Macroprudential policies and private domestic investment in developing countries: An instrumental variables approach

Bambe, Bao-We-Wal Working Paper Macroprudential policies and private domestic investment in developing countries: An instrumental variables approach IDOS Discussion Paper, No. 3/2025 Provided in Cooperation with: German Institute of Development and Sustainability (IDOS), Bonn Suggested Citation: Bambe, Bao-We-Wal (2025) : Macroprudential policies and private domestic investment in developing countries: An instrumental variables approach, IDOS Discussion Paper, No. 3/2025, ISBN 978-3-96021-246-1, German Institute of Development and Sustainability (IDOS), Bonn, https://doi.org/10.23661/idp3.2025 This Version is available at: https://hdl.handle.net/10419/313611 Standard-Nutzungsbedingungen: Die Dokumente auf EconStor dürfen zu eigenen wissenschaftlichen Zwecken und zum Privatgebrauch gespeichert und kopiert werden. Sie dürfen die Dokumente nicht für öffentliche oder kommerzielle Zwecke vervielfältigen, öffentlich ausstellen, öffentlich zugänglich machen, vertreiben oder anderweitig nutzen. Sofern die Verfasser die Dokumente unter Open-Content-Lizenzen (insbesondere CC-Lizenzen) zur Verfügung gestellt haben sollten, gelten abweichend von diesen Nutzungsbedingungen die in der dort genannten Lizenz gewährten Nutzungsrechte. Terms of use: Documents in EconStor may be saved and copied for your personal and scholarly purposes. You are not to copy documents for public or commercial purposes, to exhibit the documents publicly, to make them publicly available on the internet, or to distribute or otherwise use the documents in public. If the documents have been made available under an Open Content Licence (especially Creative Commons Licences), you may exercise further usage rights as specified in the indicated licence. https://creativecommons.org/licenses/by/4.0/ Macroprudential Policies and Private Domestic Investment in Developing Countries An Instrumental Variables Approach Bao-We-Wal Bambe IDOS DISCUSSION PAPER 3/2025 Macroprudential policies and private domestic investment in developing countries An instrumental variables approach Bao-We-Wal Bambe Bonn 2025 Bao-We-Wal Bambe is a researcher in the department “Transformation of Economic and Social Systems” at the German Institute of Development and Sustainability (IDOS) in Bonn. Email: [email protected] Published with financial support from the Federal Ministry for Economic Cooperation and Development (BMZ), based on a resolution of the German Bundestag. The institutes of the Johannes-Rau-Forschungsgemeinschaft are institutionally funded by the state of NRW. Suggested citation: Bambe, B.-W.-W. (2025). Macroprudential policies and private domestic investment in developing countries: Evidence from an instrumental variables strategy (IDOS Discussion Paper 3/2025). Bonn: German Institute of Development and Sustainability (IDOS). https://doi.org/10.23661/idp3.2025 Disclaimer: The views expressed in this paper are those of the author(s) and do not necessarily reflect the views or policies of the German Institute of Development and Sustainability (IDOS). Except otherwise noted, this publication is licensed under Creative Commons Attribution (CC BY 4.0). You are free to copy, communicate and adapt this work, as long as you attribute the German Institute of Development and Sustainability (IDOS) gGmbH and the author(s). IDOS Discussion Paper / German Institute of Development and Sustainability (IDOS) gGmbH ISSN 2751-4439 (Print) ISSN 2751-4447 (Online) ISBN 978-3-96021-246-1 (Print) DOI: https://doi.org/10.23661/idp3.2025 © German Institute of Development and Sustainability (IDOS) gGmbH Tulpenfeld 6, 53113 Bonn Email: [email protected] https://www.idos-research.de Printed on eco-friendly, certified paper. IDOS Discussion Paper 3/2025 III Abstract This paper examines the effect of macroprudential policies on private domestic investment using a panel of 87 developing countries from 2000 to 2017. Our instrumental variables strategy exploits the geographic diffusion of macroprudential policies across countries, with the idea that reforms in neighbouring countries can affect the adoption or strengthening of domestic reforms through peer pressure or imitation effects. The findings indicate that the tightening of macroprudential policies significantly reduces private domestic investment. This