Content uploaded by Wendy Janssens
Author content
All content in this area was uploaded by Wendy Janssens on Dec 07, 2020
Content may be subject to copyright.
Research Notes
The short-term economic effects of COVID-19 on low-income
households in rural Kenya: An analysis using weekly financial household
data
Wendy Janssens
a
, Menno Pradhan
b
, Richard de Groot
c,
⇑
, Estelle Sidze
d
,
Hermann Pythagore Pierre Donfouet
e
, Amanuel Abajobir
d
a
Vrije Universiteit, Amsterdam Institute for Global Health and Development and Tinbergen Institute, Netherlands
b
Vrije Universiteit, University of Amsterdam, Amsterdam Institute for Global Health and Development and Tinbergen Institute, Netherlands
c
Amsterdam Institute for Global Health and Development, Netherlands
d
African Population and Health Research Center, Kenya
e
African Population and Health Research Center (Kenya) and CREM, UMR CNRS 6211, University of Rennes 1, France
article info
Article history:
Keywords:
COVID-19 pandemic
Economic effects
Fixed-effects regressions
Risk-coping
East Africa
Kenya
abstract
This research assesses how low-income households in rural Kenya coped with the immediate economic
consequences of the COVID-19 pandemic. It uses granular financial data from weekly household interviews
covering six weeks before the first case was detected in Kenya to five weeks after during which various con-
tainment measures were implemented. Based on household-level fixed-effects regressions, our results sug-
gest that income from work decreased with almost one-third and income from gifts and remittances
reduced by more than one-third after the start of the pandemic. Nevertheless, household expenditures
on food remained at pre-COVID levels. We do not find evidence that households coped with reduced income
through increased borrowing, selling assets or withdrawing savings. Instead, they gave out less gifts and
remittances themselves, lent less money to others and postponed loan repayments. Moreover, they signif-
icantly reduced expenditures on schooling and transportation, in line with the school closures and travel
restrictions. Thus, despite their affected livelihoods, households managed to keep food expenditures at
par, but this came at the cost of reduced informal risk-sharing and social support between households.
Ó2020 The Authors. Published by Elsevier Ltd. This is an open access article under the CC BY-NC-ND license
(http://creativecommons.org/licenses/by-nc-nd/4.0/).
1. Introduction
The COVID-19 pandemic not only affects livelihoods through
increased risk of mortality, but also through containment measures
(World Health Organization, 2020). While the initial focus was on
the health effects of the pandemic, its socio-economic effects and
accompanying policy responses are receiving increased attention,
including in low-income countries.
The World Bank estimated that the COVID-19 crisis can push
between 40 and 60 million people into extreme poverty, most of
which in sub-Saharan Africa (Gerszon Mahler, Lakner, Castaneda
Aguilar, & Wu, 2020). Another report found that under a scenario
in which income and consumption contracts by 20 percent,
between 420 and 580 million people would be pushed into pov-
erty, reversing decades of decreasing poverty trends (Sumner,
Hoy, & Ortiz-Juarez, 2020). However, detailed insights in the
immediate effects of the COVID-19 crisis at the household-level
are lacking for most low-and middle-income countries. Most evi-
dence is from developed countries. In the US, the lockdown policy
reduced time spent outside the home (Gupta et al., 2020), con-
tributed to a significant decline in employment (Montenovo,
Jiang, Rojas, Schmutte, Simon, & Weinberg, 2020), and caused a
large drop in job vacancy postings (Kahn, Lange, & Wiczer, 2020).
Our study quantifies the income and expenditure loss after the
COVID-19 crisis at weekly intervals for a cohort of 328 low-income
rural households in Western Kenya. Since December 2019, house-
holds were visited on a weekly basis to collect detailed information
on each adult’s financial transactions in the past week (including
incomes, expenditures, loans, remittances and savings), allowing
for an analysis of pre- and post COVID-19 trends, unlike studies
that were initiated in response to the pandemic.
Our granular financial data enable us to assess how rural house-
holds were affected by the containment measures, how they
responded and coped financially. Have incomes from business,
employment and other sources declined? Were households forced
https://doi.org/10.1016/j.worlddev.2020.105280
0305-750X/Ó2020 The Authors. Published by Elsevier Ltd.
This is an open access article under the CC BY-NC-ND license (http://creativecommons.org/licenses/by-nc-nd/4.0/).
⇑
Corresponding author.
E-mail addresses: w.janssens@vu.nl (W. Janssens), m.p.pradhan@vu.nl
(M. Pradhan), r.degroot@aighd.org (R. de Groot), esidze@aphrc.org (E. Sidze),
hdonfouet@aphrc.org (H.P.P. Donfouet), aabajobir@aphrc.org (A. Abajobir).
World Development 138 (2021) 105280
Contents lists available at ScienceDirect
World Development
journal homepage: www.elsevier.com/locate/worlddev
to withdraw savings or sell assets? Were households able to
uphold their spending on food? These questions are of immediate
concern to policymakers seeking a balance between protecting the
health of the population by containing the spread of the virus and
implementing policies to protect people’s economic wellbeing.
