’s talents and abilities (see also Rogers,
1963). Viewed from this perspective, individuals who are provided suicient income to meet their basic survival
needs would not lose their motivation to work, but would continue to work to achieve goals at a higher level of
human aspiration. In sum, these theoretical and empirical sources all emphasize motivations to work that may
be more psychological than nancial and presumably would continue to drive individuals to work even if they
received a subsistence-level income.
While psychological theory and research on non-pecuniary motivations to work are broadly relevant to
predictions regarding labor responses to guaranteed basic income, the most direct and specic evidence is pro-
vided by empirical studies of basic income pilot programs that include one or more labor response as outcome
variables. Articles that consider the benets and drawbacks of guaranteed income often cite one or more of
these outcome studies when addressing the potential labor response to a BIG and these studies tend to report
only small labor impacts of the program (e. g., modest or no reductions in average hours of work, labor par-
ticipation rates, etc.). For instance, a World Bank report by Chaudhury, Friedman, and Onishi (2013) described
a two-year conditional cash transfer program sponsored by the government of the Philippines. In this study,
poor households with children under 14 years old received cash grants every two months ranging from 500 to
1400 local currency units depending on the number of eligible children. The results indicated that households
that received the cash transfer had a 3.1% greater drop in work hours at post-test than control households that
were equally poor but did not receive the benet.
In addition to reports from isolated studies, two multi-study evaluations of labor responses to a BIG have
been conducted in an eort to identify trends across studies. Alzúa, Cruces, and Ripani (2013) evaluated three
randomized control evaluations conducted by the World Bank and the United Nations in the Honduras, Mexico,
and Nicaragua. This work was later extended to include a set of seven randomized controlled studies in an
unpublished white paper by Banerjee, Hanna, Kreindler, and Olken (2015) with neither paper nding notable
impacts of cash transfer programs on either the propensity to work or the overall hours worked.
What is missing in the literature, and could provide a clearer indication of labor responses to a BIG, is an
analysis of all empirical evaluations of BIG trials that include an adult labor response dependent measure to
determine overall trends across the entire set of evaluation studies. The current study seeks to provide this om-
nibus assessment by reviewing all outcome studies that provide data on the impact of guaranteed basic income
on either average hours of work per week and/or changes in the labor participation rate (i. e., the percentage of
the working-age population who are currently employed or actively seeking employment.) Results indicating
substantial reductions in either measure of adult labor would provide strong empirical support for predictions
that guaranteed basic income would act as a meaningful disincentive to work. In contrast, negligible shifts in
3
Brought to you by | Loyola Marymount Law School Library
Authenticated
Download Date | 12/10/18 10:20 PM
DE GRUYTER
these labor outcomes would reinforce the claims of work reduction skeptics who emphasize that work moti-
vation is inuenced by an important set of psychological needs as well as nancial considerations. In either
case, growing public discourse about basic income, including discussions about possible labor impacts of BIG
programs, would be anchored in the most current and comprehensive data available.
Method
Searched sources and criteria for inclusion of studies
A multi-step process was undertaken to study trials for the analysis. To begin, the broadest databases from
the main social science disciplines that rely on empirical data (i. e., PsycINFO, EconLit, Sociological Abstracts,
Social Sciences Full Text from H.W. Wilson, and Worldwide Political Science Abstracts) were searched to identify
relevant English-language papers published in academic journals and publically-available reports issued by
government agencies or non-goverment organizations (NGOs) such as the World Bank or the United Nations.
For inclusion, the study trial had to contain either the word “work”or “labor”with at least one of the following
key terms in the title or abstract: basic income, cash transfer, guaranteed income, or negative income tax. These
search criteria yielded an initial set of approximately 1400 papers.
These preliminary search results were then narrowed down by determining whether 1) the study trial re-
ported data from a basic income pilot study or trial program and 2) whether at least one outcome measure in
the study trial focused on labor responses by adults (e. g., changes in hours worked, shifts in the labor partici-
pation rate, etc.). Studies were excluded if they only analyzed the impact of basic income on child or adolescent
labor or focused solely on non-work related variables such as health and educational outcomes.
