Online Annex 1.1. How Will The COVID-19 Pandemic Affect Poverty and Inequality?

CHAPTER 1 FISCAL POLICIES TO ADDRESS THE COVID-19 PANDEMIC
Online Annex 1.1. How Will the COVID-19 Pandemic Affect Poverty
and Inequality? 1
The COVID-19 pandemic could reverse decades-long progress in reducing poverty. Based on projected decline in per capita
incomes and potential changes in income distribution, an additional of 100-110 million people globally would be expected to
fall into extreme poverty relative to the pre-COVID-19 trend. The pandemic-related social assistance programs—
supporting directly the poor and cushioning the growth slowdown—may contain the projected rise in poverty by about one-
fifth. These estimates, however, are subject to a large degree of uncertainty depending on pandemic developments, designated
poverty thresholds, the growth outlook, and fiscal policy responses. A full decomposition of fiscal policy impact is difficult
given fiscal lifelines have cushioned the recession besides directly supporting the poor. At the same time, income inequality
within countries, measured by Gini coefficients, is estimated to widen by 0.03 points on average, including in many advanced
economies. Tackling rising inequality and poverty is now more urgent, which requires safeguarding essential social spending
and building a strong and resilient social protection system against future shocks and pandemics.
The economic contractions across almost all countries have imperiled decades-long progress in
reducing poverty. The global population of individuals who live in extreme poverty—defined as less than
$1.90 per day—had fallen by two-thirds since 2000 to about 650 million people in 2018 (Online Annex
Figure 1.1.1).
Online Annex Figure 1.1.1. Global Poverty Online Annex Figure 1.1.2. The
Trend, 1981-2018 Relationship between Poverty Rate and
(Millions of people,left scale; percent, right scale) Income Inequality
Poverty rates have declined over the last four (Percent and Gini coefficients)
decades. Poor countries tend to have higher income inequality.
.7
.6
.5
Gini
.4
.3
.2
0 .2 .4 .6 .8 1
Poverty
Advanced Economies Emerging Market Economies

Low Income Developing Countries
Sources: World Bank PovcalNet database; and IMF staff estimates.

The costs of the crisis are being borne disproportionately by poor and vulnerable individuals, adding to
the deep social inequities that have long afflicted many countries. The nature of the pandemic has
imposed large strains on labor-intensive, face-to-face services that employ a large share of low-income
workers. Across all income groups, countries with higher poverty rates tend to have higher income
inequality (Online Annex Figure 1.1.2) and greater poverty hardship—measured by the poverty gap, that
is, the mean distance of incomes for poor individuals from the poverty line expressed as a share of the
poverty line. This annex estimates the near-term effect of the COVID-19 pandemic on global poverty
and income inequality, noting the uncertainty surrounding the estimates.
Estimation of Global Poverty and Income Inequality during COVID-19
The analysis starts with an estimation of the trend relationship between poverty rates and per capita
output level (in purchasing power parity terms) and income inequality across 180 countries from 1981 to
1 Prepared by W. Raphael Lam, Delphine Prady, and Baoping Shang.
International Monetary Fund | October 2020 1

FISCAL MONITOR: POLICIES FOR THE RECOVERY
2018, using the poverty data from the June 2020 vintage of World Bank PovcalNet database. The
estimation uses a panel regression accounting for country fixed effects based on an unbalanced panel.
Estimates show that lower per capita income levels, slower growth, and higher income inequality are
associated with higher poverty rates. Results are robust across different poverty thresholds established by
the World Bank (for example, extreme poverty for individuals living off of $1.90 per day, and at higher
levels of income at $3.20 and $5.50 per day) (Online Annex Table 1.1.1). Additional specifications
controlling for time effects point to qualitatively similar results. Our baseline specification does not
account for the possible interactions between poverty-growth elasticity and income inequality—for
example, the poverty reduction may be less responsive to growth when inequality rises (Ravailion 1997).
The rise in poverty rates as per capita income falls is consistent with the literature that incomes of poor
households (those in the lowest two quintiles) tend to change with the aggregate growth rates (Dollar,
Kleineberg, and Kraay 2013) ), as income shares across household deciles are not statistically significant
different when growth varies over the medium term.
Online Annex Table 1.1.1. Regression Results on Relationships between Poverty and
Per Capita Income
Sources: World Bank PovcalNet database, IMF World Economic Outlook database; and IMF staff estimates.
Note: ***, **, * denote statistically significant levels at 1 percent, 5 percent, and 10 percent, respectively. Sample size spans
across all countries from 1981 to 2018. Numbers in parentheses are standard errors. PPP = purchasing power parity. The table
shows various specifications across different poverty thresholds set by the World Bank. Robustness check includes additional
control variables on the share of temporary and informal employment, and the coverage and adequacy of social protection
systems (not shown in the table). The qualitative results that per capita income and inequality affects poverty remain robust
across specifications. The regression results in the second column of the extreme poverty regression is used for projection of
global extreme poverty during the pandemic.
The estimated coefficients are then used to project poverty headcount across countries. A baseline
scenario of projections on per capita output for 2020–23 is considered along with a projected rise in
income inequality. The rise in income inequality is estimated using the pre–COVID-19 income
distribution across income deciles of countries and the estimated coefficients of negative growth shocks
on household income deciles (see below). The results suggest that the poverty rate is likely to rise during
the pandemic for the first time in this century, reversing the decades-long declining trend of extreme
poverty (Online Annex Figure 1.1.3). Under the baseline scenario in June 2020 World Economic Outlook
(WEO) Update, the COVID-19 pandemic could push around 100-110 million people into extreme
poverty (at $1.90 per day) during the pandemic, an equivalent of near 1.5 percentage point higher poverty
2 International Monetary Fund | October 2020

rates relative to pre–COVID-19 trends. 2 The estimates are comparable to those by World Bank in June
2020 that showed a rise of extreme poverty in 2020 of 70 to 100 million people, which show the rise
relative to pre–COVID-19 levels and adjusted for 2019 growth revisions, and assuming no change in
income distribution (World Bank 2020). 3 The number of people falling into poverty is subject to a wide
range—in which the ongoing revision of data in a few countries with large populations could affect the
global estimates—and could be higher if the pandemic outbreaks become more severe. Most of the
projected rise in poverty headcounts is driven by countries in sub-Saharan Africa and South Asia. The 10
countries with the highest levels of extreme poverty account for more than half of the increase in global
poverty headcounts (Online Annex Figure 1.1.4). Over three-quarters of countries are expected to see an
increase in poverty, of which one-fifth of countries could see extreme poverty rates rising by more than 2
percentage points (Online Annex Figure 1.1.5).
Even before the health crisis, income was highly concentrated at the top deciles of households,
particularly among countries with high poverty rates. In those countries, the bottom half of the
population accounts for less than one-fifth of aggregate income and consumption (Online Annex Figure
1.1.6). The pandemic has had a disproportionate effect on low-income households in many countries
because they are concentrated in the informal sectors, are more vulnerable to job losses, have lower
financial savings, and have less access to healthcare. As a result, the crisis will likely widen income
inequality within countries. In some countries, however, the pandemic tends to affect more adversely the
urban households (usually in higher-income groups) than those in the rural areas partly because the
infection rates are higher in areas with greater population density. This disproportionate impact could
reduce income inequality in the near term, particularly after accounting for social benefits.
Online Annex Figure 1.1.3. Global Online Annex Figure 1.1.4. Rising Levels
Poverty Rates of Extreme Poverty in Selected Countries
(Percent) (Millions of people)
The declining trend of global poverty rates is likely to The rise in global poverty is driven largely by countries
reverse as a result of the pandemic. in sub-Saharan Africa and South Asia.
Sources: World Bank PovcalNet database; IMF, October 2020 World Economic Outlook; and IMF staff estimates.
Note: The term “extreme poverty” refers to the threshold of $1.90 per day across all countries. Estimates are based on the
methodology in Online Annex Table 1.1 of top-down growth projections and poverty data from the World Bank PovcalNet
database, which could be different from national authorities’ estimates. For example, Uganda’s authorities estimate that the
precrisis poverty rate was 18.6 percent and could rise to 26 percent based on household income analysis. The solid lines in the
2In the baseline scenario, global extreme poverty is projected to rise by 108 million people during the pandemic relative to pre-COVID-19 trend
based on the decline of per-capita income levels, per-capita growth, and potential rise of income inequality. The estimated range of 100-110
million, underscores the uncertainty surrounding the data in a few countries as well as evolving global developments. Based on the latest global
developments as of mid-September, which has slightly improved relative to June, the global poverty estimates are likely to be about 9 million less,
putting those at the lower end of the range. However, some individual countries would have seen an increase relative to June given the pandemic
has turned more severe and growth has been revised downward. Overall, the poverty estimates vary significantly across growth outlook scenarios
and alternative poverty thresholds defined by the World Bank (for example, $3.20 per day or $5.50 per day).
3 The World Bank estimates have a wider range and higher estimates once considering changes in income distribution (World Bank 2020).

left panel refer to actual and projections based on the World Economic Outlook database, while the dotted lines refer to the
projections prior to the health crisis.
Online Annex Figure 1.1.5. Distribution of Potential Rise in Extreme Poverty

(Number of countries)
Sources: World Bank PovcalNet database, IMF World Economic Outlook, and IMF staff estimates.
Online Annex Figure 1.1.6. Income Share, Online Annex Figure 1.1.7. Estimated
by Household Income Deciles Increase in Income Inequality
(Percent of total aggerate income) (Index)
Income is highly concentrated in high-income The pandemic could widen income inequality within
households, particularly in poor countries. countries.
Sources: World Bank PovcalNet database; and IMF staff Sources: World Bank PovcalNet database; and IMF staff
estimates. estimates.
Note: Countries with a high poverty rate refer to those with rates Note: Pre-COVID-19 income inequality is based on 2018 data
higher than 50 percent of the population, while countries with a or the latest year available.
low poverty rate have rates less than 10 percent of the poverty
rate in 2015. Poverty lines vary according to country income
group as in Annex Figure 1.1.3.
The estimates suggest that median income inequality within countries would rise from 0.38 to 0.41 on
the Gini index (Annex Figure 1.1.7). Estimating the impact on income inequality starts with taking the
pre-COVID-19 income distribution across countries and then estimating the pandemic impact on the
income share of each income decile based on the changes in household income distribution in past
recessions across countries. The estimates are subject to limitations in part because governments’ social
spending in response to the health crisis could mitigate the rise of income inequality. Overall, the
magnitude of increases is comparable to those obtained through analysis using household surveys and
indices on the ability to tele-work in selected European countries (Palomino, Rodriguez, and Sebastian
2020). Other studies have used selected household surveys (Bottan, Hoffmann, and Vera-Cossio 2020)
and previous epidemic episodes (Furceri and others 2020) but applied them to a narrow set of countries.
Many countries have appropriately provided emergency lifelines to vulnerable households, which
supporting directly to the poor besides cushioning the magnitude of growth slowdown. Globally, over 1.7
billion people have received additional social assistance transfers in some form or another. Transfers for

recipients almost doubled on average from precrisis levels, representing one-third of monthly GDP per
capita (Gentilini and others 2020a). On average, countries have spent close to an additional 1 percent of
GDP on social protection and labor market measures in response to the pandemic, largely on social
assistance programs, although the increase was modest at about 0.4 percent of GDP in low-income
developing countries (Online Annex Figure 1.1.8).
Online Annex Figure 1.1.8. Additional Social Online Annex Figure 1.1.9. The Impact of
Protection and Job Measures in Response COVID-19-Related Social Assistance
to the COVID-19 Pandemic Measures on Global Extreme Poverty
(Percent of GDP) Rates
Countries on average have spent 0.9 percent of GDP on (Percent)
social protection and labor market measures, most of Social safety net lifelines have mitigated the rise in
which through social assistance programs. global extreme poverty.
Sources: Gentilini and others (2020b); and IMF staff estimates. Sources: World Bank PovcalNet database; IMF, October
Note: Weighted average by US dollar purchasing power parity 2020 World Economic Outlook; Gentilini and others (2020b);
GDP across countries with additional measures. and IMF staff estimates.
Additional social spending has helped mitigate partly the potential rise in global poverty in the face of
the pandemic. Using World Bank ASPIRE data that simulate the sensitivity of spending on social
assistance on poverty measures and Gini coefficients, Evans and Matsumoto (2020) show that 1
percentage point of GDP spending on social assistance would reduce the poverty headcount by around 6
percent on average. 4 As countries spend an additional 0.8 percent of GDP on social assistance on average
during the pandemic, this would imply an average 5 percent reduction in the poverty headcount. 5
Applying this to the global poverty estimates illustrated above and using country-level data on COVID-
19-related social spending in Gentilini and others (2020b), the estimates in this annex suggest that
COVID-19-related social assistance programs have mitigated at most one-fifth of the projected rise in
global extreme poverty rates. This implies that around 80-90 million people could fall into extreme
poverty after accounting for the direct support of social assistance programs (Online Annex Figure 1.1.9)
4A detailed methodology can be found in Yemtsov and others (2018).

