Policy Research Working Paper 10954 Revisiting Public Investment Multipliers Nonlinear Effects of the Business Cycle, Fiscal Space, Efficiency, and Capital Stock Amat Adarov Benedict Clements João Tovar Jalles Development Economics Prospects Group October 2024 Policy Research Working Paper 10954 Abstract The paper examines the macroeconomic effects of public gross domestic product raises output by 1.1 percent after investment in emerging market and developing economies. five years, on average. However, the effects are much larger To this end, the analysis develops a new measure of public when public investment spending is efficient and fiscal space investment shocks based on cyclically adjusted government is ample—reaching up to 1.6 percent over the same period. investment. Estimations using local projections based on Public investment multipliers tend to be larger during reces- a large sample of 129 countries over 1980–2019 suggest sions and in capital-scarce economies. The paper also finds that public investment can significantly boost economic that public investment can crowd in private investment, as growth: an increase in public investment by 1 percent of well as boost productivity and potential output. This paper is a product of the Prospects Group, Development Economics. It is part of a larger effort by the World Bank to provide open access to its research and make a contribution to development policy discussions around the world. Policy Research Working Papers are also posted on the Web at http://www.worldbank.org/prwp. The authors may be contacted at aadarov@worldbank.org, benedict.clements@udla.edu.ec, and joaojalles@gmail.com. The Policy Research Working Paper Series disseminates the findings of work in progress to encourage the exchange of ideas about development issues. An objective of the series is to get the findings out quickly, even if the presentations are less than fully polished. The papers carry the names of the authors and should be cited accordingly. The findings, interpretations, and conclusions expressed in this paper are entirely those of the authors. They do not necessarily represent the views of the International Bank for Reconstruction and Development/World Bank and its affiliated organizations, or those of the Executive Directors of the World Bank or the governments they represent. Produced by the Research Support Team Revisiting Public Investment Multipliers: Nonlinear Effects of the Business Cycle, Fiscal Space, Efficiency, and Capital Stock 1 Amat Adarov 2 Benedict Clements 3 João Tovar Jalles 4 Keywords: fiscal multipliers; public investment; gross fixed capital formation; local projections. JEL codes: C33; E22; E62; H30, H50 1 We are grateful to Tommy Chrimes, Antonio Fatás, Indermit Gill, Jakob de Haan, Sam Hill, Ethan Ilzetzki, Dohan Kim, Ayhan Kose, Kate McKibbin, Valerie Mercer-Blackman, Franziska Ohnsorge, Peter Pedroni, Francesco Zanetti, participants of seminars at the World Bank and 16th FIW research conference, for useful comments and feedback. Juan Felipe Serrano Ariza and Urja Singh Thapa provided excellent research assistance. We gratefully acknowledge financial support from the Policy and Human Resources Development (PHRD) fund. The findings, interpretations and conclusions expressed in this paper are entirely those of the authors and should not be attributed to the World Bank, its Executive Directors, or the countries they represent. 2 The World Bank. 1818 H Street, NW Washington, DC 20433 USA. Email: aadarov@worldbank.org 3 Universidad de Las Américas, De Los Colimes esq, Quito 170513, Ecuador. Email: benedict.clements@udla.edu.ec 4 University of Lisbon-Lisbon School of Economics and Management (ISEG), Rua do Quelhas 6, 1200-781 Lisboa, Portugal. Research in Economics and Mathematics (REM) and Research Unit on Complexity and Economics (UECE), Universidade de Lisboa-ISEG, Rua Miguel Lupi 20, 1249-078 Lisbon, Portugal. Economics for Policy, Universidade Nova de Lisboa-Nova School of Business and Economics, Rua da Holanda 1, 2775-405 Carcavelos, Portugal. IPAG Business School, 184 Boulevard Saint-Germain, 75006 Paris, France. Email: joaojalles@gmail.com 1. Introduction Public investment is often considered a powerful policy lever to foster economic growth. This is particularly the case in emerging market and developing economies (EMDEs), which face significant investment gaps to meet their sustainable development objectives in an environment where private investment has been in a broad-based and prolonged slump (World Bank 2024). 5 Public investment facilitates production; promotes the flows of capital, labor, goods and services; and provides the infrastructure needed to improve human capital and productivity. Given the non- excludable and non-rival nature of public goods and services, and substantial costs involved in the development and maintenance of public infrastructure, it often cannot be supplied effectively by the private sector. That said, public infrastructure projects are generally costly and can lead to high budget deficits and unsustainable levels of public debt, particularly when the planning and execution of investment projects is weak (Adarov and Panizza 2024; Afonso and Alves 2023; Berg et al. 2012). Government spending may suffer from inefficiency, especially in countries with governance issues, corruption, and weak fiscal monitoring and accountability mechanisms, and in the worst cases, may yield infrastructure that is unproductive and costly to maintain (Chakraborty and Dabla- Norris 2011; Dabla-Norris et al. 2012; Pritchett 2000). In countries with weak institutions, public investment may also crowd out private investment, particularly when scaling up is sizeable and fast-paced (Aschauer 1989a; Cavallo and Daude 2011). Thus, the net macroeconomic effects of public investment can also be small or even negative. In fact, the empirical estimates of public investment multipliers reported to date differ significantly. The estimates are sensitive to the sample of countries and timeline investigated, estimation methodology, incorporation of non-linear effects, and fiscal shock identification strategies (see Section 2 for discussion). Empirical work examining public investment multipliers in EMDEs is especially scarce. The main challenge in performing such research also rests in the difficulty of identifying changes in public spending that are uncorrelated with contemporaneous macroeconomic shocks, and that can thus be deemed as exogenous. In light of these caveats, in this paper we aim to contribute to this literature in a number of ways. First, we use the data on public investment for a broad sample of 129 EMDEs over a long span of time (1980-2019). The large sample is helpful to better understand implications of public investment in EMDEs as a whole, as well as explore heterogeneity across country groups. Second, we bring together many of the methodological improvements in recent empirical research focusing on the role of initial conditions in determining the macroeconomic effects of public investment. These include: (i) prevailing business cycle conditions at the time of the shock; (ii) fiscal space; (iii) public investment efficiency; and (iv) the initial public capital stock level. While these themes have been touched upon in some earlier studies, no paper has yet brought them all together and 5 Public investment in the context of the paper refers to general government gross fixed capital formation. 2 applied them to a broad sample of EMDEs. Our paper helps fill this gap in the literature. Third, we introduce a new approach to measuring public investment shocks based on the identification of the episodes of large changes in cyclically adjusted real public investment specific to each country. In contrast to many public investment shock identification frameworks, our approach relies on annual-frequency data on public investment that is publicly available and is easy to replicate. 6 Finally, in addition to the estimation of public investment multipliers, we examine the effects of public investment on private investment, productivity, and potential output, which helps to cast light on the long-run supply-side transmission channels of public investment. Our local projections estimates, using the identified public investment shocks, suggest that public investment can stimulate economic growth. In particular, the scaling up of public investment equivalent to 1 percent of GDP leads to an increase in output of 1.1 percent after five years in EMDEs, on average. We also find that the size of the public investment multiplier depends on government efficiency and fiscal space. In countries with higher public investment efficiency and ample fiscal space, an increase in public investment equivalent to 1 percent of GDP can increase output by up to 1.6 percent over the horizon of five years. 7 By contrast, in countries with low public investment efficiency and high public debt, the output effects of public investment are smaller and not statistically significant. In line with the literature, we also find that public investment multipliers tend to be greater during recessions and in capital-scarce economies. Our analysis also suggests that public investment can have significant crowding-in effects on private investment. In EMDEs, the scaling up of public investment by 1 percent of GDP leads to an increase in private investment by up to 2.2 percent over the horizon of five years, on average. Finally, the paper finds that public investment can boost productivity and potential output. In the EMDE sample with available data, a 1 percent of GDP increase in public investment can increase labor productivity by 1.9 percent and total factor productivity by 0.8 percent over five years. Potential output in response to public investment shocks increases by up to 1.1 percent over the same period. Thus, our analysis provides empirical support for both short-run demand and long- run supply-side transmission channels of public investment. The rest of the paper is structured as follows. Section 2 provides a synthesis of the literature on public investment effects. Section 3 presents the new framework for the identification of public investment shocks and the estimation methodology. Section 4 discusses the empirical results. Section 5 outlines policy implications and concludes. 6 See the comparative review of public investment shock identification strategies developed in the empirical literature in the methodology section. 7 State-dependent public investment multipliers use the dynamics of public debt as a share of GDP to proxy fiscal space and public investment and infrastructure efficiency indexes in the estimations of multipliers conditional on public investment efficiency. In more technical terms, public investment efficiency is defined as the fraction of public investment that translates into effective public capital stock (Pritchett 2000), which, in turn, depends on the strength of institutions, quality of the design and implementation of public investment projects, and effectiveness of procurement systems, among other factors (see also Kim et al. 2020). 3 2. Synthesis of the literature 2.1. Theoretical foundations The early literature examined the effects of government investment on economic growth in the context of the endogenous growth model in which public capital enters the production function as one of the productive inputs (Aschauer 1989a,b; Barro 1990; Barro and Sala-i-Martin 1992; Futagami, et al. 1993; Glomm and Ravikumar 1994; Turnovsky 1997). Subsequent studies built on this framework, incorporating features such as aid-funded public investment, infrastructure networks, and public debt accumulation to facilitate a more nuanced analysis of the transmission channels of public investment to growth (Adam and Bevan 2006; Agenor 2010; Berg et al. 2010, 2012). More recently, Chakraborty and Dabla-Norris (2011) studied the quality of public investment and distortionary effects of corruption within a general equilibrium growth model. 