The Effects of Fiscal Measures During COVID-19
Author:
Mr. Pragyan Deb
Search for other papers by Mr. Pragyan Deb in
Current site
Google Scholar
Close
,
Davide Furceri https://isni.org/isni/0000000404811396 International Monetary Fund

Search for other papers by Davide Furceri in
Current site
Google Scholar
Close
,
Mr. Jonathan David Ostry
Search for other papers by Mr. Jonathan David Ostry in
Current site
Google Scholar
Close
,
Nour Tawk
Search for other papers by Nour Tawk in
Current site
Google Scholar
Close
, and
Naihan Yang
Search for other papers by Naihan Yang in
Current site
Google Scholar
Close

This paper empirically examines the effects of fiscal policy measures during the COVID-19 pandemic, using a novel database of daily fiscal policy announcements—classified by type of fiscal measure—and high-frequency economic indicators for 52 countries from January 1 to December 31, 2020. The results suggest that fiscal policy announcements have been effective in stimulating economic activity, boosting confidence, and reducing unemployment, but their effect varies by type of measure and country characteristics. Emergency lifeline measures (which form the bulk of below-the-line measures) are more effective when containment policies are stringent, providing cashflow support to firms and households. Demand-support measures (which comprise most of above-the-line measures) are more effective when containment measures are relaxed.

Abstract

This paper empirically examines the effects of fiscal policy measures during the COVID-19 pandemic, using a novel database of daily fiscal policy announcements—classified by type of fiscal measure—and high-frequency economic indicators for 52 countries from January 1 to December 31, 2020. The results suggest that fiscal policy announcements have been effective in stimulating economic activity, boosting confidence, and reducing unemployment, but their effect varies by type of measure and country characteristics. Emergency lifeline measures (which form the bulk of below-the-line measures) are more effective when containment policies are stringent, providing cashflow support to firms and households. Demand-support measures (which comprise most of above-the-line measures) are more effective when containment measures are relaxed.

I. Introduction

Beginning in early 2020, countries worldwide launched wide-scale fiscal support measures to mitigate the unprecedented output losses from the COVID-19 pandemic (IMF, 2020). While it is understood that fiscal policies deployed during the pandemic provided timely and critical support to households, businesses, and economies overall, the effect of these measures on economic activity, unemployment, and consumer, business, and investor confidence, remains an open question. Estimating the measures’ economic effects is of paramount importance for policymakers, as they assess how much support is needed in a context of dwindling fiscal space.

A key challenge is identifying fiscal shocks in the data. Previous research used narrative or structural time series (SVAR) methods to isolate unanticipated, exogenous innovations to government spending or revenue. A problem with these approaches in our context is that they have typically been applied to annual or quarterly data, and are thus less suited if the question is to assess the effect of fiscal policy during COVID-19. Our approach instead relies on daily data of announced fiscal measures and combines the high-frequency identification (HFI) method of Gertler and Karadi (2015) applied to fiscal announcements (Auerbach and Gorodnichenko, 2016) with the narrative approach of Romer and Romer (2010), Ramey (2011), and Alesina et al. (2014).

Our paper has three main goals. The first is to quantify the average effect of fiscal policy announcements on economic activity across countries. For this purpose, we assemble a novel daily database of announcements regarding the fiscal policy interventions deployed across 52 countries throughout 2020.1 We collect information on the date of announcement and implementation, the authorizing institution (e .g. , fiscal authority, monetary authority, or a regulatory authority), the policy tool (e.g., credit guarantee vs. public investment), and the magnitude of the fiscal measure. At a daily frequency it is unlikely that fiscal announcements react to developments in the economy, which limits concerns about reverse causality. To further address endogeneity concerns, we purge the daily announcements from lagged daily measures of activity indicators that have been recently used to track the economic effects of the COVID-19 crisis—specifically Nitrogen Dioxide (NO2) emissions, international and domestic flights, mobility indicators, and daily financial variables that can help capture expectations regarding future economic activity.

Our results suggest that fiscal announcements are associated with a persistent increase in stock market indicators and, consistent with theory and evidence in Auerbach and Gorodnichenko (2013), with an appreciation of the domestic currency vis-à-vis the US dollar. At weekly frequencies, we also find that fiscal announcements are followed by an increase in the OECD economic tracker indicator of economic activity.

We then turn to effect of fiscal measures on more standard indicators of economic activity—industrial production indices, manufacturing Purchasing Managers’ Indices (PMIs), unemployment rates, the OECD’s Composite Leading Indicators (CLI) for confidence, and sovereign credit default swap (CDS) spreads—available at the monthly frequency. For this purpose, and following Gertler and Karadi (2015), we aggregate the daily shocks into monthly average shocks. The results suggest that fiscal policy announcements implemented during COVID-19 have had, on average, a significant effect on economic activity. In particular, we find that a fiscal shock of one percent of GDP increases industrial production by 0 .25 percent. In terms of the effect on GDP, back-of-the envelope calculations suggests that this effect is equivalent to a fiscal multiplier of about 0.2 for the entire sample (on average). 2

The effect of fiscal announcements is also significant on PMI, unemployment, confidence indicators, and sovereign CDS spreads: a one percent of GDP fiscal announcement leads to an increase in the PMI by 0.38 percent, a reduction in the monthly unemployment rate by 0.06 percentage points, a boost of the OECD Composite Leading Indicators (CLI) by about 0.1 percent, and a reduction in the sovereign CDS spread by 0.05 basis points.

The second goal of the paper is to assess whether the effect of (different types of) f is cal measure varies across countries depending on structural characteristics (such as the level of development, trade openness, the level of public debt before the crisis) and across time depending on the severity of the pandemic and lockdown restrictions. The results on country characteristics are consistent with previous findings of the literature (e.g., Nickel and Tudyka, 2014; Ilzetzki et al. 2013) and suggest that effects are larger for the advanced economies in our sample and in countries with lower public debt going to the crisis. In contrast, probably because of limited crosscountry heterogeneity in trade openness and exchange rate regimes, we do not find statistically significant differential effects along these two dimensions.

The third goal of this paper is to examine whether the effect of fiscal stimulus measures varies depending on the type of measure used. For this, we follow the narrative approach (Romer and Romer, 2010; Ramey 2011; Alesina et al. 2014) and read the record of each of the policy measures implemented and classify them in: (i) demand support measures; (ii) emergency lifelines measures; (iii) above-the-line; and (iv) below-the-line measures. The results show that, on average, emergency lifelines measures—such as loans to firms and households, umbrella guarantees and equity injections (which form the bulk of below-the-line measures) —have been more effective in boosting economic activity during COVID-19 than demand-support measures—such as tax cuts or payment deferrals, cash transfers and unemployment insurance. However, these results mask significant heterogeneity depending on the state of pandemic “cycle”. In particular, emergency lifeline measures appear more effective when containment is high and social mobility is low. This is consistent with the idea that when supply constraints are high, emergency lifelines—s u ch as loans to firms or credit guarantees—provide much-needed liquidity support to firms and businesses so that they may continue to operate. On the other hand, demand-support measures appear more effective when containment measures are being eased and supply constraints are lower. In other words, measures such as public investment are more likely to feed into domestic demand when supply-side restrictions have been relaxed and spending opportunities are ample.

The rest of the paper is structured as follows. We briefly survey the literature in Section II. Section III describes the data and Section IV presents some stylized facts on the fiscal measures. Section V discusses the empirical methodology and the results. Section VI concludes.