A few recent studies examine the short-term effects of the
COVID-19 pandemic in Kenya. One study conducted in Nairobi’s
five largest urban informal settlements found that 80 percent of
their sample had experienced income loss, while also experiencing
food price increases. Two-thirds had skipped a meal at least once
or eaten less in the two weeks preceding the survey. Very few
respondents had received any kind of support from the govern-
ment or other sources (Population Council, 2020). However, this
study did not collect detailed financial transaction records; instead,
respondents retrospectively reported changes in income and
expenditures. Another diaries study in Kenya, using qualitative
phone interviews with respondents, found that 88 percent of the
respondents experienced declines in income, with urban people
hit hardest (Zollmann, Ng’weno, Gachoka, & Wanjala, 2020).
Based on household-level fixed-effects regressions, our findings
show that household incomes decreased sharply by up to one-
third in the five weeks after the first preventive measures were
implemented. We particularly observe a drop in income from work,
and a strong decrease in gifts and remittances received. Whereas
most households managed to keep food expenditures stable, expen-
ditures on education and transport decreased. We also find that
households deposited less savings, but also withdrew less, suggest-
ing that they were protecting whatever savings they still had. Infor-
mal risk-sharing betweenhouseholds declined as indicated by
declining credit and loans, loan repayments, gifts, and Harambee
contributions.
2. COVID-pandemic and response measures in Kenya
In Kenya, the first COVID-19 case was detected on 13 March,
immediately followed by measures to contain the spread of the
virus. On 15 March, schools were closed, and all workers were
directed to work from home if possible. Subsequently, international
flights were suspended, and bars and restaurants were closed. A
nationwide curfew was installed on 27 March. In early April, travel
restrictions were put in place in the most affected areas, including
Nairobi, Mombasa, Kilifi and Kwala Counties (Were, 2020). As of
May 31st, there were 1,888 confirmed cases and 63 deaths on a
population of almost 50 million (Ministry of Health, 2020).
3. Data and methods
We use data from Financial and Health Diaries, collected for a
study originally designed to measure the impact of a mobile
phone-based health insurance scheme in Kakamega county
through a cluster-Randomized Controlled Trial (RCT), and free
access to public care in Kisumu county through a prospective lon-
gitudinal analysis. Both schemes were on-hold during our study
period. The study population, drawn from low-income rural vil-
lages, consists of households with either a pregnant woman or a
mother with children below four years old. First, 32 villages were
randomly selected from the catchment areas of six health facilities.
Next, in each village ten households were randomly sampled from
lists with households fulfilling the study eligibility criteria.
Weekly data collection has been on-going since December
2019. The weekly interviews record all financial transactions (in-
come from work/social schemes, expenditures, savings, gifts/
remittances, loans) among others. Data are collected from all
adults in the household, separately and in private. Diaries data col-
lection was preceded by a household survey, collecting baseline
demographic, socio-economic, and health information.
In March 2020, data collection changed from in-person to
phone-based interviewing in response to the COVID-19 social dis-
tancing measures. The built-up rapport between respondents and
enumerators ensured continuously high response rates.
1
We took
various measures to ensure continued data quality. Fieldworkers
received intensive training on phone interview techniques, including
on privacy and ethics, and were encouraged to ensure the respon-
dents’ ability to answer freely by setting appropriate timing of calls
or planning call-backs at respondents’ convenience. Overall, 94 per-
cent of households owned a phone; at the individual level, 88 (77)
percent of male (female) respondents had their own cell number.
Respondents without phones were reached through their preferred
alternative contacts, including neighbors, community health work-
ers, friends and relatives. Field team leaders were trained on remote
data quality assurance and were supported with a protocol to pro-
vide psychosocial support to fieldworkers where needed.
We focus on the short-term effects of the COVID-19 outbreak.
We note that these would be indirect effects, since as of 31 May,
only 1 COVID-19 case and no deaths were recorded in our study
areas (Ministry of Health, 2020). Nevertheless, cases could have
been underestimated because the testing capacity was extremely
limited. As pre-COVID-19 period, we take the weeks between
February and mid-March 2020, when the first case was detected
in Kenya and the first measures were taken.
2
Our post-COVID-19
period covers the five weeks between 17 March and 20 April.
Our data identify several sources of income, including income
from work, gifts/remittances, loans/credits, and savings with-
drawals. We break down income from work into sub-categories:
income from business, formal employment, casual work, crop sales
and livestock sales. For expenditures, we categorize spending into
several subgroups, including food, education, and transportation.
We also observe other money outflows, such as gifts given, money
lent out, credit repayments, and savings deposits.
Our main objective is to determine how income and spending
levels have changed since the start of the COVID-19 pandemic in
Kenya. Our unit of observation is the household-interview week
(n= 2,995; 328 households observed in 11 weeks)
3
, aggregating
incomes and expenditures over all adult household members.
4
We
first present the weekly averages in graphical format. We then test
1
Weekly response rates have slightly declined since the switch to phone
interviews with 2 to 10 percent dependent on the week. We check for selective
response by regressing all household demographic variables from Table 1 on
dummies for post-COVID weeks to determine if average household composition
changes after the COVID pandemic. Out of 145 estimates, we find 15 significant
differences at the 10% level. We are therefore confident that selective household
response is not driving our results. The post-COVID difference in weekly response
rates between respondents with and without their own phone at baseline was
relatively limited at 5 percent, although significant at the 10% level (p = 0.073).