2.2. Transmission channels Theoretical and empirical literature has identified multiple channels through which public investment influences economic growth: • Short-term aggregate demand effects. Public investment has the potential to support economic activity by boosting aggregate demand in the short term. This positive impact, however, is at least partly offset by the associated fiscal effects on the real economy, because public investment in principle is funded via taxation, debt issuance, or reallocation of government expenditure. In addition, the multiplier effect is weakened by purchases of investment goods abroad and depends on the import intensity of investment (“leakage effect”). Rapid scaling up of public investment, depending on its funding source and efficiency, may fuel fiscal imbalances and thereby undermine growth prospects (Bom and Ligthart 2014a; Romp and de Haan 2005). • Long-run aggregate supply effects. Public investment has the potential to directly increase the productive capacity of an economy by fostering enhanced productivity of private fixed capital and labor through the provision of public infrastructure. For instance, new roads and bridges can increase the overall competitiveness of an economy by enabling connectivity or reducing its cost (Aschauer 1989b; Romp and de Haan 2005; Straub 2011). • Crowding-in or crowding-out of private investment. Public investment can crowd in private investment directly by requiring the use of private capital in the implementation of an investment project, for example, via public-private partnerships. Public investment can also enable infrastructure that raises returns on private capital—for instance, roads and communications infrastructure—thereby encouraging private sector investment (Aschauer 1989a, Eden and Kraay 2014). Public investment helps to reduce uncertainty and 4 risks associated with large private investment projects, especially infrastructure projects requiring massive upfront costs but longer payback periods (IMF 2021b). However, public investment may also crowd out private investment, especially when fiscal space is limited and additional fiscal stimulus raises sovereign risk and borrowing costs for the private sector (Abiad et al. 2016; Erenburg and Wohar 1995; Huidrom et al. 2020). The net effect on the private sector depends on the balance between these opposing factors, which, in turn, is influenced by fiscal space and the quality of public investment. • Efficiency and quality of public investment. Scaling up of public investment may not necessarily result in an equivalent increase in the value of productive public capital (Pritchett 2000). Some resources are lost during the investment process because of weak governance, corruption, coordination issues, and poor design and implementation of investment projects. There may also be diminishing returns on additional public investment, though this depends on a country’s circumstances and the merits of specific projects. In the worst case, low-quality investment may yield infrastructure that is unproductive and yet requires a continuous stream of fiscal resources for maintenance, thereby hurting long-term growth prospects (Chakraborty and Dabla-Norris 2011; Dabla-Norris et al. 2012). In part, this also relates to the composition of public investment: not all projects contribute equally to growth. 8 • Public capital maintenance costs. More generally, depreciating public capital stocks require additional short- and long-run maintenance. The associated costs may lead to additional fiscal strains that undermine long-term positive growth effects. Meanwhile, inadequate or untimely maintenance of public capital could lead to even larger social and economic costs associated with infrastructure failures (Schwartz et al. 2020). 9 The strength of this channel is intertwined with the efficiency channel, as low-quality public investment is more likely to yield infrastructure that is prone to larger or more frequent upkeep costs. • Sustainability of economic growth. Public investment—though not always the only solution— can also play an important role in delivering public goods or services that may not be privately profitable, such as public healthcare and education, water and energy transmission, and national security. This type of public investment can be instrumental for facilitating sustainable and inclusive growth through its positive effects on human capital development, social 8 As a related matter, there are challenges associated with the measurement and valuation of public investment at a disaggregated level—that is, identifying infrastructure-related spending and composition of public investment by capital asset types (see also ADB 2017 and Fay et al. 2019). This hinders the assessment of the macroeconomic effects by individual categories of public investment and types of infrastructure, which are likely to be heterogeneous (see the meta-analysis in Foster et al. 2023). 9 Governments may also have stronger incentives to spend on new investment projects rather than on maintenance as the former is more visible and attractive from an electoral perspective (De Haan and Klomp 2013). 5 inclusion, and environmental impacts (Foster et al. 2023; Mazzucato and Semieniuk 2017; Turnovsky 2015; Zachmann et al. 2012). 2.3. Main approaches to constructing public investment shocks To gauge the extent to which public investment shocks impact economic growth, it is first necessary to identify changes in this spending that are independent of prevailing macroeconomic conditions. To date, the main methods deployed to tackle this identification challenge are the structural vector autoregression (SVAR) estimation with recursive identification, frameworks relying on instrumental variables, the narrative approach, and identification based on forecast errors: • SVAR with recursive identification of public spending shocks. This approach employs recursive identification schemes and other parameter restrictions to pin down unexpected public spending shocks. Specifically, the Cholesky decomposition assumes that government spending does not respond to macroeconomic shocks in the same period (Blanchard and Perotti 2002). A drawback is that this rationale becomes less compelling at an annual frequency, yet availability of higher frequency data is often constrained, especially for EMDEs. • Official lending as an instrument for exogenous public spending. This approach, pioneered by Kraay (2012, 2014), uses data on official creditor loan disbursements to identify public spending shocks in the recipient country based on the lag between loan approval and subsequent disbursements to identify a component insulated from contemporaneous macroeconomic developments. However, this framework is only applicable to countries that are recipients of official development assistance and requires the calculation of “predicted” disbursements for each loan. • Military spending as an instrument for exogenous public spending. Building on the “natural experiment” framework proposed by Barro (1981), the narrative approach, developed in Ramey and Shapiro (1998), Ramey (2011a, b), and Ramey and Zubairy (2018), uses fluctuations in governments’ military expenditures—assumed to be driven by external geopolitical factors as opposed to domestic macroeconomic conditions—to isolate exogenous changes in public spending. This approach, however, would not work well for EMDEs, in which military spending is typically less prone to fluctuation. More broadly, a pitfall of this method is that the resulting growth responses may be largely attributable to the military spending sub-component as opposed to more general fiscal stimulus. • Forecast errors in public spending as a proxy for fiscal shocks. In more recent empirical research, Auerbach and Gorodnichenko (2012, 2013) use differences between actual public spending and the level predicted by professional forecasters to identify unanticipated public 6 spending shocks. 10 The methodology has the advantage of overcoming the issue of fiscal foresight, whereby anticipated fiscal policy changes may be incorporated into current economic decisions (Forni and Gambetti 2010; Leeper et al. 2012, 2013). However, this approach relies on the availability and quality of public spending forecast data. Additional caveats relate to the nature of fiscal projections: first, they may not be fully orthogonal to past macroeconomic trends; and second, they rely on subjective, heterogeneous assumptions about future macroeconomic developments. 2.4. Estimates of public investment multipliers While the empirical work on total public spending multipliers across a broad range of countries is voluminous, the subset of the literature that distinguishes public investment multipliers is limited and primarily focuses on advanced economies. In general, there is an emerging consensus that public investment tends to have a positive growth impact in the medium term (Auerbach and Gorodnichenko 2013; Eden and Kraay 2014; Furceri and Li, 2017; Ilzetzki et al. 2013; Izquierdo et al. 2019; Leduc and Wilson 2012). However, there remains considerable variation in the estimates of the output elasticity to public investment; in a meta-analysis of 68 studies, Bom and Ligthart (2014b), report that the range is -1.7 to 2.0. In reviews of the literature, Gechert and Rannenberg (2018) and Vagliasindi and Gorgulu (2021) also report that public investment multiplier estimates range widely with an average of about 1.5. Appendix table B1 provides a comprehensive review of public investment multipliers in the literature over the past two decades. Empirical work focusing on EMDEs has been especially scarce, given data constraints and methodological challenges in identifying spending shocks. Nevertheless, limited research on EMDEs also documents important growth impacts of public investment. For instance, Miyamoto et al. (2020) report an increase in output by 0.4 percent over 4 years for a sample of 39 EMDEs. Ilzetzki et al. 2013) estimate a public investment multiplier of 0.6 on impact, increasing to 1.6 in the longer run, for a panel of 24 developing countries. Furceri and Li (2017) report smaller magnitudes for a sample of 79 EMDEs: a 10-percent increase in public investment induces growth of 0.4 percent over four years. Smaller effects are reported in Warner (2014) using a sample of 124 EMDEs: a change in public investment equivalent to 1 percent of GDP is found to spur only meagre (0.1 percent) output growth in the short run with no significant effect in the long run. The observed wide range of estimates of government spending multipliers has prompted additional inquiry into the factors that may explain these differences—discussed in the following section. 10 This approach was utilized recently in Abiad et al. (2016), Furceri and Li (2017), Honda et al. (2020), and Miyamoto et al. (2020) to estimate unconditional and state-dependent public spending multipliers. 7 2.5. Factors impacting the size of public investment multipliers Macroeconomic conditions. In a synthesis review, Izquierdo et al. (2019) note that the position of economy in the business cycle, the degree of exchange rate flexibility, debt levels, the monetary policy stance, and openness to trade are important determinants of the size of multipliers. In this regard, the literature indicates that multipliers tend to be larger during recessions (Auerbach and Gorodnichenko 2012 and 2013; Honda et al. 2020; Riera-Crichton et al. 2015). Further, the output effects tend to be larger in supply-driven recessions compared to demand-driven recessions (Ghassibe and Zanetti 2022). 