II. Literature Review

The literature examining the effects of fiscal spending relies on two main approaches for identification. The first uses Structural Vector Autoregression (SVAR) techniques developed by Blanchard and Perotti (2002), where identification is achieved by assuming that government spending is pre-determined within the quarter, using a standard Cholesky decomposition with government spending ordered first. The second identification strategy uses the narrative approach (Romer and Romer 2010, Devries et al. 2011) to identify the size, timing, and objective of fiscal shocks. Uses of the narrative approach are abundant in the literature, from narrative estimates of tax multipliers (Mertens and Ravn 2013), to the effects of fiscal adjustments (Yang, Fidrmuc and Ghosh, 2015), and estimating the effects of fiscal consolidations (Alesina et al. 2017).

A key extension of the use of the narrative approach was by Ramey (2011), who discussed the importance of timing and demonstrated how failing to account for the anticipation effect could lead to differences in empirical results. Ramey argued that government shocks predetermined in a quarter are likely anticipated quarters in advance, thus nullifying the notion that the shock is unanticipated. She created a narrative variable of estimates of changes in government spending shocks from news reports to capture the “news” component of government spending shocks. Our paper contributes to this strand of literature and follows the narrative approach through the use of a daily database of fiscal policy measures where measures are identified both at the date of their announcement and implementation.

A second strand of the literature our paper contributes to uses high-frequency data to assess the impact of monetary and fiscal policy. On the monetary policy front, Gurkaynak, Sack and Swanson (2004) were of the first to investigate the effects of US monetary policy on asset prices using high-frequency event studies. They find that monetary policy statements have a greater impact on asset prices than changes in the federal funds rate target. Gertler and Karadi (2015) show that shocks identified using high frequency surprises around policy announcements as external instruments generate responses in output and inflation which are similar to monetary VAR analyses. Auerbach and Gorodnichenko (2016) use daily data on U.S. spending news (announced and actual payments) to examine their effect on the US dollar. They find that while the dollar immediately appreciates following announcements about government spending, actual payments do not affect the exchange rate. David, Guajardo and Yepez (2019) examine the effects of fiscal consolidation announcements on sovereign spreads in emerging market economies using daily data on consolidation announcements. They find that sovereign spreads decline significantly following news that austerity measures have been approved by the legislature in countries with high spreads, due to increased confidence effects. Beetsma et al. (2015) examine the effects of fiscal consolidations on private sector confidence by constructing a monthly dataset of fiscal consolidation announcements and confidence indicators. They find that consumer and producer confidence decline around announcements of consolidation measures. This paper adds to this strand of the literature by examining the effects of daily fiscal policy announcements on high-frequency economic activity indicators, including industrial production indices, manufacturing PMIs, unemployment rates, leading composite indicators, and sovereign CDS spreads.

As for the literature on the role of fiscal policy during the COVID-19 pandemic, Bennelech and Tzur-Ilan (2020) find that a country’s credit rating is the most important determinant of its fiscal spending during the pandemic, resulting in lower income countries not being able to deploy fiscal policy tools effectively during the crisis. Gourinchas et al. (2021) study the effects of fiscal policy during the COVID-19 pandemic at the firm, sector, country and global level. They fin d that at the firm level a lack of government support increased SME failure rates, and that fiscal policy was in general poorly targeted, reaching firms which did not need it. At the global level, fiscal policy helped offset about 8% of the economic downturn, and reduced the share of demand-constrained sectors, leading to a preservation of employment in these sectors.

As for research which explores fiscal multipliers across time and countries, Ramey and Zubairy (2014) extend their narrative variable using Jordà’s (2005) local projection method to examine the state dependence of fiscal multipliers. They find no evidence that multipliers differ according to the amount of slack in the economy, nor that government spending multipliers are higher when monetary policy is near the zero-lower bound. Nickel and Tudyka (2014) find that fiscal policy has larger multipliers f or countries with lower public debt-to-GDP ratios. Iltzetki et al. (2013) find that the output effect of an increase in government consumption is larger in advanced economies, economies operating under a predetermined exchange rate, and those with low debt-to-GDP levels. This paper contributes to this strand of the literature by being the first to empirically examine the effects of fiscal policy announcements on economic activity during COVID-19 in a large set of countries, and how these effects vary across countries and the severity of the pandemic. The granularity of our database also provides a first assessment of the effectiveness of various types of policy measures.

III. Data

We assemble a comprehensive database of announced and implemented fiscal stimulus measures, economic activity indicators, financial variables, and COVID-19 infections, and containment measures on a daily basis. The daily database is complemented with monthly indicators of economic activity and confidence: industrial production indices, manufacturing PMIs, confidence CLIs, unemployment rates, and sovereign CDS spreads. In this section, we describe the collection of the fiscal measures and the construction of fiscal shocks used in the empirical analysis, while information on the other variables is provided in the Data Annex.

The main source of our data is Yale’s COVID-19 Financial Response Tracker (CFRT), which documents key intervention measures in response to the COVID-19 pandemic by fiscal authorities, central banks, and other organizations throughout the world. 3 The database provides a breakdown of intervention measures by: i) proposal date (henceforth referred to as announcement date); ii) implementation date; iii) institutional authority under which measures are announced; iv) policy tools used; and v) magnitude of intervention method. We restrict the database to include only interventions which would be considered as fiscal support measures. The information from the Yale database is supplemented by, and cross-checked with, announcements provided by the IM F Policy Tracker, the OECD Country Policy Tracker and newspaper reports. In several instances, the numbers quoted by the dataset do not match what is reported by the IMF policy tracker and other sources. On such occasions, we extensively research the policy announcements through newspaper reports and announcement links in order to report our best assessment of the date, size, and type of announced measure. In addition, and in an effort to avoid double-counting, we remove entries which appear to be duplicates, as well as any entry which is reported as financing or concessional lending by the IMF or World Bank. More detailed measures towards building our database are included in the data annex.

We extract and read the narrative of each policy record and we group the policy measures identified tools into eight sub-categories: i) asset guarantees; ii) asset purchases; iii) capital injections; iv) credit guarantees; v) grants; vi) loans; vii) payment forbearances; and viii) tax relief measures. Following the IMF’s classification of fiscal measures, grants, payment forbearances (on tax payments and pension contributions) and tax relief interventions are recorded as above-the-line measures, as these measures are likely to be reflected in the fiscal balance and government debt. 4 All other interventions fall under the scope of below-the-line measures, i.e., measures which have no immediate or upfront effect on a government’s deficit.

We also classify each measure as demand-support and emergency lifeline.5 Demand-support measures are identified as those which boost demand and household of firm disposable income, and typically include cash transfers, unemployment insurance, wage subsidies, reduction or deferral of social security or tax payments, paid sick leave, etc. In addition, public investments or healthcare spending measures are considered demand-support measures, given that their role does not entail providing cashflow support. Meanwhile, emergency lifeline measures are identified as those which provide sustained cashflow support to households and firms. Such measures include loans and umbrella guarantees to firms and households, government provision of loans, equity injections, and other liquidity support measures. From an accounting standpoint, most emergency lifelines form the bulk of below-the-line measures. Meanwhile, demand-support measures mainly fall under the above-the-line measures. Further details on the identification of measures is explained in the Data Annex.6

The dataset covers daily fiscal policy measures for 52 countries for which data on industrial production and PMI manufacturing is available, from January 1 to December 31, 2020. All data is converted to USD and then scaled to a percent of a country’s 2019 GDP.

For the empirical analysis, we purge the daily announcements from lagged daily measures of activity that have been recently used to track the economic effects of the COVID-19 crisis such as Nitrogen Dioxide (NO2) emissions (Lin and Mc Elroy , 2011), international and domestic flights (Deb, Furceri, Ostry, and Tawk, 2020), and mobility indicators (IMF 2020; Maloney and Taskin, 2020), and daily financial variables that can help capture expectations regarding future economic activity—such as bilateral exchange rate and stock market indices. Controlling for this information allows us to purge fiscal announcements of any predictable components. The residual of the regression is our shock. For robustness, several fiscal shocks are created, based on different combinations of high-frequency indicators. The results are available upon request.