5 Assumes the impact of social assistance on poverty reduction is the same across people who recently fall into poverty and those previously
living in poverty. Some evidence suggests that the people who recently fall into the pandemic would be more difficult to reach or to be eligible
for current social assistance program than people already under poverty according to the World Bank. This would imply our estimates could have
overestimated the impact of COVID-19 related social assistance spending. On the other hand, most governments rolled out a range of measures,
including progressive tax and benefit measures, which are not accounted in this estimate given it only considers social assistance program.

under the baseline. 6, 7 If we consider the 100-110 million people range above, the rise in extreme poverty
will likely be above 80-90 million people. The estimates, however, are subject to a large degree of
uncertainty depending on pandemic developments, designated poverty thresholds, the growth outlook,
and fiscal policy responses. A full decomposition of fiscal policy impact is difficult given lifelines have
partly cushioned the growth slowdown besides providing direct support to the poor.
Policies to Tackle Rising Poverty and Inequality
Tackling rising inequality and poverty has become more urgent, which requires safeguarding essential
social spending and building a strong, resilient social protection system against future shocks and
epidemics. Specific steps include the following:
• Phase out temporary lifelines cautiously, but safeguard essential social spending. The pace of phasing out
lifelines should depend on how the pandemic develops (reduction of health risks), the strength of
the recovery, and the fiscal space. Governments should strike a balance between work incentives
and adequate income support to contain poverty and mitigate inequality. Redesigned lifeline
measures can become part of stronger social protection systems, which will need to be embedded
in governments’ medium-term fiscal strategies and require sustainable financing.
• Countries with fiscal space should expand social protection and labor market programs to
improve their adequacy and coverage (Online Annex Figure 1.1.10). In economies where the
pandemic and the lockdowns weigh heavily on informal workers, governments should invest in
identification and delivery systems that will allow for reaching out to many workers and
households currently not covered by social programs. This can be complemented by supporting
local community organizations that step up timely delivery of food, medicine, and other essential
supplies to households identified as being in need.
• In countries with limited fiscal space, it will be necessary to generate additional resources to scale
up social spending. Measures could include removing inefficient and regressive subsidies and
increasing broad-based taxes. In the wake of growing social protests, solidarity on fiscal policies
becomes more important for all countries.
• Build stronger social protection systems. The crisis has exposed structural gaps in social protection
systems.8 Stronger systems will strengthen resilience against future shocks. The crisis may provide
an opportunity to strengthen these systems by both expanding coverage and increasing benefits
(Online Annex Figure 1.1.11). Priorities would be to strengthen the systems mainly along two
dimensions: (i) “reachability” to effectively disburse social protection benefits (cash or in-kind) in
a timely, secured, and adequate manner to ensure that eligible households receive the benefits to
6 Based on the projected per-capita income and changes in income inequality, the projected rise of global extreme poverty as shown before would
be expected to rise by 108 million, or around 100-110 million people as illustrated before. The COVID-19 related social assistance programs
would be expected to mitigate by at most one-fifth of the potential rise (the initial projection already reflects impact of social assistance programs
on GDP per capita).
7 Other micro-data studies point to significant contributions of COVID-19 lifelines to mitigating poverty. For example, in the United States,
federal aid through cash transfers and pandemic-related unemployment benefits have contained the rise in poverty rates across regions and
demographic groups, preventing 12 million people from falling into poverty (Han, Meyer, and Sullivan 2020; Parolin, Curran and Wimer 2020).
8 Existing gaps in social protection systems have led countries to innovate in different ways to reach vulnerable populations, leveraging existing
delivery infrastructure and finding alternatives, such as digital cashless transfers. A new cashless transfer program in Togo called Novissi has
allowed subnational governments to make transfers to targeted adult workers in the informal economy, such as taxi drivers, affected by the
lockdowns. Beneficiaries are identified through their voter IDs, which have much broader and reliable coverage than personal ID cards. Ecuador
doubled the number of cash agents in two weeks. Malaysia expanded free mobile Internet access, while Nigeria identified vulnerable informal
workers in urban areas by collaborating with mobile network operators under airtime purchase patterns. Emergency innovations enable various
social protection programs to reach vulnerable individuals, and they should be reviewed and strengthened over the medium term.

which they are entitled; and (ii) “scalability,” so that the social protection systems can be easily
expanded (either by increasing generosity or coverage) to respond to adverse shocks and mitigate
income losses.
Online Annex Figure 1.1.10. Adequacy Online Annex Figure 1.1.11. Social Assistance
and Coverage of Social Protection Programs and the Reduction of the Poverty
Programs across Regions Headcount and Inequality
(Percent, left scale; percent of GDP, right (Percent)
scale) High social safety net spending has tended to contribute to
Social protection programs have low coverage in reducing poverty and inequality across countries.
low-income developing countries and provide
insufficient benefits in emerging market developing
economies.
Sources: World Bank, ASPIRE and PovcalNet databases; Sources: World Bank, ASPIRE and PovcalNet databases; and IMF staff
and IMF staff estimates. estimates.
Note: Adequacy of benefits is the total transfers received Note: The reductions in poverty headcounts and inequality (Gini) are
by beneficiaries as a share of the pre-transfer total income driven by social assistance and likely other social insurance and labor
in the lowest income quintile of individuals. Coverage is market programs.
the share of the lowest quintile individuals who receive
social protection benefits. EM = emerging markets; EME =
emerging market economies; LIDC = low-income
developing countries; LAC = Latin America and the
Caribbean; MENAP = Middle East, North Africa,
Afghanistan, and Pakistan; SSA = sub-Saharan Africa.
In terms of reachability, strengthening the capacity to reach, target, and deliver timely benefits would
require (i) a comprehensive identification system to ensure inclusion and avoid fraud; (ii) integrated
socioeconomic data that allow for effective
Online Annex Figure 1.1.12. Main Areas to
targeting; and (iii) a delivery infrastructure to
Improve the Reach of Social Protection
disburse benefits, particularly for people in more Systems
remote and deprived areas (Online Annex Figure
1.1.12). Governments can collect, maintain, and ID system
integrate information that enables automatic Universal
Reliable
enrollment of large segments of the population
that need support when adverse shocks occur,
Socio- Delivery
including through greater use of digital solutions economic
data
infrastructure
Inclusive
to simplify application and raise the take-up of Integrated Safe
Up-to-date Transparent
social protection programs (Namibia, Pakistan,
Saudi Arabia, Togo, United States). Greater use of
Source: Prepared by IMF staff.
digital means should be accompanied by
safeguards to prevent fraud and ensure data privacy to mitigate risks.

Social protection systems can respond better to adverse shocks if they (1) have built-in triggers that
adapt the social programs to adverse shocks and deliver timely benefits to vulnerable groups (for
example, free school-meal programs can be transformed into in-house food distribution or cash transfers
for vulnerable families when school is closed under lockdowns); (2) include benefits that are
automatically conditioned to the shocks (for example, top-up of unemployment benefits in the United
States or registering ineligible applicants in social programs (Colombia); and (3) are aligned with medium-
term development needs.
At the same time, social protection systems need to limit exclusion errors to cover eligible low-income
individuals, while encouraging productive behavior. This can be done by combining broad programs
(such as child allowances or social pensions) and more complex, targeted programs (such as conditional
cash transfers) to encourage school attendance and health check-ups—a form of progressive
universalism. This can also be done by setting transparent and simple eligibility criteria for those who
cannot participate (that is, targeting out). For example, Namibia has put in place a transfer for all adult
informal workers and the unemployed that explicitly excludes formal workers and recipients of existing
social benefits.
Countries should strengthen the systems building on the lifelines extended during the COVID-19
pandemic. In some cases, this would imply refining those programs that were rolled out as “quick fixes,”
as speed was crucial to save lives and protect people’s livelihoods. Priorities are to consolidate those
inefficient, fragmented programs with duplication of benefits or high administrative costs that existed
before the crisis or were introduced in response to the pandemic. Certain provisions to expand social
protection programs can be made more permanent. In other situations, new programs may need to be
consolidated with pre-COVID-19 ones in order to align benefit levels and eligibility criteria (the Philippines,
Brazil) while preserving work incentives.

Online Annex 1.2. Smart Strategies to Contain the COVID-19

Pandemic 1
As countries reopen their economies and ease mobility restrictions, it is important to understand what
approaches work best to contain the health and economic impact of COVID-19, not only to limit new
waves of infections, but also to better prepare for future pandemics. This annex assesses the effectiveness
of various containment measures in limiting the number of fatalities, as well as the expected output and
fiscal costs associated with these measures. 2 The findings suggest that stringent containment measures
such as mobility restrictions and public health policies implemented early on are associated with better
health and expected economic outcomes. Experience from countries that successfully curbed COVID-19
at lower economic costs also suggests that a “smart” containment strategy should involve data-driven,
targeted lockdowns and quarantines, underpinned by large-scale testing, contact tracing, and public
information campaigns that promote voluntary compliance and trust.
H ealth Impact of Stringent Containment Measures
Several studies have found that containment measures have been, on average, effective in “flattening
the pandemic curve,” that is, in slowing the spread of the virus, especially when those measure are
implemented early and result in less mobility (Deb and others 2002b). Complementing these studies, this
section analyzes the effectiveness of various containment measures in limiting the number of fatalities,
accounting not only for the timing of these measures but also for their stringency and duration.
Specifically, the annex compares COVID-19 deaths in countries that implemented stringent containment
measures during the course of January through May on average versus countries that implemented such
measures early on—when the country reached 100 COVID-19 cases. The analysis controls for country
specificities, such as median age, population density, and health system preparedness (proxied by hospital
beds per 1,000 persons), as well as GDP per capita. Containment measures include internal and
international mobility restrictions, such as curfews, bans on travel, public events, or gatherings, and
closures of schools, workplaces, and transport, as well as public health policies such as large-scale testing,
contact tracing, and public information campaigns.
The evidence in Online Annex Table 1.2.1 suggests that stringent containment measures implemented
early on are associated with significantly lower fatalities. Since many containment measures are often
implemented at the same time, it is difficult to disentangle their effects. Nevertheless, assessing different
containment measures individually shows that most are associated with lower COVID-19 death rates
(Online Annex Table 1.2.2). When controlling for the overall stringency of these measures, however,
signs reverse for those measures that are highly correlated, illustrating the challenge in identifying the
relative effectiveness of different measures. Notably, in the case of international travel controls, where
the timing was often different from domestic measures, as well as cancellation of public events, the
evidence seems to be strongly in favor of these measures reducing fatalities
Country success stories in containing the virus have largely stemmed from acting early (including by
monitoring international travel), implementing large-scale testing, and contact tracing. Notably,
governments with experience in containing SARS coronavirus outbreaks (Hong Kong SAR, Singapore,
Taiwan Province of China, and Vietnam) acted fast and “smart” and were able to successfully contain
COVID-19, with a lower death toll and lower expected output and fiscal costs (see next section). An early
and smart containment strategy, based mainly on mass testing, contract tracing, and public information
campaigns allowed these governments to successfully curb the spread of the virus. In contrast with
1 Prepared by Alexandra Fotiou and Andresa Lagerborg, based on Fotiou and Lagerborg (forthcoming-a).
2 See Chapter 2 of the October 2020 World Economic Outlook for a discussion on the economic consequences of containment measures.

