Initial Output Losses from the Covid-19 Pandemic: Robust Determinants
  • 1 0000000404811396https://isni.org/isni/0000000404811396International Monetary Fund

Contributor Notes

While the COVID-19 pandemic is affecting all countries, output losses vary considerably across countries. We provide a first analysis of robust determinants of observed initial output losses using model-averaging techniques—Weighted Average Least Squares and Bayesian Model Averaging. The results suggest that countries that experienced larger output losses are those with lower GDP per capita, more stringent containment measures, higher deaths per capita, higher tourism dependence, more liberalized financial markets, higher pre-crisis growth, lower fiscal stimulus, higher ethnic and religious fractionalization and more democratic regimes. With respect to the first factor, lower resilience of poorer countries reflects the higher economic costs of containment measures and deaths in such countries and less effective fiscal and monetary policy stimulus.

Abstract

While the COVID-19 pandemic is affecting all countries, output losses vary considerably across countries. We provide a first analysis of robust determinants of observed initial output losses using model-averaging techniques—Weighted Average Least Squares and Bayesian Model Averaging. The results suggest that countries that experienced larger output losses are those with lower GDP per capita, more stringent containment measures, higher deaths per capita, higher tourism dependence, more liberalized financial markets, higher pre-crisis growth, lower fiscal stimulus, higher ethnic and religious fractionalization and more democratic regimes. With respect to the first factor, lower resilience of poorer countries reflects the higher economic costs of containment measures and deaths in such countries and less effective fiscal and monetary policy stimulus.

I. Introduction

The magnitude of the COVID-19 recession is unprecedented, and easily dwarfs the blow from the Global Financial Crisis (IMF, 2020). Initial output losses, however, vary considerably across countries. Figure 1a shows, for a sample of 60 advanced, emerging and developing economies, a density plot of growth in the first semester of 2020 minus the IMF pre-pandemic growth forecast. While all countries had a negative surprise, there is considerable variation. Unexpected growth is but a few percentage points in Korea but ranges to more than 30 percentage points in Peru. Such heterogeneity is also evident when comparing first semester growth in 2020 versus 2019 (Figure 1b).

Figure 1.
Figure 1.

Distribution of Output Performances (%)—Density Plots

Citation: IMF Working Papers 2021, 018; 10.5089/9781513567013.001.A001

What drives this heterogeneity? Because the pandemic is foremost a health crisis, a natural candidate is the severity of health-related factors measured for example by: deaths per capita; degree of health preparedness; and stringency of containment. These factors, however, explain only a small fraction of observed output performance (Figure 2), suggesting the researcher need look elsewhere for a fuller explanation.

Figure 2.
Figure 2.

Output Performances (%) and Public Health

Citation: IMF Working Papers 2021, 018; 10.5089/9781513567013.001.A001

Note: Output performance is defined the difference between the observed cumulative real GDP growth in 2020H1 and the cumulative growth that was expected before the onset of the pandemic for the same period—based on the IMF World Economic Outlook 2020 January forecast for 2020H1.

Assessing which factors drive the heterogeneous outcomes is not an easy task, for three interrelated reasons. First, the number of observations is relatively small and limited by the number of countries with available quarterly data. Second, the number of potential factors affecting economic resilience is large. Third, many of the country characteristics are correlated with one other: the level of regulation in product and financial markets is likely to be correlated with the level of development, for example (Alesina et al., 2020).

We address these issues by considering a large set of explanatory variables and analyzing all the regressors jointly by averaging outcomes for all possible combinations of regressors (more than 1.07 billion regressions) using model-averaging techniques: Weighted Average Linear Squared (WALS) developed by Magnus, Powell, and Prüfer (2010); and Bayesian Model Averaging (BMA) developed by Fernandez, Ley and Steel (2001a). WALS and BMA share similar foundations. There are two main differences. First, WALS relies on a preliminary orthogonal transformation of the auxiliary regressors and their parameters, whose advantage is to increase speed of computation. Second, while WALS uses a Laplace distribution to reduce the risk of excessive influence of the prior on final estimates, BMA uses a Gaussian distribution prior for the auxiliary parameters. Reflecting these tradeoffs, we use WALS as our baseline technique, and adopt the BMA as a robustness check.

We focus on the acute phase of the crisis because most countries are already recording positive growth in the third quarter of 2020 and the factors affecting recovery are different from those driving the downturn.1 We consider two measures of output performance: (i) actual growth in the first semester of 2020 minus the January 2020 IMF growth forecast for this period; and (ii) growth in the first semester of 2020 minus the growth rate for the first semester of 2019.

We find that larger output losses are experienced by countries with lower GDP per capita, more stringent containment, higher deaths per capita, a larger tourism share, more liberalized credit markets, higher pre-crisis growth, and more democratic regimes. We also find that lower fiscal stimulus and higher social fractionalization are positively correlated with one measure of output loss. GDP per capita is particularly important: a country at the 75th percentile of the per capita GDP distribution (such as Portugal) has a 7-percentage-point smaller growth surprise than a country at the 25th percentile (such as Bangladesh). This result reflects higher economic costs of containment and deaths in poorer countries and less effective macro policy stimulus.

Our paper contributes to two strands of the literature. The first is on resilience following major crises such as the GFC: Rose (2011); Rose and Spiegel (2010, 2011, 2012); Giannone, Lenza and Reichlin (2011); Obstfeld et al. (2009); Blanchard et al. (2010); Devereux and Dwyer (2016). In contrast to these studies, we do not find that trade and financial openness have been important drivers of output surprises in our study. The second is on use of Bayesian model-averaging techniques in the macroeconomic literature, including studies focusing on robust drivers of growth (e.g., Fernandez et al., 2001b; Brock and Durlauf, 2001; and Sala-i-Martin et al., 2004), inequality (Furceri and Ostry, 2019) and reforms (Duval, Furceri and Mieithe, 2020).

The rest of the paper is structured as followed. In Section II, we provide an overview of our empirical approach. In Section III, we introduce potential determinants of COVID-19 output losses. In Section IV, we summarize these results and provide an overall assessment of the statistical robustness of the determinants. Sections V concludes, highlighting policy implications and issues for future research.

II. Empirical Framework

Although there is a voluminous literature on the determinants of economic recessions, cross-sectional information has not been fully exploited to study the drivers of the COVID-19 recession, and theory provides little guidance on appropriate model specification. Therefore, we start from a simple linear reduced from specification:

Yi=α+βXi+μi(1)

where X is a vector of k covariates reflecting characteristics of economy i along different dimensions, and Y is a measure of output performance. Such an approach needs to confront two econometric challenges: (i) the large number of potential explanatory factors and correlation among them; and (ii) lack of an a priori “true” statistical model to test. With an unknown true model, the number of possible independent variables is very large. Depending on the model selection procedure, conclusions could vary significantly.