2
While our data allows us to go back further in time (to December 2019), there is a
large peak in expenditures and income in the financial data during the first weeks of
January due to the Christmas and New Year holidays, which would arbitrarily
increase the averages over the pre-COVID-19 period, affecting our main results
3
The potential number of household-weeks is 3,608 but we remove observations
from our analysis in cases where none of the household members were interviewed
during a particular week. Imputing these missing household-weeks with the pre- or
post-COVID average further strengthens our results. Size and significance of our
estimates increases, except for money withdrawn from savings.
4
In case one person in a household was not interviewed during a week, we
conservatively impute his or her financial transactions in the household-level total
with the average of the observed weeks. We test for robustness with two alternatives:
1) Imputing all missing values with the pre- or post-COVID average, instead of the
average across all weeks. Compared to our main results, these estimates are larger
and more often significant; 2) Reducing the sample to households with all
respondents present for a particular week. The sample size in this case drops from
2,995 to 1,663. Compared to our main results, these estimates are qualitatively
similar in size and significance, except for gifts/remittances received. Outliers, with a
value higher than the 99th percentile during an interview week, are also replaced
with the average over the other weeks. This has very little effect on our main findings.
All results available on request.
W. Janssens, M. Pradhan, R. de Groot et al. World Development 138 (2021) 105280
2
for significant differences between the mean of the pre-COVID-19 per-
iod and weekly post-COVID-19 periods using household-level fixed
effects regressions, clustering standard errors at the household level.
To investigate whether asset-rich households are better able to
cope with the lockdown than others (Dercon, 2002), we assess
heterogeneous effects by the possession of any savings, of more
than the median amount of savings; livestock, cattle or land own-
ership; and availability of agricultural products in stock at baseline.
4. Results
Table 1 presents baseline household information. Households
have on average five members with three members under 18 years.
Household heads are relatively young (37 years on average) and 24
percent are female headed. Nearly all heads are married. The
majority has received some basic education. The most common
occupation of the head is casual labor (39 percent), operating an
informal business (22 percent) or formal employment (18 percent).
About 14 percent of heads were not earning income at baseline.
Households had on average nearly Ksh 13,000 in savings, equiva-
lent to about six weeks of pre-COVID household income from work
(Ksh 2,036, see below). Households had on average about Ksh
3,900 loans outstanding and the amount of money lent to others
was Ksh 1,875. About 20 percent of households had a loan at a for-
mal institution (bank, microfinance institution) or a shop credit.
5
To contextualize our study, Table 1 also provides estimates
from Kakamega county, Kisumu county and Kenya. In comparison,
our sample is younger, with more small children, fewer elders and
younger heads, in line with our study eligibility criteria. Our sam-
ple is slightly better educated than households in Kakamega and
Kisumu, more reliant on casual labor and less on farming, espe-
cially compared to Kakamega. Finally, our sampled households
have lower amounts of loans outstanding, yet a higher share has
a loan at a formal institution, compared to the county averages.
Fig. 1 shows a decreasing trend in total cash inflows and out-
flows since mid-March. Especially income from work (e.g. revenue
from informal business, wage work), and received gifts and remit-
tances went down (Fig. 2). Table 2 shows that weekly income sig-
nificantly declined with up to Ksh 666 towards the end of April,
compared to a pre-COVID-19 average of Ksh 2,036.
Gifts and remittances comprised 22 percent of household
income before the pandemic hit (Column 2, Table 2). They
decreased considerably by Ksh 330 one month after the first case
was detected, implying a decrease of 38 percent compared to
pre-COVID-19 levels of Ksh 858 per week. This could be due to
the crisis severely affecting urban wage workers, or immigrants
in more advanced economies, who are an important source of
remittances for families in the countryside: in our data 50 percent
of remittances are from a relative outside the community.
In response to this income loss, households might resort to sev-
eral risk-coping strategies. We do not observe significant increases
in loans, loan repayments received nor incoming Harambee contri-
butions to generate additional cash (Columns 3–5, Table 2). If any-
thing, the amount borrowed since the start of the epidemic
decreased slightly. Perhaps surprisingly, households have reacted
to the crisis by withdrawing less money from their savings, with
significant declines in the last three weeks of our study period,
amounting to a decrease of Ksh 199 in the final week. This could
indicate that households are not yet willing to utilize their savings
at this stage in the crisis, but rather prepare for worse times ahead.
Table 3 breaks down income from work into sub-categories.
Most of its decline is due to decreasing income from (informal)
business (Column 1), formal employment (Column 2) and, to some
extent, income from crop sales (Column 3) and casual labor (Col-
umn 5). The first two categories also make up the largest share
of total income underscoring that households have been affected
in their main livelihoods. The insignificant results in Column 4
show that households have not increased livestock sales to raise
additional income.