IV. Stylized Facts on Fiscal Announcements

The database of announced and implemented fiscal stimulus measures confirms the unprecedented scale of the fiscal packages announced in response to the pandemic. Across the sample of 52 countries, comprising 27 advanced and 25 emerging market and developing economies, the largest announced intervention by an advanced economy was about 35 percent of 2019 GDP (Japan) while for an EMDE it was about 15 percent of GDP (Indonesia). This is in line with the overall trend where measures deployed by advanced economies were larger than those deployed by EMDEs—the mean fiscal stimulus response for advanced economies was about 12 percent of GDP, compared to a 4 percent average response in an EMDE (Figure 1).

Figure 1.
Figure 1.

Fiscal Stimulus Packages by Magnitude

(percent of 2019 GDP)

Citation: IMF Working Papers 2021, 262; 10.5089/9781557754264.001.A001

Note: the figure above reports mean and maximum fiscal package size across 52 countries (27 advanced economies and 27 emerging market and developing economies) scaled as a percent of their respective 2019 GDP over the year 2020.

Turning to the timing of the measures, Figure 2 reports the announced fiscal measures on a daily basis. It shows that responses were most frequent and significant in the months of March and April, which represent the acute first stage of the COVID-19 pandemic for most countries. Advanced economies responded faster, with 10 advanced economies announcing stimulus measures on March 16, 2020. They were quickly followed by emerging market economies in the ensuing days. Responses continued well into June 2020, but then diminished petered out, owing to dwindling fiscal space but also the relaxation of containment measures towards the end of 2020Q2 in many countries, that prompted the authorities to use more targeted measures.

Figure 2.
Figure 2.

Fiscal Stimulus Measures During the COVID-19 Pandemic (percent of 2019 GDP)

Citation: IMF Working Papers 2021, 262; 10.5089/9781557754264.001.A001

Note: the figure above reports daily announced fiscal stimulus measures of 52 countries (27 advanced economies and 27 emerging market and developing economies) scaled as a percent of their respective 2019 GDP over the year 2020.

Figure 3 reports the evolution of emergency lifelines and demand-support measures. The rollout of emergency lifelines was also concentrated during March and April 2020—the most stringent phase of the lockdown—but they continued to be of use throughout the year. And while lifelines were used together with demand support measures in the initial phase of the lockdown, the latter—which includes measures to boost households and firms’ disposable income—seemed to pick up towards the end of the second quarter of 2020 (Figure 3), which coincides with the beginning of reopening in many countries.

Figure 3.
Figure 3.

Fiscal Stimulus Measures by Type During the COVID-19 Pandemic (percent of 2019 GDP)

Citation: IMF Working Papers 2021, 262; 10.5089/9781557754264.001.A001

Note: the figure above reports daily announced emergency lifeline and demand support measures of 52 countries (27 advanced economies and 25 emerging market and developing economies) scaled as a percent of their respective 2019 GDP over the year 2020.

Across policy tools, loans to households and businesses, as well as credit guarantees, made up the bulk of lifeline measures used during 2020. They were more frequently used in EMDEs. Grants (including cash transfers, wage subsidies, healthcare spending) and tax relief measures (ta x cuts, tax payment deferrals) made up most of demand support measures (Figure 4). In terms of magnitude, the largest fiscal policy tools in the case of AEs were credit guarantees, grants (including cash transfers, wage subsidies, healthcare spending), tax relief measures (tax cuts, tax payment deferrals) and loans. For EMDEs, grants, payment forbearances and loans were the largest three types of instruments used (Figure 5).

Figure 4.
Figure 4.

Fiscal Measures, by Policy Tool

(frequency of measure used)

Citation: IMF Working Papers 2021, 262; 10.5089/9781557754264.001.A001

Note: the figure above reports the frequency of fiscal policy tools used during 2020 by 27 AEs and 25 EMDEs.
Figure 5.
Figure 5.

Fiscal Measures, by Type and Magnitude

(percent of 2019 GDP)

Citation: IMF Working Papers 2021, 262; 10.5089/9781557754264.001.A001

Note: the figure above reports maximum fiscal policy tool size used during 2020 by 27 AEs and 25 EMDEs, scaled as a percent of their respective 2019 GDP over the year 2020.

Finally, about 70 percent of stimulus measures for both AEs and EMDEs were below-the-line measures, likely reflecting the importance of limited fiscal space in determining the size and nature of stimulus packages. Countries facing tighter constraints (measured by higher debt to GDP levels) relied more on below-the-line measures, while countries with relatively more space favored above the line measures (Figure 6).

Figure 6.
Figure 6.

Fiscal Measures by Type and Fiscal Space Considerations

Citation: IMF Working Papers 2021, 262; 10.5089/9781557754264.001.A001

Note: The charts depict the relationship between above-the-line and below-the-line measures announced, and debt-to-GDP levels. Fiscal measures and debt-to-GDP are reported as a percent of each country’s 2019 GDP.

V. Analysis

A. Effects of fiscal shocks at daily and weekly frequency

We follow Jordà (2005) to assess the dynamic effect of fiscal policy announcements at the daily frequency on bilateral exchange rates and stock prices, and at weekly frequencies on the OECD weekly tracker. The methodology (also used by Auerbach and Gorodnichenko (2013, 2016), Ramey and Zubairy (2018), and Alesina et al. (2019) among others) does not impose the dynamic restrictions embedded in vector autoregressions and is particularly suited to estimating nonlinearities in the dynamic response. The first regression we estimate is:

Si,t+h=ui+=0θh,Fi,t+Xi,tΓh+=1ψh,ΔSi,t+εi,t+h(1)

where Si,t+h represents the logarithm of the daily (weekly) economic indicator (stock prices, bilateral exchange rates, the OECD economic activity tracker) in country i observed at date t; Fi,t is the daily (weekly) fiscal announcements; ui are country-fixed effects to account for time-invariant country-specific characteristics; X is a vector of control variables which includes lags of the containment measures, the amount of number of COVID-19 infections and deaths in country i observed at date t, and country time-trends.7

Equation (1) is estimated for each day h=0,..,30 (15 for weekly indicators). Impulse response functions are computed using the estimated coefficients θh, and the 90 and 95 percent confidence bands associated with the estimated impulse-response functions are obtained using the estimated standard errors of the coefficients θh. Equation (1) is estimated using Driscoll-Kraay (1998) standard errors to account for cross-sectional and time dependence in the error term Figure 7 shows the estimated dynamic response of stock market prices and bilateral exchange rate to fiscal stimulus measures over the 30-day period following their announcement, together with the 90 and 95 percent confidence interval around the point estimate. The left-hand panel shows the cumulative response of bilateral exchange rates, while the right-hand panel shows the cumulative response of stock market prices.

Figure 7.
Figure 7.

Effect of fiscal announcements on bi lateral exchange rates and stock market prices

Citation: IMF Working Papers 2021, 262; 10.5089/9781557754264.001.A001

Note: Impulse response functions are estimated for a sample of 42 countries using daily data from January 1, 2020 to December 31, 2020. The graph shows the response and confidence bands at 90 and 95 percent. The horizontal axis shows the response x days after the announcement of fiscal stimulus measures. Estimates based on Si,t+h=ui+Σ=0θh,Fi,t+Xi,t'Γh+Σ=1ψh,ΔSi,t+εi,t+h where Si,t is the logarithm of the bilateral exchange rate or stock market prices in country i observed at date t. The model is estimated at each horizon h = 0,1, ...H, with a lag structure = 1,2 ...ℒ Fit denotes the announced fiscal measures as a share of GDP; X is a matrix of time varying control variables and country specific time trends. (For the bilateral exchange rate, increase denotes appreciation).