11 Multipliers are found to be larger during periods of monetary policy easing and elevated macroeconomic uncertainty, particularly when nominal interest rates are at very low levels, i.e. reach the “zero lower bound” (Christiano et al. 2011; Gbohoui 2021). Country structural characteristics. Multipliers tend to be greater in countries with a fixed exchange rate regime, low levels of debt, and lower economic informality (Colombo et al. 2022; Honda et al. 2020; Huidrom et al. 2020). Financial development and economic development are also positively associated with the size of spending multipliers (Ilzetzki et al. 2013; Koh 2017). Initial stock of public capital. Public investment multipliers are greater in countries with a lower level of initial public capital stock (Izquierdo et al. 2019). Excessive levels of public capital stock and investment may be detrimental for growth if resources are diverted away from more productive uses or crowd out private investment (Canning and Pedroni 2008; Devarajan et al. 1996). However, Honda et al. (2020) find that in low-income countries, output effects of public investment are greater in economies with higher initial capital stock, and conjecture that in low-income countries the private sector may not be responsive to fiscal policy shocks when the initial capital stock is too low. Public investment efficiency. The quality of the public investment management process is important in order to reap the positive macroeconomic effects of public capital. The public investment management process encompasses multiple aspects, including project development, implementation, monitoring, and evaluation—the efficiency of which are generally difficult to quantify (Dabla-Norris et al. 2012). Gupta et al. (2014) compute “efficiency adjusted” public capital stocks and find that the growth effects of efficient public investment are higher. The positive effects of higher spending efficiency has been documented in other empirical research (Berg et al. 2013; Cavallo and Daude 2011; Furceri and Li 2017; Izquierdo et al. 2018; Leduc and Wilson 2012; Leeper et al. 2010; Miyamoto et al. 2020). In sum, the literature is generally in agreement that public investment tends to have a positive impact on economic growth, particularly in the longer run. The wide dispersion of the magnitudes of public investment multipliers, however, motivated research that sought to reconcile this 11 Ramey (2019) notes that these results may not be fully robust given their sensitivity to the sample composition and the methodology. 8 heterogeneity. Findings suggest that country-level characteristics—including public spending efficiency and capital scarcity—are relevant, as well as macroeconomic factors. Methodological and sample differences may also explain some variation in results across studies. That said, the evidence on the impact of public investment to growth in EMDEs is relatively meagre, owing in large part to associated data constraints. Thus, further research—including that provided in this paper—is needed to better inform how public investment can stimulate economic growth and mobilize private investment. 3. Methodology and data 3.1. Data and sample The database used in the estimation of the macroeconomic effects of public investment draws from several sources. Public investment, private investment, and capital stock data are from the International Monetary Fund’s Investment and Capital Stock Dataset (IMF 2021a). Public debt data are retrieved from the World Bank’s Fiscal Space Database (Kose et al. 2022). Public investment efficiency data are obtained from several sources: IMF (2021b) Fiscal Monitor database, Devadas and Pennings (2018), and Dabla-Norris et al. (2012). Potential output data are sourced from Kilic Celik et al. (2023). GDP and inflation series are from the IMF’s World Economic Outlook database. Labor productivity and total factor productivity data are from Penn World Table 10.01. The resulting dataset comprises up to 129 EMDEs, spanning the period 1980- 2019 (see Appendix table A1 for the sample composition). 12 3.2. A new approach to identify public investment shocks based on cyclically adjusted public spending In this paper we apply a new approach to identify public investment shocks. The methodology builds on the work of Alesina et al. (1998), Alesina and Ardagna (2010), and related studies that assess the macroeconomic effects of changes in cyclically adjusted fiscal variables. Conceptually, the approach is consistent with the literature arguing that large and apparent scaling up of public investment tends to reflect exogenous decisions by the public authorities (Deleidi et al. 2020; Warner 2014). The shock identification framework involves four steps: • First, we estimate output elasticities of public investment for each country in the sample by regressing the logarithm of real public investment on the logarithm of real GDP. Estimates are done using ordinary least squares for each country with at least 20 continuous observations of both variables. 13 12 The sample composition and time coverage are determined entirely by the data availability. In particular, the current version of the Investment and Capital Stock Dataset (IMF 2021a), providing consistent disaggregated series for public and private investment, covers the period 1960-2019. 13 Countries with fewer than 20 continuous observations were not included in the sample going forward. 9 • Second, we obtain measures of potential output via a Hodrick-Prescott (HP) filter as the baseline estimate. We also use Baxter-King, Christiano-Fitzgerald Random Walk and the Hamilton (2018) filters as a robustness check. • Third, we compute cyclically adjusted real public investment (CAPI) as follows: = � � (1) where PI is real public investment and is the estimated elasticity of public investment. • Fourth, for each country i we define public investment shocks (ℎ , ) as a measure taking the value of one when a country’s first difference of CAPI exceeds its country-specific mean by one standard deviation: ��������� + , 0 otherwise ℎ , = 1 > (2) Focusing on country-level public investment adjustments greater than one standard deviation is in the spirit of Alesina and Ardagna (2010), who argue that focusing on large fiscal adjustments helps identification of changes in fiscal variables that are induced by discretionary policy, rather than influenced by the business cycle. Some examples of the episodes identified using this approach include the rapid scaling up of public investment in Poland in 2005-06 and 2018, in Brazil in 2007, and Morocco in 2008. In Poland, the episodes followed significant EU fund inflows and reforms as part of its EU integration. In Brazil, the episode followed the launch of the Growth Acceleration Program—a major infrastructure program including investment projects and policies to boost growth. In Morocco, the episode involved major public investment in infrastructure projects, such as the Green Morocco Plan to bolster the agricultural sector and the expansion of the Tanger Med port. This approach to the identification of public investment shocks has several advantages. Given the focus only on large episodes of public investment increases, the results are more robust to imperfections in measuring the effect of the business cycle on fiscal variables, as small changes in cyclically adjusted public spending are excluded from the estimation. The proposed framework eschews certain limitations of existing identification methods that rely on data that are not publicly available (for instance, methods based on government spending forecast errors) or yield estimates for a limited set of countries (for instance, frameworks relying on narrative shock identification or quarterly-frequency data). As such, identification of disaggregated public investment shocks can be undertaken for a broad sample of countries with available annual data. The large sample, in turn, facilitates estimation of multipliers conditional on country characteristics. 10 In contrast to one-size-fits-all approaches, this framework accounts for heterogeneity across countries by considering the magnitude of public investment shocks within country-specific historical contexts. This is an important feature in the analysis of EMDEs, which may exhibit fiscal procyclicality (commodity exporters) or budgetary process-driven volatility of public spending (low-income countries). The approach can be expanded to allow for time-varying or state- dependent thresholds. At the same time, a few caveats should be noted. First, issues related to endogeneity and fiscal foresight may persist, despite focusing on large cyclically adjusted public spending innovations to mitigate business cycle effects. Second, the methodology relies on measures of potential output, which are generally estimated with a certain degree of imprecision. Third, the measure of the public investment shock is a binary variable and does not yield an estimate of a multiplier directly from a regression. Rather, the output effects need to be interpreted in the context of the average change in public investment for the effective sample subject to the shock, rescaling to obtain the public investment multiplier value. 3.3. Methodological framework for the estimation of public investment multipliers Responses of real GDP to public investment shocks are estimated using the local projections method proposed by Jordà (2005). This approach has a number of distinct advantages over other estimation frameworks, including VAR models. 14 First, it does not impose dynamic restrictions and obviates the need to estimate the equations for dependent variables other than the variable of interest, thereby economizing on the number of estimated parameters. Second, it is well-suited to estimating nonlinear effects of public investment conditional on country-characteristics (state- dependent multipliers). Third, it is relatively simple to deal with correlation in error terms—a likely complication in cross-country analysis. The following baseline specification is estimated: 15 ( ,+ ) − (,−1 ) = + + ℎ, + , + ε, (3) in which k = 0,...,5 is the forecast horizon in years; (,+ ) − (,−1 ) represents the cumulative change in real GDP (in percent) over the forecast horizon; , are country and time fixed effects to account for time-invariant country heterogeneity and global factors (such as the world business cycle or oil price movements); , is a set of control variables including—as in Abiad et al. (2016) and Furceri and Li (2017)—two lags of the shocks and two lags of real GDP 14 See Plagborg-Møller and Wolf (2021) for a discussion on the trade-offs between VARs and local projections. 15 Similar local projections specifications have been widely used in empirical literature on public spending multipliers, for instance, Abiad et al. (2015); Furceri and Li (2017); Honda et al. (2020); Miyamoto et al. (2020). 11 growth. 16 To control for outliers, data points above the 99th percentile and below the 1st percentile are dropped in the estimations. The coefficient denotes the response of output in each period k to a public investment shock at t = 0 (ℎ , , identified using the methodology described in the previous section). Specifically, it measures the average cumulative real GDP change in period t+k relative to period t-1 (in percent), in response to the public investment shock for the effective sample—the sample of countries used in the estimation. To ease interpretation, the estimated coefficients are scaled by the average change in public investment as a percent of GDP for the effective sample that experienced the public investment shock, so that the impulse responses can be interpreted as the change in output (in percent) in response to a 1 percent of GDP increase in public investment.17 The model is estimated for the broadest sample of countries available for robustness. For some exercises, however, the sample size is much smaller (for instance, for potential output, productivity estimations, and subgroups of EMDEs), and thus the results are not directly comparable and should be interpreted with caution. Descriptive statistics summarizing average changes in output and public investment during public investment shock episodes are summarized in the Appendix table A3. Impulse response functions are obtained by plotting the estimated multipliers for k = 0,...,5, with 90-percent confidence bands computed using robust standard errors clustered at the country level. To examine heterogeneity across country groups (for instance, categorized using income levels, commodity exporter status, degree of public spending efficiency) and discrete macroeconomic states (negative and positive economic growth periods), the model is estimated separately for each subsample. State-dependent multipliers, conditional on the values of continuous time-varying variables, are estimated using a local projections framework with a smooth transition function: 18 ( ,+ ) − (,−1 ) = + + (, )ℎ, + (1 − (, ))ℎ, + , + , (4) exp (− ) with �, � = 1+exp (− , ) , > 0. , 16 Among other robustness checks, the model is also estimated with additional control variables to examine the omitted variable bias, dropping lagged real GDP growth, and using the generalized method of moments (GMM) estimator to address possible bias arising from the lagged dependent variable. 