others, successful governments (1) quickly imposed international travel restrictions, closed schools, and
cancelled public events; (2) proactively implemented stronger health policies such as testing, contact
tracing, and public information campaigns; and (3) only later implemented stay-at-home orders, closures
of workplaces and transport, and restrictions on gatherings and internal travel. Because of past pandemic
experience, public information campaigning seems to have been enough for people to act cautiously.3
Nepal and Korea are also success stories, adopting response plans similar to the rest of the Asian region. 4
Interestingly, Korea was one of the countries highly affected by MERS outbreaks in 2015. A set of smaller
countries, including Cyprus, Georgia, Greece, Malta, and the Slovak Republic, were also effective in containing
the virus by acting fast. Learning from other countries’ experiences, and given their concerns with limited
healthcare capacity, these countries imposed international travel restrictions and bans on public event
early on.
The lack of experience from other epidemics is associated with less preparedness not only by
governments, but also by citizens. In countries with no experience, governments had to force people to
stay at home and restrict their mobility, imposing stricter stay-at-home orders and bans on gatherings .
Evidence from the Ebola experience shows that supporting the health system in a smart way can make a
difference. The impact of the Ebola outbreak was profound in both its human toll and severe
socioeconomic effects that included job and education losses. Health systems were severely compromised
by overwhelming demand, healthcare worker deaths, resource diversion, and closure of health facilities.
These, together with the fear of getting infected, lowered trust in health systems, with large reductions in
health care utilization. Since the Ebola experience, studies have suggested that the best strategy to
successfully control an outbreak involves early, aggressive, and supportive healthcare, including contact
tracing, preventive initiatives, active surveillance, effective isolation and quarantine procedures, and timely
response to patients (Kalra and others 2014; Wojda and others 2015). These measures, combined with
public health education, point-of-care diagnostics, a vaccine, and coordinated efforts from the
international community, made a big difference in the fight against Ebola.
Economic Impact of Stringent Containment Measures
Stringent containment measures can have potentially large macroeconomic costs. 5 This section assesses
the expected output and fiscal costs of containment measures. The analysis is based on the revision of
GDP and primary balance projections for 2020 between the vintages of the IMF’s World Economic
Outlook database in October 2019 and July 2020. 6 The regressions suggest that countries that have more
successfully contained COVID-19 (specifically, certain countries in Asia) may also experience lower
output losses and milder deteriorations in fiscal balances on average (Online Annex Tables 1.2.3 and
1.2.4). On the other hand, advanced economies and countries with larger fiscal stimulus packages also
saw larger downward revisions in GDP growth and fiscal balances. 7
3Chen and others (forthcoming) provide supportive evidence on how the voluntary decision to social distance and take precautions matters
compared with de jure non-pharmaceutical interventions.
4 A country is classified as successful if the number of deaths per 1 million population is below 20, and if the number of tests per 1 million
population is above 10,000.

5 See Deb and others (2020a), Coibion, Gorodnichenko, and Weber (2020), and Gourinchas (2020).
6See Fotiou and Lagerborg (forthcoming-a) for more details regarding the analysis, for robustness checks when looking into January 2020 versus
July 2020 projections , and for effects on the debt-to-GDP ratio.
7 No relation is observed between stronger COVID-19 containment measures, on average, and the size of fiscal packages (Fotiou and Lagerborg,
forthoming-b). The evidence suggests, however, that countries that acted swiftly in putting in place strong COVID-19 containment measures
ultimately deployed smaller fiscal packages.

Countries with stricter containment measures, on average, saw larger downward revisions to growth
and primary balance projections for 2020 between October 2019 and July 2020. 8 However, countries that
put stringent containment measures in place earlier (rather than later and for longer) saw smaller
downward revisions (Online Annex Figure 1.2.1). This applies mainly to mobility restrictions (such as
workplace closures and stay-at-home orders), while stronger public health policies are associated with
better expected outcomes for GDP growth and fiscal balances regardless of these policies being
implemented earlier or later (Online Annex Figure 1.2.2). Large downward revisions to GDP growth in
2020 will be only partially offset by a stronger expected rebound in 2021 for countries with stricter
average containment measures. In the medium term, the implications of costly containment measures are
expected to be long-lasting particularly in the case of public debt (Online Annex Figure 1.2.3, panel 1).
Online Annex Figure 1.2.1. Stringency of Containment Measures and World
Economic Outlook Database Revisions to GDP Growth and the Primary Balance,
2020
(Percentage points)
1. GDP Growth Revision versus Average Stringency 2. GDP Growth Revision versus Early Stringency
3. Primary Balance Revision versus Average Stringency 4. Primary Balance Revision versus Early Stringency
Source: Fotiou and Lagerborg (forthcoming-a).

Note: WEO = IMF, World Economic Outlook database.
Online Annex Figure 1.2.4 shows that the governments that successfully contained the virus, but
without past SARS experience, had a larger downward revision of GDP growth and primary balances
(Cyprus, Georgia, Greece, Malta, New Zealand) than governments in the Asian region with past SARS
experience (Hong Kong SAR, Korea, Nepal, Singapore, Taiwan Province of China, Vietnam). 9 The former were
governments that acted early but implemented stricter stay-at-home orders and on average announced
larger fiscal stimulus packages. Governments with less experience in containing epidemics, which usually
were less prepared to implement effective testing policies and contact tracing, in many cases resorted to
8 This is in line with evidence on Sweden’s containment strategy from Bricco, Misch, and Solovyeva (forthcoming), who find that a less-stringent
strategy can soften economic effects temporarily but have unclear medium-term effects.
9Large revisions of the non-SARS sample could also correspond to other drivers, such as a sudden stop in tourism, which is a key economic
sector for Cyprus and Greece, for example.

stronger mobility restrictions over a longer period as a way to save lives. These measures have come with
substantial economic and fiscal costs. Debt-to-GDP ratios for these countries were drastically revised in
July 2020 compared with October 2019. 10 Large revisions correspond to the high costs of virus
containment measures and fiscal stimulus for economic revival after strict lockdown measures are eased.
Online Annex Figure 1.2.2. Stringency of Individual Containment Measures and
World Economic Outlook Database Revisions to GDP Growth, 2020
(Percentage points)
1. Correlation of Average Containment Measures 2. Correlation of Early Containment Measures with
with GDP Growth Revisions GDP Growth Revisions
Sources: Fotiou and Lagerborg (forthcoming-a); OxCGRT Database; IMF staff calculations; and IMF World Economic
Outlook (WEO) database.
Note: The correlations refer to estimated coefficients when regressing World Economic Outlook database projection
revisions on individual containment measures normalized on a [0,1] scale, one at a time, controlling for GDP per capita
(purchasing power parity in US dollars). Health policy corresponds to the principal component of tracing, testing, and public
information campaigns. These panels are produced from a standard ordinary least squares analysis in which the regression
includes a constant variable. The constant is negative and significant, implying on average a downward revision of GDP
growth for all countries. In the panels, a negative correlation means that the expected average downward revision is larger
compared with the October 2019 World Economic Outlook database forecasts. If the correlation is instead positive, it means
that the downward revision is smaller. *** p < 0.01, ** p < 0.05, * p < 0.10.
10 Results are robust to considering projection revisions between January and July 2020.

Online Annex Figure 1.2.3. Stringency of Containment Measures and World Economic
Outlook Database Revisions to Medium-Term Forecasts, 2020—24
(Percentage points)
1. Effect of Average and Early Mobility Restrictions on GDP Growth, Primary Balance, and Public Debt Forecasts
2. Effect of Average and Early Public Health Policies on GDP Growth, Primary Balance, and Public Debt
Forecasts

Note: p.p. = percentage point; WEO = IMF, World Economic Outlook database.
Online Annex Figure 1.2.4. Average World Economic Outlook Database Forecast
Revisions for Different Country Groups, 2020—24
(Percentage points)
1. SARS-Experienced Countries: Average Forecast Revision, 2. Other Successful Countries: Average Forecast Revision,
July 2020 – October 2019 July 2020 – October 2019

Online Annex Table 1.2.1. Effect of Containment Measures on Death Rates: Average
versus Early Stringency
(1) (2) (3) (4)
Deaths/Population Deaths/Population Deaths/Population Deaths/Population
Stringency (average) -0.991

(1.11)
Stringency and health policies (average) -0.993
(1.069)
Stringency (at 100 cases) -1.205***
(0.416)
Stringency and health policies (at 100 cases) -1.268**
(0.49)
Median age 7.612*** 7.582*** 6.603*** 6.572***
(1.838) (1.843) (1.925) (1.897)
Hospital beds / 1,000 population -13.574** -13.798** -12.221** -12.425**
(5.468) (5.452) (5.599) (5.539)
GDP per capita (USD PPP) 0.001 0.001 0 0
(0.001) (0.001) (0.001) (0.001)
Constant -127.796** -125.640** -54.552 -52.056
(52.382) (53.318) (62.647) (59.459)
Observations 135 136 123 124

R-squared 0.234 0.231 0.267 0.274
Sources: Fotiou and Lagerborg (forthcoming-a); and IMF staff estimates.
Note: Standard errors in parentheses. USD PPP = purchasing power parity in US dollars. *** p < 0.01, ** p < 0.05, * p <
0.10.
Online Annex Table 1.2.2. Effect of Individual Containment Measures (One at a

Time) on Death Rates: Average versus Early Stringency
Average Response Early Response

Deaths/Pop. Deaths/Pop. Deaths/Pop. Deaths/Pop. Stringency
(1) (2) (3) (4) Correlation
Containment measures:
School closures -29.096 -20.519 -27.892*** -17.43 0.67
Workplace closures 26.051 76.028** -15.282* 8.788 0.71
Cancellations of public events -18.069 -19.96 -54.756*** -46.806** 0.69
Restrictions on gatherings 12.612 35.289 -7.34 12.169 0.66
Public transport closures -39.421 -38.819 -17.088 13.701 0.63
Stay-at-home requirements 5.176 38.623 -8.877 26.633* 0.72
Restrictions on internal movement 6.975 43.553 -8.579 37.718** 0.62
International travel controls -46.014*** -56.767*** -31.482*** -27.360** 0.40
Public information campaigns -14.863 -0.888 -39.475* -18.043 0.55
Testing policy -26.57 -23.436 -10.699 -1.952 0.38
Contact tracing -26.493 -23.813 12.484 15.956 0.28
Controls:
Overall stringency No Yes No Yes
Median age Yes Yes Yes Yes
Hospital beds / 1,000 population Yes Yes Yes Yes
GDP per capita (USD PPP) Yes Yes Yes Yes
Constant Yes Yes Yes Yes

Note: Standard errors in parentheses. USD PPP = purchasing power parity in US dollars. *** p < 0.01, ** p < 0.05, * p <
0.10.