With these exclusion criteria, a nal pool of 16 trial programs conducted across the globe in the last half-
century were identied, each of which provided empirical data assessing the relationship between guaranteed
income and patterns of adult labor. As a cross-check to conrm that all major BIG trials had been identied
via the search criteria, websites for the following primary organizations disseminating information related to a
BIG were reviewed: Basic Income Earth Network (www.basicincome.org); United States Basic Income Guaran-
tee Network (www.usbig.net); Basic Income Canada (www.basicincomecanada.org); and Basic Income Action
(www.basicincomeaction.org). In addition, a popular online forum dealing with BIG issues (the sub-Reddit
on Basic Income, https://www.reddit.com/r/BasicIncome) was examined and a general Google search was
conducted using the same terms previous used to search the academic databases. The examination of these
additional sources identied a set of basic income trial programs that are currently in the early implementation
or planning stages in Canada, Finland, India, Kenya, the Netherlands, Scotland, Spain, Uganda, and the United
States. These programs were noted for possible inclusion in an expanded analysis should they be completed
and disseminate labor response ndings at a later point.
Coding the trials
Reported variables
The primary manuscript(s) for each completed study trial were coded to obtain information in three broad
categories.
In the category of Study Trial Program Identication and Source of Information, the country in which the study
trial was conducted; the name of the program; the type of trial (i. e., Government Program and/or Randomized
Experiment); the years the program was undertaken, and the author(s) and year(s) of the primary manuscript(s)
providing data about the trial were noted and summarized in Table 1.
Table 1: Study Trial Program Identication and Source of Information
Country Program Name Type1Program Years Source of Information2
Bangladesh Female Secondary Education Program
Stipend (FESP)
GP 1994–2002 Shamsuddin (2015)
Brazil Brazil Bolsa Familia Cash Transfer
Programs
GP 2004-Current Soares, Ribas, & Osorio (2010)
Canada Manitoba Basic Annual Income
Experiment
GP 1975–1978 Hum and Simpson (1993)
4Brought to you by | Loyola Marymount Law School Library
Authenticated
Download Date | 12/10/18 10:20 PM
DE GRUYTER
Honduras Programa del Asignacion Familiar
(PRAF)
RE/GP 2001 Galiani & McEven (2013)
India The Madhya Pradesh Unconditional
Cash Transfer Project (MPUCT)
RE/GP 2011–2012 Schjoedt (2016); Standing (2013);
Arya, A., Kapoor, A., & Konwar,
D. (2014)
Mexico Programa de Apoyo Alimentario (PAL) RE/GP 2004–2005 Skouas, Unar, and de Cossio
(2013)
Namibia Basic Income Grant (BIG) Coalition GP 2008–2009 Perkio (2014); Kaufmann (2010)
Nicaragua Red de Protection Social (RPS) RE 2000–2004 Maluccio & Flores (2004)
Peru Peruvian Juntos Program GP 2005-Current Fernandez & Saldarriaga (2014)
Philippines Pantawid Pamilyang Pilopino Program RE/GP 2009–2011 Chaudhury, Freidman, & Onishi
(2013)
Uganda Youth Opportunities Program (YOP) of
the Northern Ugandan Social Action
Fund (NUSAF)
RE/GP 2008–2012 Blattman, Fiala, & Martinez
(2013)
USA The Alaska Permanent Fund Dividend
Program (PFD)
GP 1982-Current Goldsmith (2010); Knapp (1984)
USA Gary Income Maintenance Experiment
(GIME)
RE/GP 1971–1974 Moit (1976); United States
Department of Health & Human
Services (1972); Keher,
McDonald, & Moitt (1980)
USA New Jersey Graduated Work Incentive
Program
RE/GP 1968–1971
1969–1972
Kershaw & Skidmore (1974)
USA Rural Income Maintenance Program
(RIME)
RE/GP 1969–1971 Lee, Harrar, and Kerachsky
(1976)
USA Seattle-Denver Income Maintenance
Experiment
GP 1970–1977 United States Department of
Health & Human Services (1983)
1With regard to the type of study used, GP refers to a government program, RE refers to a randomized experiment.
2See Reference section for full citations.