17 It is a standard approach in the empirical literature on spending multipliers to use an ex-post conversion using the average government spending ratio to GDP. As argued in Ramey and Zubairy (2016), this may however introduce a bias if government spending-to-GDP ratio varies significantly over the sample period. 18 The same approach was used to estimate state-dependent public spending multipliers in Abiad et al. (2015), Furceri and Li (2017), Miyamoto et al. (2020), and Honda et al. (2020). 12 in which , is the value of a conditioning variable, normalized to have zero mean and unit variance. 19 The coefficients and capture the output impact of public investment shocks at each horizon k for the state characterized by low values of a conditioning variable, �, � ≈ 1 when z goes to minus infinity; and the state characterized by high values of a conditioning variable, 1 − �, � ≈ 1 when z goes to plus infinity. This approach is equivalent to the smooth transition autoregressive model developed by Granger and Teräsvirta (1993). The advantage of this methodology is twofold. First, it permits a direct test of whether the effect of government spending varies across high and low levels of a given conditioning variable. Second, it allows the effect of government spending shocks to change smoothly between the levels of a conditioning variable by considering a continuum of states to estimate the impulse response functions, thus making the responses more stable and precise. To compute multipliers conditional on the public capital scarcity and fiscal space, equation 4 is estimated using the following conditioning variables for ( ): (i) Gross government debt as a share of GDP as a proxy for fiscal space; and (ii) Public capital stock as a share of GDP to examine the implications of capital scarcity. Variable definitions are reported in Appendix table A2. 4. Empirical results 4.1. Stylized facts about public investment Before examining the estimated unconditional and state-dependent public investment multipliers, in this section we provide several key observations about public investment patterns in EMDEs. Public investment tends to play a greater role in EMDEs than in advanced economies. Over the past decade, public investment comprised about 7 percent of GDP, on average, versus about 4 percent of GDP in advanced economies (figure 1.A). Public investment amounted to almost 30 percent of total investment in EMDEs, with a greater share in low-income countries (about 40 percent). By contrast, in advanced economies public investment comprised less than a fifth of total investment in the past decade (figure 1.B). This pattern is consistent with the fact that the private sector tends to be weaker in EMDEs than in advanced economies, and often lacks the capacity to invest in large infrastructure projects. EMDEs thus tend to rely more on the public sector to deliver necessary infrastructure. In addition, creditors often see EMDE governments as more creditworthy than local private investors, given their power 19 The weights assigned to each regime vary between 0 and 1 according to the weighting function so that �, � can be interpreted as the probability of being in a given economic state. Following the literature that uses a similar approach, (Abiad et al. 2015 and Furceri and Li 2017), the parameter is set to 1.5, while the results do not change materially when other values are used. 13 to tax, their ownership of significant assets that can serve as collateral, and their capacity to pool the resources needed to execute large-scale projects (Martinez et al. 2023). Figure 1. Public investment in EMDEs A Public investment, percent of GDP B. Public investment, percent of total investment Percent of GDP Percent of total investment 12 50 10 40 8 30 6 20 4 2 10 0 0 EMDEs AEs LICs Com. Com. EMDEs AEs LICs Com. Com. Exp. Imp. Exp. Imp. C. Public investment growth D. Public investment growth by EMDE group Percent 1990-2022 average Percent 1990-2022 average 12 20 16 8 12 8 4 4 0 0 1990-99 2000-09 2010-22 1990-99 2000-09 2010-22 1990-99 2000-09 2010-22 1990-99 2000-09 2010-22 1990-99 2000-09 2010-22 1990-99 2000-09 2010-22 EMDEs EMDEs excl. World LICs Commodity Commodity China exporters importers Sources: Haver Analytics; IMF Investment and Capital Stock Dataset (IMF 2021a); WDI (database); World Bank. Note: AEs = advanced economies; EMDEs = emerging market and developing economies; Com. Exp. = commodity- exporting EMDEs; Com. Imp. = commodity-importing EMDEs; LICs = low-income countries. A.B. Bars show means and whiskers show interquartile ranges for 2010-22 by group. Sample includes up to 36 advanced economies and 126 EMDEs. C.D. Average annual public investment growth calculated with countries’ real public investment in constant international dollars as weights. Sample includes up to 162 economies, of which 126 are EMDEs, 23 are LICs, 76 are commodity exporting EMDEs, and 50 are commodity-importing EMDEs. Over the past three decades, public investment patterns in EMDEs have evolved notably. In the 1990s, public investment in EMDEs was growing at a rapid pace largely on account of robust growth in China. This was followed by exceptionally high public investment growth in the 2000s which witnessed a period of macroeconomic stability, rapid economic integration, and reduction of poverty, amid elevated commodity prices. The latter also resulted in accelerated public investment growth in commodity-exporting EMDEs (figures 1.C and 1.D). 14 However, public investment growth plunged after the global financial crisis. Average annual public investment growth in EMDEs halved, dropping from 10 percent in the 2000s to 5 percent in the 2010s—the slowest average pace over the past three decades (figure 1.C). This slowdown was associated with multiple factors: weaker economic growth in EMDEs in the aftermath of the global financial crisis; the worsening of the global macroeconomic environment resulting in the slowdown of trade and capital flows; heightened economic uncertainty; geopolitical tensions; tight financial conditions; elevated debt levels, eroding fiscal space; and commodity price volatility. In particular, public investment growth in commodity-exporting EMDEs—which was much higher than that in other EMDEs through the 2000s—slumped as global commodity prices declined, adversely affecting these countries’ public finances (figure 1.D). The COVID-19 recession further worsened the macroeconomic backdrop. Government expenditures to contain the pandemic and provide support to vulnerable population groups and the private sector were prioritized, resulting in cutbacks and delays in public investment spending. 4.2. Baseline unconditional public investment multipliers The estimations using the new methodological framework outlined in section 3 suggest that public investment shocks lead to positive output responses that remain highly statistically significant at the horizon of five years (Figure 2.A). In particular, an increase in public investment equivalent to 1 percent of GDP is associated with a gradual increase in output from 0.4 percent after one year, reaching 1.1 after five years. The output effects of public investment tend to be smaller in the short run but increase over the long term as the supply-side effects on productivity and productive capacity fully manifest themselves, consistent with the literature (Leduc and Wilson 2012; Ramey 2021). In the short run, offsetting fiscal effects, the impact of leakage through imports, possible transitory crowding out of private investment, private sector capacity constraints, and the time needed to adjust consumption and production may dampen the effects of public investment. 20 The estimated effects are broadly in line with public investment multipliers reported in past empirical work (see section 2 and Appendix B). A noteworthy aspect of the results is the high degree of heterogeneity across EMDEs. In higher- income EMDEs, positive public investment shocks lead to a strong and persistent impacts on output. In low-income countries, however, the effects on output are characterized by a wide dispersion, which translates to much lower statistical significance of public investment multipliers. Small sample size may play a part, but this could also be the result of lower efficiency in their public investment. That said, the average effect tends to be larger in low-income countries than in higher-income EMDEs, reaching up to 1.7 percent over the horizon of five years after a public investment shock equivalent to 1 percent of GDP. Public investment effects are slightly lower in commodity-exporting EMDEs than in other EMDEs (Appendix figure A1). 20 Similar results are reported in recent empirical studies (for instance, Abiad et al. 2016; Furceri and Li 2017; Ilzetzki et al. 2013). 15 Figure 2. Macroeconomic impacts of public investment in EMDEs A. Impact on output B. Impact on potential output Percent Percent 2.0 2.0 1.6 1.6 1.2 1.2 0.8 0.8 0.4 0.4 0.0 0.0 -1 0 1 2 3 4 5 -1 0 1 2 3 4 5 Year Year C. Medium-run impact on productivity and inflation D. Impact on private investment Percent Percent 5 4 4 3 3 2 2 1 1 0 0 -1 -1 -2 -3 -2 TFP Labor Inflation -1 0 1 2 3 4 5 productivity Year Source: Authors’ estimates. Note: Responses of variables (cumulative change in year t relative to year t = -1, in percent) to a public investment shock equivalent to 1 percent of GDP; t = 0 is the year of the shock. Shaded areas denote 90-percent confidence bands, based on standard errors clustered at the country level. A. Response of real GDP to a public investment shock. B. Response of real potential GDP to a public investment shock. C. Responses of total factor productivity (TFP), labor productivity, and consumer price index to a public investment shock after five years based on local projections. Bars indicate the point estimates, whiskers indicate 90-percent confidence intervals. D. Response of real private investment to a public investment shock. The analysis suggests that potential output also increases steadily in response to public investment. In a sample of EMDEs with the available data, a 1 percent of GDP rise in public investment leads to an increase in potential output peaking at about 1.1 percent over five years (Figure 2.B). 21 This effect is associated with a concurrent boost in productivity—by up to 0.8 percent for total factor productivity and 1.9 percent for labor productivity over the medium term (Figure 2.C). The impact of public investment surges on output does not lead to a corresponding increase in inflation. These 21 The sample size for the exercises with potential output and productivity is smaller because of data availability, and is not directly comparable to the baseline results, which use the full EMDE sample. 16 findings support the hypothesis that public investment can increase output through both short-term aggregate demand and longer-run aggregate supply channels, thereby boosting potential output (Ramey 2021). An important impact of public investment occurs via the crowding-in effect on private investment. An increase in public investment equivalent to 1 percent of GDP induces an increase in private investment by up to 2.2 percent at the horizon of five years (Figure 2.D). The estimates also suggest a possible crowding-out effect on impact; however, the effect is small, not statistically significant from zero, and is reversed within a year. The crowding-in effect on private investment is in line with the estimates reported in the literature (see, for instance, Eden and Kraay 2014; Furceri and Li 2017). In this regard, the results provide empirical support for policies to mitigate the slowdown of private investment through a scaling up of public investment. This effect could operate through several transmission channels. An increase in public capital can raise the return on private capital by facilitating connectivity (for instance, roads and bridges), thereby facilitating private sector investment (Aschauer 1989b; Eden and Kraay 2014). Public investment reduces uncertainty and risks associated with private investment in large infrastructure projects and may also directly crowd in private investment via public-private partnerships (IMF 2021b). 