For outflows (Table 4), we observe a sharp and rising drop in
household expenditures of up to Ksh 569 in the final study week,
compared to a pre-COVID-19 average of Ksh 2,414 (Column 1),
i.e. a 24 percent decrease in weekly expenditures over a one-
month period. Gifts and remittances also decreased with 36 per-
cent compared to pre-pandemic levels (Column 2). Other strategies
to cope with the income loss are to reduce lending to others (Col-
umn 3), postpone repayment of outstanding loans and credit (Col-
umn 4), and reduce outgoing Harambee contributions (Column 5).
Also, households have reduced savings, especially immediately
after the first COVID-19 case was detected and in the last study
week (Column 6).
A major concern is whether families have been able to uphold
their food expenses since the lockdown measures were put in
place. Despite the decline in income, families spend the same
amount of money on food as before the crisis struck (Fig. 3 and Col-
umn 1 of Table 5). However, households spend much less on edu-
cation and transportation, in line with the school closures and
travel restrictions (Columns 2–3, Table 5).
6
This may have given
households some room in their budget to maintain food expendi-
tures. Whereas spending on communication has started to decline
(Column 4, Table 5), we do not observe a significant change in spend-
ing on recreation, ceremonies or funerals (Column 5, Table 5), nor on
other consumption groups – including short-term business and agri-
cultural investments.
7
Table 6 investigates households’ preparedness to cope with this
shock. Three quarters of households (75.9 percent) had some for-
mal or informal savings at baseline, 77.4 percent owned livestock,
and 39.3 percent owned cattle.
8
Households also possessed other
assets that could be sold during times of need, 68.5 percent owning
land and 44.2 percent having agricultural produce in stock. Most
households (59.8 percent) had a loan outstanding at baseline.
Households with savings at baseline withdraw less money from
their savings than the households without savings at baseline. This
heterogeneous effect appears to be independent of the level of sav-
ings since withdrawals are not significantly related to having sav-
ings above or below the median amount of savings. A potential
explanation is that households generally try to hold on to their sav-
ings as much as possible in the early weeks of the pandemic, and
that those with prior savings habits are better able to do so. House-
holds with an outstanding loan at baseline appear less able to
maintain their income when the pandemic hits than those without
outstanding loans, suggesting an association between longer-term
limited income-generating capacity and vulnerability.
For baseline ownership of cattle, land and agricultural stock, we
do not find evidence of heterogeneous effects on income, expendi-
tures, and other financial flows. This is consistent with the findings
that households did not resort to selling livestock (Table 5). House-
holds that owned livestock at baseline were better able to maintain
food expenditure; however, they did not differ systematically on
other financial flows.
5
This share is 34 percent among households with any loan.
6
There is a peak in transportation expenditure in the week of 13 March –
potentially related to individuals traveling home in anticipation of the lockdown
(Figure 3), which might increase the pre-COVID average. Excluding this week from the
analysis yields slightly smaller, yet still significantly negative estimates for four of the
five post-COVID weeks.
7
Results available upon request.
8
Note that the baseline occurred in November 2019, four months before the COVID
19-pandemic struck, and so may not provide an accurate picture of savings and assets
right before COVID-19 pandemic.
W. Janssens, M. Pradhan, R. de Groot et al. World Development 138 (2021) 105280
3
5. Discussion
This paper uses detailed high-frequency data to estimate the
short-term economic effects of the COVID-19 pandemic on low-
income rural households in Kenya, and their strategies to cope
with declining incomes. Our results suggest that income from work
decreased with almost one third in the five weeks following the
first confirmed case in Kenya. Gifts and remittances – a
major source of income – reduced by more than one third.
Because of the limited spread of COVID-19 by the end of the study
period, the decline in income is primarily a result of the lockdown
measures and subsequent economic downturn.
These results are unlikely to be driven by independent seasonal
effects. A similar Financial Diaries study conducted pre-COVID
among 120 households in a neighboring rural area (Nandi county)
showed a similarly slight decrease in income and expenditures
until the end of February; but a steady increase in income and
expenditures from early March onwards until at least the end of
April (Geng, Janssens, Kramer, & van der List, 2018,Fig. 1).
We do not find evidence that households coped through
increased borrowing or withdrawing savings. Instead, households
reduced the gifts and remittances given to others; they also lent
out less money, postponed loan repayments, and deposited less
savings. Households also significantly reduced expenditures,
Table 1
Baseline characteristics of households in the sample.