effect holds for both instruments targeting borrowers and those targeting financial institutions, and is subject to heterogeneity depending on several economic and institutional factors. The transmission channel analysis highlights that the negative impact of macroprudential policies on investment is primarily driven by a reduction in credit supply and financial inclusion. Keywords: Macroprudential policies; private domestic investment; developing countries; instrumental variables JEL Classification: E22; E44; G28 Acknowledgments I thank Kathrin Berensmann, Clara Brandi, Jean-Louis Combes, Tim Roethel, Christoph Sommer and Yabibal Walle for their valuable comments and suggestions to improve the paper. Usual disclaimers apply. IDOS Discussion Paper 3/2025 IV Contents Abstract Acknowledgments Abbreviations 1 Introduction 1 2 Background and theoretical predictions 2 3 Data and stylised facts 3 3.1 Data 3 3.2 Stylised facts 4 4 Methodology and main findings 8 4.1 Instrumental variables strategy 8 4.2 Econometric specification and main results 8 5 Robustness checks 9 5.1 Alternative specifications and additional controls 9 5.2 Alternative subsamples and measures 13 6 Heterogeneity 14 7 Mechanisms 17 8 Conclusions and policy recommendations 18 References 20 Appendix A: Further robustness 25 GMM estimates 25 Three-year window 25 Table A1: Macroprudential policies (MPI) and private domestic investment: alternative subsamples and measures 26 Table A2: Alternative subsamples and measures: first stage results 27 Table A3: Macroprudential policies and private domestic investment: System-GMM and IV-three-year window estimates 28 Table A4: Heterogeneity: first stage results 29 Appendix B: Sample and descriptive statistics 30 Table B1: Summary statistics of the baseline model variables 30 Table B2: Sample 31 Table B3: Sources of variables 32 IDOS Discussion Paper 3/2025 V Figures Figure 1: Trends in private domestic investment by income level 5 Figure 2: Trends in MPI by income level 6 Figure 3: Trends in MPI by category 6 Tables Table 1: Global macroprudential policy instruments survey 7 Table 2: The effect of macroprudential policies on private domestic investment 11 Table 3: The effect of macroprudential policies on private domestic investment: first stage results 12 Table 4: The effect of macroprudential policies on private domestic investment: heterogeneity 16 Table 5: Macroprudential policies and private domestic investment: channels 18 IDOS Discussion Paper 3/2025 VI Abbreviations CG credit growth CONC concentration limits DTI debt-to-income FC foreign currency loans FX foreign exchange GDP gross domestic product GMM generalised method of moments IMF International Monetary Fund ICRG International Country Risk Guide INTER interbank exposures IV instrumental variables LEV leverage LTV loan-to-value MPI macroprudential policy index SIFI systemically important financial institutions WDI World Bank’s World Development Indicators IDOS Discussion Paper 3/2025 1 1 Introduction Macroprudential policies are increasingly used in advanced and developing economies, especially since the 2008–2009 global financial crisis and following Basel III, which introduced a comprehensive framework aimed at reforming banking oversight, regulation and risk control – such as enhanced capital adequacy requirements, improved leverage management, new capital and liquidity buffers, and leverage ratio limits (Rubio & Carrasco-Gallego, 2016).1 The role of macroprudential policies in promoting financial stability by regulating credit cycles is all the more crucial, as studies show that financial crises are more likely to occur when they are preceded by private credit booms (Gertler et al., 2020; Schularick & Taylor, 2012). Empirical evidence from a growing body of literature suggests that macroprudential policies tend to reduce credit procyclicality, credit growth, and house prices (e.g. see Alam et al., 2019; Cerutti et al., 2017; De Schryder & Opitz, 2021; Gómez et al., 2020; Kuttner & Shim, 2016; Lim et al., 2011; Teixeira & Venter, 2023). Another strand of the literature shows that by moderating credit and asset price cycles, macroprudential tools help constrain financial cycles and systemic risks (for instance, see Altunbas et al., 2018; Belkhir et al., 2022; Bianchi & Mendoza, 2018; Claessens et al., 2013; Gertler et al., 2020; Fernandez-Gallardo, 2023). While