To meet these concerns, the literature has turned to model-averaging techniques.2 Model-averaging addresses the challenges by: (i) running the maximum combination of models; and (ii) providing estimates that take into account the performance of each potential driver not only in the final “reported” model but over the whole set of possible specifications. Formally, assuming that we are faced with M different models and that βx is the coefficient related to variable X in each model, a final estimate of βx is computed as βx=Σ1Mωiβx,i, where the weights a)t denote a measure of goodness of fit of each model.

In this paper, we rely on two model-averaging techniques: Weighted Average Linear Squares (WALS) developed by Magnus, Powell, and Prüfer (2010), and Bayesian Model Averaging (BMA) developed by Fernandez, Ley and Steel (2001a). WALS and BMA share similar foundations. There are, however, two main differences. First, WALS relies on a preliminary orthogonal transformation of the auxiliary regressors and their parameters. The key advantage of this transformation is that the space over which model selection is performed rises linearly rather than exponentially with model size as in BMA (2K2 where K2 is the number of “auxiliary” regressors to be tested). Second, while WALS uses a Laplace distribution to reduce risks of excessive influence of the prior on final estimates, BMA uses a Gaussian prior for the auxiliary parameters (see Annex B). Reflecting these considerations, we use WALS as our baseline technique, and adopt BMA as a robustness check.

To decide which regressors are robust determinants of output loss, we follow the literature. For WALS, Magnus, Powell, and Prüfer (2010) suggest using a threshold value of the t-statistic—greater than 1 (in absolute value)—to determine that a regressor is robust. Using such a threshold means including regressors which improve the model fit (measured by the adjusted R2) and the precision of the estimators measured by the MSE. For BMA, the procedure involves estimating the posterior probability that a given variable belongs in the “true” model and selecting variables with high posterior probabilities as the robust determinants.

While model averaging addresses model uncertainty and omitted variable bias, it does not address reverse-causality issues—where event studies may be appropriate. While reverse causality is not an issue for many of the more structural characteristics used in our analysis, it may be a valid concern for policies implemented in response to the pandemic.

III. Potential Determinants

Variable selection is driven to an important extent by data availability. Given the small number of quarterly GDP growth observations (96), we constrain the choice and number of variables so that we are left with enough degrees of freedom for estimation. The set of regressors in the baseline includes 30 variables grouped into six categories: (i) Public health; (ii) Sectoral composition; (iii) Fiscal and monetary response; (iv) Macroeconomic characteristics; (v) Regulation; and (vi) Development level, Demographics and Institutions. In the robustness section, where we extend the set of regressors to 34, the results based on the more limited (full-model) sample of 48 observations are qualitatively similar but less precise. Data sources and key descriptive statistics are reported in Table A1 of Annex A.

A. Public Health Indicators

Countries with higher per capita deaths should experience greater output losses through reduced labor supply and greater demand-reducing social distancing (Maloney and Taskin, 2020). Hasell (2020) finds a negative relationship between deaths per capita and year-over-year growth in the second quarter of 2020, supporting this prior. Stringency of non-pharmaceutical (containment) measures, designed to avoid overwhelming the medical system while effective treatments and vaccines are developed, is associated with short-term output loss. Main measures include: (i) school closures; (ii) workplace closures; (iii) cancellation of public events; (iv) restrictions on size of gatherings; (v) closures of public transport; (vi) stay-at-home orders; (vii) restrictions on internal movement; (viii) restrictions on international travel.3 Likewise, countries with better health systems in terms of epidemic management and prevention are expected to suffer smaller economic losses (Deb et al. 2020b).

To test the empirical relevance of these factors, we use the following three variables: (i) log of deaths per capita—cumulative deaths as of June 30 relative to population; (ii) the containment stringency index from the Oxford Coronavirus Government Response Tracker, normalized from 0 to 1;4 and (iii) the Global Health Security Index from Johns Hopkins University.5 Figure 2 presents scatter plots between these measures and our first measure of output loss (the second measure is shown in the Appendix). Output loss is larger for countries with higher mortality and containment, while no relation is found with the Health Security Index, or any of its sub-indicators. OLS and WALS regressions confirm these findings (Table 1).

Table 1.

Regression Results of Public Health, Ols and Wals

article image
Note: Output performance 1 is the difference between the observed cumulative real GDP growth in 2020H1 and the cumulative growth that was expected before the onset of the pandemic for the same period—based on the IMF World Economic Outlook 2020 January forecast for 2020H1; Output performance 2 is the difference in cumulative real GDP growth between the first half of 2020 (2020H1) and the first half of 2019 (2019H1). t-statistic reported in parentheses. In bold those regressors that can be considered “robust”—that is, with a t-value in absolute value greater than 1. * p<0.05, ** p<0.01, *** p<0.001.

B. Industry Shares

Recessions tend to have heterogenous effects across industries. Evidence from past recessions and financial crises in advanced economies suggests that finance and manufacturing tend to contract more than other sectors during downturns (Aaronson, Rissman, and Sullivan, 2004; Furceri et al., 2020), while services tend to be more resilient (Kopelman and Rosen, 2016). However, because this crisis is foremost a health crisis and has been met with strong containment measures, high-contact sectors (such as tourism and retail) and non-teleworkable industries (mining, manufacturing, and construction) have been the ones to experience relatively large drops in activity (Stephany et al., 2020).

To test the role of sectoral composition, we consider three indicators (for 2019) from the World Development Indicators: shares of services, manufacturing, and tourism in value added (we exclude agriculture to avoid perfect collinearity). The scatter plot in Figure 3, as well as the OLS and WALS results reported in Table 2, confirm that services, and particularly tourism, have been hit the hardest during this crisis.

Figure 3.
Figure 3.

Output Performances (%) and Sectoral Composition

Citation: IMF Working Papers 2021, 018; 10.5089/9781513567013.001.A001

Note: Output performance is defined the difference between the observed cumulative real GDP growth in 2020H1 and the cumulative growth that was expected before the onset of the pandemic for the same period—based on the IMF World Economic Outlook 2020 January forecast for 2020H1.
Table 2.

Regression Results of Sectorial Composition, Ols and Wals

article image
Note: Output performance 1 is the difference between the observed cumulative real GDP growth in 2020H1 and the cumulative growth that was expected before the onset of the pandemic for the same period—based on the IMF World Economic Outlook 2020 January forecast for 2020H1; Output performance 2 is the difference in cumulative real GDP growth between the first half of 2020 (2020H1) and the first half of 2019 (2019H1). t-statistic reported in parentheses. In bold those regressors that can be considered “robust”—that is, with a t-value in absolute value greater than 1. * p<0.05, ** p<0.01, *** p<0.001.