Online Annex Table 1.2.3. Effect of COVID-19 Containment Measures on World Economic Outlook
Database Revisions to GDP Growth, 2020
(1) (2) (3) (4) (5) (6) (7) (8)
GDP GDP GDP GDP GDP GDP GDP GDP
Growth Growth Growth Growth Growth Growth Growth Growth
Stringency (average) -0.067** -0.104*** -0.089*** -0.088***

(0.03) (0.026) (0.025) (0.026)
Health policies (average) 0.059 0.578*** 0.321* 0.350*
(0.188) (0.174) (0.185) (0.186)
Stringency (at 100 cases) 0.026*** -0.009 -0.007 -0.008
(0.008) (0.009) (0.008) (0.008)
Early health policies 0.186 0.501*** 0.343** 0.368**
(0.186) (0.165) (0.163) (0.16)
GDP per capita (USD PPP) -0.007 -0.005 -0.007 -0.007 -0.008 -0.01
(0.012) (0.012) (0.012) (0.012) (0.011) (0.011)
Advanced economy = 1 -3.581*** -3.709*** -2.842*** -3.575*** -3.770*** -2.797***
(0.607) (0.59) (0.696) (0.596) (0.571) (0.648)
SARS experience = 1 3.195** 3.643*** 3.258*** 3.228***
(1.237) (1.363) (1.099) (1.215)
SARS-mimicking Asia 3.993** 3.757** 3.852** 3.650**
(1.635) (1.624) (1.484) (1.449)
COVID-19 fiscal measures -0.084** -0.101***
(0.037) (0.034)
Constant -4.232*** -1.884* -2.579** -2.362** -8.335*** -5.160*** -5.370*** -4.963***
(1.16) (1.032) (1.023) (1.043) (0.544) (0.701) (0.671) (0.668)
Observations 153 152 152 149 134 133 133 131

R-squared 0.04 0.331 0.379 0.405 0.091 0.363 0.43 0.468
Note: Standard errors in parentheses. Negative coefficients imply a stronger downward revision compared with the October 2019 World Economic
Outlook database projections. USD PPP = purchasing power parity in US dollars. *** p < 0.01, ** p < 0.05, * p < 0.10.
Online Annex Table 1.2.4. Effect of COVID-19 Containment Measures on World Economic Outlook
Database Revisions to Primary-Balance-to-GDP Ratio, 2020
(1) (2) (3) (4) (5) (6) (7) (8)
Primary Primary Primary Primary Primary Primary Primary Primary
Balance Balance Balance Balance Balance Balance Balance Balance
Stringency (average) -0.07 -0.093** -0.088** -0.082**

(0.043) (0.039) (0.039) (0.038)
Health policies (average) -0.456 0.119 -0.036 0.021
(0.28) (0.266) (0.28) (0.275)
Stringency (at 100 cases) 0.042*** 0.004 0.004 0.004
(0.012) (0.014) (0.014) (0.013)
Early health policies -0.457* -0.123 -0.201 -0.187
(0.273) (0.259) (0.265) (0.263)
GDP per capita (USD PPP) -0.065*** -0.060*** -0.063*** -0.064*** -0.062*** -0.063***
(0.018) (0.018) (0.018) (0.019) (0.019) (0.019)
Advanced economy = 1 -1.649* -1.784** -0.869 -1.399 -1.558* -0.657
(0.881) (0.881) (1.023) (0.925) (0.927) (1.067)
SARS experience = 1 1.831 1.623 1.165 1.1
(2.423) (2.372) (2.4) (2.383)
SARS-mimicking Asia 4.094* 3.799 3.85 3.663
(2.429) (2.38) (2.398) (2.384)
COVID-19 fiscal measures -0.085 -0.093*
(0.054) (0.056)
Constant -1.575 0.949 0.629 0.653 -7.090*** -3.033*** -3.130*** -2.766**
(1.698) (1.567) (1.571) (1.537) (0.798) (1.089) (1.091) (1.105)
Observations 147 147 147 146 129 129 129 129

R-squared 0.078 0.279 0.295 0.31 0.091 0.245 0.262 0.279

Note: Standard errors in parentheses. Negative coefficients imply a stronger downward revision compared with the October 2019 World Economic
Outlook database projections. USD PPP = purchasing power parity in US dollars. *** p < 0.01, ** p < 0.05, * p < 0.10.

Online Annex 1.3. From Lockdown to Recovery: Spending Measures

to Support Livelihoods during the COVID-19 Crisis 1
Many countries have swiftly introduced a diverse range of measures to support incomes for households
and protect jobs as unemployment rises and incomes fall. As economies start to reopen, the design of
spending measures needs to be geared more toward facilitating recovery and building resilience against
future shocks, including the possibility of renewed infection waves and lockdown restrictions.
General Considerations
• Countries should have clear objectives to inform policy responses. The primary objective during the
lockdown phase is to provide lifelines to maintain employment and business links.2 Priorities should
focus on a speedy response, while economic incentives such as work disincentives are less of a
concern. However, as economies reopen, the design of the measures will need to be refined to ensure
appropriate incentives.
• Countries should prioritize measures that are consistent with their medium-term development needs.
In countries where existing social protection systems are underdeveloped, the crisis provides an
opportunity to strengthen these systems (see Online Annex 1.1).
• Targeting policy responses to specific households and firms involves tradeoffs that need to be
carefully managed. Targeted measures can provide better protection for a given spending envelope or
help contain fiscal costs. They can contribute to supporting aggregate demand, since the most
vulnerable are often those who have a higher propensity to consume. However, targeted measures
can inadvertently exclude some who are in need of support, and they require more time to design and
implement. Targeting may distort work incentives by increasing the implicit marginal tax rate when
benefits are withdrawn as earnings rise.
• Policy measures should be closely aligned with existing infrastructure to accelerate deployment.
Governments in advanced economies can take advantage of well-developed tax and benefits systems
and governments in emerging market developing economies can take advantage of existing social
program structures. Mobile or digital payment systems may provide a broad and timely mechanism
for delivering the support in some countries.
Fiscal Support during the Lockdown
The unprecedented fiscal policy responses to the pandemic have mostly been on the expenditure side.
Country experiences have provided various lessons (Shang, Evans, and An 2020):
• Countries have relied on a broad range of measures. The designs vary across countries, reflecting the
severity of the pandemic, the availability of fiscal space, and the level of administrative capacity.
Variations include the following: (1) the overall spending responses are much larger in advanced
economies; (2) liquidity support accounts for a relatively larger share in advanced economies; and (3)
advanced economies and emerging and developing Europe rely more on social insurance and labor
market measures, while sub-Saharan Africa relies more on social assistance measures (Online Annex
Figure 1.3.1).
1 Prepared by Baoping Shang, Brooks Evan, Delphine Prady, and Zhiyong An.
2As defined in Chapter 1, the three phases of the pandemic are broadly classified into (1) the Great Lockdown; (2) partial and gradual reopening;
and (3) pandemic under full control with effective vaccines and treatment.

Annex Figure 1.3.1. Spending Measures to Online Annex Figure 1.3.2. Average
Support Workers and Households Number of Measures to Support
(Number of measures as a percent share of total) Workers and Households
(Number of measures)
Sources: Gentilini and others (2020b); and IMF staff estimates.
• Short-time work programs or job retention schemes are often used to protect jobs during
recessions. 3 They are common in Europe and are being used in other emerging market economies
during the COVID-19 crisis (Brazil, Egypt, Uruguay). These programs covered from one-fourth to
one-third of private sector employees in several European countries as of mid-2020. Evidence points
to large positive effects of such programs on reducing layoffs of permanent workers, and the
programs can be cost-effective. For example, one-fifth of jobs are saved for every worker on short-
term work, while the cost per saved job is only 7 percent of the average labor cost, which outweighs
the fiscal cost when workers lose their jobs (25 percent of the average labor cost). 4
• Providing support to people in the informal sector has been a challenge, as these workers are not
covered by unemployment benefits and they are difficult to reach. Some governments have taken
innovative approaches and channeled support to informal sector firms by working with existing
institutions that serve these groups, such as micro-credit institutions and informal sector
organizations. However, in many emerging market developing economies, given the large informal
sector and limited fiscal space, these workers can be more effectively covered by social assistance
programs.
• Capacity constraints have made it difficult to rapidly expand existing social assistance programs in
some countries. In such cases, countries often resort to alternative (less effective) ways of targeting,
including cash transfers targeted at specific regions or population groups (that is, the elderly, families
with children, or informal sector workers, as in India and Bolivia), or subsidies for key goods and
services such as food, health, transportation, and utilities (Jordan). Some countries identify
beneficiaries by using databases maintained by various government entities and private organizations
or by distributing benefits through local governments and community organizations (Rwanda, Nepal,
Egypt, Peru). Universal (or near-universal) transfer programs have only been used as one-offs because
3These initiatives compensate for the reduced working hours while maintaining employment links: employees receive a wage subsidy from the
government proportional to their lost working hours. This allows firms hoarding labor to keep specific firm human capital and avoid costly
separation and then rehiring and training as recovery takes place. For workers, it avoids search and layoffs.
4 See Cahuc, Kramarz, and Nevouss (2018), Guipponi and Landais (2018), and Kopp and Siegenthaler (2019).

of the difficulties of ensuring adequate support for the most vulnerable within a reasonable fiscal cost
(Japan, Singapore, Korea, Malaysia, Bolivia) (Prady 2020).
• Rapid expansion of existing Online Annex Figure 1.3.3. Country Fiscal Measures in
and new programs may have Selected Areas
resulted in duplication of (Percent of total)
benefits and high
administrative costs. Many
countries have utilized
multiple cash transfer and in-
kind benefit programs for
income support (Online
Annex Figure 1.3.2) and
support for workers (Online
Annex Figure 1.3.3). For
example, in Indonesia, in
addition to expanding the
existing social assistance Sources: OECD (2020); and IMF, Fiscal Monitor database of Fiscal Policy
program, the food assistance Measures in Response to COVID-19.
program, and the pre-work Note: Results are based on a sample of 69 countries, including all Organization for Economic Co-
operation and Development economies, G20 economies, and selected economies in the Fiscal
card program, central and Monitor database.
local governments have also
introduced two new cash transfer programs, electricity fee reductions, public works, and a number of
programs targeted at the local levels. Ensuring that eligibility rules are clearly communicated and
aligned with programs that were functioning well before the COVID-19 crisis would reduce
administrative costs and help enhance effectiveness. 5
Improving Economic Incentives as Economies Reopen
As activity resumes, the focus switches to redesigning measures to better support the return to work.
Measures such as waiving the need for job search, training, and other requirements for unemployment
benefits (Austria, Canada, United States), increasing benefit levels (Australia, Belgium, United States), and
extending the duration of benefit eligibility (Greece, Japan, United States) can be changed to promote greater
job search efforts. For example, at the outbreak of the pandemic in the United States, many unemployed
were eligible for benefits larger than their lost earnings (Xie 2020), 6 while in Canada, the application for
unemployment benefits were substantially simplified with a flat benefit of CAD$2,000 per month,
resulting in high replacement rates for low-wage earners.
Support measures need to gradually move away from protecting current jobs to protecting workers. For
example, keeping short-time work schemes in place for too long could prevent the reallocation of labor
from low-productive firms to growing ones as the recovery gains steam, and eventually lead to
inefficiency and output losses. Among various options countries can:
5As the COVID-19 crisis unfolded, the government in Brazil first leveraged its conditional cash transfer, Bolsa Familia, to include 1.2 million
new families. It also introduced a temporary program called Auxilio Emergencial targeted mainly at informal and own-account workers that
provides a three-month transfer more generous than the Bolsa Familia benefit (later extended by two months). Ninety-five percent of Bolsa
Familia beneficiaries opted to temporarily enroll in the more generous Auxilio Emergencial.
6 The US Congressional Budget Office estimates that roughly five of every six recipients would receive benefits that exceeded the weekly
amounts they could expect to earn from work during those six months.