Study sample Kakamega Kisumu Kenya
County
Kakamega (residence of households) 0.697
Kisumu (residence of households) 0.303
Household composition
# of household members 5.070 4.788 3.995 3.979
# of members age 0–5 years 1.382 0.836 0.648 0.649
# of members age 6–12 years of age 1.092 1.022 0.816 0.787
# of members age 13–18 years of age 0.538 0.823 0.599 0.567
# of members age 19–64 years of age 2.000 1.908 1.804 1.822
# of members age 65 and over 0.058 0.198 0.129 0.155
Characteristics of the head
Age in years 37.086 46.834 46.681 44.557
Female head 0.235 0.349 0.308 0.323
Head is married 0.920 0.758 0.715 0.693
Educational status
No schooling 0.037 0.115 0.030 0.135
Incomplete primary 0.321 0.398 0.305 0.276
Complete primary 0.327 0.152 0.236 0.169
Incomplete secondary 0.092 0.107 0.101 0.092
Complete secondary or higher 0.223 0.228 0.328 0.328
Main occupation
None 0.144 0.045 0.022 0.049
Own business 0.220 0.241 0.372 0.245
Farm owner 0.061 0.331 0.083 0.264
Casual labor (including casual farm work) 0.391 0.096 0.132 0.123
Wage work 0.183 0.287 0.392 0.320
Livelihood activity
Any member in household engaged in own business 0.294 0.339 0.515 0.324
Any member in household engaged in own farm work 0.073 0.479 0.158 0.362
Any member in household engaged in casual labor 0.459 0.134 0.193 0.164
Any member in household engaged in wage work 0.242 0.519 0.504 0.415
Total amount of savings 12,996.7
Loan amount outstanding 3,898.6 8,169.0 5,019.9 17,020.5
Household has loan at formal institution
a
0.205 0.077 0.061 0.116
Amount of money lent out 1,874.8
Observations 328 493 497 21,658
Notes: Estimates for Kakamega , Kisumu and Kenya are derived from the Kenya Integrated Household Budget Survey 2015/16 and sampling weights are used to account for
the sampling design of the survey. Information on savings and money lent out was not available in the KIHBS.
a
includes banks, microfinance institutions and shop credit.
Fig. 1. Average cash inflow (left) and outflow (right), with linear trend. Notes: The dots represent the weekly average, with a 95% confidence interval.
W. Janssens, M. Pradhan, R. de Groot et al. World Development 138 (2021) 105280
4
especially on schooling and transportation, in line with preventive
measures such as school closures and travel restrictions. These
strategies enabled households to keep their food expenditures at
par in the short-term, but at the cost of reduced informal risk-
sharing between households. We find little evidence that baseline
savings and assets increased household resilience. However,
households with an outstanding loan at baseline seemed more vul-
nerable to income loss during the lockdown than others.
Few other studies give such a detailed picture of household
finances before and after the onset of the current pandemic, espe-
Fig. 2. Trends in selected inflow variables, income from work (left) and gifts/remittances (right). Notes: The dots represent the weekly average, with a 95% confidence
interval.
Table 2
Change in weekly cash inflow variables after 17 March, compared to the period 4 February 16 March.
(1) (2) (3) (4) (5) (6)
Income from
work
Gift/ remittance
received
Money
borrowed
Loan/credit repayment
received
Harambee contribution
received
Savings
withdrawn
17 March23 March 187.1 83.7 60.3** 42.2 7.8 20.2
(117.1) (63.6) (23.4) (47.2) (5.1) (73.8)
24 March30 March 524.0*** 189.4*** 26.9 31.7 3.9 15.3
(101.7) (61.2) (43.9) (28.3) (4.1) (57.8)
31 March6 April 354.9** 296.7*** 199.8 10.2 2.8 195.1**
(157.2) (57.5) (228.8) (23.2) (4.2) (88.7)
7 April13 April 354.2*** 130.0** 50.3** 35.8 88.0 211.0***
(117.2) (60.3) (23.8) (40.1) (89.2) (63.9)
14 April20 April 666.1*** 329.9*** 3.5 39.1 28.6 198.8***
(117.6) (64.2) (32.6) (27.0) (27.4) (70.0)
Mean between 4 February16
March
2036.1 858.3 94.8 138.3 18.8 699.1
Household FE YES YES YES YES YES YES
Observations 2,995 2,995 2,995 2,995 2,995 2,995
Notes: This table presents estimates from household fixed-effects (FE) regressions, with standard errors clustered at the household level. * p< 0.1; ** p< 0.05; *** p< 0.01.
Table 3
Change in weekly income after 17 March, by source of income, compared to the period 4 February16 March.
(1) (2) (3) (4) (5) (6)
Business income
(revenue)
Income from
employment (salary)
Income from farming
(crop sales)
Income from
livestock sales
Income from casual
labour
Other
income
17 March23 March 120.9 56.4 11.7 24.8 –23.4 49.6
(100.4) (133.5) (25.6) (23.0) (18.2) (38.7)
24 March30 March 219.9** –233.3*** 51.3*** 21.8 47.4** 7.7
(92.0) (57.6) (14.1) (22.1) (19.5) (12.0)
31 March6 April 300.9*** 12.0 –23.3 6.4 85.1*** 51.6
(111.8) (117.4) (24.3) (28.6) (21.9) (41.8)
7 April13 April 106.1 79.5 57.9*** 24.8 –33.8 12.4
(114.3) (82.6) (13.7) (22.5) (29.7) (21.7)
14 April20 April 249.3*** 312.9*** 41.0* 26.5 39.5 6.0
(91.5) (73.1) (24.6) (23.8) (38.5) (11.9)
Mean between 4
February16 March
1254.5 470.3 108.6 26.1 186.2 27.7
Household FE YES YES YES YES YES YES
Observations 2,995 2,995 2,995 2,995 2,995 2,995
Notes: This table presents estimates from household fixed-effects (FE) regressions, with standard errors clustered at the household level. * p< 0.1; ** p< 0.05; *** p< 0.01.
W. Janssens, M. Pradhan, R. de Groot et al. World Development 138 (2021) 105280
5
cially from low-income countries.