C. Fiscal and Monetary Policies

Governments and central banks have implemented unprecedented support measures in response to the pandemic. As of June 30, 2020, more than 90 countries had announced fiscal packages ranging in size from 1 to 23 percent of GDP (IMF’s Covid-19 Policy Tracker). In addition, monetary policy rates have been cut in 97 countries from December 2019 to-June 2020 and many central banks have deployed unconventional tools. Preliminary evidence suggest that these measures have been effective in reducing the depth of the recession, especially in advanced economies where fiscal multipliers are higher and monetary policy transmission is more effective (Faria-e-Castro, 2020; Jinjarak et al., 2020; Fornaro and Wolf; Bayer et al., 2020).

To test the role of policy stimulus, we use the IMF’s Covid-19 Policy Tracker measures of: (i) total fiscal stimulus (above and below the line) deployed (or announced); (ii) the cumulative change in the policy interest rate from December 2019 to June 2020; (iii) the amount of liquidity (in percent of GDP) injected by central banks from December 2019 to June 2020. Figure 4 shows that only policy rate cuts seem to be associated with lower output loss. Moreover, none of the variables is statistically significant when performing OLS and WALS regressions (Table 3). While lack of significance could be due omitted variable bias or reverse causality—as countries may provide more support in response to weak activity, it could also reflect the lack of a causal impact for two reasons: first, some of the fiscal measures have been announced but not yet implemented; and second, it may take time for policy stimulus to affect activity. In addition, it is likely that impacts are heterogeneous across countries, an issue explored below.

Table 3.

Regression Results of Fiscal and Monetary Response, Ols and Wals

article image
Note: Output performance 1 is the difference between the observed cumulative real GDP growth in 2020H1 and the cumulative growth that was expected before the onset of the pandemic for the same period—based on the IMF World Economic Outlook 2020 January forecast for 2020H1; Output performance 2 is the difference in cumulative real GDP growth between the first half of 2020 (2020H1) and the first half of 2019 (2019H1). t-statistic reported in parentheses. In bold those regressors that can be considered “robust”—that is, with a t-value in absolute value greater than 1. * p<0.05, ** p<0.01, *** p<0.001.
Figure 4.
Figure 4.

Output Performances (%) and Fiscal and Monetary Response

Citation: IMF Working Papers 2021, 018; 10.5089/9781513567013.001.A001

Note: Output performance is defined the difference between the observed cumulative real GDP growth in 2020H1 and the cumulative growth that was expected before the onset of the pandemic for the same period—based on the IMF World Economic Outlook 2020 January forecast for 2020H1.

D. Regulation

Labor and product market regulations can affect realized output losses given the shifts of labor across industries in response to the pandemic. Evidence from past recessions and financial crises suggest that countries with more flexible product and labor market regulations are more resilient (Eichhorst et al., 2010; Artha and de Haan, 2011; Bernal-Verdugo et al., 2012; Bluedorn et al., 2019).6 The relationship between resilience and credit market regulation is less settled. While liberalized markets contribute to financial depending and lower volatility (Beck and Demirguc-Kunt, 2009), in the short term they may amplify volatility: Caprio and Honohan (2002) find that banking systems less subject to monitoring exhibit more procyclicality; Giannone et al. (2011) find a negative correlation between credit market liberalization and output growth during the GFC.

To test the role of regulatory variables, we consider the most recent observation (typically, 2019) for the following indicators from the Fraser Institute Index of Economic Freedom: (i) credit market deregulation, which includes ownership of banks, competition, and extension of credit; (ii) labor market deregulation, a composite index of hiring and firing practices; (iii) business deregulation, which assesses difficulty in starting a new business, including administrative rules and government bureaucracy.7 The indicators range from 0 to 10, with higher values indicating less regulation.8 The scatter plots in Figure 5, as well as the WALS results in Table 4, confirm that countries with freer financial markets are less resilient. In contrast, we do not find robust significant relationship between other regulatory measures and both measures of output performance.

Figure 5.
Figure 5.

Output Performances (%) and Regulation

Citation: IMF Working Papers 2021, 018; 10.5089/9781513567013.001.A001

Note: Output performance is defined the difference between the observed cumulative real GDP growth in 2020H1 and the cumulative growth that was expected before the onset of the pandemic for the same period—based on the IMF World Economic Outlook 2020 January forecast for 2020H1.
Table 4.

Regression Results of Regulation, Ols and Wals

article image
Note: Output performance 1 is the difference between the observed cumulative real GDP growth in 2020H1 and the cumulative growth that was expected before the onset of the pandemic for the same period—based on the IMF World Economic Outlook 2020 January forecast for 2020H1; Output performance 2 is the difference in cumulative real GDP growth between the first half of 2020 (2020H1) and the first half of 2019 (2019H1). t-statistic reported in parentheses. In bold those regressors that can be considered “robust”—that is, with a t-value in absolute value greater than 1. * p<0.05, ** p<0.01, *** p<0.001.

E. Macroeconomic characteristics

Macroeconomic fundamentals can play a substantial role in mitigating output losses during a crisis. One reason is that crises usually come with excess volatility and increases in uncertainty (Ahir et al., 2018) which can lead to significant outflows of capital in countries with large imbalances (McQuade and Schmitz, 2017; Aizenman and Pasricha, 2010). Domestic imbalances, such as a high debt-to-GDP ratios, can also affect resilience by reducing fiscal space, and constraining counter-cyclical policies (Ostry et al., 2010; Kim and Ostry, 2018).

Financial markets can also affect resilience. On the one hand, financial depth can foster risk-sharing across economic agents, enhance consumption smoothing and dampen the effect of cyclical shocks (Beck et al. 2009; Ostry et al., 2009). On the other, excess leverage can lead to larger output losses during periods of financial stress (Feldkircher, 2014; Berkmen et al., 2012; Devereux and Dwyer, 2016; Frankel and Saravelos, 2010; Cecchetti et al., 2011; Babecký, 2012, Babecký, 2013; Caprio et al., 2014).