• Transition gradually away from job retention schemes. This should be coupled with in-work benefits or
hiring subsidies to help get people back to work. For, example, the United Kingdom unveiled measures
to pay employers a bonus per each reinstated furloughed worker and similar incentives to train and
hire apprentices.
• Gradually reduce the generosity of unemployment benefits to preserve work incentives if their
replacement rates are too high relative to precrisis work incomes (Fang, Nie, and Xie 2020). In the
event that the precrisis social protection benefits were inadequate, making some provisions of
temporary expansion more permanent would help strengthen the social protection systems and build
resilience to future shocks. For example, in Canada, when the government restores capacity to
process unemployment benefits, the flat lump-sum unemployment benefits could be replaced by
formula-based benefits.
• Strengthen activation policies for unemployment benefits. For example, training and job search requirements
that were relaxed during the lockdown phase should be reinstated, and the expansion of benefit
periods rolled back.
• Target active labor market policies (for example, training, employment services and public works) to specific groups.
There may be opportunities for public works to be directed towards “green jobs” such as
reforestation, soil and water conservation, and flood protection.
Policy Responses to Potential Second Waves
The course of the pandemic is still highly uncertain. Governments will need to prepare for potential
renewed waves of infections. Some considerations include the following:
• Policy responses should be targeted at the specific localities where the resurgence occurs helping
limit the adverse impact on economic activity and contain fiscal costs.
• Refining measures to restore economic incentives may need to be postponed. Short-time work
schemes will again be useful, especially for workers in the sectors most affected. The duration of
unemployment benefits can be extended.
• This, however, does not imply that countries should restore the same lifelines used during the initial
phase. The design of the fiscal support should draw on lessons learned and adapt to changing
circumstances. For example, limited budget resources imply that targeted measures are preferred, and
improving the institutional capacity to reach the vulnerable and deliver benefits is even more
important. The focus will be on achieving wide coverage, not just increasing benefit levels. A
stronger social safety net will help build resilience against future shocks.

Online Annex 1.4. Determining the Size of Fiscal Stimulus for

Sustained Recovery
The COVID-19 crisis has generated both a deep downturn and increased debt, intensifying the tension
between supporting growth and debt sustainability. A protracted recession has substantial costs. It can
damage human capital (as the unemployed lose their skills) and physical capital (as firms scale down their
investment plans). This may warrant a debt-financed fiscal stimulus, but elevated debt would increase
risks of future fiscal crisis (Moreno-Badia and others 2020), which are associated with output losses
(Medas and others 2018; Trebesch and Zabel 2017). This annex thus explores the appropriate balance
between stabilization and debt sustainability in determining the fiscal stance. The first section briefly
describes the model used, and the next section presents an illustrative example of the optimal fiscal
stance for a country in the post-COVID recovery.
A Buffer-Stock Model for the Government
The analysis uses the “Buffer-Stock” model to analyze the tradeoffs governments face when pursuing
economic stabilization and ensuring debt sustainability. Specifically, governments maximize utility under a
debt constraint (Fournier 2019). For this purpose, governments choose the fiscal stance, defined as a
change in the structural primary balance, through discretionary policies. Economic output is affected by
exogenous shocks, which can persist for some time. The government can loosen the fiscal stance to
boost output at the cost of eroding its fiscal buffers, or tighten it to build buffers at a cost of lower
output. Adverse recessionary shocks reduce potential output, reflecting human and physical capital losses
(hysteresis effect). The model thus features a permanent negative effect of large negative output gaps on
the level of potential output. Countercyclical fiscal policy is constrained by a one-year decision-making
delay, as annual budgets are voted on before the year starts, as well as by adverse effects of higher debt
such as higher borrowing costs, and greater risks of losing market access. Low debt levels enable the
government to borrow without substantially raising borrowing costs or risk of debt distress.
This Buffer-Stock model shows Online Annex Figure 1.4.1. Output Losses during a
that the choice of fiscal stance Crisis: Temporary and Permanent Effects
would depend on both debt levels
and short-term economic
fluctuations. The utility
maximization framework provides
a normative view on the fiscal Permanent loss
stance. It recommends a higher
fiscal surplus at higher debt levels
to preserve sustainability, and a
countercyclical fiscal stance to Temporary loss
smooth fluctuations. Rising interest
rates add to the motive to reduce
debt. At low debt levels, hysteresis
Pre-crisis trend No counter-cyclical policy
reinforces the motive to counter Best possible trend Counter-cyclical policy
negative shocks.
Sources: Fournier (2019); and IMF staff estimates.
• Importantly, fiscal stimulus
should focus on stabilizing the cyclical downturns (shown in green in Online Annex Figure 1.4.1) to
mitigate the downturn and the subsequent potential permanent costs from hysteresis (the yellow line
above the blue line). However, part of the long-term costs could be induced by a supply shock that
cannot be offset (the difference between the gray line and the yellow line). As such, fiscal policies will

have to adjust to permanent output losses. In practice, governments may not be able to fully
differentiate the cyclical component in real time—this is especially true in cases of large shocks such
as the current pandemic. Policies will need to adjust the size of stimulus to the changing
circumstances and may rely more on automatic stabilizers. Many governments have implemented
rule-based measures that are directly linked to economic outcomes, such as larger unemployment
insurance or support to firms triggered by a significant drop in turnover.
• Highly indebted governments should react less to shocks. The debt buffer (difference between the
current debt level and levels at which sustainability is at risk) has an insurance value—it is the
“reserve” of debt that the government can issue to smooth shocks. When the buffer is small, the
probability of market stress is high. This implies that when debt is high, the optimal countercyclical
fiscal stimulus to offset a negative shock is smaller than when debt is low. Beyond these stylized
cases, the model can be calibrated with country-specific parameters to tailor fiscal stance
recommendations (Fournier and Lieberknecht, 2020; IMF 2020b).
Fiscal Stance Advice for the Post-COVID Recovery
In countries with fiscal space, high unemployment in the wake of the COVID-19 crisis warrants large
and sustained fiscal stimulus to avoid long-term scars from the crisis. In many countries, the benefits of a
stimulus outweigh the cost of increasing public debt. A baseline analysis is applied to a hypothetical
advanced economy, using unweighted average data to calibrate parameters and macroeconomic
conditions. 1 Magnitudes are thus illustrative, showing the reasoning and providing directions. The
negative output gap is assumed to be about 7 percent of potential GDP in 2020, and the public-debt-to-
GDP ratio is close to 80 percent. The interest rate is below the growth rate for a protracted period of
time and goes up to marginally above the growth rate in the long run. In the absence of stimulus, the
phasing out of emergency lifelines would imply a fiscal tightening of almost 3 percent of GDP.
Model-based analysis shows that governments with fiscal space should take advantage of low interest
rates to borrow to finance stimulating the economy. An additional 4 percent of GDP support would not
only offset phasing out of emergency measures but also provide a timely boost to sustain the recovery.
This fiscal expansion would not necessarily increase the debt-to-GDP ratio in 2021 because of higher
projected growth. Given the depth of the health crisis, the fiscal expansion can continue but gradually
wind down so that cumulative stimulus over 2021–24 is about 9 percent of GDP. This sustained effort to
avoid a protracted recession or a double dip from renewed lockdowns would shift the debt level up, as
the boost to growth would eventually fade. The results are highly dependent on the depth of the negative
output gap. For example, a 1 percentage point higher negative output gap would suggest that a
cumulative stimulus should be larger—on the order of 11 percent instead of 9 percent of GDP.
The sustained stimulus is optimal to avoid long-term costs of hysteresis. In an alternative scenario in
which this channel is shut down, the recommended cumulative discretionary fiscal stimulus would be
shorter (three years instead of the five years in the baseline) and smaller (less than 3 percent of GDP),
resulting in a lower medium-term debt-to-GDP ratio (Online Annex Figure 1.4.2.).
However, countries with high debt and interest rates should take a more cautious fiscal stance to
preserve debt sustainability given the risks of debt distress would be higher. In an alternative scenario
with debt at about 140 percent instead of 80 percent of GDP and all else being equal, the model suggests
that the fiscal stimulus for 2021 should only be 2 percent of GDP and that the country should turn to a
fiscal consolidation starting in 2023 (Online Annex Figure 1.4.3). In this high-debt scenario, the priorities
would be to reduce the debt-to-GDP ratio by 4.5 percent of GDP by 2025 to prevent a large surge of
1 The calibration follows Fournier and Lieberknecht (2020).

interest payments and reduce the risks of a future fiscal crisis. At this high debt level, with an interest rate
level 1 percent above the baseline, a progressive consolidation plan should start sooner, in 2022.
Online Annex Figure 1.4.2. Simulated Fiscal Adjustments across Countries with Different
Levels of Debt and Interest Rates
1. Normative Structural Primary Balance 2. General Government Gross Debt
(Percent of potential GDP) (Percent of GDP)
3. Output Gap 4. Output Growth

(Percent of GDP) (Percent change)
Source: IMF staff estimates.

Note: Panel 1 shows a normative fiscal adjustment path with discretionary stimulus in the first few years for an advanced economy
with an average debt level (baseline) at 80 percent of GDP.
The presence of hysteresis does not always warrant an expansionary fiscal stance, particularly in cases of
high, and expensive, debt (Online Annex Figure 1.4.3, panel 2). If the costs from protracted recessions
are large—implying lower potential economic growth—it would make it harder to manage elevated debt
levels and increase the risk of a fiscal crisis. As a result, the more debt is high and costly, the less
governments should stimulate the economy even in the presence of hysteresis. In 2021, hysteresis
warrants about 1 additional GDP percentage point of stimulus—compared to a scenario without
hysteresis—in a country with debt at about 140 percent of GDP and an interest rate 1 percent above the
baseline, against 3 additional GDP percentage points of stimulus to counter hysteresis effects in the
baseline with debt at about 80 percent of GDP. For countries already facing prohibitive interest rates
and high foreign exchange exposure, the scope of further fiscal stimulus would be more limited. The
focus should be on restoring macroeconomic stability (Fournier 2019).

Online Annex Figure 1.4.3. Discretionary Fiscal Support, Hysteresis, and Debt
1. Illustrative Pace of Fiscal Adjustment in the Case 2. Cumulative Change in Structural Primary Balance
of Long-Term Scarring from the Pandemic relative to a No-Policy Change Scenario over 2021–23
(Structural primary balance in percent of potential (Percent of potential GDP)
GDP)
4 Higher debt and higher

Average debt levels Baseline Higher debt interest rate
No policy decision
4
2 High debt
High debt, high interest rates
2
0
0
-2
-2
-4
-4
-6
-6
-8
With hysteresis Without hysteresis
2015 17 19 21 23 25 -8
Sources: Fournier (2019); and IMF staff estimates.