Consistent with the literature on the effect of fiscal announcements at the daily level (Auerbach and Gorodnichenko 2016), the results provide evidence that fiscal stimulus announcements in 2020 led to an appreciation of bilateral exchange rates of about 5 percent 30 days after their announcement. Similarly, fiscal stimulus announcements led to stock market cumulative gains of 40 percent logs following their announcement. At the weekly frequency, the results in Figure 8 also highlight that announced fiscal measures significantly boosted economic activity as proxied by the OECD weekly tracker.

Figure 8.
Figure 8.

Effect of fiscal announcements on the OECD weekly tracker

Citation: IMF Working Papers 2021, 262; 10.5089/9781557754264.001.A001

Note: Impulse response functions are estimated for a sample of 46 countries using weekly data from January 1, 2020 to December 31, 2020. The graph shows the response and confidence bands at 90 and 95 percent. The horizontal axis shows the response x days after the announcement of fiscal stimulus measures. Estimates based on Si,t+h=ui+Σ=0Lθh,Fi,t+Xi,t'Γh+Σ=1Lψh,ΔSi,t+εi,t+h where Si,t is the weekly economic tracker (in percent) in country i observed at date t. The model is estimated at each horizon h = 0,1, ...H, with a lag structure ℓ = 1,2 ... ℒ Fi,t denotes the announced fiscal measures as a share of GDP; X is a matrix of time varying control variables and country specific time trends.

B. Effects of fiscal shocks at the monthly frequency

We then turn to standard indictors of economic activity such as industrial production, unemployment, etc. The advantage of using these indicators it that it allows to infer about the fiscal multiplier associated to these measures and relate our results to those presented in the previous literature. The main drawback is the limited time dimension of the sample (12 observations per country) which prevents us to look into dynamic effects.

To estimate the (unconditional) impact of fiscal shocks on monthly economic activity indicator, we use the following equation, for an unbalanced sample of 52 economies from January to December 2020:

ΔYi,t=α+ui+βShocki,t+θXi,t+εi,t(2)

where Yi,t is the indicator of economic activity—as typically in the literature of the effectiveness of monetary policy, we use industrial production as preferred indictor, but we also consider alternative measures of economic activity such unemployment, manufacturing PMI, the OECD CLI and sovereign CDS spread—of country i in month t. ΔYi,t alternately denotes the growth rate of industrial production (year on year), the growth rate of the composite leading indicator (year-on-year), PMI at its level given that it is already a percent variable, the change in unemployment rate (defined as the difference between the current and previous month), and the level of CDS spreads. Shocksi,t denotes the fiscal shock at the monthly frequency—following Gertler and Karadi (2015), we aggregate the daily shocks into monthly average shocks to study the effect of fiscal measures on a range of economic activity indicators. In the baseline, we use the simple average of all the daily measures. The results are similar and not statistically different when we use a weighted average of daily shocks, with higher weights for shocks occurring at the beginning of the month. ui are country-fixed effects to account for time-invariant specific factors. Xi,t is a vector of control variables including containment measures, retail, mobility the number of COVID-19 infections and fatalities, monetary policy announcements and policy rate changes; and εi,t is the error term. Equation (2) is estimated using Driscoll-Kraay (1998) standard errors to account for cross-sectional and time dependence in the error term.

The results from the baseline monthly regressions are summarized in Table 1. They show that fiscal measures announced by countries during COVID-19 have been effective in boosting economic activity and confidence, as well as reducing unemployment and compressing CDS spreads. In particular, we find that a one percent of GDP fiscal shock has led, on average, to a 0.25 percent increase in the growth rate industrial production (year-on-year). To compare this result with the literature on fiscal multipliers, back of the envelope calculations based on the historical relationship between industrial production and GDP would imply a fiscal multiplier of about 0.2.8

Table 1.

Baseline effect of announced fiscal measures on economic activity

article image
Note: results reported are based on a sample of 52 countries using daily data from January 1, 2020 to December 30, 2020. Estimates are based on ΔYi,t = α +ui + βShocki,t + θXi,t + εi,t , Where ΔYit is the growth rate of industrial production, manufacturing PMI, the CLI, the unemployment rate, or sovereign CDS spreads of country i at month t; Shocki,t denotes the fiscal shock, ui are country-fixed effects to account for time-invariant specific factors; Xi,t is a vector of control variables, and εi,t is the error term. Driscoll-Kraay standard errors are presented in parentheses. *** ** * denote statistical significance at 1 percent, 5 percent, and 10 percent, respectively.

Beyond industrial production, we find that fiscal measures had economically and statistically significant effects on other measures of economic activity such as manufacturing PMI (0.39 percent increase) and confidence indicators such as OECDs composite leading indicator (about 0.1 percent). The effect is also significant on unemployment and sovereign CDS spreads, with a one percent of GDP fiscal shock leading to a decline in the monthly unemployment rate by 0.06 percentage point, and a narrowing of sovereign CDS spreads by about 0.05 bps. The reduction in the CDS spreads following the announcement of fiscal measures suggests that the measures were seen by the markets as positive news for debt sustainability. However, and in line with the findings of Beetsma et al. (2015) on the role of institutional quality in shaping the response of confidence indicators to fiscal consolidation, we find that the reduction in CDS spread is statistically significant only for advanced economies, where fiscal credibility is more established.9

We perform several robustness checks of our results. To further address endogeneity concerns related to the possibility that fiscal announcements are correlated with other factors and policy measures (such as monetary policy actions) affecting economic activity, we extend the set of controls to include monetary policy announcements and policy rate changes, COVID-19 fatalities and retail mobility, and expectations of economic activity based on monthly consensus GDP growth forecasts. These results are reported in Table 2 (for industrial production) and Annex Tables A14 (for other economic indicators) and are robust to these controls. We also replicated the results with the extended set of controls to include time fixed effects (Table 3, as well Annex Tables A14). As a further step, we also include the lagged dependent variable in each of our regressions (Table 4) to explicitly control for the time-dependence in the monthly shocks.

Table 2.

Robustness check: additional controls

article image
Note: results reported are based on a sample of 52 countries using daily data from January 1, 2020 to December 30, 2020. Estimates are based on ΔYi,t = α +ui + βShocki,t + θXi,t + εi,t , Where ΔYi,t is the growth rate of industrial production, manufacturing PMI, the CLI, the unemployment rate, or sovereign CDS spreads of country i at month t; Shocki,t denotes the fiscal shock, ui are country-fixed effects to account for time-invariant specific factors; Xi,t is a vector of control variables, and εi,t is the error term. Driscoll-Kraay standard errors are presented in parentheses. *** ** * denote statistical significance at 1 percent, 5 percent, and 10 percent, respectively.
Table 3.

Robustness check: with time fixed effects

article image
Note: results reported are based on a sample of 52 countries using daily data from January 1, 2020 to December 30, 2020. Estimates are based on ΔYi,t = α +ui + βShocki,t + θXi,t + εi,t , Where ΔYi,t is the growth rate of industrial production, manufacturing PMI, the CLI, the unemployment rate, or sovereign CDS spreads of country i at month t; Shocki,t denotes the fiscal shock, ui are country-fixed effects to account for time-invariant specific factors; Xi,t is a vector of control variables, and εi,t is the error term. Driscoll-Kraay standard errors are presented in parentheses. *** ** * denote statistical significance at 1 percent, 5 percent, and 10 percent, respectively.
Table 4.