Trade and financial linkages have also played an amplification role in past crises. Blanchard et al. (2010) find that the economic performance in trading partners was a strong predictor of output loss during the GFC, while Claessens et al. (2012) find that the GFC exerted a larger impact on trade-dependent firms; Groot et al. (2011), Ho (2010), Levchenko et al. (2010) and Demir and Javorchik (2020) also stress the role of the trade channel. In a similar vein, Rodrik (1998), Bhagwati (1998) and Stiglitz (2002) argue that financial integration induces volatility in times of recession and endangers financial stability (see Kose et al., 2009), while Rose (2011) and Rose and Spiegel (2012) show that greater financial exposure to the United States was not associated with larger output losses in the GFC. Exchange rate flexibility may also affect resilience in the face of external shocks (Ghosh, Ostry and Wolf, 1997; Ghosh, Ostry and Tsangarides, 2011; Ghosh, Ostry and Qureshi, 2015).

To test the role of these factors, we consider the most recent pre-crisis value of the following variables: (i) current account balance as a share of GDP; (ii) general government debt-to-GDP ratio; (iii) financial system deposits as a percent of GDP; (iv) bank concentration; (v) domestic credit as a percent of GDP; (vi) and (vii) trade and financial globalization indices developed by the KOF Swiss Economic Institute; and (viii) an exchange rate regime variable— which assumes 1 for fixed; 2 for intermediate and 3 for flexible—from the IMF. We also consider the three-year average GDP growth preceding the COVID-19 crisis to control for crosscountry heterogeneity in pre-crisis growth. Figure 6 presents scatter plots between output loss and each of these variables. Output performance seems weaker for countries with higher pre-crisis growth, greater financial development and openness, higher pre-crisis debt-to-GDP ratios and current account deficits. Among these variables, however, only the debt-to-GDP ratio and pre-crisis growth appear to be robust determinants of output performance (Table 5).

Figure 6.
Figure 6.

Output Performances (%) and Macroeconomics Characteristics

Citation: IMF Working Papers 2021, 018; 10.5089/9781513567013.001.A001

Note: Output performance is defined the difference between the observed cumulative real GDP growth in 2020H1 and the cumulative growth that was expected before the onset of the pandemic for the same period—based on the IMF World Economic Outlook 2020 January forecast for 2020H1.
Table 5.

Regression Results of Macroeconomics Characteristics, Ols and Wals

article image
Note: Output performance 1 is the difference between the observed cumulative real GDP growth in 2020H1 and the cumulative growth that was expected before the onset of the pandemic for the same period—based on the IMF World Economic Outlook 2020 January forecast for 2020H1; Output performance 2 is the difference in cumulative real GDP growth between the first half of 2020 (2020H1) and the first half of 2019 (2019H1). t-statistic reported in parentheses. In bold those regressors that can be considered “robust”—that is, with a t-value in absolute value greater than 1. * p<0.05, ** p<0.01, *** p<0.001.

F. Development level, Demographics and Institutions

Development level (per capita GDP) may influence resilience, as containment measures may be costlier in poorer countries because of limited social safety nets and larger shares of financially-constrained households and firms. In addition, there is evidence that fiscal and monetary policy are more effective in advanced economies (Ilzetki et al., 2013 on fiscal policy; Brandao-Marques et al., 2020 on monetary policy). On the other hand, resilience could be enhanced in poorer economies as larger informal sectors reduce nominal rigidities (Mithra, 2013).

Income distribution may affect resilience: Wright et al. (2020) find that shelter-in-place policies are more effective in reducing virus spread in richer countries. Weill et al. (2020) show that social distancing measures reduce mobility more in wealthier areas. In addition to its effects through compliance with social distancing, inequality can affect resilience if more inequal societies have larger shares of vulnerable workers.

Turning to demographic characteristics, the pandemic is more serious in terms of symptoms and death for the elderly: Ioannidis et al. (2020) finds that 88–96% of people dying with or because of COVID-19 are 65 or above. In addition, the effect of the crisis on labor force dropouts is larger for older workers. Thus, countries with older populations are likely to suffer more from job loss due to injury, death or labor force dropout. Country size may also play a role: smaller economies are typically more volatile (Furceri and Karras 2007) while larger economies may find it more difficult and costly to manage public health services (Alesina et al. 2005). Finally, virus spread runs through social proximity, which is why high population density is associated with high case numbers.

Other factors we consider include: remittances; social fractionalization; and the nature of the political regime. The effect of remittances on resilience during a crisis is unclear as they tend to be countercyclical in the worker’s country of origin and procyclical in the migrant’s host country (Frankel 2011). Ethnic and religious fractionalizations can also affect output performance during a crisis by impairing the quality of the government and its policy response (Alesina et al., 2003). Finally, the type of political regime may shape both pandemic management and readiness of the public health system: democratic countries tend to have better public health systems (Ruger, 2005; Sen, 1999) which can give them an edge in fighting the disease (Kavanagh and Singh, 2020), but authoritarian governments may react faster and adopt drastic policy measures without fearing popular resistance. Cepaluni et al. (2020) show that more democratic countries face higher per capita deaths than less democratic countries do.

To assess the role of these factors, we consider the most recent pre-crisis value of: (i) the (log) of GDP per capita from the World Development Indicators (WDI); (ii) the Gini coefficient of after-tax income from the Standardized World Income Inequality Database; (iii) the share of population over 65; (iv) and (v) (log of) population and population density from the World Development Indicators; (vi) the share of remittances to GDP from the World Bank Financial Structure Database; (vii) a composite indicator of ethnic and religious fractionalization from Alesina et al. (2003); (viii) an indicator of informality from WDI; (vii) the level of democracy from Polity IV. Figure 7 suggests larger output losses in countries with lower GDP per capita, higher inequality, larger informal sectors, higher remittances and more democratic regimes. The evidence for inequality and informality are confirmed by the WALS results (Table 6).

Figure 7.
Figure 7.

Output Performances (%) and Development, Demographic and Institutions

Citation: IMF Working Papers 2021, 018; 10.5089/9781513567013.001.A001

Note: Output performance is defined the difference between the observed cumulative real GDP growth in 2020H1 and the cumulative growth that was expected before the onset of the pandemic for the same period—based on the IMF World Economic Outlook 2020 January forecast for 2020H1.
Table 6.

Regression Results of Development, Demographic and Institutions, Ols and Wals

article image
Note: Output performance 1 is the difference between the observed cumulative real GDP growth in 2020H1 and the cumulative growth that was expected before the onset of the pandemic for the same period—based on the IMF World Economic Outlook 2020 January forecast for 2020H1; Output performance 2 is the difference in cumulative real GDP growth between the first half of 2020 (2020H1) and the first half of 2019 (2019H1). t-statistic reported in parentheses. In bold those regressors that can be considered “robust”—that is, with a t-value in absolute value greater than 1. * p<0.05, ** p<0.01, *** p<0.001.