4.3. State-dependent public investment multipliers The role of the business cycle. Public investment multipliers, on average, are greater in magnitude during recessions than during expansions. A 1 percent of GDP increase in public investment yields an increase in output by 1.1 percent in times of expansion after five years. An equivalent public investment shock in recessions leads to an increase in output by up to 1.6 percent over the same period. However, the estimates during recessions are characterized by notable heterogeneity across countries, resulting in wider confidence bands (figure 3). These results are consistent with the empirical literature reporting larger government spending multipliers in recessions. 22 The position of an economy in the business cycle may affect the size of the multiplier for several reasons. In expansions, public spending stimulus may be less effective, because if the economy is operating close to full capacity, an additional increase in public spending is less likely to crowd in private sector resources. 23 In contrast, economic slack during recessions enables public investment to mobilize unused private sector capacity (Batini et al. 2014). Public spending during recessions may also help mitigate unemployment and improve market confidence and is less likely to be 22 Larger public investment multipliers in recessions are reported in Auerbach and Gorodnichenko (2012 and 2013); Caggiano et al. (2015); Furceri and Li (2017); Honda et al. (2020); Riera-Crichton et al. (2015). That said, such estimates may not be fully robust, with significant heterogeneity across countries (Ramey 2019). 23 That said, during expansions public investment also has a positive effect on output in EMDEs. This is in line with the view that EMDEs often have underutilized capacity because of infrastructure gaps, limited access to finance constraining the ability of the private sector to expand production capacity, and unused available labor resources, which can be engaged in expansions through public investment. 17 accompanied by increasing inflation and interest rates (Auerbach and Gorodnichenko 2012; Ghassibe and Zanetti 2022). In practice, however, EMDEs often have limited fiscal resources for public investment projects during recessions and crises. In fact, public investment tends to contract during economic distress. “Shovel-ready” investment projects may help revive economic activity and crowd in private investment during economic downturns as long as they are well-planned and executed, and do not undermine fiscal sustainability; such projects and conditions, however, may not always be present. Figure 3. Effects of public investment on output conditional on the business cycle A. Impact on output in recessions B. Impact on output in expansions Percent Baseline response Percent Baseline response 3.0 3.0 2.5 2.5 2.0 2.0 1.5 1.5 1.0 1.0 0.5 0.5 0.0 0.0 -0.5 -0.5 -1.0 -1.0 -1 0 1 2 3 4 5 -1 0 1 2 3 4 5 Year Year Source: Authors’ estimates. Note: Responses of real GDP (cumulative change in year t relative to year t = -1, in percent) to a public investment shock equivalent to 1 percent of GDP in recessions and expansions, defined as periods of negative and positive real GDP growth, respectively. Dashed lines indicate the baseline unconditional responses. Shaded area denotes 90 percent confidence bands, based on standard errors clustered at the country level. Implications of fiscal space. EMDEs with larger fiscal space, as measured by public-debt-to-GDP ratios, experience much stronger positive impacts of public investment: output increases by up to 1.6 percent five years after a public investment shock equivalent to 1 percent of GDP. Conversely, public investment in countries with high and rising debt (implying limited fiscal space) appears to be ineffective: the estimated public investment multipliers are lower and not statistically significant (figure 4). 24 While changes in public-debt-to-GDP ratios only partly reflect fiscal space dynamics, these results nevertheless imply that the effect of public investment on output in 24 These results thus are in line with the literature, arguing that in countries with high debt, public spending multipliers can be insignificant or even negative (Huidrom et al. 2020; Ilzetzki et al. 2013). 18 countries with large fiscal space is up to 1 percentage point higher than in countries with small fiscal space, on average. 25 Figure 4. Effects of public investment on output conditional on fiscal space A. Large fiscal space B. Small fiscal space Percent Baseline response Percent Baseline response 2.5 2.5 2.0 2.0 1.5 1.5 1.0 1.0 0.5 0.5 0.0 0.0 -0.5 -0.5 -1 0 1 2 3 4 5 -1 0 1 2 3 4 5 Year Year Source: Authors’ estimates. Note: Responses of real GDP (cumulative change in year t relative to year t = -1, in percent) to a public investment shock equivalent to 1 percent of GDP, conditional on fiscal space. t = 0 is the year of the shock. High-efficiency and low-efficiency samples are based on the top and bottom quartiles of the IMF (2021b) public infrastructure efficiency index, which ranges from 0 to 100 (the values above 81 and below 47, respectively). Fiscal space influences the output effects of public investment through two channels. The first is associated with the effects on private sector, as additional public spending in countries with weak fiscal positions may lead to lower disposable income of liquidity-constrained households, as well as increased tax burdens for the private sector in the future (which may also be anticipated). The second channel relates to the interest rate effect, as scaling up of government expenditures in countries with high levels of debt may lead to higher international interest rate spreads, on account of higher sovereign risk and inflation, thus increasing borrowing costs for the private sector (Blanchard 1990; Huidrom et al. 2020; Sutherland 1997). Infrastructure investment projects, given their large upfront costs and long time horizons, are often financed by borrowing rather than from current government revenues. Larger fiscal space implies that the sovereign has more capacity to service its borrowing and therefore is more creditworthy, allowing it to finance such investment at a lower interest rate. 25 The high-debt and low-debt states are defined using the smooth transition function that reflects the historical dynamics of public-debt-to-GDP ratios on a country-by-country basis. For a median EMDE, a low-debt state over the sample period corresponds to about 30 percent of GDP and a high-debt state corresponds to about 80 percent of GDP for EMDEs that experienced a public investment shock. 19 Implications of public investment efficiency. The efficiency of public investment plays a crucial role in driving its growth effects. 26 The estimates suggest a greater effect on GDP in response to public investment shocks in EMDEs with the highest efficiency, culminating in an increase in output of about 1.6 percent after five years—one-half percentage point higher than the effect of public investment in EMDEs with the lowest efficiency (figure 5). In countries with the lowest efficiency, the effects of public investment are lower and not statistically significant (albeit still positive). Figure 5. Effects of public investment on output conditional on public investment efficiency A. High investment efficiency B. Low investment efficiency Percent Baseline response Percent Baseline response 3.0 3.0 2.5 2.5 2.0 2.0 1.5 1.5 1.0 1.0 0.5 0.5 0.0 0.0 -0.5 -0.5 -1 0 1 2 3 4 5 -1 0 1 2 3 4 5 Year Year Source: Authors’ estimates. Note: Responses of real GDP (cumulative change in year t relative to year t = -1, in percent) to a public investment shock equivalent to 1 percent of GDP, conditional on fiscal space. t = 0 is the year of the shock. Dashed line indicates the baseline unconditional response. Shaded area denotes 90 percent confidence bands, based on standard errors clustered at the country level. High-efficiency and low-efficiency samples are based on the top and bottom quartiles of the IMF (2021b) public infrastructure efficiency index, which ranges from 0 to 100 (the values above 81 and below 47, respectively). These results are consistent with empirical studies using other samples and methods and provide support for the argument that low public investment efficiency is problematic. 27 Poor design, evaluation, and implementation of investment projects, including issues with corruption and governance, can deplete valuable fiscal resources without necessarily increasing the quantity or quality of public infrastructure that supports growth (Dabla-Norris et al. 2012; IMF 2014; Pritchett 26 The analysis uses the IMF (2021b) public infrastructure efficiency index, which is a cross-sectional index available for 120 countries (including 93 EMDEs), produced using the data envelopment analysis. The index ranges from 0 to 100, with higher values indicating better efficiency. The model was also estimated using the Devadas and Pennings (2018) infrastructure efficiency index and the Dabla-Norris et al. (2012) public investment management index. Estimations using alternative measures also suggest statistically insignificant and lower output effects of public investment in low-efficiency economies (see the results in the robustness section). 27 See, for instance, Cavallo and Daude (2011); Furceri and Li (2017); IMF (2014); Izquierdo et al. (2019); Leduc and Wilson (2012); and Leeper et al. (2010). 20 2000). Therefore, well-designed public investment management processes are essential to ensure the effectiveness of public investment. Initial public capital stock. The impact of public investment on output also varies with the initial level of the public capital stock (figure 6). The magnitude and statistical significance of the public investment multiplier tends to decrease with the level of public capital stock relative to GDP, consistent with expectations of diminishing marginal returns to capital. Specifically, a 1 percent of GDP increase in public investment is associated with a 1.7 percent increase in GDP after five years in capital-scarce countries. This contrasts with 0.9 percent (not statistically significant in the medium term) when the public-capital-stock-to-GDP ratio is high. Similar results are found in empirical studies using other samples and methods (for instance, Izquierdo et al. 2019). Figure 6. Effects of public investment on output conditional on public capital stock A. Small public capital stock B. Large public capital stock Percent Baseline response Percent Baseline response 2.5 2.5 2.0 2.0 1.5 1.5 1.0 1.0 0.5 0.5 0.0 0.0 -0.5 -0.5 -1 0 1 2 3 4 5 -1 0 1 2 3 4 5 Year Year Source: Authors’ estimates. Note: Responses of real GDP (cumulative change in year t relative to year t = -1, in percent) to a public investment shock equivalent to 1 percent of GDP, conditional on fiscal space. t = 0 is the year of the shock. Dashed line indicates the baseline unconditional response. Shaded area denotes 90 percent confidence bands, based on standard errors clustered at the country level. Small public capital stock and large public capital stock responses are based on local projections with the smooth transition function that uses public-capital-stock-to-GDP ratio as the conditioning variable to capture historically lowest and highest capital stock ratios of a given country in public investment shock years (the values of public capital stock of 68 and 113 percent of GDP for a median EMDE, respectively). 