Note: Average debt levels are assumed to be 80 percent of GDP (simple average across advanced economies). In the
simulations, a high debt level is assumed to be about 140 percent of GDP (75th percentile of the weighted distribution among
advanced economies), and high interest rates refers to an addition of 1 percent more than in the baseline.
Countries with a large share of foreign-currency-denominated debt will also be constrained because of
possible effects of a currency depreciation. The value of public debt as a share of GDP could change
significantly during the crisis. This
Online Annex Figure 1.4.4. Discretionary Fiscal Support
concern is common for emerging and Foreign-Currency-Denominated Debt
markets, where fiscal pressures can (Illustrative pace of fiscal adjustment, structural primary balance
be accompanied by pressures in in percent of potential GDP)
the balance of payments and large
currency depreciations. To reflect
this risk, alternative simulations
with a calibration reflecting
emerging markets assume that the
public-debt-to-GDP ratio is
affected by a depreciation when
the government loses market
access, on top of the effect of
economic activity on the debt-to-
GDP ratio. When the change in
value because of currency effects
is moderate and debt is close to
the emerging market average, then
the government should stimulate
the economy (Online Figure Source: IMF staff calculations based on the Buffer-Stock model (Fournier 2019).
1.4.4). If debt is higher or the risk Note: The emerging market baseline reflects an average emerging market country
with initial debt at 64 percent of GDP. Higher debt refers to public debt at the 75th
of depreciation is higher, percentile of emerging markets, 16 percent of GDP above the baseline. Higher
government would have limited foreign exchange exposure is associated with an increase of public debt by 10
scope for additional stimulus. And percent if a fiscal crisis occurs, against 5 percent in the baseline. EM = emerging
market.

if high debt is combined with depreciation risk, governments would need to tighten fiscal policy early to
reduce risks of a future crisis.

Online Annex 1.5. Policy Options to Support the Economic Recovery

This annex discusses the revenue or expenditure measures, or mix of measures, that countries can
adopt to help support the recovery from the COVID-19 pandemic in a cost-effective manner. The focus
is on measures better directed to support the recovery once the pandemic is under control (phase 3) but
some could potentially be applied earlier. 1 In phase 3, fiscal policies should rebalance to facilitate recovery
if fiscal space permits. The main objectives of fiscal policy are to facilitate recovery from a deep
recession, restore debt sustainability, and protect the most vulnerable groups. With reduced or limited
fiscal space, it is important to choose among measures that are cost-effective, that is, that have large
output multipliers. In addition, policies will need to help protect the most vulnerable groups, since the
poor and low-skilled workers have been hit the hardest (Furceri and others 2020; Kikuchi, Kitao, and
Mikoshiba 2020). In developing countries with little fiscal space for stimulus, fiscal policy should strive to
provide essential public services and support the vulnerable. Both of these measures are crucial to
maintaining economic stability (Loayza and Pennings 2020).
Cost-effectiveness of Various Spending and Revenue Measures
To assess the cost-effectiveness of various policy tools, this annex compares multipliers of different
spending and tax measures. 2 A model-based simulation is conducted using a revised model based on
Traum and Yang (2015) to quantify the multipliers of various spending and revenue measures that are
commonly used to facilitate economic recovery. 3 The model features two types of households: those that
are liquidity-constrained and consume all the disposable income each period, representing the poorer
income group; and the higher-income group, which is comprised of asset holders who have both labor
and capital income. 4 As the health and economic crisis has resulted in a drop in the living standard of the
poor with little savings and wealth (Bangham and Leslie 2020), the analysis examines the effects of
measures (spending and revenue) that target liquidity-constrained households. These are compared to
untargeted transfers and tax cuts, which also benefit higher-income households that have accumulated
savings. The simulations are mainly focused on those instruments that can be deployed quickly.
The baseline scenario (without additional fiscal measures) largely captures qualitatively the current
economic and macroeconomic policy conditions of most economies: a severe recession, and monetary
policy constrained at the effective lower bound. 5 For illustrative purposes, the model is calibrated to a
hypothetical economy, and the public-debt-to-GDP ratio before the recession is calibrated to be 83
percent of GDP, which is the weighted average of the world economy in 2019. The policy scenarios
inject a stimulus measure, one at a time, to the baseline scenario. For illustrative purposes, the size of a
1As defined in Chapter 1, the three phases of the pandemic are (1) the outbreak with lockdowns; (2) partial reopening; and (3) a high degree of
control of the virus through medical advances.
2 Multipliers calculated in this simulation are the discounted, cumulative change of output or consumption over a horizon per US dollar change in
government spending or tax revenue over the same horizon.

3 Key revisions relative to the model in Traum and Yang (2015) include (1) modeling fiscal measures to target liquidity-constrained households,
(2) allowing for the binding of the effective lower bound in monetary policy, and (3) delaying fiscal adjustment, as governments are less likely to
address high-debt issues in the early stage of recovery.
4 The model calibration closely follows those used in the Chapter 2 of April 2020 Fiscal Monitor for the global economy. Some calibration
deviations include a stickier nominal wage, as nominal wage rigidity is likely to prevent a large adjustment in the nominal wage in the recovery
stage from the pandemic-led recession. Also, reflecting quick action on interest rate policy in response to the pandemic, the output coefficient in
the policy interest rate is revised upward to drive the economy immediately to the effective lower bound during the first quarter when the
macroeconomic shock hits.
5 The model does not feature a health shock. Instead, it relies on a negative demand-side shock that has a negative effect on consumption, private
investment, and labor. While the pandemic also disrupts the supply side of the economy during the lockdown phase, the weak demand is likely to
dominate the macroeconomic dynamics in the gradual-reopening and recovery phases. The effective lower bound is modeled as the zero lower
bound, as in many advanced economies. In emerging market economies and developing countries, the effective lower bound is typically above
the zero policy rate.

stimulus is set to be about 1 percentage point of GDP for two years and then to gradually unwind as the
economy recovers from a deep recession. To ensure that the public debt increase is limited over time, the
simulations assume that the consumption tax rate adjusts slightly upward in two years after a stimulus is
implemented, but that the adjustment is insufficient to lower the public-debt-to-GDP ratio.
The simulations show that all Online Annex Figure 1.5.1. Cumulative Multipliers of
the measures can help raise Various Fiscal Measures
output, but they differ greatly in 1.60
Short-run output Long-run output
the size of their multipliers Short-run consumption (liquidity constrained) Long-run consumption (liquidity constrained)
(Online Annex Figure 1.5.1). 6

1.20
Government purchases and

transfers to liquidity-constrained
dollars
0.80
households stand out to be the
most effective measures to boost
0.40
output in the short run: the two-
year cumulative output multipliers
are about 1.4 and 1.3, 0.00
Government Targeted Untargeted Capital tax Consumption Targeted labor Uuntargeted
respectively; that is, a cumulative purchase transfers transfers tax tax labor tax
$1 increase in government
purchases and targeted transfers Source: IMF staff estimates.
Note: Original tax multipliers are negative for expansionary effects. To facilitate
will lead to a $1.4 and $1.3 comparison, the original tax multipliers are multiplied by -1 as plotted here; thus, a
increase, respectively, in output positive number indicates an increase in a variable as a result of the tax stimulus.
cumulatively over two years. 7 As
monetary policy is at the effective lower bound, an increase in government purchases does not have the
typical crowding-out effect on private investment and hence the output multipliers can be significantly
above 1. 8
Between the two measures that have large multipliers, an increase in targeted transfers has additional
benefits because it is effective in raising the income and consumption of liquidity-constrained households
(yellow bars, Online Annex Figure 1.5.1). Another instrument suitable to support liquidity-constrained
households is a targeted labor income tax cut. In addition to directly boosting the income of liquidity-
constrained households, both measures increase labor demand because of higher demand for goods. 9
In contrast, untargeted transfers and labor income tax cuts have much smaller multipliers of around 0.2
to 0.3, as a large part of additional income is saved, rather than spent, by the higher-income households.
This result is consistent with the empirical literature, which generally finds that output multipliers for
broad-based transfers are small (Gechert and Rannenberg 2018). 10 In practice, however, targeted
6 The multipliers are calculated based on the response differences between the baseline scenario (a severe recession without additional fiscal
stimulus) and a policy scenario (a recession with one of the fiscal stimulus measures).
7 This result is consistent with Coenen and others (2012), who find that government purchases and targeted transfers to liquidity-constrained
households are most effective in boosting output when monetary policy is expected to remain accommodative for a prolonged period. The
average multipliers from several structural models used in other institutions (including the US Federal Reserve, the European Central Bank, and
the Organization for Economic Co-operation and Development) in Coenen and others (2012) are also in line with the simulation results here.
8 Note that the multipliers simulated in the analysis may not be directly applicable to country-specific stimulus measures because they may differ
in size, design, and macroeconomic conditions from those assumed in the model and policy specifications.
9 The labor income tax cut has an additional positive supply-side effect of enhancing work incentives. This effect on labor supply, however,
reduces the magnitude of the real wage increase, generating an overall smaller increase in households’ income and hence produces a slightly
smaller output multiplier than the increase in targeted transfers.
10The relatively small output multiplier also holds for broad-based transfers provided in recessions. Shapiro and Slemrod (2003) find that only 22
percent of households report a spending increase from the 2001 US federal income tax rebate in response to the dot-com bubble crisis. Agarwal,
Liu, and Souleles (2007) estimate that about 40 percent of this rebate was consumed. Kan, Peng, and Wang (2017) estimate that the 2009

measures unavoidably exclude some of those in need of support, especially when administrative capacity
is low, as in many developing countries (IMF 2020a). Thus, their benefits in terms of cost-effectiveness
need to be weighed against the possibility of insufficient coverage and delayed distribution of funds.
Among all the measures, a capital income tax cut is the least effective in supporting liquidity-
constrained households and has relatively small output multipliers, particularly in the near term. Different
from normal times, when cutting capital income tax rates provides strong incentives to invest, the trickle-
down effect of a capital income tax cut is particularly weak in a deep recession, especially when it is
expected to last a long time. This result is driven by suppressed demand. As demand gradually recovers,
the output multipliers can increase over time. Under the consumption tax rate cut, the output multipliers
are slightly above 1 in the short run and increase more in the longer run as demand recovers further. 11
Since a consumption tax cut also benefits the liquidity-constrained households, it is reasonably effective
in boosting their consumption, helping to alleviate poverty.
The effectiveness of these measures will also depend on the timing, especially if countries decide to
adopt them while the pandemic is not fully under control and social distancing restrictions are still in
place (or may be tightened because of new infections). The simulations here assume that there are no
supply constraints as a consequence of social distancing (including lockdowns). Some measures—such as
increasing government purchases and targeted transfers to liquidity-constrained households—may also be
appropriate during gradual reopening, although the multipliers are likely to be smaller for the following
reasons:
• As many sectors remain closed during the gradual reopening phase, the effectiveness of broad-based
measures could be significantly lower. For example, labor income tax cuts will fail to encourage labor
supply in those sectors that are affected by social distancing restrictions or where demand remains
depressed because of health concerns. Also, before effective vaccines are available and widely
accessible, renewed infection waves can result in new restrictions and lockdowns, undermining the
stimulus measures.
• Increases in government purchases, given the difficulty in only increasing demand in those sectors
that do not have production constraints, can drive up production costs of those sectors with supply
constraints. In this circumstance, a higher production cost reduces private demand of the same
goods, thus offsetting the effectiveness of government purchases in boosting aggregate demand
(Baqaee and Farhi 2020).
• Because of precautionary saving motives in a deep recession with high uncertainty about the recovery
and concerns regarding employment, transfers to households in general may have a lower impact
during the gradual reopening stage (Auerbach, Gorodnichenko, and Murphy 2020). However,
transfers to those in need during the crisis serve an important objective—to save lives—and should
be implemented if necessary.
• The power of supply-side stimulus such as a labor income tax cut is unlikely to work well because
labor inputs in some sectors can be determined by governments’ or firms’ considerations based on
public health concerns during the gradual reopening stage.
shopping vouchers distributed in Taiwan Province of China in response to the global financial crisis had a marginal propensity to consume of
0.25.
11 The consumption tax cut has been adopted by several countries, including Germany and the United Kingdom, in response to COVID-19. The
sizes of multipliers implied by the simulations here, however, may not be directly applicable to country-specific consumption tax cut measures, as
the policy design and country conditions may not be captured by the simulations and model calibration.