In the category of Program Methodology, the country and area where the sample of participants was obtained
(i. e., city/town, state/province, region, or nation) was recorded. In addition to the spatial aspects of the trial,
specic design characteristics used in the study were coded including: the unit of distribution of the BIG (to
individuals or families/households), the total number of units in the sample at the end of the trial, and whether
the cash transfer was unconditional or conditional.
With regard to conditionality, an income guarantee was considered conditional if, after a participant meets
any eligibility requirements, the cash transfer could be eliminated or reduced if certain behavioral or economic
conditions were not met. For example, in conditional cash transfer programs where funds are disbursed to fam-
ilies/households, behavioral conditions may involve ensuring designated levels of school attendance and/or
timely immunization of children, or regular attendance at community-based psycho-educational programs.
Similarly, a number of basic income programs contain provisions to reduce or eliminate the cash transfer if
participants’earnings go above a designated ceiling during the trial (i. e., the condition of “means testing”). In
cases of conditional transfer programs, the types of conditions attached to preserving the guarantee were also
noted. All variables coded in the category of Program Methodology are summarized in Table 2.
Table 2: Program Methodology: Characteristics of Included Study Trials
Country Area of
Sample
Unit of Distribution Units at Time
of Post-test
Conditionality Type of
Conditionality1
Bangladesh Nation Individual 4,193,352 Conditional E, UM
Brazil Nation Family/Household 11,000,000 Conditional MT, E, HC
Canada State Family/Household 1,300 Conditional MT
Honduras Nation Family/Household 5,748 Conditional E, HC
India State Individual 15,000 Unconditional -
Mexico (PAL) Nation Family/Household 2,829 Unconditional -
Namibia City/Town Individual 398 Unconditional -
Nicaragua Region Family/Household 1,581 Conditional E, HC
Peru Nation Family/Household 1,087 Conditional E, HC
Philippines Nation Family/Household 1,418 Conditional E, HC
Uganda Region Individual 11,288 Unconditional -
USA (Alaska) State Individual 1,016 Unconditional -
USA (GIME) City/Town Family/Household 1,792 Conditional MT
USA (NJ/PA) State Family/Household 1,357 Conditional MT
5
Brought to you by | Loyola Marymount Law School Library
Authenticated
Download Date | 12/10/18 10:20 PM
DE GRUYTER
USA (RIME) Region Family/Household 729 Conditional -
USA (SE/DNV) Region Family/Household 4,800 Conditional MT
1Type of conditionality is abbreviated as follows: MT –means tested, E –education (regular school attendance and/or suicient test
scores is required to receive cash transfer), HC –healthcare (regular doctor visits are required to receive cash transfer), S –stop receiving
benets from pre-existing social welfare programs, UM –unmarried (participants must remain unmarried to receive the cash transfer).
Finally, information related to the Amount and Benchmarking of the Income Guarantee was also coded and
summarized in Table 3. With respect to the guarantee itself, a number of variables were recorded: year of the
post-test evaluation, the name of the local currency, and the amount of the guarantee expressed as an annualized
gure in local currency units for standardization. Thus, if a trial called for monthly cash transfers of 1000 units
of local currency, the amount of the guaranteed would be recorded as an annualized amount of 12,000 units.
In cases where a range of cash transfer payments was provided to participants in the program rather than a
xed amount (e. g., depending on the number of young children in the household), the mid-point of the range
expressed as an annualized gure was entered as the guarantee.