the literature has widely examined the effects of macroprudential policies on credit growth or procyclicality and financial stability, our study complements existing studies by assessing the side effects of these policies on private domestic investment in developing countries.2 The effect of macroprudential policies on private domestic investment is not so clear-cut. If successful, they should improve financial stability, thus contributing to reducing economic uncertainty and improving private sector investment over time – especially as financial crises are found to hamper economic growth, employment and investment (e.g. see Barro, 2001; Reinhart & Reinhart, 2015). On the other hand, certain macroprudential policies such as countercyclical capital buffers, liquidity tools, or those targeting borrowers (such as loan-to-value or debt-to-income ratios) have restrictive effects on credit, thereby worsening financial inclusion (Aiyar et al., 2014; Ayyagari et al., 2018; Deléchat et al., 2021). Lower credit supply resulting from macroprudential policies may penalise private sector investment, especially in developing countries, where numerous firms already face significant challenges in accessing adequate financing (Beck et al., 2005; Chauvet & Jacolin, 2017; Harrison et al., 2004). This paper examines the effect of macroprudential policies on private domestic investment in developing countries. Macroprudential policies are likely to be correlated with (in)observable factors that could also affect the economy’s overall performance, including domestic investment. Therefore, to mitigate endogeneity, we draw on existing studies to exploit an exogenous source of variation, instrumenting macroprudential policies by the average regional macroprudential policy index. The results from our instrumental-variables (IV) strategy suggest that macroprudential policies reduce private domestic investment. The effects are statistically and economically significant, and robust to various tests. Heterogeneity analyses show that the negative effect of macroprudential policies on private investment is observed for both policies targeting 1 The International Monetary Fund (IMF) defines macroprudential policy as “the use of primarily prudential tools to limit systemic risk or system-wide financial risk” (IMF, 2011). See Clement (2010) for the origins and evolution of the term. 2 We focus on developing countries, i.e., emerging and low-income economies, for two main reasons. First, the latter have experienced a surge in macroprudential tools to improve financial stability in recent years, and in contrast to their advanced counterparts, the private sector in these regions is severely penalized by low levels of investment, thus creating a further challenge to achieving development goals. Second, focusing on developing countries allows us to have a relatively homogeneous sample of countries, for instance in terms of several economic, structural, and institutional factors (e.g. per capita income, vulnerability to external shocks, quality of institutions, access to financial markets, etc.). IDOS Discussion Paper 3/2025 8 4 Methodology and main findings 4.1 Instrumental variables strategy Macroprudential policies may be endogenous, as their adoption or tightening may be associated with other alternative measures, creating an identification bias due to unobservables. In other words, estimating a causal effect is challenging, as it is difficult to determine whether the observed effect is genuinely due to macroprudential policies or rather to alternative policies. Hence, purging macroprudential policy actions to resolve potential endogeneity problems is crucial. At the same time, finding a relevant and valid instrument is widely acknowledged as a challenging task. Many studies examining the effect of reforms explore the regional diffusion of these reforms as an instrumental variable, in the idea that structural reforms often occur as regional waves. The underlying intuition is that reforms in neighbouring countries can have a strong spillover effect in the adoption or strengthening of domestic reforms – via simple imitation mechanisms, peer pressure, learning or competition (see Buera et al., 2011; Dobbin et al., 2007; Huntington, 1991; Shipan & Volden, 2008). Studies exploring the influence of regional democratic reforms as a driving force or instrument for