Uncertain Fiscal Multipliers

In the current crisis, assessing fiscal multipliers is subject to an unusually high degree of uncertainty. On
the one hand, fiscal multipliers are expected to be larger in the recovery from a recession (Auerbach and
Gorodnichenko 2012, 2013; Canzoneri and others 2016) and also when monetary policy is
accommodative, with advanced economies at the effective lower bound (Christiano, Eichenbaum, and
Rebelo 2011; Erceg and Lindé 2014). 12 This is because a demand stimulus under accommodative
monetary policy does not drive up the real interest rate as in normal times. On the other hand, high levels
of public debt are likely to offset the expansionary effects, possibly because of expectations of fiscal
consolidation or rising sovereign risk premia (Ilzetzki, Mendoza, and Végh 2013; Bi, Shen, and Yang
2016; Fotiou, Shen, and Yang 2020; Huidrom and others, forthcoming).
These factors are further intertwined with unfamiliar and unpredictable trajectories of virus
transmission, which could magnify precautionary saving motives in anticipation of a worsening and
prolonged health and economic crisis. Historically, strong precautionary saving motives following a
pandemic tend to suppress the real interest rate (Jordà, Singh, and Taylor 2020). Although this provides
some breathing room for governments with constrained fiscal space, precautionary savings dampen the
expansionary effects of fiscal stimulus. During the lockdown phase, the saving rates in the euro area and
the United States soared. Although part of the savings was “forced” because people could not consume as
in normal times, some of it was likely a result of precautionary savings, which can persist if
unemployment declines slowly, as observed after the global financial crisis (Mody, Ohnsorge, and Sandrei
2012). If, however, effective treatment and vaccines become available quickly, repressed consumption
could be unleashed, producing an initial boost to the recovery.
Facilitating Economic Recovery with Limited or No Fiscal Space
After the delivery of unprecedented fiscal firepower in response to the COVID-19 pandemic, global
gross public debt is projected to reach similarly unprecedented levels approaching 100 percent of GDP in
2020. A challenge for many governments is to find additional resources to support the recovery and
avoid a too-abrupt fiscal adjustment as the exceptional lifelines are eliminated. For countries with limited
or no fiscal space, an option is to consider a productive stimulus measure that can be jointly implemented
with an adjustment measure in order to achieve dual objectives of facilitating the economic recovery in
the short run while keeping debt under control over the medium term. For example, to revive the
economy, the government of Spain proposed using a €10 billion fund to increase the capital of companies
in strategic sectors that are considered viable after the pandemic. The measure is combined with a tax
reform that will increase taxes on larger companies.
To illustrate the potential benefits of combining instruments, the impact of a targeted transfer increase
combined with an increase in progressive labor income taxes is simulated. The targeted transfer increases
by about 1 percentage point for the first two years in a deep recession and then declines over time as the
economy improves. The fiscal adjustment via the progressive tax increase is imposed on the higher-
income group under two adjustment speeds (Online Annex Figure 1.5.2): the slow-adjustment scenario
assumes the tax rate is increased three years after the stimulus by a relatively small magnitude; and the
fast-adjustment scenario assumes the tax rate is increased immediately by a bigger magnitude. For
computing multipliers, a third scenario without a transfer increase is also simulated. 13 In this scenario, the
12The empirical evidence that supports higher multipliers in low interest rates or at the effective lower bound includes Miyamoto, Nguyen, and
Sergeyev (2018) and Amendola and others (2020). See also Chapter 2 of the April 2020 World Economic Outlook.
13 Thisthird scenario is similar to the baseline scenario simulated earlier, which has a macroeconomic shock that generates a deep recession and
no fiscal stimulus. The baseline simulated earlier, however, has the consumption tax rate as the fiscal adjustment instrument. The scenario here,
instead, has the labor income tax rate on the higher-income group as the fiscal adjustment instrument.

progressive labor income tax rate does not increase within the first 10 years (although an adjustment
would eventually be needed to contain debt).
Online Annex Figure 1.5.2. Different Fiscal Adjustment Plans: Targeted
Transfers

Note: Output is in percent deviations from the path without the recession and additional transfer stimulus. The
x axes are years.
The two adjustment speeds have similar short-run expansionary effects: the one-year output multipliers
are 1.3 to 1.4, respectively (Online Annex Figure Online Annex Figure 1.5.3. Cumulative Output
1.5.3). With a fast speed, the labor income tax rate Multipliers for Targeted Transfers under Different
on the higher-income group begins to rise earlier Adjustment Speeds
by a larger magnitude (panel 2, Online Annex (Percent of GDP)
Figure 1.5.2), which discourages labor supply to a
greater degree. Over a longer horizon, the output
benefits from higher demand of liquidity-
constrained households are largely offset by a fast
adjustment speed because of a higher labor
income tax rate on the higher-income group,
leaving a longer-run cumulative output multiplier
at about zero. A fast adjustment speed puts public
debt on a downward path in the medium term
(red line in panel 4, Online Annex Figure 1.5.2).
Meanwhile, fiscal adjustments through a
progressive income tax help protect the income of
the vulnerable and mitigate after-tax income
inequality. The slow adjustment speed, on the other hand, produces a multiplier around 1 in the longer

horizon (black bars, Online Annex Figure 1.5.3) because the increase in the labor income tax rate is
delayed and relatively small. The public-debt-to-GDP ratio, however, is higher and could require further
adjustment in the long run (blue line in panel 4, Online Annex Figure 1.5.2).

References
Agarwal, S., C. Liu, and N.S. Souleles. 2007. “The Reaction of Consumer Spending and Debt to Tax
Rebates—Evidence from Consumer Credit Data.” Journal of Political Economy 115: 986–1019.
Amendola, A., M. Di Serio, M. Fragetta, and G. Melina. 2020. “The Euro-Area Government Spending
Multiplier at the Effective Lower Bound.” European Economic Review 127: 1–17.
Auerbach, A., and Y. Gorodnichenko. 2012. “Measuring the Output Responses to Fiscal Policy.”
American Economic Journal: Economic Policy 4 (2): 1–27.
Auerbach, A., and Y. Gorodnichenko. 2013. “Fiscal Multipliers in Recession and Expansion.” In Fiscal
Policy After the Financial Crisis, edited by A. Alesian and F. Giavazzi. Chicago: University of
Chicago Press.
Auerbach, A.J., Y. Gorodnichenko, and D. Murphy. 2020. “Fiscal Policy and COVID-19 Restrictions in a
Demand-Determined Economy.” NBER Working Paper 27366, National Bureau of Economic
Research, Cambridge, MA.
Bangham, G., and J. Leslie. 2020. “Rainy Days: An Audit of Household Wealth and the Initial Effects of
the Coronavirus Crisis on Saving and Spending in Great Britain.” Resolution Foundation and
Standard Life Foundation (June).
Baqaee, D., and E. Farhi. 2020. “Supply and Demand in Disaggregated Keynesian Economies with an
Application to the Covid-19 Crisis.” NBER Working Paper 27152, National Bureau of
Economic Research, Cambridge, MA.
Bi, H., W. Shen, and S.-C.S. Yang. 2016. “Debt-Dependent Effects of Fiscal Expansions.” European
Economic Review 56: 389–410.
Bottan, N., B. Hoffmann, and D. Vera-Cossio. 2020. “The Unequal Burden of the Coronavirus
Pandemic: Evidence from Latin America and the Caribbean.” Unpublished.
Bricco, J., F. Misch and A. Solovyeva. Forthcoming. “What Are the Economic Effects of Containment
Policies? Evidence from Sweden.” IMF Working Paper, International Monetary Fund,
Washington, DC.
Cahuc, P., F. Kramarz, and S. Nevoux. 2018. “When Short-Time Work Works.” CEPR Discussion Paper
13041, Centre for Economic Policy Research, London.
Canzoneri, M., F. Collard, H. Dellas, and B. Diba. 2016. “Fiscal Multipliers in Recessions.” The Economic
Journal 126: 75–108.
Chen, S., D.O. Igan, N. Pierri, and A.F. Presbitero. Forthcoming. “Tracking the Economic Impact of
COVID-19 and Mitigation Policies in Europe and the United States.” IMF Special Series on
COVID-19 (May 6), International Monetary Fund, Washington, DC.
Christiano, L.J., M. Eichenbaum, and S. Rebelo. 2011. “When Is the Government Spending Multiplier
Large?” Journal of Political Economy 1: 78–121.
Coenen, G., C.J. Erceg, C. Freedman, D. Furceri, M. Kumhof, R. Lalonde, D. Laxton, D., J. Linde, A.
Mourougane, D. Muir, S. Mursula, C. De Resende, J. Roberts, W. Roeger, S. Snudden, M.
Trabandt, and J. Veld. 2012. “Effects of Fiscal Stimulus in Structural Models.” American Economic
Journal: Macroeconomics 4: 22–68.
Coibion, O., Y. Gorodnichenko, and M. Weber. 2020. “The Cost of the COVID-19 Crisis: Lockdowns,
Macroeconomic Expectations, and Consumer Spending.” NBER Working Paper 27141,
National Bureau of Economic Research, Cambridge, MA.