9
An exception is a comparable
diaries study in Bangladesh, which follows 60 households daily,
showing a 75 percent drop in daily earnings in the first week of the
lockdown in Bangladesh (Hrishipara Daily Diaries, 2020). This is a
stronger decline than our findings, potentially due to the (peri-)
urban nature of their sample compared to our rural mostly self-
employed population.
Our findings on income effects provide depth to the emerging
results from one-time phone surveys across the African continent
and other low-and middle-income regions. In Senegal, 86 percent
of respondents reported a below-average income during the early
stages of the pandemic (Le Nestour, Mbaye, Sandefur, &
Moscoviz, 2020). In a nine-country study, BRAC International,
2020 found that income loss is common across study sites, ranging
from 47 percent of respondents in Myanmar to 93 percent in
Liberia reporting that their income ‘reduced a lot’, or ‘completely
stopped’. The biggest losses were reported by those engaged in
small businesses or casual work, like our findings. In contrast, they
find strong negative effects on food consumption with most
respondents in Liberia, Philippines, Uganda, and Rwanda reporting
‘a lot’ less food consumption than before the COVID-19 crisis,
Fig. 3. . Trends in weekly expenditures for selected sub-groups: Food, education, and transportation. Notes: The dots represent the weekly average, with a 95% confidence
interval.
Table 4
Change in weekly outflow variables after 17 March, compared to the period 4 February16 March.
(1) (2) (3) (4) (5) (6)
Expenditures Gift/ remittance
given
Money lent
out
Loan/credit repayment
given
Harambee
contribution
Savings
deposited
17 March23 March 31.2 2.2** 60.9** 108.8*** 4.9 117.8**
(169.1) (0.9) (24.7) (31.9) (5.2) (53.0)
24 March30 March 266.2** 2.4 59.8* 136.1*** 1.2 170.2***
(111.1) (1.6) (30.8) (29.9) (3.9) (46.8)
31 March6 April 286.6* 2.8* 93.6*** 144.5*** 1.9 99.5
(161.5) (1.6) (30.7) (34.8) (2.3) (76.2)
7 April13 April 285.3** 2.8* 19.9 164.3*** 4.8*** 65.8
(132.7) (1.5) (35.8) (29.1) (1.7) (52.4)
14 April20 April 568.7*** 3.5** 104.3*** 188.4*** 4.4** 240.3***
(127.6) (1.6) (30.0) (33.5) (1.7) (51.1)
Mean between 4 February16
March
2413.5 4.1 203.6 518.8 6.7 654.5
Household FE YES YES YES YES YES YES
Observations 2,995 2,995 2,995 2,995 2,995 2,995
Notes: This table presents estimates from household fixed-effects (FE) regressions, with standard errors clustered at the household level. * p< 0.1; ** p< 0.05; *** p< 0.01.
9
Baker et al. (2020) and Carvalho et al. (2020) show high-detail pre- and post-
COVID-19 transaction data from the US and Spain, respectively.
W. Janssens, M. Pradhan, R. de Groot et al. World Development 138 (2021) 105280
6
potentially due to the more urban nature of their samples. A major
disadvantage of these surveys is that they lack pre-COVID-19 data
and rely on retrospective self-reported changes in income or food
consumption. Moreover, our data allow us to assess the financial
coping strategies of households dealing with the economic
downturn.
Some caution is warranted though. Our results on food expen-
ditures focus on spending – not on consumption since the diaries
only capture financial transactions. This might partially explain
differences with surveys on self-reported food consumption. Given
the rural nature of our sample, this most likely strengthens our
results on short-term food security, to the extent that households
could consume home-grown foods or purchase food at lower
prices. Lower food prices in food-producing areas could result from
a breakdown in the food markets because of transportation restric-
tions. However, this also underscores that our findings should not
be extrapolated to an urban population. Likewise, the sample con-
sists of families with small children – young single adults or old-
age people might have fared differently, dependent on their
employment status and dependence on remittances.
Another potential limitation concerns the shift to phone-based
interviewing: we cannot preclude some underreporting occurred
during the first week or two of the transition. However, the trust
built up during several months of weekly meetings since 2019
combined with intensive training of field workers and supervisors,
helped in easing the transition, reducing concerns about low data
quality after the pandemic. Attrition remained low and, to the
extent it occurred, there was no systematic bias in which house-
holds dropped out of the sample.
What do our findings imply for government relief measures in
Kenya? If the crisis and preventive measures persist, households’
income could continue to decline. While households were able to
maintain food spending during the initial weeks after the first mea-
sures, this could become more difficult as time passes. Income
levels need to be protected to avoid food insecurity. Our findings
on reduced gifts, remittances, and informal borrowing and lending
also suggest that most households were in it alone – at least during
the initial stages of the pandemic. Whereas in other times rural
households rely on informal credit (Udry, 1994) or support from
their social networks in case of need – especially when dealing
with idiosyncratic shocks such as illness or injury (De Weerdt &
Dercon, 2006; Fafchamps & Lund, 2003; Geng et al., 2018), these
informal risk-sharing arrangements faltered when households
throughout the region were affected. Previous research suggests
that households may sell productive assets to smooth consumption
during covariate shocks (Nguyen, White, & Ma, 2019; Rosenzweig
& Wolpin, 1993), but we find no evidence that households resorted
to selling assets such as livestock in the short-run.