Robustness check: with lagged dependent variable

article image
Note: results reported are based on a sample of 52 countries using daily data from January 1, 2020 to December 30, 2020. Estimates are based on ΔYi,t = α +ui + βShocki,t + θXi,t + εi,t , Where ΔYi,t is the growth rate of industrial production, manufacturing PMI, the CLI, the unemployment rate, or sovereign CDS spreads of country i at month t; Shocki,t denotes the fiscal shock, ui are country-fixed effects to account for time-invariant specific factors; Xi,t is a vector of control variables, and εi,t is the error term. Driscoll-Kraay standard errors are presented in parentheses. *** ** * denote statistical significance at 1 percent, 5 percent, and 10 percent, respectively.

In the baseline we aggregated daily fiscal shocks into monthly using simple average. To check the robustness of our results, we compute the fiscal shocks based on alternative weighting schemes. In particular, instead of computing shocks as a monthly average of the daily shocks (equal weights to all days of the month), we assign higher weights to announcements in earlier weeks of the month, with the rationale that shocks which take place earlier in the month may have larger effect, as there is more time for their impact to materialize on monthly economic indicators. Specifically, we create a fiscal shock where 45 percent of its weight is assigned for the announcements made in the first ten days of the month, followed by 35 percent of the weight being assigned for announcements the next 10 days, and 20 percent for the remainder of the days of the month. The results obtained with this alternative weighing scheme are presented in Table 5 and are similar to, and not statistically, different from the baseline.10

Table 5.

Robustness check: Alternative weights

article image
Note: results reported are based on a sample of 52 countries using daily data from January 1, 2020 to December 30, 2020. Estimates are based on ΔYi,t = α +ui + βShocki,t + θXi,t + εi,t , Where ΔYi,t is the growth rate of industrial production, manufacturing PMI, the CLI, the unemployment rate, or sovereign CDS spreads of country i at month t; Shocki,t denotes the fiscal shock, ui are country-fixed effects to account for time-invariant specific factors; Xi,t is a vector of control variables, and εi,t is the error term. Fiscal shocks are aggregated from a daily to monthly frequency by assigning higher weights to the earlier weeks of the month (as opposed to a simple monthly average). Driscoll-Kraay standard errors are presented in parentheses. ***, **, * denote statistical significance at 1 percent, 5 percent, and 10 percent, respectively.

The results presented so far are based on announcements as forward-looking agents are expected to respond to announcements rather than fiscal actions (Ramey 2011; Auerbach and Gorodnichenko, 2013). In the case of the COVID-19 crisis, however, fiscal measures were rolled out quite quickly, with many of implemented only few days they were announced.11 But given the presence of some announcements with longer lags for a few countries, and in order to test the robustness of our results, we repeated the analysis constructing fiscal shocks based on implementations dates. The results reported in Table 6 show are broadly similar to those obtained with announcements.12

Table 6.

Baseline effect of implemented fiscal measures on economic activity

article image
Note: results reported are based on a sample of 52 countries using daily data on implemented fiscal measures from January 1, 2020 to December 30, 2020. Estimates are based on ΔYi,t = α +ui + βShocki,t + θXi,t + εi,t , Where ΔYi,t is the growth rate of industrial production, manufacturing PMI, the CLI, the unemployment rate, or sovereign CDS spreads of country i at month t; Shocki,t denotes the fiscal shock, ui are country-fixed effects to account for time-invariant specific factors; Xi,t is a vector of control variables, and εi,t is the error term. Driscoll-Kraay standard errors are presented in parentheses. *** ** * denote statistical significance at 1 percent, 5 percent, and 10 percent, respectively.

C. Country Characteristics

We analyze whether the impact of fiscal shocks vary depending on country-specific conditions, such as countries’ s structural characteristics (e.g., the level of development, trade openness, the level of public debt before the crisis) and the severity of the pandemic and lockdown restrictions. For this purpose, we extend Equation (2) as follows:

ΔYi,t=α+ui+F(zit)βLShocki,t+(1F(zit))βHShocki,t+θXi,t+εi,t(3)
withF(zit)=expγzit(1expγzit),γ=1.5(4)

where z is a country-specific characteristic normalized to have zero mean and a unit variance. The approach allows us to exploit cross-country variation by looking at country-specific characteristics (such as debt levels), as well as within-country variation, by examining how fiscal shocks differ depending on a country’s containment level for instance. The weights assigned to each regime vary between 0 and 1 according to the weighting function F(zit), so that F(zit) can be interpreted as the probability of being in a given regime. For instance, F(zit) = 1 would correspond to a country with very high social mobility , while F(zit) = 0 would correspond to a country with very low social mobility. This approach is equivalent to the smooth transition model developed by Granger and Teravistra (1993).

Our results on how country characteristics affect the impact of fiscal shocks are summarized in Table 7. Consistent with most of the evidence in the literature, we find that fiscal shocks have been more effective in advanced economies (Ilzetkietal., 2011). Second , we fin d that fiscal shocks are likely to be much more effective in countries with lower public debt levels before the crisis, as the crowding out effects on private investment and consumption are typically larger in countries with high debt (Nickel and Tudyka 2014; Furceri and Zdzienicka, 2020). The results are confirmed for other economic indicators. In particular, and consistent with the view that fiscal actions were seen by the market as important stabilization tools, we see stronger and statistically significant decline in CDS spreads for countries with lower public debt. 13

Table 7.

Effect of announced fiscal shocks, by country characteristics

article image
Note: results reported are based on a sample of 52 countries using daily data from January 1, 2020 to December 30, 2020. Estimates are based on ΔYi,t = α + ui + F(zitLShocki,t + (l — F(zit))βHShocki,t + θXi,t + εi, , with F(zit)=expγzit(1expγzit), y = 1.5. z is a country-specific characteristic normalized to have zero mean and a unit variance. The weights assigned to each regime vary between 0 and 1 according to the weighting function F (zit), so that F(zit) can be interpreted as the probability of being in a given regime. ΔYi,t is the growth rate of industrial production, manufacturing PMI, the CLI, the unemployment rate, or sovereign CDS spreads of country i at month t; Shocki,t denotes the fiscal shock, ui are country-fixed effects to account for time-invariant specific factors; Xi,t is a vector of control variables, and εi,t is the error term. Driscoll-Kraay standard errors are presented in parentheses. *** ** * denote statistical significance at 1 percent, 5 percent, and 10 percent, respectively.

In contrast, and somewhat unsurprisingly, we do not find evidence that effects vary with the degree of trade openness. This is likely due to the limited cross-country variation in our sample since most of the countries are relatively open and with flexible exchange rates, and the fact that trade was generally anemic during the crisis.

D. Type of Fiscal Measures and the Pandemic Cycle

Finally, we extend the baseline specification to test whether the impact of fiscal measures varies by type of measure deployed. In particular, we extend Equation (2) as follows:

ΔYi,t=α+ui+βLLi,t+γDSi,t+θXi,t+εi,t(5)
ΔYi,t=α+ui+βBLi,t+γALi,t+θXi,t+εi,t(6)

where LLi,t is an emergency lifeline dummy interacted by the fiscal shock to represent an emergency lifelines package, while DSi,t is demand-support dummy interacted by the fiscal shock to represent a demand-support measures package. We follow Alesina, Favaro and Giavazzi (2014) who study the effects of tax-driven or expenditure-based austerity by distinguishing whether fiscal plans are expenditure based or taxed based depending on whether the largest component of a fiscal correction was an increase in taxes, or a decrease in expenditure.14 In that context, we set the emergency lifeline (demand-support) dummy as equal to 1 (zero) for any given country if emergency lifelines (demand-support measures) make up more than 50 percent of its entire fiscal stimulus package in 2020. Similarly, BLi,t is a below-the-line dummy interacted by the fiscal shock to represent a below-the-line package, while ALi,t is an above-the-line dummy interacted by the fiscal shock to represent a fiscal package of above-the-line measures.