Deb, P., D. Furceri, J.D. Ostry, and N. Tawk. 2020a. “The Economic Effects of COVID-19
Containment Measures.” COVID Economics 24: 32–75.
Deb, P., D. Furceri, J. D. Ostry, and N. Tawk. 2020b. “The Effect of Containment Measures on the
COVID-19 Pandemic.” COVID Economics 19: 53–86.
Dollar, D., T. Kleineberg, and A. Kraay. 2013. “Growth Still Is Good for the Poor.” Policy Research
Working Paper 6568, World Bank, Washington, DC.
Erceg, C., and J. Lindé. 2014. “Is There a Fiscal Free Lunch in a Liquidity Trap?” Journal of the European
Economic Association 12: 73–107.
Evans, B., and R. Matsumoto. 2020. “Social Protection and Labor Toolkit Guidance Note.” International
Monetary Fund, Washington, DC.
Fang, L., J. Nie, and Z. Xie. 2020. “Unemployment Insurance during a Pandemic.” Federal Reserve Bank
of Atlanta Working Paper 2020-13.
Fotiou, A., W. Shen, and S.-C.S. Yang. 2020. “The Fiscal State-Dependent Effects of Capital Income Tax
Cuts.” Journal of Economic Dynamics and Control 117: 1–21.
Fotiou, A., and A. Lagerborg. Forthcoming-a. “Early and Smart Containment: Lessons from Countries
with Past Experience.” IMF Working Paper, International Monetary Fund, Washington, DC.
Fotiou, A., and A. Lagerborg. Forthcoming-b. “Designing the COVID-19 Fiscal Response: Did Country
Precrisis Conditions Matter?” IMF Working Paper, International Monetary Fund, Washington,
DC.
Fournier, J-M. 2019. “A Buffer-Stock Model for the Government: Balancing Stability and Sustainability.”
IMF Working Paper No. 19/159, International Monetary Fund, Washington, DC.
Fournier, J-M. and P. Lieberknecht. 2020. “A Model-based Fiscal Taylor Rule and a Toolkit to Assess the
Fiscal Stance.” IMF Working Paper 20/33, International Monetary Fund, Washington, DC.
Furceri, D., P. Loungani, J.D. Ostry, and P Pizzuto. 2020. “COVID-19 Will Raise Inequality if
Past Pandemics Are a Guide.” VoxEU.org, March 8.
Gechert, S., and A. Rannenberg. 2018. “Which Fiscal Multipliers Are Regime-Dependent? A Meta-
Regression Analysis.” Journal of Economic Surveys 32: 1160–182.
Gentilini, U., M. Almenfi, P. Dale, A.V. Lopez, I. Mujica, R. Quintana, and U. Zafar. 2020a.
“Social Protection and Jobs Responses to COVID-19.” Living Paper, Version 11 (June
12), World Bank, Washington, DC.
Gentilini, U., M. Almenfi, P. Dale, A.V. Lopez, I. Mujica, R. Quintana, and U. Zafar. 2020b.
“Global Database on Social Protection and Jobs Responses to COVID-19.” Living
Paper, Version 11 (June 12), World Bank, Washington, DC.
Giupponi, G., and C. Landais. 2018. “Subsidizing Labor Hoarding in Recessions: The Employment and
Welfare Effects of Short-Time Work.” CEPR Discussion Paper No.13310, Centre for Economic
Policy Research, London.
Gourinchas, P. 2020. “Flattening the Pandemic and Recession Curves.” In Mitigating the COVID Economic
Crisis: Act Fast and Do Whatever It Takes, edited by Richard Baldwin and Beatrice Weder di Mauro.
London: Centre for Economic Policy Research.
Han, J., B Meyer, and J. Sullivan. 2020. “Income and Poverty in the COVID-19 Pandemic.” University of
Chicago Working Paper (June).
Huidrom, R., M.A. Kose, J. J. Lim, and F.L. Ohnsorge. Forthcoming. “Why Do Fiscal Multipliers
Depend on Fiscal Positions?” Journal of Monetary Economics.

Ilzetzki E., E.G. Mendoza, and C.A. Végh. 2013. “How Big (Small?) Are Fiscal Multipliers?” Journal of
Monetary Economics 60: 239–54.
International Monetary Fund (IMF). 2020a. “Expenditure Policies in Support of Firms and Households.”
Special Series on Fiscal Policies to Respond to COVID-19, IMF Fiscal Affairs Department,
Washington, DC.
International Monetary Fund (IMF). 2020. “United States 2020 IMF Article IV Consultation Staff
Report.” IMF, Washington, DC.
Jordà, Ò., S.R. Singh, and A.M. Taylor. 2020. “Longer-Run Economic Consequences of Pandemics.”
Working Paper 2020-09, Federal Reserve Bank of San Francisco.
Kalra, S., D. Kelkar, S.C. Galwankar, T.J. Papadimos, S.P. Stawicki, B. Arquilla, B.A. Hoey, R.P. Sharpe,
D. Sabol, and J.A. Jahre. 2014. “The Emergence of Ebola as a Global Health Security Threat:
From ‘Lessons Learned’ to Coordinated Multilateral Containment Efforts.” Journal of Global
Infectious Diseases 6 (4): 164.
Kan, K., S.-K. Peng, and P. Wang. 2017. “Understanding Consumption Behavior: Evidence from
Consumers’ Reaction to Shopping Vouchers.” American Economic Journal: Economic Policy 9: 137–
53.
Kikuchi, S., S. Kitao, and M. Mikoshiba. 2020. “Heterogeneous Vulnerability to the COVID-19 Crisis
and Implications for Inequality in Japan.” Discussion Paper No. 20039, Research Institute of
Economy, Trade and Industry.
Kopp, D. and M. Siegenthaler. 2019. “Short-Time Work and Unemployment In and After the Great
Recession.” KOF Swiss Economic Institute Working Paper 462.
Loayza, N.V. 2020. “Costs and Trade-Offs in the Fight Against the COVID-19 Pandemic: A Developing
Country Perspective.” Research & Policy Brief 35, World Bank Group, Washington, DC.
Loayza, N.V., and S. Pennings. 2020. “Macroeconomic Policy in the Time of COVID-19: A Primer for
Developing Countries.” Research and Policy Brief (March 26), World Bank Group, Washington,
DC.
Medas, P., T. Poghosyan, Y. Xu, J. Farah-Yacoub, and K. Gerling. 2018. “Fiscal Crises.” Journal of
International Money and Finance vol 88, issue C.
Miyamoto, W., T., L. Nguyen, and D. Sergeyev. 2018. “Government Spending Multipliers under the Zero
Lower Bound: Evidence from Japan.” American Economic Journal: Macroeconomics 10: 247–77.
Mody, A., F. Ohnsorge, and D. Sandri. 2012. “Precautionary Savings in the Great Recession.” IMF
Economic Review 60: 114–38.
Moreno-Badia, M., P. Medas, P. Gupta, and Y. Xiang. 2020. “Debt is Not Free.” IMF Working Paper
20/1, International Monetary Fund, Washington, DC.
Palomino, J., J. Rodriguez, and R. Sebastian. 2020. “Wage Inequality and Poverty Effects of Lockdown
and Social Distancing in Europe.” COVID Economics 25: 186–229.
Parolin, Z., M.A. Curran, and C. Wimer. 2020. “The CARES Act and Poverty in the COVID-19 Crisis.”
Columbia University Center on Poverty and Social Policy Poverty and Social Policy Brief 4 (8).
Prady, D. 2020. “Managing the Impact on Households: Assessing Universal Transfers (UT).” IMF
Special Series on Fiscal Policies to Respond to COVID-19, International Monetary Fund,
Washington, DC.

Shang, B., B. Evans, and Z. An. 2020. “Expenditure Policies in Support of Firms and Households.” IMF
Special Series on Fiscal Policies to Respond to COVID-19, International Monetary Fund,
Washington, DC.
Shapiro, M.D., and J. Slemrod. 2003. “Consumer Response to Tax Rebates.” American Economic Review 93:
381–96.
Traum, N., and S.-C.S. Yang. 2015. “When Does Government Debt Crowd Out Investment?” Journal of
Applied Econometrics 30: 24–45.
Trebesch, C., and M. Zabel .2017. “The Output Cost of Hard and Soft Sovereign Default.” European
Economic Review 92 (C): 416–32.
Wojda, T.R., P.L. Valenza, K. Cornejo, T. McGinley, S.C. Galwankar, D. Kelkar, R.P. Sharpe, T.J.
Papadimos, and S.P. Stawicki. 2015. “The Ebola Outbreak of 2014–2015: From Coordinated
Multilateral Action to Effective Disease Containment, Vaccine Development, and
Beyond.” Journal of Global Infectious Diseases 7(4): 127.
World Bank. 2020. “Projected Poverty Impacts of COVID-19 (Coronavirus).” World Bank Brief, June
2020. https://www.worldbank.org/en/topic/poverty/brief/projected-poverty-impacts-of-
COVID-19
Xie, Z. 2020. “Changes in State Unemployment Insurance Rules During the COVID-19 Outbreak.”
Federal Reserve Bank of Atlanta Policy Hub 04-2020.
Yemtsov, R., M. Honorati, B. Evans, Z. Sajaia, and M. Lokshin. 2018. “Measuring the Effectiveness of
Social Protection: Concepts and Applications. Streamlined Analysis with ADePT Software.”
World Bank, Washington, DC.

Online Annex 1.1. How Will The COVID-19 Pandemic Affect Poverty and Inequality?

Uploaded by

Copyright:

Available Formats

Online Annex 1.1. How Will The COVID-19 Pandemic Affect Poverty and Inequality?

Uploaded by

Document Information

Original Description:

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Online Annex 1.1. How Will The COVID-19 Pandemic Affect Poverty and Inequality?

Uploaded by

Copyright:

Available Formats

CHAPTER 1 FISCAL POLICIES TO ADDRESS THE COVID-19 PANDEMIC

Advanced Economies Emerging Market Economies

Sources: World Bank PovcalNet database; and IMF staff estimates.

1 Prepared by W. Raphael Lam, Delphine Prady, and Baoping Shang.

International Monetary Fund | October 2020 1

2 International Monetary Fund | October 2020

International Monetary Fund | October 2020 3

Online Annex Figure 1.1.5. Distribution of Potential Rise in Extreme Poverty

4 International Monetary Fund | October 2020

4A detailed methodology can be found in Yemtsov and others (2018).

International Monetary Fund | October 2020 5

6 International Monetary Fund | October 2020

International Monetary Fund | October 2020 7

8 International Monetary Fund | October 2020

Online Annex 1.2. Smart Strategies to Contain the COVID-19

International Monetary Fund | October 2020 9

population is above 10,000.

10 International Monetary Fund | October 2020

Source: Fotiou and Lagerborg (forthcoming-a).

International Monetary Fund | October 2020 11

12 International Monetary Fund | October 2020

Source: Fotiou and Lagerborg (forthcoming-a).

Source: Fotiou and Lagerborg (forthcoming-a).

International Monetary Fund | October 2020 13

Stringency (average) -0.991

Observations 135 136 123 124

Online Annex Table 1.2.2. Effect of Individual Containment Measures (One at a

Average Response Early Response

Sources: Fotiou and Lagerborg (forthcoming-a); and IMF staff estimates.

14 International Monetary Fund | October 2020

Stringency (average) -0.067** -0.104*** -0.089*** -0.088***

Observations 153 152 152 149 134 133 133 131

Stringency (average) -0.07 -0.093** -0.088** -0.082**

Observations 147 147 147 146 129 129 129 129

Sources: Fotiou and Lagerborg (forthcoming-a); and IMF staff estimates.

International Monetary Fund | October 2020 15

Online Annex 1.3. From Lockdown to Recovery: Spending Measures

16 International Monetary Fund | October 2020

Sources: Gentilini and others (2020b); and IMF staff estimates.

International Monetary Fund | October 2020 17

18 International Monetary Fund | October 2020

International Monetary Fund | October 2020 19

Online Annex 1.4. Determining the Size of Fiscal Stimulus for

20 International Monetary Fund | October 2020

1 The calibration follows Fournier and Lieberknecht (2020).

International Monetary Fund | October 2020 21

3. Output Gap 4. Output Growth

Source: IMF staff estimates.

22 International Monetary Fund | October 2020

4 Higher debt and higher

Sources: Fournier (2019); and IMF staff estimates.

International Monetary Fund | October 2020 23

24 International Monetary Fund | October 2020

Online Annex 1.5. Policy Options to Support the Economic Recovery

government spending or tax revenue over the same horizon.

International Monetary Fund | October 2020 25

(Online Annex Figure 1.5.1). 6

Government purchases and

Stringency (average) -0.067 -0.104* -0.089* -0.088*

Stringency (average) -0.07 -0.093 -0.088 -0.082**