Table 6
Heterogeneity analysis, comparing change in financial transactions after 17 March, compared to the period 4 February16 March, by subgroup.
(1) (2) (3) (4) (5) (6) (7)
Baseline
mean
Income Expenditures Food Money
borrowed
Income from animal
sales
Money
withdrawn
Difference: Any savings at baseline 0.759 41.8 136.6 28.7 185.7 26.7 224.9**
(194.3) (173.5) (56.2) (159.0) (30.1) (93.1)
Difference: More than median savings (Ksh 2530) at
baseline
0.500 37.8 33.8 67.6 68.9 34.5 131.6
(157.6) (155.4) (57.3) (80.8) (44.5) (85.8)
Difference: Any loan at baseline 0.598 299.3* 140.2 40.8 128.1 42.8 32.1
(164.6) (158.4) (54.8) (97.9) (38.4) (83.3)
Difference: Household owns any livestock 0.774 95.5 128.2 162.3** 27.9 –23.7 7.5
(156.8) (164.3) (65.5) (63.6) (29.2) (105.2)
Difference: Household owns any cattle 0.393 51.9 66.4 73.8 109.6 46.6 104.7
(158.0) (158.3) (57.9) (68.0) (39.2) (90.4)
Difference: Household owns any land 0.686 101.6 143.0 100.1 169.1 28.3 29.5
(197.5) (199.1) (71.2) (126.2) (32.8) (110.9)
Difference: Household has any agricultural product
in stock
0.442 191.4 12.7 38.1 93.3 39.8 6.0
(156.2) (154.5) (60.3) (72.2) (52.0) (88.2)
Notes: This table presents results from a difference-in-difference analysis, comparing the mean before and after the first COVID-19 case, for each subgroup specified. The
estimate shown in the table is the interaction between a post-COVID dummy and the subgroup dummy. The regression includes household fixed-effects, with standard errors
clustered at the household level. * p< 0.1; ** p< 0.05; *** p< 0.01.
Table 5
Change in weekly expenditures for selected sub-groups after 17 March, by category, compared to the period 4 February16 March.
(1) (2) (3) (4) (5)
Food Education Transportation Communication Recreation/ ceremonies/ offerings/ funerals
17 March23 March 5.0 82.1*** 7.9 2.5 13.7
(45.5) (15.9) (18.2) (4.6) (16.4)
24 March30 March 28.4 71.2*** 35.2*** 2.4 11.3
(46.2) (14.1) (12.1) (4.8) (9.0)
31 March6 April 21.9 76.0*** 35.5*** 4.1 18.3
(50.1) (11.6) (13.2) (5.5) (24.4)
7 April13 April 48.9 68.6*** 43.1*** 12.0*** 303.1
(50.5) (13.0) (14.0) (3.6) (324.1)
14 April20 April 58.6 70.8*** 61.3*** 9.1** 103.9
(46.5) (10.6) (13.6) (4.5) (89.4)
Mean between 4 February16 March 939.9 109.6 132.8 52.9 77.4
Household FE YES YES YES YES YES
Observations 2,995 2,995 2,995 2,995 2,995
Notes: This table presents estimates from household fixed-effects regressions, with standard errors clustered at the household level. * p< 0.1; ** p< 0.05; *** p< 0.01.
W. Janssens, M. Pradhan, R. de Groot et al. World Development 138 (2021) 105280
7
The Kenyan government has announced several economic sup-
port measures, including tax relief, reduction of VAT, and a reduc-
tion of income and business tax (Were, 2020). As most of our
households are employed in the informal sector, these relief mea-
sures may not reach them. The most relevant measure seems to be
the cash transfer to the elderly, orphans, and other vulnerable
members, distributing 8,000 shillings to more than 1 million ben-
eficiaries during the week of 20 April.
10
The latest figures show that
36 percent of Kenyans still live below the poverty line of US$1.90 per
day, amounting to more than 19 million citizens (Awiti, Dennis,
Mutie, Sanya, Angelique, & Wankuru, 2018). To protect consumption
and food security of all the poor and vulnerable, safety net measures
need to be expanded. Mobile money transfers can play a major role
in this regard, but challenges such as a lack of centralized household
registries remain an issue (Chacha et al., 2020). At the other end, the
food supply chain should be protected to keep down food inflation
and ensure households are able to purchase food at their preferred
markets.
Declaration of Competing Interest
The authors declare that they have no known competing finan-
cial interests or personal relationships that could have appeared
to influence the work reported in this paper.
Acknowledgements
We are grateful for funding from the Health Insurance Fund,
Amref Health Africa and PharmAccess Amsterdam through the i-
PUSH programme (National Postcode Lottery) and the Joep Lange
Institute. We would like to thank the two anonymous reviewers
and the editor for their constructive comments.
References
Awiti, C. A., Dennis, A. C. K., Mutie, C. K., Sanya, S. O., Angelique, U., Wankuru, P. C.,
et al. (2018). Kenya economic update: Policy options to advance the Big 4 -
unleashing Kenya’s private sector to drive inclusive growth and accelerate
poverty reduction (No. 125056; pp. 1–88). The World Bank.