While on average, fiscal measures seemed effective in boosting economic activity and confidence, their effectiveness may differ depending on the type of measure—above-the-line , below-the-line, emergency lifeline, or demand support measure—that was used. For example, it is reasonable to assume that demand support measures (the bulk of which are classified as above-the-line measures) are likely to be less effective during lockdowns when supply constraints are binding. In contrast, emergency lifelines that are aimed to reduce supply constraints, keep firms operating and workers employed are likely to be more effective during lockdowns.

The results summarized in Table 8 seems confirm this intuition, with emergency lifelines having a larger impact out than demand-support measures during our COVID-19 sample, on average. Similar results are obtained when we look at below- and above-the-line measures. Th e reason is that a big bulk of lifelines measures—which include loans, credit guarantees, provisions provided to firms, etc.—have been typically categorized as below-the-line measure (indeed the correlation between below-the-line measures and emergency lifelines is around 0.8).

Table 8.

Effect of announced fiscal shocks on industrial production, by type of measure

article image
Note: results reported are based on a sample of 52 countries using daily data from January 1 , 2020 to December 30, 2020 . Estimates are based on ∆Yi,t = α + ui + βLLi,t + γDSi,t + θXi,t+ εi,t Where ∆Yi,t is the growth rate of industrial production, manufacturing PMI, the CLI, the unemployment rate, or sovereign CDS spreads of country i at month t; ui are country-fixed effects to account for time-in variant specific factors; Xi,t is a vector of control variables, and εi,t is the error term. LLi,t is an emergen cy lifeline dummy interacted by the fiscal shock to represent an emergency lifelines package, while DSi,t is demand-support dummy interacted by the fiscal shock to represent a demand-support measures package. The emergency lifeline (demand-support) dummy is equal to 1 (zero) for any given country in case emergency lifelines (demand-support measures) make up more than 50 percent of its entire fiscal stimulus package in 2020. The same approach is applied to above-the-line and below-the-line packages. Driscoll-Kraay standard errors are presented in parentheses. ***, **, * denote statistical significance at 1 percent, 5 percent, and 10 percent, respectively.

Similar results hold for other economic indicators (Annex Tables A58) and when choosing different thresholds to differentiate between emergency lifelines and demand-support measures (see Annex Table A9 for 60 percent as a threshold).15

These results, however, mask important heterogeneity with respect to the pandemic cycle. In particular, we find that emergency lifelines were more effective when containment measures were high, as they provided much needed cashflow and liquidity support during the constrained months of economic activity (Table 9). This supports our earlier result that lifeline policies have been more effective on average, since the bulk of the fiscal policy announcement came during times of strict lockdown. In contrast, our results show that the effect of demand-support measures are much stronger when containment measures are being eased, which allows for more opportunities of increased domestic consumption. Similar results are obtained when we look at mobility—lifelines are more effective when mobility is low, while and the coefficient for demand-support is high with high mobility, though the result is not statistically significant (Table 10). The results reported in Table A10 are also robust to alternative thresholds.

Table 9.

Effect of announced fiscal shocks, interaction with containment measures

article image
Note: results reported are based on a sample of 52 countries using daily data from January 1, 2020 to December 30, 2020. Estimates are based on ΔYi,t = α + ui + F(zitLShocki,t + (l — F(zit))βHShocki,t + θXi,t + εi,t , with F(zit)=expγzit(1expγzit), y = 1.5 . z is a country-specific characteristic normalized to have zero mean and a unit variance. The weights assigned to each regime vary between 0 and 1 according to the weighting function F (zit), so that F(zit) can be interpreted as the probability of being in a given regime. ΔYi,t is the growth rate of industrial production, manufacturing PMI, the CLI, the unemployment rate, or sovereign CDS spreads of country i at month t; Shocki,t denotes the fiscal shock, ui are country-fixed effects to account for time-invariant specific factors; Xi,t is a vector of control variables, and εi,t is the error term. Driscoll-Kraay standard errors are presented in parentheses. *** ** * denote statistical significance at 1 percent, 5 percent, and 10 percent, respectively.
Table 10.

Effect of announced fiscal shocks, interaction with mobility

article image
Note: results reported are based on a sample of 52 countries using daily data from January 1, 2020 to December 30, 2020. Estimates are based on ΔYi,t = α + ui + F(zitLShocki,t + (l — F(zit))βHShocki,t + θXi,t + εi,t , with F(zit)=expγzit(1expγzit), y = 1.5 . z is a country-specific characteristic normalized to have zero mean and a unit variance. The weights assigned to each regime vary between 0 and 1 according to the weighting function F (zit), so that F(zit) can be interpreted as the probability of being in a given regime. ΔYi,t is the growth rate of industrial production, manufacturing PMI, the CLI, the unemployment rate, or sovereign CDS spreads of country i at month t; Shocki,t denotes the fiscal shock, ui are country-fixed effects to account for time-invariant specific factors; Xi,t is a vector of control variables, and εi,t is the error term. Driscoll-Kraay standard errors are presented in parentheses. *** ** * denote statistical significance at 1 percent, 5 percent, and 10 percent, respectively.

Again, we get similar results when differentiating between above- and below-the-line measures given the high correlation between above-the-line and demand-support measures, and below-the-line and lifelines.

VI. Conclusions

Countries worldwide launched historically unprecedented fiscal support measures to offset the economic fallout of the COVID-19 pandemic. To quantify the economic effects of these measures, we assemble a novel daily database of announcements regarding the fiscal policy interventions implemented across 52 countries throughout 2020—the use of daily frequency is key to reduce the risks that announcements react to developments in the economy.

The results demonstrate that announced fiscal policy measures in response to the COVID-19 pandemic have, on average, been effective in stimulating economic activity, as captured by a range of high-frequency indicators, namely industrial production, manufacturing PMI, unemployment, confidence indicators, and sovereign CDS spreads. The effects are economically and statistically significant and robust to alternative econometric specifications. In addition, emergency lifelines measures such as loans to firms and households, umbrella guarantees and equity injections (which form the bulk of below-the-line measures) were more effective in boosting economic activity during period of lockdown and supply side shocks, while demand-support measures—including tax cuts or payment deferrals, cash transfers and unemployment insurance— were less significant, on average.

Country characteristics play an important role in the effectiveness of fiscal policy, with larger effects for advanced economies and countries with lower public debt levels. In addition, a country’s stage in the pandemic cycle is critical, with emergency lifelines more effective when containment measures are high and supply is constrained, while demand-support measures are more effective when containment measures are being eased, and domestic consumption opportunities are more plentiful.

Our findings highlight the important role played by fiscal policy during the COVID-19 crisis. By supplementing the literature with quantitative empirical estimates, our paper can help policy makers make informed decisions on fiscal spending during the pandemic. One primary policy lesson derived from this analytical exercise is that emergency lifelines should not be withdrawn prematurely as they have been vital to support the economy during the COVID-19 crisis. But as supply constraints from containment measures abate, demand-support policies— including through investment in digital and green infrastructure—will be more effective and can rep lace emergency lifelines. In countries with high debt, fiscal policy is less effective and monetary policy might have to play a larger role to support economic activity.