IV. Robust Drivers

Scatter plots and estimates based on a few covariates suggest that several factors are associated with output losses. But which factors are robust? To answer this question, we report WALS estimates of the most robust drivers for the two alternative measures of output loss mentioned earlier. To remind, a regressor is considered robust if the t-statistic in absolute value is larger than 1—broadly speaking, this corresponds to a statistically significant increase in the adjusted R2 due to the inclusion of that variable.

The results confirm the associations highlighted earlier. Output performance is negatively related to: more stringent containment measures; higher deaths rates; a larger tourism share; less stringent credit market regulation; higher pre-crisis growth; and more democratic political regimes. Lower fiscal stimulus and higher social fractionalization are negatively correlated with at least one measure of output performance (Table 7).

Table 7.

Robust Drivers of Output Performance Across Countries—Wals

article image
Note: t-statistic reported in the table. In bold those regressors that can be considered “robust”. Estimates based on equation (1).

In Figure 8, we present the effect on output performance from moving from the 25th to the 75th percentile of each variable’s distribution—that is, we multiply the WALS coefficient by the inter-quartile range. The results show that GDP per capita is quantitatively the largest player in driving output loss: a country at the 25th percentile of the GDP per capita distribution (such as Bangladesh) has, on average, a 7 percentage point lower-than-expected output growth than a country at the 75th percentile (such as Portugal). The next two positions in the ranking are containment measures and deaths, with similar magnitudes: an increase from the 25th to the 75th percentile of their distributions is associated with an increase in output losses of 4½-5 percentage points. While these results confirm the large economic cost associated with containment, they also highlight the close relation between “saving lives” and “saving the economy.”

Figure 8.
Figure 8.

Robust Drivers of Output Performance Across Countries, Magnitude of the Effects

Citation: IMF Working Papers 2021, 018; 10.5089/9781513567013.001.A001

Note. Output performance 1 is the difference between the observed cumulative real GDP growth in 2020H1 and the cumulative growth that was expected before the onset of the pandemic for the same period—based on the IMF World Economic Outlook 2020 January forecast for 2020H1; Output performance 2 is the difference in cumulative real GDP growth between the first half of 2020 (2020H1) and the first half of 2019 (2019H1);. The chart shows the differential effect on output performance moving the level of the variable from the 25th percentile to the 75th percentile of its distribution, based on the coefficients of the variables that are robust in column (I-II) of Table 7. – (+) denotes a negative (positive) effect on output. Estimates based on equation (1).

Two other economically important drivers of resilience are tourism dependence and the democracy score. Countries with a large share of tourism—notably Caribbean and Pacific islands—experience a 3½ percentage point smaller than expected output growth along the interquartile range. In contrast, countries that score low in the Polity IV democracy index (such as China and Vietnam) are associated with significantly higher resilience (reductions of about 4 percentage points in output growth losses) than more democratic countries. As suggested by Cepaluni et al. (2020), this may reflect better enforcement of containment measures and compliance with social distancing as well as faster interventions to pandemic outbreaks.

A. Robustness checks

Outliers

Do outlier observations influence the results? To check, we winsorize the upper and lower 5 percentiles of the distribution of the dependent variables, and show in Table 8 (Figure A7) results to be broadly in line with the baseline in Table 7.9 For the first measure of output performance (column I), we confirm the same robust drivers, but higher fiscal stimulus becomes a robust driver of output performance. The results are also similar for the second output performance measure (column II), except that higher government debt is also robustly associated with lower output performance.

Table 8.

Robust Drivers of Output Performance Across Countries—Controlling for Outliers

article image
Note: t-statistic reported in the table. In bold those regressors that can be considered “robust”. Estimates based on equation (1).

BMA

As discussed earlier, WALS is theoretically superior to BMA because, while BMA uses a Gaussian prior for the auxiliary parameters, WALS uses a Laplace distribution which reduces the risk of the prior overly influencing the final estimates. WALS is also practically superior because the space over which model selection is performed increases linearly rather than exponentially with size. At the same time, a key advantage of BMA is the larger number of models considered. To check robustness of our results, we repeat the analysis using BMA and consider the entire model space (230 models). In Table 9 we report the posterior inclusion probability of each regressor—that is, the probability that a variable belongs to the true model. Similar to the baseline, the variables with the highest posterior probabilities are containment stringency, tourism and deaths per capita. Other variables that would enter in at least 10 percent of the 230 (1,073,741,824) models are typically those found to be robust in WALS such as, pre-crisis growth, credit market regulation, the level of GDP per capita and democracy. In addition, we find that the share of the elderly seems to be robust in BMA—with a 35 percent posterior probability to enter in all models for the second output loss measure.

Table 9.

Robust Drivers of Output Performance Across Countries—Bma

article image
Note: posterior-inclusion-probability reported in the table. In bold those regressors with a posterior inclusion probability above 0.50; in italic those with a posterior inclusion probability above 0.1. Estimates based on equation (1). – (+) denotes a negative (positive) effect on output.

Additional determinants

As mentioned earlier, the selection of the variables is partly dictated by data availability. To check robustness to the inclusion of additional factors, we expanded the set of controls to include: additional measures of regulation pertaining to trade, the current and capital accounts, and indicators of rule of law; the share of non-performing loans (NPL); a measure of poverty and the amount of central bank reserves in percent of imports. The results with this larger set of (36) controls, based on a more limited sample of observations (and 11 degrees of freedom), confirm the baseline findings that GDP per capita, containment measures, deaths, and tourism are the most robust determinants of output performance (Table 9 and Figure A8). Additionally, we also find that countries with higher rule of law, higher debt-to-GDP ratios and smaller fiscal stimulus suffer higher output losses. In contrast, democracy is no longer significant, reflecting that in this restricted sample most countries have a similar democracy score and the democracy variable has a high negatively correlation with the rule of law indicator.

Alternative period

For many countries the peak of the economic crisis has been observed in the second quarter of 2020. It is useful to check, therefore, the validity of our results when considering only economic performance in the second quarter alone. Results reported in Table 11 (Figure A9) confirm our previous findings, and also suggest that countries with higher debt-to-GDP ratio, larger current account surpluses and more flexible exchange rates tend to experience weaker economic performance.

Table 10.

ROBUST DRIVERS OF OUTPUT PERFORMANCE ACROSS COUNTRIES—ADDITIONAL COVARIATES

article image
Note: t-statistic reported in the table. In bold those regressors that can be considered “robust”. Estimates based on equation (1).
Table 11.

Robust Drivers of Output Performance Across Countries—Using Only Q2 Data

article image
Note: t-statistic reported in the table. In bold those regressors that can be considered “robust”. Estimates based on equation (1).