Baker, S. R., Farrokhnia, R. A., Meyer, S., Pagel, M., & Yannelis, C. (2020). How does
household spending respond to an epidemic? consumption during the 2020 covid-
19 pandemic (No. w26949). National Bureau of Economic Research.
BRAC International. (2020). Rapid food and income security assessment Round 2:
How are BRAC International volunteers and programme participants coping
with COVID-19. BRAC International.
Carvalho, V. M., Hansen, S., Ortiz, A., Garcia, J. R., Rodrigo, T., Rodriguez Mora, S.,
et al. (2020). Tracking the Covid-19 crisis with high-resolution transaction data
(No. DP14642). Centre for Economic Policy Research.
Chacha, P. W., Angelique, U., Sienaert, A., Mutie, C. K., Afram, G. G., Belli, P., et al.
.Kenya Economic Update: Turbulent Times for Growth in Kenya - Policy Options
during the COVID-19 Pandemic, 148076, 1–70.
De Weerdt, J., & Dercon, S. (2006). Risk-sharing networks and insurance against
illness. Journal of Development Economics, 81(2), 337–356.
Dercon, S. (2002). Income risk, coping strategies, and safety nets. The World Bank
Research Observer, 17(2), 141–166.
Fafchamps, M., & Lund, S. (2003). Risk-sharing networks in rural Philippines. Journal
of Development Economics, 71(2), 261–287.
Geng, X., Janssens, W., Kramer, B., & van der List, M. (2018). Health insurance, a
friend in need? Impacts of formal insurance and crowding out of informal
insurance. World Development, 111, 196–210.
Gerszon Mahler, D., Lakner, C., Castaneda Aguilar, R. A., & Wu, H. (2020). The impact
of COVID-19 (Coronavirus) on global poverty: Why Sub-Saharan Africa might be
the region hardest hit. https://blogs.worldbank.org/opendata/impact-covid-19-
coronavirus-global-poverty-why-sub-saharan-africa-might-be-region-hardest.
Gupta, S., Nguyen, T. D., Rojas, F. L., Raman, S., Lee, B., Bento, A., et al. (2020).
Tracking Public and Private Response to the Covid-19 Epidemic. Evidence from
State and Local Government Actions. National Bureau of Economic Research.
Hrishipara Daily Diaries. (2020). Hrishipara Daily Diaries—Corona virus. https://
sites.google.com/site/hrishiparadailydiaries/home/corona-virus.
Kahn, L.B., Lange, F., & Wiczer, D.G. (2020). Labor Demand in the Time of Covid-19:
Evidence from Vacancy Postings and Ui Claims. National Bureau of Economic
Research.
Le Nestour, A., Mbaye, S., Sandefur, J., & Moscoviz, L. (2020). Phone survey on the
Covid crisis in Senegal. Harvard Dataverse.
Ministry of Health. (2020). COVID-19 outbreak in Kenya. Daily situation report 66.
Ministry of Health.
Montenovo, L., Jiang, X., Rojas, F.L., Schmutte, I.M., Simon, K.I., Weinberg, B.A., et al.
(2020). Determinants of Disparities in Covid-19 Job Losses. National Bureau of
Economic Research.
Nguyen, G. T. H., White, B., & Ma, C. (2019). When Faced with Income and Asset
Shocks, Do Poor Rural Households in Vietnam Smooth Food Consumption or
Assets? The Journal of Development Studies, 55(9), 2008–2023.
Population Council. (2020). Kenya: COVID-19 knowledge, attitudes and practices—
Responses from Second Round of Data Collection in Five Informal Nairobi
Settlements (Kibera, Huruma, Kariobangi, Dandora, Mathare) [COVID-19
Research & Evaluations Brief]. Population Council.
Rosenzweig, M. R., & Wolpin, K. I. (1993). Credit Market Constraints, Consumption
Smoothing, and the Accumulation of Durable Production Assets in Low-Income
Countries: Investments in Bullocks in India. Journal of Political Economy, 101(2),
223–244.
Sumner, A., Hoy, C., & Ortiz-Juarez, E. (2020). Estimates of the impact of COVID-19
on global poverty (WIDER Working Paper 2020/43). UNU-WIDER.
Udry, C. (1994). Risk and Insurance in a Rural Credit Market: An Empirical
Investigation in Northern Nigeria. The Review of Economic Studies, 61(3),
495–526.
Were, M. (2020). COVID-19 and socioeconomic impact in Africa (WIDER
Background Note 2020/3). UNU-WIDER.
World Health Organization. (2020). WHO Coronavirus Disease (COVID-19)
Dashboard. https://covid19.who.int/.
Zollmann, J., Ng’weno, A., Gachoka, A., & Wanjala, C. (2020). When hustling fails:
The impact of coronavirus mitigation efforts on ordinary people’s livelihoods.
https://fsdkenya.org/blog/when-hustling-fails/
10
The Star Kenya: https://www.the-star.co.ke/news/2020–04-19-inua-jamii-bene-
ficiaries-to-get-sh8000-each-from-monday/
W. Janssens, M. Pradhan, R. de Groot et al. World Development 138 (2021) 105280
8