4.4. Robustness checks A range of robustness checks were carried out, and results were corroborative of baseline findings. These included testing alternatives for public investment shock identification, sensitivity checks to the choice of statistical filters, robustness checks to the sample period, and model specifications (selected results are reported in Table 1): 21 Table 1. Selected additional results and robustness checks Public investment Model multiplier t=1 t=5 Baseline specification 0.4*** 1.1*** IV (2SLS) estimation using public investment shocks to instrument public investment 0.4*** 1.0*** GMM estimation 0.5** 0.8** Dropping country fixed effects 0.4*** 1.5*** Dropping lagged real GDP growth variable 0.3** 1.1*** Pre-global financial crisis period only (1980-2007) 0.3* 0.9*** Additional control variables: two lags of inflation and trade-to-GDP ratio 0.4** 1.1*** Alternative fiscal space specification Large increase in debt-to-GDP ratio (upper quartile = above 3.7) 0.3 0.7 Large decrease in debt-to-GDP ratio (lower quartile = below -3.2) 0.5* 1.4*** Alternative public investment efficiency measures Low efficiency: Dabla-Norris et al. (2012) PIMI below the sample mean 0.1 0.9 High efficiency: Dabla-Norris et al. (2012) PIMI above the sample mean 0.4*** 1.2*** Low efficiency: Bottom quartile of Devadas and Pennings (2018) Infrastructure Efficiency index 0.2 0.3 High efficiency: Top quartile of Devadas and Pennings (2018) Infrastructure Efficiency index 0.2* 1.0*** Low efficiency: Bottom quartile of CPIA Public Sector Management and Institutions index 0.3 0.6 High efficiency: Top quartile of CPIA Public Sector Management and Institutions index 0.4** 1.1*** Source: Authors’ estimates. Note: 2SLS = two-stage least squares; CPIA = Country Policy and Institutional Assessment; GMM = generalized method of moments; IV = instrumental variables approach; PIMI = Public Investment Management Index. The table shows responses of real GDP (cumulative change in year t relative to year t = -1, in percent) to a public investment shock equivalent to 1 percent of GDP; t = 0 is the year of the shock. ***, **, and * indicate statistical significance at the 1, 5, and 10 percent levels, respectively. Sensitivity checks for public investment shock parameterization and statistical filters. Given that the identification of public investment shocks may be sensitive to the choice of the statistical filters or the cut-off level to isolate large changes in public investment, alternative threshold levels and filters were explored, including the Baxter-King, Christiano-Fitzgerald, and Hamilton filters. The results are not statistically different from the baseline results. Our focus only on large changes in cyclically adjusted public investment also mitigates concerns regarding the imprecision in the estimation of potential output. Using higher threshold levels to identify public investment shocks would come at the cost of a lower number of shock episodes, resulting in less precise estimates. Sensitivity to country fixed effects. A possible bias from estimating the baseline model using country fixed effects stems from the fact that the error term may have a non-zero expected value on account of the interaction between fixed effects and country-specific developments (Teulings and Zubanov 2014). Estimates excluding country fixed effects are similar to the baseline results. Omitted variables and the choice of estimator. The baseline model was estimated with additional variables, introduced to control for inflation and trade openness. The results indicate no 22 large differences relative to the baseline. As an additional check, the model was estimated dropping the lagged dependent variable. Results are also robust to using generalized method of moments as the estimator. As an alternative, the identified public investment shocks were also used as an instrument for a change in public investment as a share of GDP, in two-stage least squares (2SLS) estimation, yielding very similar results. Sensitivity to the sample period and the sample composition. In order to examine whether the effects of public investment may have changed in the aftermath of the global financial crisis, the multipliers were estimated also for the pre-2007 period, with the findings confirming the baseline results. As a sensitivity check, the estimations were also carried out using the same common sample of countries across all empirical exercises, however this results in less reliable estimates on account of a much smaller sample, with larger error bands. 5. Conclusion and policy implications In this paper, we contribute to the literature by introducing a new methodology to identify public investment shocks that can be easily replicated and extended by other researchers going forward. We use the identified public investment shocks to examine the macroeconomic implications of public investment for EMDEs using a broader sample than employed in the earlier literature, which allows a more granular analysis of cross-country heterogeneity and non-linear effects associated with prevailing macroeconomic conditions and country structural characteristics. The estimation results confirm the findings of previous research and expand the empirical evidence to a broad sample of EMDEs. Inter alia, our paper provides new insights into the heterogeneity of public investment multipliers across country groups and helps to cast light on the factors that explain the varying growth effects of public investment. In this regard, a particularly noteworthy aspect of our results is that the positive effects of public investment on economic growth can be significantly bolstered under certain conditions that are feasible to achieve with robust policy interventions. In particular, fiscal space and public spending efficiency are crucial for bolstering the positive effects of public investment. While the scaling up of public investment by 1 percent of GDP on average raises output by 1.1 percent after 5 years, in countries with ample fiscal space and high public investment efficiency, the effects are much higher—reaching 1.6 percent over the same period. These results help inform policy debates on the macroeconomic effects of public spending, particularly in the context of developing economies, which generally have larger infrastructure gaps and unmet funding needs to address the Sustainable Development Goals. The positive effects of public investment reflect both the beneficial aggregate demand and supply impacts, including through boosting the productive capacity of the economy and crowding in of private investment. The latter is especially important for EMDEs, given the prolonged weakness in private investment growth they have experienced over the past decade and their massive private capital mobilization needs to address investment gaps. 23 Furthermore, the results suggest that in countries with low fiscal space, public investment multipliers tend to be lower and statistically insignificant. Thus, limited fiscal space not only impedes the ability of a government to scale up public investment, but also undermines its effectiveness. This underscores the importance of boosting fiscal space during good times to help provide room for maneuvering during recessions. If public debt ratios do not decrease during times of economic expansion, countries may be unable to avail themselves of countercyclical spending increases to help dig themselves out of recessions. As the paper also shows, the scaling up of public investment is especially effective during recessions. Redoubled policy efforts to boost fiscal space—such as reforms to improve tax collection efficiency, enhance fiscal frameworks, and curtail unproductive expenditures—are particularly important in light of the significant debt accumulated over the past decade. Finally, our results accentuate the large payoffs that countries could realize by strengthening public investment efficiency. Policy interventions that can improve the efficiency of public investment range from broad structural reforms to improve the quality of institutions—tackling corruption, poor governance, and limited capacity of fiscal administration—to policies that focus more narrowly on enhancing public investment project management frameworks, such as planning, allocation, implementation, and monitoring of public investment projects. Low-income countries that face particularly deep structural challenges, vast infrastructure gaps, and limited fiscal resources, have much lower capacity to undertake such reforms on their own. Therefore, extensive coordinated support from the global community, including both financial support and technical assistance, is imperative to help these countries address their pressing development needs. 24 References Abiad, A., D. Furceri, and P. Topalova. 2016. “The Macroeconomic Effects of Public Investment: Evidence from Advanced Economies.” Journal of Macroeconomics 50: 224-40. Adam, C., and D. Bevan. 2006. “Aid and the Supply Side: Public Investment, Export Performance, and Dutch Disease in Low-Income Countries.” The World Bank Economic Review 20 (2): 261- 90. Adarov, A., and U. Panizza. 2024. “Public Investment Quality and Its Implications for Sovereign Risk and Debt Sustainability.” Policy Research Working Paper 10877. Washington, DC: World Bank. Afonso, A., and J. Alves. 2023. “Does Government Spending Efficiency Improve Fiscal Sustainability?” European Journal of Political Economy, 102403. Agenor, R. 2010. “A Theory of Infrastructure-Led Development.” Journal of Economics Dynamics and Control 34 (5): 932-50. Alesina, A., and S. Ardagna, 2010. “Large Changes in Fiscal Policy: Taxes versus Spending.” In J. Brown (ed.), Tax Policy and the Economy (University of Chicago Press), 24. Alesina, A., S. Ardagna, and J. Galí. 1998. “Tales of Fiscal Adjustment.” Economic Policy 13 (27): 489-545. Aschauer, D. 1989a. “Does Public Capital Crowd Out Private Capital?” Journal of Monetary Economics 24 (2): 171-88. Aschauer, D. 1989b. “Is Public Expenditure Productive?” Journal of Monetary Economics 23 (2): 177-200. Asian Development Bank. 2017. Meeting Asia’s Infrastructure Needs. Manila: Asian Development Bank. Auerbach, A. J., and Y. Gorodnichenko. 2012. “Measuring the Output Responses to Fiscal Policy.” American Economic Journal: Economic Policy 4 (2): 1-27. Auerbach, A. J., and Y. Gorodnichenko. 2013. “Fiscal Multipliers in Recession and Expansion.” In Fiscal Policy after the Financial Crisis, edited by A. Alesina and F. Giavazzi, 63-102. Chicago: University of Chicago Press. Barro, R. J. 1981. “Output Effects of Government Purchases.” Journal of Political Economy 89 (December): 1086-121. Barro, R. J. 1990. “Government Spending in a Simple Model of Endogenous Growth.” Journal of Political Economy 95: 103-26. Barro, R., and X. Sala-I-Martin. 1992. “Public Finance in Models of Economic Growth.” The Review of Economic Studies 59 (4): 645-61. Batini, N., L. Eyraud, A. Weber, and L. Forni. 2014. “Fiscal Multipliers: Size, Determinants, and Use in Macroeconomic Projections.” IMF Technical Notes and Manuals 2014/004, International Monetary Fund, Washington, DC. Berg, A., J. Gottschalk, R. Portillo, and L. Zanna. 2010. “The Macroeconomics of Medium-Term Aid Scaling-Up Scenarios.” IMF Working Paper 10/160, International Monetary Fund, Washington, DC. Berg, A., R. A. Portillo, E. F Buffie, C. A. Pattillo, and L. Zanna. 2012. “Public Investment, Growth, and Debt Sustainability: Putting Together the Pieces Together.” IMF Working Papers 2012/144, International Monetary Fund, Washington, DC. Berg, A., Portillo, R., and S. C. Yang, 2013. “Public Investment in Resource-Abundant Developing Countries.” IMF Economic Review 61: 92-129. 25 Blanchard, O. J., and R. Perotti. 2002. “An Empirical Characterization of the Dynamic Effects of Changes in Government Spending and Taxes on Output.” Quarterly Journal of Economics 117 (4): 1329-68. Blanchard, O. J. 1990. “Comment: can severe fiscal contractions be expansionary? Tales of two small European countries.” NBER Macroeconomics Annual 1990, 5 : 111-16. Bom, P., and J. Ligthart. 2014a. “Public Infrastructure Investment, Output Dynamics, and Balanced Budget Fiscal Rules.” Journal of Economic Dynamics and Control 40: 334-54. Bom, P., and J. Ligthart. 2014b. “What Have we Learnt from Three Decades of Research on the Productivity of Public Capital?” Journal of Economic Surveys 28 (5): 889-916. Caggiano, G., E. Castelnuovo, V. Colombo, and G. Nodari. 2015. “Estimating Fiscal Multipliers: News from a Nonlinear World.” Economics Journal 125 (584): 746-76. Canning, D., and P. Pedroni. 2008. “Infrastructure, Long-Run Economic Growth and Causality Tests for Cointegrated Panels.” The Manchester School 76 (5): 504-27. Cavallo, E., and C. Daude. 2011. “Public Investment in Developing Countries: A Blessing or a Curse?” Journal of Comparative Economics 39: 65-81. Chakraborty, S., and E. Dabla-Norris. 2011. “The Quality of Public Investment.” The B.E. Journal of Macroeconomics 11 (1): 1-29. Christiano, L., M. Eichenbaum, and S. Rebelo, 2011. “When Is the Government Spending Multiplier Large?” Journal of Political Economy 119 (1): 78-121. Colombo, E., D. Furceri, P. Pizzuto, and P. Tirelli. 2022. “Fiscal Multipliers and Informality.” IMF Working Paper No. 2022/082, International Monetary Fund, Washington, DC. Corsetti, G., K. Kuester, A. Meier, and G. J. Müller. 2013. “Sovereign Risk, Fiscal Policy, and Macroeconomic Stability.” Economic Journal 123 (566): F99-F132. Dabla-Norris, E., J. Brumby, A. Kyobe, et al. 2012. “Investing in Public Investment: an Index of Public Investment Efficiency.” Journal of Economic Growth 17: 235-66. De Haan, J., and J. Klomp. 