References

  • Alesina, A., Favero, C. and Giavazzi, F., 2014. The output effect of fiscal consolidation plans. Journal of International Economics, 96, pp.S19S42.

    • Search Google Scholar
    • Export Citation
  • Alesina, Alberto, Carlo Favero, Francesco Giavazzi, Omar Barbiero, and Matteo Paradisi. 2017. Working Paper. “The Effects of Fiscal Consolidations: Theory and Evidence”.

    • Search Google Scholar
    • Export Citation
  • Auerbach AJ, Gorodnichenko Y (2013) Fiscal multipliers in recession and expansion. In: Alesina A, Giavazzi F (eds) Fiscal policy after the financial crisis, National Bureau of Economic Research Conference Report. University of Chicago Press, Chicago, pp 6398.

    • Search Google Scholar
    • Export Citation
  • Auerbach, Alan J, and Yuriy Gorodnichenko. 2016. “Effects of Fiscal Shocks in a Globalized World,” IMF Economic Review, Palgrave Macmillan, Vol. 64(1), pages 177215

    • Search Google Scholar
    • Export Citation
  • Beetsma, Roel, Cimadomo, Jacopo, Furtuna, Oana, and Giuliodori, Massimo. (2015). The confidence effects of fiscal consolidations. Economic Policy. 30. 439489.

    • 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 Paper 27461.

    • Search Google Scholar
    • Export Citation
  • Blanchard, O. and Perotti, R., 2002. Anempirical characterization of the dynamic effects of changes in government spending and taxes on output. The Quarterly Journal of Economics, 117(4), pp.13291368.

    • Search Google Scholar
    • Export Citation
  • David, Antonio C., Guajardo, Jaime, and Juan F. Yepez. 2019. “The Rewards of Fiscal Consolidation: Sovereign Spreads and Confidence Effects”. IMF Working Paper, WP/19/141.

    • Search Google Scholar
    • Export Citation
  • Deb, Pragyan, Davide Furceri, Jonathan D. Ostry, and Nour Tawk. 2020a. “The Economic Effects of COVID-19 Containment Measures.” IMF Working Paper 20/158, International Monetary Fund, Washington, DC.

    • Search Google Scholar
    • Export Citation
  • Devries, P., Guajardo, J., Leigh, D. and Pescatori, A., 2011. A new action-based dataset of fiscal consolidation. IMF Working Papers, pp.190.

    • Search Google Scholar
    • Export Citation
  • Driscoll, J.C. and Kraay, A. C., 1998. Consistent covariance matrix estimation with spatially dependent panel data. Review of Economics and Statistics, 80(4), pp.549560.

    • Search Google Scholar
    • Export Citation
  • Furceri, D., Zdzienicka, A. Twin Deficits in Developing Economies. Open Econ Rev 31, 123 (2020). https://doi.org/10.1007/s11079-019-09575-1.

    • Search Google Scholar
    • Export Citation
  • Gertler, M. and Karadi, P., 2015. Monetary policy surprises, credit costs, and economic activity. American Economic Journal: Macro economics, 7(1), pp.4476.

    • Search Google Scholar
    • Export Citation
  • Gourinchas P., Kalemli-Ozcan S., Penciakova V., and Nick Sander, 2021. “Fiscal Policy in the Age of COVID” Does it “ get in all of the Cracks?”, CEPR Discussion Paper DP16576.

    • Search Google Scholar
    • Export Citation
  • Granger, Clive W. J. & Terasvirta, Timo, 1993. “Modelling Non-Linear Economic Relationships,” OUP Catalogue, Oxford University Press, number 9780198773207.

    • Search Google Scholar
    • Export Citation
  • Gürkaynak, R. S., Sack, B.P. and Swanson, E. T., 2004. Do actions speak louder than words? The response of asset prices to monetary policy actions and statements. The Response of Asset Prices to Monetary Policy Actions and Statements (November 2004).

    • Search Google Scholar
    • Export Citation
  • International Monetary Fund. 2020. ‘The Great Lockdown: Dissecting the Economic Effects’. World Economic Outlook, Chapter 2.

  • 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
  • Jordà, Òscar. “Estimation and Inference of Impulse Responses by Local Projections.” The American Economic Review, Vol. 95, No. 1, 2005, pp. 161182.

    • Search Google Scholar
    • Export Citation
  • Lin, J. T. and McElroy, M. B.. 2011. “Detection from Space of a Reduction in Anthropogenic Emissions of Nitrogen Oxides during the Chinese Economic Downturn. Atmospheric Chemistry and Physics. M

    • Search Google Scholar
    • Export Citation
  • Maloney W. and Taksin T., 2020. Determinants of social distancing and economic activity during COVID-19: A global view. Covid Economics: Vetted and Real-Time Papers. 2020(13), pp.157177.

    • Search Google Scholar
    • Export Citation
  • Mertens, Karel and Ravn, Morten. (2012). A Reconciliation of SVAR and Narrative Estimates of Tax Multipliers. Journal of Monetary Economics. 68.10.1016/j.jmoneco.2013.04.004.

    • Search Google Scholar
    • Export Citation
  • Nickel, C. and Tudyka, A., 2014. Fiscal stimulus in times of high debt: Reconsidering multipliers and twin deficits. Journal of Money, Credit and Banking, 46(7), pp.13131344.

    • Search Google Scholar
    • Export Citation
  • OECD (2021), Composite leading indicator (CLI) (indicator). doi: 10.1787/4a174487-en (Accessed on 14 June 2021)

  • Ramey, Valerie A. 2011. “Identifying Government Spending Shocks: It’s all in the Timing,” The Quarterly Journal of Economics, Oxford University Press, Vol. 126(1), pages 150.

    • Search Google Scholar
    • Export Citation
  • Ramey, Valerie A., and Sarah Zubairy, 2018. “Government Spending Multipliers in Good Times and in Bad: Evidence from US Historical Data”. Journal of Political Economy 2018 126:2,850901

    • Search Google Scholar
    • Export Citation
  • Romer, Christina D., and David H. Romer. 2010. “The Macroeconomic Effects of Tax Changes: Estimates Based on a New Measure of Fiscal Shocks.” American Economic Review, 100 (3): 763801.

    • Search Google Scholar
    • Export Citation
  • Yang Weonho, Fidrmuc Jan, and Sugata Ghosh, (2015), Macroeconomic effects of fiscal adjustment: A tale of two approaches, Journal of International Money and Finance, 57, (C), 3160.

    • Search Google Scholar
    • Export Citation

ANNEX

Table A1.

Effect of announced fiscal measures on manufacturing PMIs, additional controls

article image
Note: results reported are based on a sample of 52 countries using daily data from January 1, 2020 to December 30, 2020. Estimates are based on ΔYi,t = α +ui + βShocki,t + θXi,t + εit , Where ΔYi,t is the growth rate of industrial production, manufacturing PMI, the CLI, the unemployment rate, or sovereign CDS spreads of country i at month t; Shocki,t denotes the fiscal shock, ui are country-fixed effects to account for time-invariant specific factors; Xi,t is a vector of control variables, and εi,t is the error term. Driscoll-Kraay standard errors are presented in parentheses. *** ** * denote statistical significance at 1 percent, 5 percent, and 10 percent, respectively.
Table A2.

Effect of announced fiscal measures on Composite Leading Indicator, additional controls

article image
Note: results reported are based on a sample of 52 countries using daily data from January 1, 2020 to December 30, 2020. Estimates are based on ΔYi,t = α +ui + βShocki,t + θXi,t + εit , Where ΔYi,t is the growth rate of industrial production, manufacturing PMI, the CLI, the unemployment rate, or sovereign CDS spreads of country i at month t; Shocki,t denotes the fiscal shock, ui are country-fixed effects to account for time-invariant specific factors; Xi,t is a vector of control variables, and εi,t is the error term. Driscoll-Kraay standard errors are presented in parentheses. *** ** * denote statistical significance at 1 percent, 5 percent, and 10 percent, respectively.
Table A3.