B. Cross-Country Heterogeneity: Mediating Channels

The results suggest that differences in GDP per capita are the most robust and important drivers of cross-country differences in output loss. What drives this result? Potential mediating channels could be the higher economic costs of health crises and less effective macroeconomic stimulus in poorer countries. To shed light on this, we extend the specification to include interaction terms between three alternative measures of economic development and deaths per capita, the stringency of containment measures, and the monetary and fiscal policy response variables. The measures of economic development we consider are: (i) the level of GDP per capita; (ii) a dummy which takes value 1 for countries with a level of GDP per capita above the sample average; (iii) a dummy which takes value 1 for advanced economies (IMF definition). The WALS results in Table 12 (Figure A10) highlight mediating channels consistent with our priors. First, output costs associated with containment measures and deaths are larger in lower-income countries, probably because of more limited social safety nets and larger shares of financially-constrained households and firms. Second, monetary stimulus—specially liquidity provisions—has been less effective in poorer countries, consistent with the literature on the more limited transmission of monetary policy in emerging market and developing economies. Third, there is some evidence that effectiveness of fiscal stimulus is lower in poorer countries.

Table 12.

Robust Drivers of Output Performance Across Countries—Interaction with Income Level

article image
Note: t-statistic reported in the table. In bold those regressors that can be considered robust”. Estimates bases on equation (1). Dummy 1 uses the average of GDP per capita as reference, 1 denotes above average, otherwise 0; Dummy 2 uses the definition of income level in the World Economic Outlook, 1 is advanced economies, otherwise 0; OP stands for output performance.

V. Conclusion

This paper has explored the factors that drive heterogeneity of output losses across countries in the first phase of the Covid-19 recession. Using model-averaging techniques to address model uncertainty, we find that countries experiencing smaller output losses are those with: higher GDP per capita; less stringent containment measures; smaller number of deaths per capita; smaller tourism sectors; less flexible credit markets; lower pre-crisis growth; higher fiscal stimulus; less social (ethnic and religious) fractionalization; and less democratic regimes. Among these factors, the level of GDP per capita has the largest quantitative effect on resilience among the robust factors: a country at the 75th percentile of the GDP per capita distribution (such as Portugal) has, on average, a 7 percentage point smaller output loss than a country at the 25th percentile (such as Bangladesh). Our analysis suggests two key reasons why less-developed economies may be less resilient: the higher economic costs of containment measures—probably because of more limited social safety nets—and less effective fiscal and monetary policy stimulus.

We also find that death rates and containment stringency have similar effects on resilience, which suggests that rollback of containment should be implemented in a way that minimizes health risks. This implies relaxing containment only when new infections are declining and implementing strong testing and contact tracing policies. Second, fiscal stimulus has helped to reduce economic losses, underscoring that premature withdrawal of such stimulus is self-defeating. Our results indicate that monetary stimulus enhanced resilience more in advanced than non-advanced economies, underscoring the criticality of improving transmission in the latter. Third, reflecting that this is foremost a health crisis, the economic fallout has been particularly acute in high-contact sectors such as tourism and retail. This underscores the need for targeted rather than generalized support, particularly in the later stages of the crisis.

Our findings also speak to the more general literature on resilience. In contrast to studies on the GFC, we do not find that trade and financial openness have been important drivers of output loss during the pandemic. Whether such factors will play a key role going forward, including during the recovery, is an important question for future research.

References

  • Aaronson, D., Rissman, E.R. and Sullivan, D. G., 2004. Can sectoral reallocation explain the jobless recovery?. Economic Perspectives-Federal Reserve Bank of Chicago, 28, pp. 3649.

    • Search Google Scholar
    • Export Citation
  • Ahir, Hites, Bloom, N., and Furceri, D., 2018. “World Uncertainty Index,” Stanford, mimeo.

  • Aizenman, J., and Pasricha, G. K., 2010. Selective swap arrangements and the global financial crisis: Analysis and interpretation. International Review of Economics & Finance, 19(3), pp. 353365.

    • Search Google Scholar
    • Export Citation
  • Alesina, A., Devleeschauwer, A., Easterly, W., Kurlat, S. and Wacziarg, R., 2003. Fractionalization. Journal of Economic growth, 8(2), pp. 155194.

    • Search Google Scholar
    • Export Citation
  • Alesina, A., Spolaore, E., and Wacziarg, R., 2005. “Trade, Growth and the Size of Countries,” Handbook of Economic Growth, ed. by P. Aghion, and S. Durlauf, 1 (23), pp.14991542. Elsevier.

    • Search Google Scholar
    • Export Citation
  • Alesina, A. F., Furceri, D., Ostry, J. D., Papageorgiou, C. and Quinn, D. P., 2020a. “Structural Reforms and Elections: Evidence from a World-Wide New Dataset (No. w26720).” National Bureau of Economic Research.

    • Search Google Scholar
    • Export Citation
  • Artha, I. K. D. S., and de Haan, J., 2011. Labor market flexibility and the impact of the financial crisis. Kyklos, 64(2), pp. 213230.

  • Babecký, J., Havranek, T., Mateju, J., Rusnák, M., Smidkova, K., and Vasicek, B., 2012. Banking, debt and currency crises: early warning indicators for developed countries. No 1485, Working Paper Series, European Central Bank.

    • Search Google Scholar
    • Export Citation
  • Babecký, J., Havránek, T., Matějů, J., Rusnák, M., Šmídková, K., and Vasiček, B., 2013. Leading indicators of crisis incidence: Evidence from developed countries. Journal of International Money and Finance, 35, pp. 119.

    • Search Google Scholar
    • Export Citation
  • Baek, I. and Bouzinov, M., 2020. Does democratic progress deter terrorist incidents? European Journal of Political Economy, pp. 101951.

  • Baker, Scott R., Farrokhnia, R. A., Meyer, S., Pagel, M., and Yannelis, C., 2020. “How Does Household Spending Respond to an Epidemic? Consumption during the 2020 COVID-19 Pandemic.” NBER Working Paper 26949, National Bureau of Economic Research.

    • Search Google Scholar
    • Export Citation
  • Bayer, C., Born, B., Luetticke, R., and Müller, G. J., 2020. The Coronavirus Stimulus Package: How large is the transfer multiplier? CEPR Discussion Papers 14600, C.E.P.R. Discussion Papers.

    • Search Google Scholar
    • Export Citation
  • Beck, T. and Demirguc-Kunt, A., 2009. Financial institutions and markets across countries and over time-data and analysis. The World Bank.

    • Search Google Scholar
    • Export Citation
  • Béland, L. P., Brodeur, A. and Wright, T., 2020. The short-term economic consequences of Covid-19: exposure to disease, remote work and government response.