2013. “Conditional Political Budget Cycles: A Review of Recent Evidence.” Public Choice 157: 387-410. Deleidi, M., F. Iafrate, and E. S. Levrero. 2020. “Public Investment Fiscal Multipliers: An Empirical Assessment for European Countries.” Structural Change and Economic Dynamics 52: 354-65. Demetriades, P. O., and T. P. Mamuneas. 2000. “Intertemporal Output and Employment Effects of Public Infrastructure Capital: Evidence from 12 OECD Economies.” Economic Journal 110 (465): 687-712. Devadas, S., and S. Pennings. 2018. “Assessing the Effect of Public Capital on Growth: An Extension of the World Bank Long-Term Growth Model.” Policy Research Working Paper WPS 8604, World Bank Group, Washington, DC. Devarajan, S., V. Swaroop, and H. Zou. 1996. “The Composition of Public Expenditure and Economic Growth.” Journal of Monetary Economic 37 (2): 313-44. Eden, M., and A. Kraay. 2014. ““Crowding in” and the Returns to Government Investment in Low- Income Countries.” Policy Research Working Paper Series 6781, The World Bank, Washington, DC. Elkhdari, M., M. Souissi, and M. A. Jewell. 2018. “Empirical Estimation of Fiscal Multipliers in MENA Oil-Exporting Countries with an Application to Algeria.” IMF Working Papers 2018/124, International Monetary Fund, Washington, DC. Erenburg, S. J., and M. E. Wohar. 1995. “Public and Private Investment: Are There Causal Linkages?” Journal of Macroeconomics 17 (1): 1-30. 26 Espinoza, R. A., and A. S. Senhadji. 2011. “How Strong are Fiscal Multipliers in the GCC? An Empirical Investigation.” IMF Working Paper, WP/11/16. International Monetary Fund, Washington, DC. Fay, M., H. Lee, M. Mastruzzi, S. Han, and M. Cho. 2019b. “Hitting the Trillion Mark—A Look at How Much Countries Are Spending on Infrastructure.” Policy Research Working Paper Series 8730, The World Bank, Washington, DC. Forni, M., and L. Gambetti. 2010. “Fiscal Foresight and the Effects of Government Spending.” CEPR Discussion Paper, 049. Foster, V., N. Gorgulu, D. Jain, S. Straub, and M. Vagliasindi. 2023. “The Impact of Infrastructure on Development Outcomes: A Meta-Analysis.” Policy Research Working Paper Series 10350, The World Bank, Washington, DC. Furceri, D., and B. Li. 2017. “The Macroeconomic (and Distributional) Effects of Public Investment in Developing Economies.” IMF Working Paper 17/217, Washington, DC. Futagami, K., Y. Morita, and A. Shibata. 1993. “Dynamic Analysis of an Endogenous Growth Model with Public Capital.” The Scandinavian Journal of Economics 95 (4): 607-25. Gbohoui, W. 2021. “Uncertainty and Public Investment Multipliers: The Role of Economic Confidence.” IMF Working Papers 2021/272, International Monetary Fund, Washington, DC. Gechert, S., and A. Rannenberg. 2018. “Which Fiscal Multipliers are Regime-Dependent? A meta- regression analysis.” Journal of Economic Surveys 32: 1160-82. Ghassibe, M., and F. Zanetti. 2022. “State Dependence of Fiscal Multipliers: the Source of Fluctuations Matters.” Journal of Monetary Economics 132: 1-23. Glomm, G., and B. Ravikumar. 1994. “Public Investment in Infrastructure in a Simple Growth Model.” Journal of Economic Dynamics and Control 18 (6): 1173-87. Gonzales-Garcia, J., A. Lemus, and M. Mrkaic, 2013. “Fiscal Multipliers in the ECCU,” IMF Working Paper No. 2013/117. International Monetary Fund, Washington, DC. Granger, C. W. J., and T. Terasvirta. 1993. “Modeling Nonlinear Economic Relationships.” New York: Oxford University Press. Gupta, S., A. Kangur, C. Papageorgiou, and A. Wane. 2014. “Efficiency-Adjusted Public Capital and Growth.” World Development 57: 164-78. Hamilton, J. D. 2018. “Why You Should Never Use the Hodrick-Prescott Filter.” The Review of Economics and Statistics 100 (5): 831-43. Huidrom, R., M. A. Kose, J. Lim, and F. Ohnsorge. 2020. “Why Do Fiscal Multipliers Depend on Fiscal Positions?” Journal of Monetary Economics 114 (1): 109-25. Honda, J., H. Miyamoto, and M. Taniguchi. 2020. “Exploring the Output Effect of Fiscal Policy Shocks in Low Income Countries.” IMF Working Papers 2020/012, International Monetary Fund, Washington, DC. Ilzetzki, E., E. G. Mendoza, and C. A. Végh. 2013 “How Big (Small?) are Fiscal Multipliers?” Journal of Monetary Economics, 60 (2): 239-54. IMF (International Monetary Fund). 2014. World Economic Outlook, October 2014: Legacies, Clouds, Uncertainties.” International Monetary Fund, Washington, DC. IMF (International Monetary Fund). 2021a. “Investment and Capital Stock Dataset.” Macroeconomic and Financial Data, International Monetary Fund, Washington, DC. https://data.imf.org/?sk=1ce8a55f-cfa7-4bc0-bce2-256ee65ac0e4 IMF (International Monetary Fund). 2021b. Fiscal Monitor, April 2021: A Fair Shot. International Monetary Fund, Washington, DC. 27 Izquierdo, A., R. E. Lama, J. Medina, J. P. Puig, D. Riera-Crichton, C. Vegh, and G. Vuletin. 2019. “Is the Public Investment Multiplier Higher in Developing Countries? An Empirical Investigation.” NBER Working Papers 26478. Jong-A-Pin, R., and J. de Haan, 2008. “Time-varying Impact of Public Capital on Output: New evidence based on VARs for OECD countries.” EIB Papers 3/2008. Jordà, O. 2005. “Estimation and Inference of Impulse Responses by Local Projections.” American Economic Review 95: 161-82. Kilic Celik, S., M. A. Kose, F. Ohnsorge, and F. U. Ruch. 2023. “Potential Growth: A Global Database.” Policy Research Working Paper 10354, World Bank, Washington, DC. Kim, J. H., J. A. Fallov, and S. Groom. 2020. Public Investment Management Reference Guide. Washington, DC: World Bank. Koh, W. C. 2017. “Fiscal Multipliers: New Evidence from a Large Panel of Countries.” Oxford Economic Papers 69 (3): 569-90. Kose, A. M., S. Kurlat, F. Ohnsorge, and N. Sugawara. 2022. “A Cross-Country Database of Fiscal Space.” Journal of International Money and Finance, 128, 102682. Kraay, A. 2012. “How Large Is the Government Spending Multiplier? Evidence from World Bank Lending.” The Quarterly Journal of Economics 127 (2): 829-87. Kraay, A. 2014. “Government Spending Multipliers in Developing Countries: Evidence from Lending by Official Creditors.” American Economic Journal: Macroeconomics 6 (4): 170-208. Leduc, S., and D. Wilson. 2012. “Roads to Prosperity or Bridges to Nowhere? Theory and Evidence on the Impact of Public Infrastructure Investment.” NBER Working Papers 18042. Leeper, E. M., A. W. Richter, and T. B. Walker. 2012. “Quantitative Effects of Fiscal Foresight.” American Economic Journal: Economic Policy 4 (2): 115-44. Leeper, E. M., T. B. Walker, and S. Yang. 2013. “Fiscal Foresight and Information Flows.” Econometrica 81 (3): 1115-45. Leeper, E., T. Walker, and S. Yang. 2010. “Government Investment and Fiscal Stimulus.” Journal of Monetary Economics 57: 1000-12. Martinez, L., F. Roch, F. Roldán, and J. Zettelmeyer. 2023. “Sovereign Debt.” In Research Handbook of Financial Markets, edited by R. S. Gürkaynak and J. H. Wright. 378-405. Northampton, MA: Edward Elgar Publishing. Mazzucato, M., and G. Semieniuk. 2017. “Public Financing of Innovation: New Questions.” Oxford Review of Economic Policy 33 (1): 24-48. Minea, A., and L. Mustea. 2015. “A Fresh Look at Fiscal Multipliers: One Size Fits it All? Evidence from the Mediterranean Area.” Applied Economics 47 (26): 2728-44. Miyamoto, H., et al. 2020. “Growth Impact of Public Investment and the Role of Infrastructure Governance.” In Well Spent, edited by Schwartz, F. et al. Washington, DC: International Monetary Fund. Petrović, P., M. Arsić, and A. Nojković. 2021. “Increasing Public Investment Can be an Effective Policy in Bad Times: Evidence from Emerging EU Economies.” Economic Modelling 94: 580- 97. Plagborg-Møller, M., and C. K. Wolf. 2021. “Local Projections and VARs Estimate the Same Impulse Responses.” Econometrica 89: 955-80. Pritchett, L. 2000. “The Tyranny of Concepts: CUDIE (Cumulated, Depreciated, Investment Effort) is Not Capital.” Journal of Economic Growth 5(4): 361-84. Puig, J. 2014. “Multiplicador del Gasto Público en Argentina.” Económica 60 (2014): 188-210. 28 Rafiq, S., and A. Zeufack. 2012. “Fiscal Multipliers over the Growth Cycle: Evidence from Malaysia.” World Bank Policy Research Working Paper, 5982. World Bank, Washington, DC. Rajaram, A., K. Kaiser, T. M. Le, J. H. Kim, and J. Frank, eds. 2014. The Power of Public Investment Management: Transforming Resources into Assets for Growth. Washington, DC: World Bank. Ramey, V. A. 2011a. “Can Government Purchases Stimulate the Economy?” Journal of Economic Literature, 49 (3): 673-85. Ramey, V. A. 2011b. “Identifying Government Spending Shocks: It’s All in the Timing.” Quarterly Journal of Economics, 126 (1): 51-102. Ramey, V. A. 2019. “Ten Years after the Financial Crisis: What Have We Learned from the Renaissance in Fiscal Research?” Journal of Economic Perspectives 33 (2): 89-114. Ramey, V. A. 2021. “The Macroeconomic Consequences of Infrastructure Investment.” In Economic Analysis and Infrastructure Investment, edited by Glaeser E. L., and J. M. Poterba, 219-76, Chicago: University of Chicago Press. Ramey, V. A., and M. D. Shapiro. 1998. “Costly Capital Reallocation and the Effects of Government Spending,” in Carnegie-Rochester Conference Series on Public Policy 48: 145-94. Ramey, V. A., and S. Zubairy. 2018. “Government Spending Multipliers in Good Times and in Bad: Evidence from US Historical Data.” Journal of Political Economy 126 (2): 850-901. Riera-Crichton, D., C. A. Vegh, and G. Vuletin. 2015. “Procyclical and Countercyclical Fiscal Multipliers: Evidence from OECD Countries.” Journal of International Money and Finance 52 (C): 15-31. Romp, W. E., and J. de Haan. 2005. “Public Capital and Economic Growth: A Critical Survey.” European Investment Bank (EIB) Papers, Luxembourg 10 (1): 41-70. Schwartz, G., M. Fouad, T. Hansen, and G. Verdier, eds. 2020. Well Spent: How Strong Infrastructure Governance Can End Waste in Public Investment. Washington, DC: International Monetary Fund. Straub, S. 2011. “Infrastructure and Development: A Critical Appraisal of the Macro-level Literature” The Journal of Development Studies 47 (5): 683-708. Sutherland, A. 1997. “Fiscal Crises and Aggregate Demand: Can High Public Debt Reverse the Effects of Fiscal Policy?” Journal of Public Economics 65: 147-62. Teulings, C. N., and N. Zubanov. 2014. “Is Economic Recovery a Myth? Robust Estimation of Impulse Responses.” Journal of Applied Econometrics 29 (3): 497-514. Turnovsky, S. 1997. “Fiscal Policy in a Growing Economy with Public Capital.” Macroeconomic Dynamics 1 (3): 615-39. Turnovsky, S. 2015. “Economic Growth and Inequality: The Role of Public Investment.” Journal of Economic Dynamics and Control 61: 204-21. Vagliasindi, M., and N. Gorgulu. 2021. “What Have We Learned about the Effectiveness of Infrastructure Investment as a Fiscal Stimulus? A Literature Review.” Policy Research Working Paper Series 9796, The World Bank. Warner, A. 2014. “Public Investment as an Engine of Growth.” IMF Working Paper 14/148, International Monetary Fund, Washington, DC. World Bank. 2024. Global Economic Prospects. June. Washington, DC: World Bank. World Bank—WDI (World Development Indicators) database. “World Development Indicators.” https://databank.worldbank.org/source/world-development-indicators Zachmann, G., E. Calthrop, A.-D. Riess, and A. Kolev. 2012. Investment and Growth in the Time of Climate Change. Brussels, Belgium: Bruegel. 29 APPENDIX A Table A1. Sample composition Albania Costa Rica Lao PDR Romania Algeria Côte d’Ivoire Lebanon Russian Federation Angola Djibouti Lesotho Rwanda Antigua and Barbuda Dominica Liberia São Tomé and Príncipe Argentina Dominican Republic Libya Saudi Arabia Armenia Ecuador Madagascar Senegal Azerbaijan Egypt, Arab Rep. Malawi Serbia Bahamas, The El Salvador Malaysia Seychelles Bahrain Equatorial Guinea Maldives Sierra Leone Bangladesh Eritrea Mali South Africa Barbados Eswatini Mauritania Sri Lanka Belize Ethiopia Mauritius St. Kitts and Nevis Benin Fiji Mexico St. Lucia Bhutan Gabon Moldova St. Vincent and the Bolivia Gambia, The Mongolia Grenadines Bosnia and Herzegovina Georgia Montenegro Sudan Botswana Ghana Morocco Syrian Arab Republic Brazil Grenada Mozambique Tajikistan Brunei Darussalam Guatemala Myanmar Tanzania Bulgaria Guinea Namibia Thailand Burkina Faso Guinea-Bissau Nepal Togo Burundi Guyana Nicaragua Tunisia Cabo Verde Haiti Niger Uganda Cambodia Honduras Nigeria Ukraine Cameroon Hungary North Macedonia United Arab Emirates Central African Republic India Oman Uruguay Chad Indonesia Pakistan Uzbekistan Chile Iran, Islamic Rep. Panama Viet Nam China Iraq Papua New Guinea Yemen, Rep. Colombia Jordan Paraguay Zambia Comoros Kazakhstan Peru Congo, Dem. Rep. Kenya Philippines Congo, Rep. Kuwait Poland Source: Authors. 