Table 3: Amount and Benchmarking of the Income Guarantee
Country Year of
Post-test
Local
Currency
Amount
of BIG
(annual-
ized)
Per
Capita
GDP
(Total
popula-
tion)
Household
GDP (Total
population)
% of
BIG to
Per
Capita/-
House-
hold
GDP
Ratio
Transfer/
Consumption
Bangladesh 2002 Takas 940 16,536 x*4,84 = 80,528 6 -
Brazil 2006 Real 1,128 7,377 x*3,50 = 25,818 4 -
Canada 1978 CA Dollars 4,800 9,672 x*3.10 = 29,983 16 -
Honduras 2001 Lempiras 6,552 16,866 x*5.02 = 84,665 8 4 %
India 2012 Rupees 9,189 72,997 x*4.13 =
301,480
13 -
Mexico (PAL) 2005 Pesos 1,800 161,193 x*4.00 =
644,772
0 11.5 %
Namibia 2009 Namibian
Dollars
1,200 23,454 x*4.70 =
110,235
5 -
Nicaragua 2002 Cordoba 3,500 14,194 x*5.50 = 78,069 4 20 %
Peru 2009 Nuevos Solaes 1,200 12,157 x*3.91 = 47,535 3 -
Philippines 2011 PhP 11,400 62,541 x*4.60 =
287,689
4 11 %
Uganda 2012 Ugandan
Shillings
96 1,329,637 x*4.70 =
6,249,296
0 -
USA (Alaska) 1984 USD 331 13,514 x*2.60 = 35,136 2 -
USA (GIME) 1974 USD 3,800 7,242 x*2.97 = 21,510 18 -
USA (NJ/PA) 1971 USD 1,196 5,624 x*3.11 = 17,489 7 -
USA (RIME) 1971 USD 1,500 5,624 x*3.11 = 17,489 9 -
USA
(SE/DNV)
1971 USD 1,576 5,624 x*3.11 = 17,489 9 -
Note: Transfer/Consumption data is as reported according to Banerjee et al. (2015).
Additionally, in order to evaluate the nancial signicance of the guarantee provided to participants in the
study, each annualized cash transfer was benchmarked as a percentage of either the per capita or average house-
hold income (depending on the unit of distribution) in the country and year of the post-test evaluation. For each
country and year, per capita income was computed by dividing GDP data expressed in national currency from
the International Monetary Fund (http://data.imf.org/) by national population gures provided by the World
Bank (http://data.worldbank.org/indicator). Household GDP was determined by multiplying per capita GDP
by the average household size for each country and year. Information on average household size usually was
not available via a single source and had to be collected via a set of online government reports accessed via
a Google search using the country and year of the trial and “average household size”as search terms. Thus,
if individuals received an annualized guarantee of 8,000 units of local currency in a particular year, and the
per capita income of the country where the trial was occurring was 100,000 units in that year, the BIG would
represent 8% of per capita income. If the same 8000 units of currency were distributed to households in that
country, with an average number of four persons per household, the income guarantee would represent 2%
6Brought to you by | Loyola Marymount Law School Library
Authenticated
Download Date | 12/10/18 10:20 PM
DE GRUYTER
of average household income. This process of benchmarking the cash transfer against average national income
gures for the country and year of the trial provides a rational basis to compare the magnitude and signicance
of guaranteed income across trials conducted in a diverse set of countries, time periods, and cultural contexts;
and examines whether there is an association between the proportional size of a BIG and subsequent labor
impacts.
Finally, data that addressed whether there were any changes in adult labor response during the trial pe-
riod were recorded. For trials that only included a treatment group, this involved coding pre-post changes in
participants’average hours of work or labor force participation rates (i. e., the number of participants actively
involved in the labor force divided by the total number of participants in the study multiplied by 100 to obtain
a percentage). These data are reported in the Results section.
The process of coding
To establish the reliability of coding, two members of the research team independently coded the manuscript(s)
containing information and data for each trial. For all 16 trials, the lead author served as one of these coders,
with the third through fth authors serving as an additional or secondary coder. Prior to coding, a secondary
coder received instruction in the criteria for coding each variable and coded a sample trial. After coding the
sample trial, the primary and secondary coder met to discuss any areas of confusion to ensure that there was a
shared understanding regarding all decision rules for coding. Following this discussion, the initial secondary
coder worked independently on a set of trails that were assigned one at a time based on a random sequence
of numbers from 1 to 16. When this coder graduated after coding ve trials, the next coder followed the same
process and coded the next seven trials in the random sequence, with the nal coder completing the last four
numbered trials.