initiating national democratic reforms include, among others, Acemoglu et al., 2019; Giuliano et al., 2013; Kalenborn & Lessmann, 2013; Persson & Tabellini, 2009. Similarly, other works exploit fiscal rules in neighbouring countries, with the idea that countries draw on the experience of their neighbours when introducing such reforms (e.g. see Altunbaş & Thornton, 2017; Apeti et al., 2024a; Balvir, 2024; Caselli & Reynaud, 2020). In the case of monetary reforms, for instance, Balima et al. (2017) use the proportion of neighbouring countries that have adopted inflation targeting as an instrumental variable for a country’s decision to adopt the monetary framework. Based on the literature discussed above, we use the average of macroprudential policies in regional countries as an instrumental variable for domestic macroprudential policies, to provide a source of exogenous variation. There are several reasons why a country may adopt or strengthen its macroprudential policies, drawing on the experience of its neighbouring countries. For example, economic integration may lead a country to strengthen its macroprudential regulations due to pressure from its peers, to reduce systemic risks. The spread of macroprudential policies in the Eurozone, under the ECB’s influence, is a striking example. The same applies in Africa, where macroprudential policies have spread among the member countries of monetary unions to strengthen the resilience of banking systems in response to economic shocks. In Asia, many countries adopted or strengthened their macroprudential tools after the 1997 Asian financial crisis to prevent future financial instability. Our identifying hypothesis is therefore that macroprudential regulations in regional peer countries can play an important role in strengthening domestic macroprudential policies, without directly affecting private sector investment in the domestic country – conditional on the vector of controls. However, we recognise that if macroprudential policies successfully reduce systemic risks, they may (indirectly) affect regional economic performance. To address this limitation of our IV methodology, we include regional banking crises and regional economic growth as additional controls in our main regression for robustness. 4.2 Econometric specification and main results We estimate the effect of macroprudential policies on private domestic investment based on the following econometric specification Yit = αi + βXit + ηZit + µi + ψt + εit (1) where Yit represents private domestic investment (as a percentage of GDP) for a country i in year t. Xi,t is the macroprudential policy index, and Zit is the set of control variables of the baseline model. µi and ψt denote country and time-fixed effects, respectively. Country-fixed effects IDOS Discussion Paper 3/2025 9 capture unobserved country-specific and time-invariant factors; and time-fixed effects account for common time-varying shocks correlated with macroprudential policies and private domestic investment. εit is the usual residual error term. Column [1] of Table 2 reports the main results, including the baseline model controls and country and year fixed effects. The coefficient on the macroprudential policy index is negative and significant at the 1% threshold, suggesting that macroprudential policies are associated with a significant drop in private sector investment in developing countries. Specifically, a one-unit increase in the Cerutti et al. (2017) index is associated with a drop of roughly 1.4 percentage points in private domestic investment. More importantly, statistical tests show that this result is not due to a lack of relevance of the instrument. Indeed, the Kleibergen-Paap F-statistic of the baseline model gives a value well above the value of 10 of the rule of thumb of Staiger and Stock (1997), suggesting that, in our exactly identified model, the weak instrument bias is rather low (Angrist & Pischke, 2009). Another way to check the instrument’s relevance is to refer to the results of the first-stage equation reported in Table 3. These results show a positive and significant effect of regional macroprudential policies on domestic macroprudential policies, reinforcing the relevance of the instrument. Lastly, with regard to the baseline model controls, the results show that employment, corruption control and per capita income enhance investment, while government durability is negatively associated. Another important question relates to the economic size of the main estimates. In our sample and over our study period, we report an average private domestic investment of 13% of GDP. Consequently, the main results suggest that for an average country in the sample, a one-unit increase in the Cerutti et al. (2017) index — or MPI — reduces private investment by around 11%, indicating an economically significant effect. 