30 Table A2. Definitions of data used and sources Variable Definition Source Real public investment General government investment (gross fixed capital formation) in Investment and Capital Stock billions of national currency deflated using the GDP deflator Dataset (IMF 2021a) Real private investment Private investment (gross fixed capital formation), in billions of Investment and Capital Stock national currency deflated using the GDP deflator Dataset (IMF 2021a) Real GDP Gross domestic product, in billions of national currency deflated IMF World Economic Outlook using the GDP deflator Database Potential GDP Index derived from real potential output growth estimated using Potential growth database (Kilic the production function approach Celik et al. 2023) Inflation Growth rate of consumer price index, in percent IMF World Economic Outlook Database Labor productivity Real GDP per average annual hours worked by persons engaged Penn World Table 10.01 Total factor productivity Total factor productivity in constant national prices (2017 = 1) Penn World Table 10.01 Public debt General government debt, percent of GDP World Bank’s Fiscal Space Database (Kose et al. 2022) Public infrastructure efficiency Public infrastructure efficiency index constructed based on the IMF (2021b) index (IMF 2021b) data envelopment analysis using the volume and quality of infrastructure as output, and public capital stock and per capita GDP as input variables Public infrastructure efficiency Infrastructure efficiency index constructed as a weighted average Devadas and Pennings (2018) index (Devadas and Pennings of the quality of electricity, water, and road infrastructure 2018) Public investment management Index based on country performance scores in public investment Dabla-Norris et al. (2012) index (PIMI) project appraisal, selection, implementation, and evaluation Public capital stock General government capital stock, percent of GDP Investment and Capital Stock Dataset (IMF 2021a) Source: Authors. Table A3. Summary statistics for public investment shocks Variable Value Number of economies 129 Public investment shock = 1 Number of observations 557 Mean real GDP growth (percent) 4.48 Mean public investment (percent of GDP) 8.11 Mean change in public-investment-to-GDP ratio (percentage points) 2.68 Public investment shock = 0 Number of observations 3804 Mean real GDP growth (percent) 3.79 Mean public investment (percent of GDP) 6.76 Mean change in public-investment-to-GDP ratio (percentage points) -0.43 Source: Authors. Note: Summary statistics for the sample of 129 EMDEs used in the estimation of public investment multipliers, differentiating between periods with and without public investment shocks. 31 Figure A1. Impact of public investment on real GDP by EMDE country groups A. EMDEs excluding low-income countries B. Low-income countries Percent Percent 1.6 3.5 1.4 3.0 1.2 2.5 1.0 2.0 1.5 0.8 1.0 0.6 0.5 0.4 0.0 0.2 -0.5 0.0 -1.0 -1 0 1 2 3 4 5 -1 0 1 2 3 4 5 Year Year C. Commodity-importing EMDEs D. Commodity-exporting EMDEs Percent Percent 2.0 2.0 1.5 1.5 1.0 1.0 0.5 0.5 0.0 0.0 -0.5 -0.5 -1 0 1 2 3 4 5 -1 0 1 2 3 4 5 Year Year Source: Authors’ estimates. Note: EMDEs = emerging market and developing economies; LICs = low-income countries. Responses of real GDP (cumulative change in year t relative to year t = -1, in percent) to a public investment shock equivalent to 1 percent of GDP; t = 0 is the year of the shock. Shaded areas denote 90 percent confidence bands, based on standard errors clustered at the country level. 32 APPENDIX B Table B1. Output effects of public investment: Summary of the literature Public investment Study Sample Methodology Notes multiplier Output growth in response to a 1 percent of GDP increase in 17 OECD economies; Local projections with Abiad et al. (2016) 0.4 (year 0) – 1.4 (year 4) public investment. Larger multipliers in low-growth episodes and 1985-2013 forecast error shocks in countries with high spending efficiency. Cumulative output effect in dollars in response to a 1 dollar Auerbach and 2.12 (peak response over 20 the United States; SVAR with forecast increase in public investment. Larger multipliers in recessions that Gorodnichenko (2012) quarters) 1947-2008 error shocks in expansion. SVAR with Blanchard Cumulative change in output in percent in response to a 1 percent 0.16 (quarter 1) – 1.10 Barry et al. (2018) Cameroon; 1999 -2015 and Perotti (2002) increase in public investment. Statistically insignificant for most (quarter 12) identification of the forecast horizon and in the longer run. Meta-regression estimates of output elasticity of public capital Bom and Ligthart 0.08 (short run) – 0.12 (long 68 studies over 1983– Meta-regression based on studies utilizing a production-function approach. Impact (2014b) run) 2008 analysis on output growth in response to a 1 percent increase in public capital. SVAR with Blanchard Output growth in percent, in response to a 1 percent increase in 0.1 (quarter 0) – 2.1 (quarter David (2017) Paraguay; 1998-2015 and Perotti (2002) public investment. Larger multipliers for public investment than 20) identification for public consumption. Output growth in response to a 1 percent increase in public [0.9 – 1.2] in year 0; 11 euro area countries; Deleidi et al. (2020) Local projections investment. Smaller multipliers in the pre-2007 period. The ranges [1.9 – 3.4] in year 6 1970-2016 of individual country estimates are in parentheses. Output elasticities. Impact on output of a 1 percent increase in Demetriades and [0.36 – 2.06] in the short run; 12 OECD countries; Simultaneous equations public capital. The ranges of individual country estimates are in Mamuneas (2000) [0.36 – 1.97] in the long run 1972-91 system parentheses. 39 low-income Eden and Kraay 2SLS with Kraay Output increase in dollars in response to a 1 dollar increase in 1.5 (year 1) countries (IDA (2014) (2012) identification public investment. Lower multiplier using OLS (0.2). borrowers) 0.3 (year 1) – 1.2 (year 5) Algeria; 2008-15 SVAR with Blanchard Output growth in response to a 1 percent increase in capital Elkhdari et al. (2018) 9 MENA countries; and Perotti (2002) expenditures. Larger multipliers during periods with negative 0.5 (year 1) – 1.8 (year 5) identification output gaps. 2000-15 Gulf Cooperation Output growth in response to a 15 percent increase in capital Espinoza and 0.2-0.3 in the short term; Council countries; Panel models expenditures. Larger multipliers for public investment than for Senhadji (2011) 0.6-1.1 in the long term 1975-2009 public consumption. Output growth in response to a 10 percent increase in public Local projections with investment. Larger multipliers during economic slack, in closed Furceri and Li (2017) 0.2 (year 1) – 0.4 (year 4) 79 EMDEs; 1990-2013 forecast error shocks economies, in countries with fixed exchange rates, lower public debt, and higher investment efficiency. 33 Table B1 (continued) Public investment Study Sample Methodology Notes multiplier Advanced economies; Output growth in response to a 1 percent of GDP increase in 0.55 (year 0) – 0.07 (year 2) Local projections with Gbohoui (2021) 1996- 2019 public investment. Larger multipliers during heightened forecast error shocks 0.22 (year 0) – 0.56 (year 2) EMDEs; 1990-2019 uncertainty. Output growth in response to a 1 percent of GDP increase in Gechert and 0.6 98 empirical studies Meta-regression analysis public investment. Larger multipliers during economic Rannenberg (2018) downturns. Cumulative output growth in response to a 1 percent of Eastern Caribbean SVAR with Blanchard and Gonzales-Garcia et al. 0.12 (year 0) – 0.44 (after 4 GDP increase in public investment. Larger multipliers for Currency Union; 1994- Perotti (2002) (2013) years) public investment than for public consumption (the latter are 2009 identification not statistically significant). Output growth in response to a 1 percent of GDP increase in 42 low-income countries; Local projections with public investment. Larger multipliers in recessions, under a Honda et al. (2020) 0.1 (year 1) – 0.2 (year 2) 1995-2017 forecast error shocks fixed exchange rate regime, in countries with better institutions. High income countries; Cumulative output response to public investment shocks. 0.4 (quarter 0) – 1.5 (quarter 20) SVAR with Blanchard and 1985–2013 Larger multipliers in countries with fixed exchange rates, Ilzetzki et al. (2013) Perotti (2002) Developing countries; closed economies, countries with low debt (not stat. 0.6 (quarter 0) – 1.6 (quarter 20) identification 1985–2013 significant in high-debt sample). Output response to a 1 percent of GDP increase in public Advanced economies; Local projections with 0.4 (year 0) – 1.5 (year 4) investment. Larger multipliers during low growth and in 1985–2013 forecast error shocks countries with higher spending efficiency. IMF (2014) Local projections with Output response to a1e percent of GDP increase in public (a) 0.3 in year 0 – 0.5 in year 4; Developing economies; shocks based on Corsetti investment. Shock identification: (a) Corsetti et al. (2013) (b) 0.5 in year 0 – 0.9 in year 4 1990–2013 et al. (2013) and Kraay approach based on fiscal rules; (b) Kraay (2012) (2012) methodology based on official development assistance. Local projections with Output growth in response to a 1 percent increase in public 31 European countries; Blanchard and Perotti investment. Larger multipliers in countries with low public Izquierdo et al. (2019) 0.2 – 1.4 after two years U.S. states; Argentine (2002), forecast error, and capital stock; statistically insignificant multipliers in provinces IV identification countries with low spending efficiency. Output elasticity of public capital at the horizon of 20 years. Jong-A-Pin and de 21 OECD countries over Ranging from about -2.5 to 2.5 VAR Output growth in response to a 1 percent increase in public Haan (2008) 1960-2001 capital. Cumulative output growth in response to a 1 percent Minea and Mustea Mediterranean countries; increase in public investment. Heterogeneous multipliers 0.53 (year 1) – 1.18 (year 10) PVAR (2015) 1980-2012 across country groups within the sample: larger in Asian, smaller in African countries. 34 Table B1 (continued) Public investment Study Sample Methodology Notes multiplier Output growth in response to a 1 percent of GDP increase 0.2 (year 0) – 1.2 (year 4) 17 advanced economies Local projections with in public investment. Statistically insignificant in low- Miyamoto et al. (2020) forecast error shocks income countries. Larger multipliers in countries with 0.2 (year 0) – 0.5 (year 4) 39 EMDEs better governance. Local projections and SVAR Cumulative output growth in percent in response to a 1 10 Central and Eastern Petrović et al. (2021) 0.7 – 0.8 (after one year) with Blanchard and Perotti percent increase in public investment. Larger multipliers European countries; (2002) identification in low-growth periods. Increase in output in dollars in response to a 1 dollar Puig (2014) 1.03 over two years Argentina; 1993-2012 SVAR increase in public investment. Greater impact of public investment than public consumption. Output growth in response to a 1 percent increase in Rafiq and Zeufack 2.7 in recessions; SVAR with Blanchard and Malaysia; 1981-2004 public investment. Greater impact of public investment (2012) 2.0 in expansions (year 1) Perotti (2002) identification than public consumption. Output per capita growth in response to a 1 percent 124 EMDEs; Warner (2014) 0.14 (year 0) OLS increase in public investment. Insignificant impact in the 1960-2011 long run. Source: Authors. Note: 2SLS = two-stage least squares; EMDEs = emerging market and developing economies; IDA = International Development Association; IV = instrumental variables approach; MNA = Middle East and North Africa; OECD = Organisation for Economic Co-operation and Development; OLS = ordinary least squares; PVAR = panel vector autoregression; SVAR = structural vector autoregression; VAR = vector autoregression. 35