When a secondary coder completed their set of trials, he or she met with the primary coder to discuss
each of the coded trials, resolve any coding disagreements, and arrive at a nal coding sheet. Because of the
straightforward nature of most of the variables (e. g., timing for disbursing the guarantee: weekly, monthly,
quarterly, annually, or via a lump sum; income guarantee issued to individuals or households), and the simple
decision rules for coding, few discrepancies occurred in a coded trial and these were almost always resolved in
a brief discussion between the two coders. In only two instances was an agreement about a coding decision hard
to reach. In these cases, one of the other secondary coders was brought in to make an independent judgment
and establish a consensus code.
Results
Two primary outcomes were assessed: number of hours worked per week (hours worked) and labor participation
rate (LPR). LPR represents the portion of working-age adults currently employed or actively seeking work. It
is calculated by dividing the number of people actively participating in the labor force by the total number of
people eligible to participate in the labor force, with the resulting quotient multiplied by 100 to get a percentage.
Trials were included in quantitative analyses if the trial reported pre- and post-trial outcome data or provided
a dierence score between pre- and post-trial data (e. g., dierence in hours worked or change in LPR).
Seven trials reported suicient quantitative data to calculate the dierence in hours worked between pre-
and post-trial intervention, and the weighted mean dierence in average hours worked was M= 0.86 (SD =
7.39). A one-sample t-test measured whether this average increase was signicantly dierent from zero, t(6) =
0.32, p= 0.76, 95%CI [−5.94, 7.72]. Results showed that the dierence in average hours worked per week did
not signicantly dier from zero, suggesting that the BIG intervention had no eect on average hours worked.
Nine trials reported suicient quantitative data to calculate the dierence in LPR between pre- and post-
trial intervention; the weighted mean dierence of LPR was M= 1.84 (SD = 3.75). A one-sample t-test measured
whether this LPR weighted mean dierence was signicantly dierent from zero, t(8) = 1.47, p= 0.18, 95%CI
[−1.04, 4.72]. Results showed the dierence in LPR did not signicantly dier from zero, suggesting that BIG
interventions had no eect on LPR
.
.
In addition to conducting these statistical tests, descriptive data was used to evaluate whether reported
changes in labor responses were consistent with the hypothesis that a basic income program would have a
limited impact on work activity. Across the set of studies, we computed 15 data points that reected changes
in hours of work (8 pre-post dierences in the treatment group and 7 post-test dierences in treatment and
control groups) and arranged them according to 4 criterion to support the prediction of limited impact on work
7
Brought to you by | Loyola Marymount Law School Library
Authenticated
Download Date | 12/10/18 10:20 PM
DE GRUYTER
hours. As shown in Table 4, the data indicate that the number of changes in work hours consistent with the
hypothesis of limited impact rose steadily from 40% when the criterion was set at an increase or zero change;
to 67 % and 87 % at criterion levels of a 0.1 to 2.0 % and a 2.1 to 5.0 % decline in work; and reached 100% when
the criterion for support was set at a 5.1 to a 10 % drop in weekly work hours. Using a 40-hour workweek as a
standard, these data indicate that 87 % of the reported ndings found a reduction of less than 1 to 2 hours of
work per week and 100% of the ndings found reductions of less than 1 to 4 hours of work. Table 5 presents
the same descriptive data to evaluate whether reported changes in labor participation rates were consistent
with the hypothesis that a basic income program would have a limited impact on work activity, and a similar
pattern of ndings emerged. Of the 14 reported changes in labor participation rates (9 pre-post dierences in
the treatment group and 5 post-test dierences in treatment and control groups), 71% supported the prediction
of limited impact when the criterion was set as an increase or zero change in labor participation following a BIG
program; 93% supported the prediction when the criterion range was set as at −0.1 to −2.0 %; and 100 % of
the ndings were consistent with the hypothesis when the criterion level was a 2.1% to 5% decline in labor
participation.