5 Robustness checks In the previous section, we have established a statistically and economically significant negative effect of macroprudential policies on private domestic investment in developing countries. In this section, we conduct a series of robustness tests. Specifically, we re-estimate Equation 1 by including additional controls, considering alternative subsamples and measures of macroprudential policies, and using the System-GMM method, respectively. 5.1 Alternative specifications and additional controls First, we consider alternative specifications by lagging the instrumental variable by one and two years respectively. This allows us to account for potential lags in convergence dynamics between regional and domestic macroprudential policies. In other words, we consider that the regional diffusion of reforms could increase over time, with potential influence on the baseline model estimates. Although the new coefficients reported in Columns [2] and [3] of Table 2 increase slightly compared to that of the baseline model, their magnitude remains very similar. The same applies to the coefficients of the first-stage regression reported in Columns [2] and [3] of Table 3. Second, instead of the KOF Globalisation Index used in the main model, in Column [4], we consider the trade openness variable from the World Bank’s WDI database, measured as the sum of exports and imports as a percentage of GDP. The results hold. We further augment the baseline model using additional controls, which might affect our dependent variable. First, in addition to the trade globalisation variable included in the main model, we also control for financial openness. Next, we include the terms of trade to capture the potential influence of costly shocks on private domestic investment. We further consider monetary factors such as the inflation rate, the exchange rate regime and the real effective exchange rate. We expect inflation to reduce domestic investment, via the macroeconomic uncertainty and volatility it generates (e.g. see Bambe et al., 2024; Bloom et al., 2007; Dixit & IDOS Discussion Paper 3/2025 10 Pindyck, 1994; Lucas Jr, 1967; Nickell, 1974). The exchange rate regime may equally be a key determinant of inflation performance and macroeconomic volatility, with potentially important side effects on private sector investment decisions (e.g. see Edwards, 1993; Ghosh et al., 1996). Appreciations in the real effective exchange rate may dampen investment by exacerbating competition challenges in the export sector. Fourth, we consider public investment and remittances, since they can exert significant upward or downward effects on private domestic investment (Borensztein et al., 1998; Buiter, 1977; Chauvet & Jacolin, 2017; Dash, 2023; Fry, 1993). Fifth, we include natural resources, which are probably a key determinant of privatesector investment, particularly in the manufacturing sector. Indeed, the literature shows that natural resource booms in highly dependent countries tend to generate exchange rate appreciations, with adverse effects on the competitiveness of the non-extractive sector (see Corden, 1984 and Sachs & Warner, 2001 for pioneering work on the literature dealing with Dutch disease). Sixth, we complement the institutional variables of the baseline model by including the V-Dem (Varieties of Democracy) “property rights” index and the QOG (Quality of Government) “political pressures and controls on the media index”. Seventh, we consider other reforms that could also impact domestic investment, including IMF programmes and an index of economic freedom that includes 12 quantitative and qualitative variables – from property rights to financial freedom – capturing government reforms.8 Lastly, to reduce the risk that the effect observed in the baseline model is biased by the potential influence of the instrument on regional performance, we account for regional GDP growth and regional