Effect of announced fiscal measures on unemployment, additional controls

article image
Note: results reported are based on a sample of 52 countries using daily data from January 1, 2020 to December 30, 2020. Estimates are based on ΔYi,t = α +ui + βShocki,t + θXi,t + εit , Where ΔYi,t is the growth rate of industrial production, manufacturing PMI, the CLI, the unemployment rate, or sovereign CDS spreads of country i at month t; Shocki,t denotes the fiscal shock, ui are country-fixed effects to account for time-invariant specific factors; Xi,t is a vector of control variables, and εi,t is the error term. Driscoll-Kraay standard errors are presented in parentheses. *** ** * denote statistical significance at 1 percent, 5 percent, and 10 percent, respectively.
Table A4.

Effect of announced fiscal measures on sovereign CDS spreads, additional controls

article image
Note: results reported are based on a sample of 52 countries using daily data from January 1, 2020 to December 30, 2020. Estimates are based on ΔYi,t = α +ui + βShocki,t + θXi,t + εit , Where ΔYi,t is the growth rate of industrial production, manufacturing PMI, the CLI, the unemployment rate, or sovereign CDS spreads of country i at month t; Shocki,t denotes the fiscal shock, ui are country-fixed effects to account for time-invariant specific factors; Xi,t is a vector of control variables, and εi,t is the error term. Driscoll-Kraay standard errors are presented in parentheses. *** ** * denote statistical significance at 1 percent, 5 percent, and 10 percent, respectively.
Table A5.

Effect of announced fiscal shocks on manufacturing PMI, by type of measure

article image
Note: results reported are based on a sample of 52 countries using daily data from January 1 , 2020 to December 30, 2020 . Estimates are based on ∆Yi,t = α + ui + βLLi,t + γDSi,t + θXi,t + εi,t Where ∆Yi,t is the growth rate of industrial production, manufacturing PMI, the CLI, the unemployment rate, or sovereign CDS spreads of country i at month t; ui are country-fixed effects to account for time-invariant specific factors; Xi,t is a vector of control variables , and εi,t is the error term . LLi,t is an emergency lifeline dummy interacted by the fiscal shock to represent an emergency lifelines package, while DSi,t is demand-support dummy interacted by the fiscal shock to represent a demand-support measures package. The emergency lifeline (demand-support) dummy is equal to 1 (zero) for any given country in case emergency lifelines (demand-support measures) make up more than 50 percent of its entire fiscal stimulus package in 2020. The same approach is applied to above-the-line and below-the-line packages. Driscoll-Kraay standard errors are presented in parentheses. ***, **, * denote statistical significance at 1 percent, 5 percent, and 10 percent , respectively.
Table A6.

Effect of announced fiscal shocks on CLI, by type of measure

article image
Note: results reported are based on a sample of 52 countries using daily data from January 1 , 2020 to December 30, 2020 . Estimates are based on ∆Yi,t = α + ui + βLLi,t + γDSi,t + θXi,t + εi,t Where ∆Yi,t is the growth rate of industrial production, manufacturing PMI, the CLI, the unemployment rate, or sovereign CDS spreads of country i at month t; ui are country-fixed effects to account for time-invariant specific factors; Xi,t is a vector of control variables , and εi,t is the error term . LLi,t is an emergency lifeline dummy interacted by the fiscal shock to represent an emergency lifelines package, while DSi,t is demand-support dummy interacted by the fiscal shock to represent a demand-support measures package. The emergency lifeline (demand-support) dummy is equal to 1 (zero) for any given country in case emergency lifelines (demand-support measures) make up more than 50 percent of its entire fiscal stimulus package in 2020. The same approach is applied to above-the-line and below-the-line packages. Driscoll-Kraay standard errors are presented in parentheses. ***, **, * denote statistical significance at 1 percent, 5 percent, and 10 percent , respectively.
Table A7.

Effect of announced fiscal shocks on unemployment, by type of measure

article image
Note: results reported are based on a sample of 52 countries using daily data from January 1 , 2020 to December 30, 2020 . Estimates are based on ∆Yi,t = α + ui + βLLi,t + γDSi,t + θXi,t + εi,t Where ∆Yi,t is the growth rate of industrial production, manufacturing PMI, the CLI, the unemployment rate, or sovereign CDS spreads of country i at month t; ui are country-fixed effects to account for time-invariant specific factors; Xi,t is a vector of control variables , and εi,t is the error term . LLi,t is an emergency lifeline dummy interacted by the fiscal shock to represent an emergency lifelines package, while DSi,t is demand-support dummy interacted by the fiscal shock to represent a demand-support measures package. The emergency lifeline (demand-support) dummy is equal to 1 (zero) for any given country in case emergency lifelines (demand-support measures) make up more than 50 percent of its entire fiscal stimulus package in 2020. The same approach is applied to above-the-line and below-the-line packages. Driscoll-Kraay standard errors are presented in parentheses. ***, **, * denote statistical significance at 1 percent, 5 percent, and 10 percent , respectively.
Table A8.

Effect of announced fiscal shocks on sovereign CDS spreads, by type of measure

article image
Note: results reported are based on a sample of 52 countries using daily data from January 1 , 2020 to December 30, 2020 . Estimates are based on ∆Yi,t = α + ui + βLLi,t + γDSi,t + θXi,t + εi,t Where ∆Yi,t is the growth rate of industrial production, manufacturing PMI, the CLI, the unemployment rate, or sovereign CDS spreads of country i at month t; ui are country-fixed effects to account for time-invariant specific factors; Xi,t is a vector of control variables , and εi,t is the error term . LLi,t is an emergency lifeline dummy interacted by the fiscal shock to represent an emergency lifelines package, while DSi,t is demand-support dummy interacted by the fiscal shock to represent a demand-support measures package. The emergency lifeline (demand-support) dummy is equal to 1 (zero) for any given country in case emergency lifelines (demand-support measures) make up more than 50 percent of its entire fiscal stimulus package in 2020. The same approach is applied to above-the-line and below-the-line packages. Driscoll-Kraay standard errors are presented in parentheses. ***, **, * denote statistical significance at 1 percent, 5 percent, and 10 percent , respectively.
Table A9.

Effect of announced fiscal shocks on industrial production, by type of measure, higher threshold

article image
Note: results reported are based on a sample of 52 countries using daily data from January 1 , 2020 to December 30, 2020 . Estimates are based on ∆Yi,t = α + ui + βLLi,t + γDSi,t + θXi,t + εi,t Where ∆Yi,t is the growth rate of industrial production, manufacturing PMI, the CLI, the unemployment rate, or sovereign CDS spreads of country i at month t; ui are country-fixed effects to account for time-invariant specific factors; Xi,t is a vector of control variables , and εi,t is the error term . LLi,t is an emergency lifeline dummy interacted by the fiscal shock to represent an emergency lifelines package, while DSi,t is demand-support dummy interacted by the fiscal shock to represent a demand-support measures package. The emergency lifeline (demand-support) dummy is equal to 1 (zero) for any given country in case emergency lifelines (demand-support measures) make up more than 50 percent of its entire fiscal stimulus package in 2020. The same approach is applied to above-the-line and below-the-line packages. Driscoll-Kraay standard errors are presented in parentheses. ***, **, * denote statistical significance at 1 percent, 5 percent, and 10 percent , respectively.