Table 4: Are Percentage Changes in Hours of Work Consistent with the Hypothesis of Limited Impact on Work? Results
at Four Criterion Levels
8 Within-Group Changes (Pre-Post Dierences for the Treatment Group)
+ to 0 -0.1 to
-2.0
-2.1 to
-5.0
-5.1 to
-10.0
+ to 0 0.1 to
-2.0
-2.1 to
-5.0
-2.1 to
-5.0
-5.1 to
-10.0
Country
Bangladesh
Brazil −1.3
Canada −1.0
Honduras
India 0.0 0.0
Mexico
Namibia
Nicaragua + 5.3 −8.3
Perua−2.5
Philippines 0.0 −3.1
Uganda +15.0 +17.0
USA
(Alaska)
USA
(GIME)
−4.7
USA
(RIME)
−1.0
USA
(NJ/PA)b
−2.0
USA
(SE/DNV)
−7.3
TOTALS 50 %
(4/8)
75 %
(6/8)
100 %
(8/8)
29 %
(2/7)
57 %
(4/7)
71 %
(5/7)
100 %
(7/7)
aOnly 7 Trials had reportable control data for changes in Hours of Work.
bPeru evaluated changes in paid and unpaid work, we reported paid work only.
cUsed weighted means for hours worked for 3 demographic groups – “Whites, Blacks, Spanish.”
Table 5: Are Percentage Changes in Labor Participation Rates Consistent with the Hypothesis of Limited Impact on
Work? Results at Four Criterion Levels
9 Within-Group Changes (Pre-Post
Dierences for the Treatment Group)
5 Between Group Changes (Dierences
between the Treatment and Control Groups
at Post-test)a
+ to 0 −0.1 to
−2.0
−2.1 to
−5.0
−5.1 to
−10.0
+ to 0 −01 to
−2.0
−2.1
to
−5.0
−5.1
to
−10.0
−2.1
to
−5.0
Country
Bangladeshb,c +2.2
Brazil +1.7
8Brought to you by | Loyola Marymount Law School Library
Authenticated
Download Date | 12/10/18 10:20 PM
DE GRUYTER
Canada
Honduras 0.0
India 0.0 0.0
Mexico −0.1 0.0
Namibia +11.1
Nicaragua
Perud0.0
Philippines 0.0
Uganda
USA (Alaska)e−1.0
USA (GIME) - −3.8
USA (RIME) −0.5
USA (NJ/PA) +0.2
USA
(SE/DNV)
TOTALS 67 % (6/9) 89 %
(8/9)
100 %
(9/9)
80 %
(4/5)
100 %
(5/5)
aOnly 5 Trials had reportable control data for changes in Labor Participation Rates.
bAll data reects changes in LPR for Primary Earners. Except for Bangladesh, the primary earner was the male.
cBangladesh data reects post-test data at 5 or less years after the BIG trail (not 5 or more) to be similar to other post-test durations.
dPeru evaluated changes in paid and unpaid work, we reported paid work only.
Overall, when all 29 reported changes in work activity, either hours of work or labor participation, are com-
bined, the results indicate that 27 out of 29 results, or 93 %, support the prediction that a Basic Income program
would have a limited impact on work activity when the criterion is set at between a 2.1 and 5% decline in labor
force participation or less than a 1 to 2 hour reduction in a standard 40-hour work week.
Discussion
The current study provides a comprehensive review of basic income trial programs that report empirical data
on adult labor outcomes in order to address one of the central controversies related to Basic Income programs:
Whether or not the provision of a subsistence-level income guarantee would act as signicant disincentive
to work. The study examined 16 trials conducted in the past half-century in 12 nations in the developed and
developing world with a cumulative sample of over 105,000 BIG recipients and found no evidence of signif-
icant reductions in either hours of work or labor participation rates in response to these programs. In fact,
increases or zero change in work activity occurring in multiple programs undertaken in developing nations
(e. g., Bangladesh, Brazil, India, Namibia, and Uganda) oset modest reductions in work activity in developed
nations such as the United States and Canada and resulted in small average increases in hours worked per week
and labor participation across the entire set of trials. As a possible explanation for this increased work activ-
ity, Schjoedt (2016) notes that, rather than using the subsistence guarantee to fund idleness and leisure, many
recipients used the additional income to invest in goods that expanded their capacity for employment (e.g., a
mobile phone to more easily communicate with prospective employers or customers, transportation to attend
job interviews and meetings, clean clothing, etc.) or to purchase tools and commodities such as seeds, fertilizer,
or yarn that enabled them to shift from low wage paid labor to self-employed work activities. Taken as a whole,
the current ndings provide strong support for the prediction of work reduction skeptics that guaranteed ba-
sic income would not serve as a major disincentive to work. In addition, the results are consistent with work
motivation theories and research that emphasize that individuals are psychologically fullled by work-related
activities (Gagné & Deci, 2005) and motivations for work extend beyond nancial rewards (Latham & Pinder,
2005; Maslow, 1954)
While the current results support the view that basic income programs are not a meaningful disincentive
to work, and in some cases expand work opportunities, a number of design and interpretive issues limit the
ability to draw denitive conclusions from the set of evaluations. With respect to design, less than half of the
studies had a control groups, let alone randomly assigned participants to intervention and control conditions.