    • Search Google Scholar
    • Export Citation
  • Benmelech Efraim and Nitzan Tzur-Ilan, 2020. “The Determinants of Fiscal and Monetary Policies During the Covid-19 Crisis,” NBER Working Papers 27461.

    • Search Google Scholar
    • Export Citation
  • Berkmen, S. P., Gelos, G., Rennhack, R., and Walsh, J. P., 2012. The global financial crisis: Explaining cross-country differences in the output impact. Journal of International Money and Finance, 31(1), pp. 4259.

    • Search Google Scholar
    • Export Citation
  • Bernal-Verdugo, Lorenzo E., Furceri, D., and Guillaume, D., 2013. “Banking crises, labor reforms, and unemployment,” Journal of Comparative Economics, Elsevier, 41(4), pp. 12021219.

    • Search Google Scholar
    • Export Citation
  • Bhagwati, J., 1998. The capital myth: the difference between trade in widgets and dollars. Foreign Affairs, pp. 712.

  • Blanchard, O. J., Faruqee, H., Das, M., Forbes, K. J., and Tesar, L. L., 2010. The initial impact of the crisis on emerging market countries [with comments and discussion. Brookings papers on economic activity, pp. 263323.

    • Search Google Scholar
    • Export Citation
  • Bluedorn, J., Aiyar, S., Duval, R., Furceri, D., Garcia-Macia, D., Ji, Y., Malacrino, D., Qu, H., Siminitz, J., and Zdzienicka, A., 2019. “Strengthening the Euro Area; The Role of National Structural Reforms in Building Resilience,” IMF Staff Discussion Notes, 2019/005.

    • Search Google Scholar
    • Export Citation
  • Brandao-Marques, L., Gelos, G., Harjes, T., Sahay, R. and Xue, Y., 2020. Monetary policy transmission in emerging markets and developing economies. IMF Working Papers 2020/035, International Monetary Fund.

    • Search Google Scholar
    • Export Citation
  • Brock, W. A., and Durlauf, S. N., 2001. “What have we learned from a decade of empirical research on growth? Growth empirics and reality,” The World Bank Economic Review, 15(2), pp. 229272.

    • Search Google Scholar
    • Export Citation
  • Cacciatore, M., Duval, R., Furceri, D., and Zdzienicka, A., 2020. “Fiscal Multipliers and Job-Protection Regulation,” European Economic Review, forthcoming.

    • Search Google Scholar
    • Export Citation
  • Caprio, G. and Honohan, P., 2002. Banking policy and macroeconomic stability: An exploration. The World Bank.

  • Caprio Jr., G., D’Apice, V., Ferri, G., and Puopolo, G. W., 2014. Macro-financial determinants of the great financial crisis: Implications for financial regulation. Journal of Banking & Finance, 44, 114129.

    • Search Google Scholar
    • Export Citation
  • Carvalho, V. M., Hansen, S., Ortiz, A., García, J. R., Rodrigo, T., Mora, S. R., and Ruiz, J., 2020. “Tracking the COVID-19 Crisis with High-Resolution Transaction Data.” CEPR Discussion Paper 14642, Centre for Economic Policy Research.

    • Search Google Scholar
    • Export Citation
  • Cecchetti, S. G., King, M., and Yetman, J., 2011. Weathering the financial crisis: good policy or good luck? BIS Working Papers 351, Bank for International Settlements.

    • Search Google Scholar
    • Export Citation
  • Cepaluni, G., Dorsch, M., and Branyiczki, R., 2020. Political Regimes and Deaths in the Early Stages of the COVID-19 Pandemic. Social Science Research Network 3586767.

    • Search Google Scholar
    • Export Citation
  • Chernozhukov, V., Kasahara, H. and Schrimpf, P., 2020. Mask mandates and other lockdown policies reduced the spread of COVID-19 in the US.

    • Search Google Scholar
    • Export Citation
  • Chronopoulos, D. K., Lukas, M. and Wilson, J. O., 2020. Consumer Spending Responses to the COVID-19 Pandemic: An Assessment of Great Britain. Social Science Research Network 3586723.

    • Search Google Scholar
    • Export Citation
  • Claessens, S., Tong, H., and Wei, S. J., 2012. From the financial crisis to the real economy: Using firm-level data to identify transmission channels. Journal of International Economics, 88(2), pp. 375387.

    • Search Google Scholar
    • Export Citation
  • Coibion, O., Gorodnichenko, Y. and Weber, M., 2020. The cost of the covid-19 crisis: Lockdowns, macroeconomic expectations, and consumer spending No. w27141. National Bureau of Economic Research.

    • Search Google Scholar
    • Export Citation
  • Correia, S., Luck, S., and Verner, E., 1918. Pandemics Depress the Economy, Public Health Interventions Do Not: Evidence from the 1918 Flu. Public Health Interventions do not: Evidence from the 1918.

    • Search Google Scholar
    • Export Citation
  • Cuaresma, J. C., and Feldkircher, M., 2012. Drivers of output loss during the 2008–09 crisis: a focus on emerging Europe. Focus on European Economic Integration, 2(12), pp. 4664.

    • Search Google Scholar
    • Export Citation
  • Deb, P., Furceri, D., Ostry, J. D., and Tawk, N., 2020a. The economic effects of Covid-19 containment measures. IMF Working Paper No. 20/158.

    • Search Google Scholar
    • Export Citation
  • Deb, P., Furceri, D., Ostry, J. D., and Tawk, N., 2020b. The effect of containment measures on the COVID-19 pandemic. IMF Working Paper No. 20/159.

    • Search Google Scholar
    • Export Citation
  • Demir, B., and Javorcik, B., 2020. Trade finance matters: evidence from the COVID-19 crisis. Oxford Review of Economic Policy, 36(Supplement_1), S397S408.

    • Search Google Scholar
    • Export Citation
  • Demirguc-Kunt, A., Lokshin, M. and Torre, I., 2020. The sooner, the better: The early economic impact of non-pharmaceutical interventions during the COVID-19 pandemic. World Bank Policy Research Working Paper, 9257.

    • Search Google Scholar
    • Export Citation
  • Devereux, J., and Dwyer, G. P., 2016. What determines output losses after banking crises?. Journal of International Money and Finance, 69, pp. 6994.

    • Search Google Scholar
    • Export Citation
  • Devereux, M. B., Young, E. R., and Yu, C., 2015. A new dilemma: Capital controls and monetary policy in sudden stop economies, No. w21791. National Bureau of Economic Research.

    • Search Google Scholar
    • Export Citation
  • Didier, T., Hevia, C., and Schmukler, S. L., 2012. How resilient and countercyclical were emerging economies during the global financial crisis?. Journal of International Money and Finance, 31(8), pp. 20522077.