The limited inclusion of control conditions makes it diicult to rule out the eect of variables other than the
receipt of basic income that could have aected work activity, such as general uctuations in the national or
local economy.
In addition, with respect to drawing conclusions from the current data, a number of issues of external valid-
ity (i.e., the ability to generalize the ndings of these limited trials to ongoing national programs) are important
to consider. First, all of the data derived from BIG pilot programs are subject to what could be termed a dura-
tion or transitory bias in which results from a time-limited trial may not generalize to a permanent program.
9
Brought to you by | Loyola Marymount Law School Library
Authenticated
Download Date | 12/10/18 10:20 PM
DE GRUYTER
Specically, reductions in labor responses to a program known to be temporary may understate the long-term
impact of guaranteed income because signicant reductions in work activity may only occur when individuals
have a more enduring sense of personal and nancial security.
A second issue of external validity relates to the size and signicance of the income guarantees found in
the BIG trials. According to the benchmarking data reported in Table 3, the average income guarantee across
the 16 trials was approximately 6.8% of the average annual income for individuals or households in the year
and country in which the trial was conducted. It is possible that the modest size of the average guarantee
represents an insuiciency bias in which the level of basic income found in evaluation studies is inadequate to
produce changes in work behavior that might be found in programs that supply a more generous benet. As a
base of comparison, if a basic income guarantee of one dollar above the oicial poverty line for an individual
were implemented in the United States in 2016, the ratio of the guarantee ($11,771) to annual per capita income
($52,195) would be approximately 22.5%. As this ratio is signicantly higher than that found in most evaluation
studies, it might lead to more signicant work reductions. Currently, there is no agreed upon ratio of guaranteed
income to annual income or consumption/expenditure, or any other consensus benchmarking measure, to
establish what constitutes a suicient guarantee and assess its impact on work activity.
In sum, continued progress in understanding the relationship between basic income and work activity
would benet from additional controlled studies, assessing long-term programs, and establishing standards
for benchmarking the income guarantee. At the same time, the current body of empirical data, in conjunction
with associated psychological research on non-pecuniary motivations to work, does not support predictions
that guaranteed subsistence income constitutes a substantial work disincentive.
As a new era of highly capable and intelligent technologies increasingly perform physical, cognitive, and
social-emotional tasks that have been the exclusive domain of human labor for many years, concerns have
grown regarding whether the number of paid activities for physical workers that are eliminated will exceed
those that are created by emerging markets. In the face of this automation anxiety, an increasing number of
policy makers in developed and developing nations are giving serious attention to the idea of guaranteed sub-
sistence income. As part of this consideration, social science has an important role to play. As reected in the
current work, social science research can contribute to the public discussion by providing a thorough analysis
of available data that bears on central issues related to this policy, such as the impact of guaranteed subsistence
income on the motivation to work.
References
–
–
–
– ’
’ ’
’
–
10 Brought to you by | Loyola Marymount Law School Library
Authenticated
Download Date | 12/10/18 10:20 PM
DE GRUYTER
–
–
–
–
’
’
–
’
’ –
–
–
’
–
–
’
11
Brought to you by | Loyola Marymount Law School Library
Authenticated
Download Date | 12/10/18 10:20 PM
DE GRUYTER
‘’
–
’ –
–
–
’
–
’
’
’
–
–
12 Brought to you by | Loyola Marymount Law School Library
Authenticated
Download Date | 12/10/18 10:20 PM