    • Search Google Scholar
    • Export Citation
  • Duval, R., Furceri, D., and Miethe, J., 2020. Robust political economy correlates of major product and labor market reforms in advanced economies: Evidence from BAMLE for Logit Models. Journal of Applied Econometrics.

    • Search Google Scholar
    • Export Citation
  • Eichhorst, W., Feil, M. T., and Marx, P., 2010. Crisis, what crisis? Patterns of adaptation in European labor markets. Applied Economics Quarterly (formerly: Konjunkturpolitik), Duncker & Humblot, Berlin, Vol. 61(Supplement), pages 2964.

    • Search Google Scholar
    • Export Citation
  • Faria-e-Castro, M., 2020. Fiscal policy during a pandemic. FRB St. Louis Working Paper, (2020–006).

  • Feldkircher, M., 2014. The determinants of vulnerability to the global financial crisis 2008 to 2009: Credit growth and other sources of risk. Journal of international Money and Finance, 43, pp. 1949.

    • Search Google Scholar
    • Export Citation
  • Fernandez, C., Ley, E., and Steel, M. F., 2001a. Benchmark priors for Bayesian model averaging. Journal of Econometrics, 100(2), pp. 381427.

    • Search Google Scholar
    • Export Citation
  • Fernandez, C, Ley, E., and Steel, M. F., 2001b. Model uncertainty in cross‐country growth regressions. Journal of applied Econometrics, 16(5), pp. 563576.

    • Search Google Scholar
    • Export Citation
  • Fornaro, L. and Wolf, M., 2020. Covid-19 coronavirus and macroeconomic policy. CREI/UPF and University of Vienna

  • Frankel, J. A., 2011. “Are Bilateral Remittances Countercyclical?,” Open Economies Review, 22(1), pp. 116.

  • Frankel, J. A. and Saravelos, G., 2010. Are leading indicators of financial crises useful for assessing country vulnerability? Evidence from the 2008–09 global crisis (No. w16047). National Bureau of Economic Research.

    • Search Google Scholar
    • Export Citation
  • Frankel, J, and Saravelos, G., 2012. Can leading indicators assess country vulnerability? Evidence from the 2008–09 global financial crisis. Journal of International Economics, 87(2), pp. 216231.

    • Search Google Scholar
    • Export Citation
  • Furceri, D. and Ostry, J., 2019. Robust determinants of income inequality. Oxford Review of Economic Policy, 35(3), pp. 490517.

  • Furceri, D., Celik, S. K., Jalles, J. T., and Koloskova, K., 2020. Recessions and Total Factor Productivity: Evidence from Sectoral data. Economic modelling.

    • Search Google Scholar
    • Export Citation
  • Furceri, D. and Karras, G., 2007. “Country size and business cycle volatility: Scale really matters,” Journal of the Japanese and International Economies, 21(4), pp. 424434.

    • Search Google Scholar
    • Export Citation
  • Ghosh, A. R., Ostry, J., and Qureshi, M. S., 2015. “Exchange Rate Management and Crisis Susceptibility: A Reassessment.” IMF Economic Review, 63(1), pp. 238276.

    • Search Google Scholar
    • Export Citation
  • Ghosh, A. R., Ostry, J., and Tsangarides, C. G., 2011. “Exchange Rate Regimes and the Stability of the International Monetary System,” IMF Occasional Paper 270.

    • Search Google Scholar
    • Export Citation
  • Ghosh, A. R., Ostry, J., and Wolf, H., 1997. “Does the Nominal Exchange Rate Regime Matter?NBER Working Paper 5874.

  • Giannone, D., Lenza, M., and Reichlin, L., 2011. “Market freedom and the global recession,” IMF Economic Review, 59(1), pp. 111135.

    • Search Google Scholar
    • Export Citation
  • Gopinath, G., 2020. Global liquidity trap requires a big fiscal response. Viewed 10 November 2020, <https://www.ft.com/content/2e1c0555-d65b-48d1–9af3–825d187eec58>.

    • Search Google Scholar
    • Export Citation
  • Groot, S. P., Möhlmann, J. L., Garretsen, J. H., and de Groot, H. L., 2011. The crisis sensitivity of European countries and regions: stylized facts and spatial heterogeneity. Cambridge Journal of Regions, Economy and Society, 4(3), pp. 437456.

    • Search Google Scholar
    • Export Citation
  • Gupta, S., Montenovo, L., Nguyen, T. D., Rojas, F. L., Schmutte, I. M., Simon, K. I., Weinberg, B.A. and Wing, C., 2020. Effects of social distancing policy on labor market outcomes No. w27280. National Bureau of Economic Research.

    • Search Google Scholar
    • Export Citation
  • Hasell, J., 2020. Which countries have protected both health and the economy in the pandemic. Our World in Data, 1.

  • Ho, T. K., 2010. Looking for a Needle in a Haystack: Revisiting the Cross-Country Causes of the 2008–09 Crisis. Economica.

  • Ilzetzki, E., Mendoza, E. G. and Végh, C. A., 2013. How big (small?) are fiscal multipliers?. Journal of monetary economics, 60(2), pp. 239254.

    • Search Google Scholar
    • Export Citation
  • International Monetary Fund (IMF). 2020. World Economic Outlook: A Long and Difficult Ascent. Washington, DC, October.

  • Ioannidis, J. P., Axfors, C., and Contopoulos-Ioannidis, D. G., 2020. Population-level COVID-19 mortality risk for non-elderly individuals overall and for non-elderly individuals without underlying diseases in pandemic epicenters. medRxiv.

    • Search Google Scholar
    • Export Citation
  • Jinjarak, Y., Ahmed, R., Nair-Desai, S., Xin, W., and Aizenman, J., 2020. Pandemic shocks and fiscal-monetary policies in the Eurozone: COVID-19 dominance during January-June 2020 No. w27451. National Bureau of Economic Research.

    • Search Google Scholar
    • Export Citation
  • Kavanagh, M. M., and Singh, R., 2020. Democracy, Capacity, and Coercion in Pandemic Response—COVID 19 in Comparative Political Perspective. Journal of Health Politics, Policy and Law.

    • Search Google Scholar
    • Export Citation
  • Kim, J.I. and Ostry, J., 2018. “Boosting Fiscal Space: The Role of GDP-Linked Debt and Longer Maturities,” IMF Discussion Paper 18/04. Forthcoming Economic Policy.

    • Search Google Scholar
    • Export Citation
  • Kopelman, J.L. and Rosen, H. S., 2016. Are Public Sector Jobs Recession-Proof? Were They Ever?. Public Finance Review, 44(3), pp. 370396.