banking crises. Columns [5]-[18] of Table 2 include the new controls independently, and the last column considers them in the same regression. The new coefficients for the variable of interest remain strongly robust. The same holds for the baseline model controls, except for institutional variables, probably due to the high correlation with the new institutional indices. Regarding the new controls, we find that capital openness increases private domestic investment, while natural resources are negatively associated. More importantly, accounting for IMF programmes and economic freedom does not alter our results, suggesting that the effects obtained are likely due to macroprudential policies and not to alternative economic reforms or policies. Similarly, accounting for regional economic growth and regional banking crises does not affect our results, which is reassuring, as it reinforces the exclusion condition stipulated above. 8 The terms of trade, inflation, natural resources, and remittances are from the World Bank’s World Development Indicators (WDI) database. Capital openness is from Chinn and Ito (2008) and varies approximately between -2 and 2 (higher values indicate greater openness). The exchange rate regime is computed based on Ilzetzki et al. (2019)’s classification. We construct a dummy equal to 1 if country i is classified as having a fixed exchange rate regime in the year t, and to 0 otherwise. The real effective exchange rate variable is from Darvas (2012), 2007 is the reference year with a base of 100. Public investment is proxied by government gross fixed capital formation (as a percentage of GDP) and is from the IMF’s Investment and Capital Stock dataset. IMF programmes are captured by a dummy variable equal to 1 if a country i has benefited from any type of IMF-supported programme in the year t, and to 0 otherwise. The variable is from Dreher (2006b). The V-Dem property rights index ranges between 0.1 and 0.9 in our sample, where higher values indicate better performance. The QOG index of political pressures and controls on the media index ranges between 2 and 40 in our sample, where 2 indicates better performance. The economic freedom/reform index can range from 0 to 100 and is from the Heritage Foundation. Following Furceri and Loungani (2018), we identify major reform episodes using a dummy equal to 1 when, for a given country at a given time, the annual change in the index exceeds the overall average annual change across all observations by two standard deviations, and 0 otherwise. Banking crises are captured by a dummy equal to 1 during times of crisis, and 0 otherwise, and are from Laeven and Valencia (2020). Table 2: The effect of macroprudential policies on private domestic investment (1) (2) (3) (4) (5) (6) (7) (8) (9) (10) (11) (12) (13) (14) (15) (16) (17) (18) (19) MPI -1.430 *** -2.071 *** -3.104 *** -1.327 *** -1.494 *** -1.600 *** -1.397 *** -1.701 *** -1.560 *** -1.407 *** -1.499 *** -1.526 *** -1.384 *** -1.675 *** -1.462 *** -1.416 *** -1.025 ** -1.410 *** -1.597 *** (0.446) (0.566) (0.819) (0.416) (0.485) (0.477) (0.451) (0.493) (0.438) (0.432) (0.438) (0.451) (0.434) (0.490) (0.466) (0.442) (0.445) (0.449) (0.526) Log. Employment 10.406*** 11.298*** 14.588*** 11.861*** 12.777*** 11.006*** 10.825*** 9.604*** 11.528*** 10.329*** 11.023*** 10.778*** 11.580*** 9.825*** 13.016*** 10.446*** 10.591*** 10.316*** 14.964*** (2.248) (2.593) (3.547) (2.257) (2.294) (2.402) (2.342) (2.441) (2.335) (2.213) (2.322) (2.247) (2.252) (2.399) (2.268) (2.259) (2.097) (2.248) (2.509) Trade globalisation -0.007 -0.011 -0.026 -0.018 -0.012 -0.008 0.003 0.001 -0.006 -0.002 0.008 -0.012 0.003 -0.013 -0.006 -0.005 -0.006 0.027 (0.017) (0.018) (0.023) (0.018) (0.017) (0.017) (0.017) (0.018) (0.017) (0.017) (0.017) (0.017) (0.017) (0.017) (0.017) (0.016) (0.017) (0.020) Corruption control 0.515*** 0.591*** 0.457* 0.408** 0.387** 0.515*** 0.434** 0.623*** 0.508*** 0.516*** 0.189 0.534*** 0.348* 0.604*** 0.431** 0.516*** 0.487*** 0.517*** 0.054 (0.182) (0.202) (0.240) (0.175) (0.190) (0.188) (0.183) (0.193) (0.184) (0.182) (0.190) (0.183) (0.191) (0.200) (0.182) (0.182) (0.175) (0.182) (0.206) Log. Government durability -0.297*** -0.304** -0.285* -0.237** -0.343*** -0.285** -0.367*** -0.249** -0.151 -0.298*** -0.181 -0.285*** -0.314*** -0.257** -0.352*** -0.303*** -0.270** -0.295*** -0.077 (0.109) (0.119) (0.147) (0.119) (0.116) (0.113) (0.110) (0.118) (0.114) (0.108) (0.116) (0.110) (0.116) (0.118) (0.110) (0.109) (0.105) (0.109) (0.143) Log. GDP per capita 5.970*** 6.626*** 8.006*** 5.252*** 5.967*** 6.153*** 5.971*** 6.174*** 5.170*** 5.927*** 7.579*** 5.855*** 6.254*** 6.160*** 5.933*** 5.968*** 5.800*** 5.972*** 6.916*** (1.183) (1.293) (1.663) (1.393) (1.209) (1.199) (1.221) (1.201) (1.331) (1.189) (1.267) (1.192) (1.242) (1.193) (1.186) (1.182) (1.124) (1.176) (1.419) Observations 1412 1412 1329 1316 1350 1398 1373 1330 1395 1412 1380 1412 1361 1330 1378 1412 1412 1412 1178 R-squared 0.785 0.752 0.681 0.793 0.783 0.777 0.783 0.782 0.780 0.786 0.791 0.783 0.787 0.783 0.785 0.786 0.802 0.786 0.806 KleibergenPaap LM stat (p-value) 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 KleibergenPaap F-stat 58.51 44.35 26.01 68.13 50.28 54.53 57.21 51.13 64.19 63.88 58.79 58.28 62.74 51.32 54.09 59.29 54.81 58.29 43.84 This table reports estimates of the effect of macroprudential policies on private domestic investment, using the instrumental variables (IV). The instrument (Contiguity) is the average macroprudential policy index in regional countries. Column [1] displays the main results. In Columns [2] and [3], we consider alternative specifications by lagging the instrumental variable by one and two years, respectively. In Column [4], instead of the KOF Globalisation Index used in the main model, we consider the trade openness variable from the World Bank’s WDI database, measured as the sum of exports and imports as a percentage of GDP. Columns [5]-[18] include the following additional controls independently, and the last column considers them in the same regression: capital openness, log. terms of trade, inflation, exchange rate regime, real effective exchange rate, public investment, remittances, log. natural resources, property rights, political pressures and controls on media, IMF programmes, economic freedom/reforms, regional economic growth and regional banking crises. Regarding the new controls, we find that capital openness increases private domestic investment, while natural resources are negatively associated. Robust standard errors are in parentheses. All regressions include the constant, not reported in the table. * p < 0.1, ** p < 0.05, *** p < 0.01 Table 3: The effect of macroprudential policies on private domestic investment: first stage results (1) (2) (3) (4) (5) (6) (7) (8) (9) (10) (11) (12) (13) (14) (15) (16) (17) (18) (19) Contiguity 0.851 *** 0.941 *** 0.803 *** 0.828 *** 0.852 *** 0.851 *** 0.893 *** 0.884 *** 0.859 *** 0.839 *** 0.893 *** 0.856 *** 0.821 *** 0.854 *** 0.859 *** 0.852 *** 0.861 *** (0.107) (0.110) (0.109) (0.108) (0.108) (0.114) (0.107) (0.106) (0.108) (0.106) (0.108) (0.115) (0.107) (0.107) (0.112) (0.107) (0.124) Log. Employment 0.874 0.945 1.586 ** 0.664 0.989 0.961 0.958 0.508 0.492 0.772 0.871 0.918 0.853 0.367 1.175 * 0.889 0.888 0.869 0.003 (0.659) (0.673) (0.711) (0.680) (0.641) (0.667) (0.678) (0.668) (0.667) (0.644) (0.675) (0.653) (0.674) (0.655) (0.637) (0.667) (0.663) (0.662) (0.662) Trade globalisation -0.004 -0.006 -0.009 ** -0.007 * -0.005 -0.003 -0.001 -0.006 -0.004 -0.003 -0.002 -0.002 -0.001 -0.006 -0.004 -0.004 -0.004 0.000 (0.004) (0.004) (0.005) (0.004) (0.004) (0.004) (0.004) (0.004) (0.004) (0.004) (0.004) (0.004) (0.004) (0.004) (0.004) (0.004) (0.004) (0.005) Corruption control 0.051 0.066 0.045 0.009 0.037 0.053 0.037 0.054 0.059 0.054 0.039 0.053 0.060 0.076 * 0.041 0.052 0.051 0.051 0.052 (0.042) (0.043) (0.045) (0.044) (0.044) (0.043) (0.042) (0.041) (0.041) (0.041) (0.043) (0.042) (0.045) (0.042) (0.042) (0.042) (0.042) (0.042) (0.046) Log. Government durability 0.013 0.011 0.014 -0.044 0.012 0.015 0.001 0.028 -0.020 0.011 0.005 0.015 0.014 0.027 0.009 0.012 0.014 0.014 -0.031 (0.034) (0.034) (0.035) (0.038) (0.036) (0.035) (0.035) (0.037) (0.036) (0.034) (0.037) (0.034) (0.037) (0.036) (0.035) (0.035) (0.035) (0.034) (0.045) Log. GDP per capita 0.326 0.373 0.498 * 0.765 ** 0.312 0.353 0.327 0.278 0.529 * 0.262 0.424 0.304 0.332 0.278 0.333 0.327 0.326 0.327 0.550 * (0.259) (0.261) (0.285) (0.302) (0.262) (0.259) (0.267) (0.263) (0.276) (0.264) (0.271) (0.259) (0.271) (0.262) (0.260) (0.259) (0.259) (0.259) (0.300) t-1 Contiguity 0.768 *** (0.111) t-2 Contiguity 0.624 *** (0.117) Observations 1412 1412 1329 1316 1350 1398 1373 1330 1395 1412 1380 1412 1361 1330 1378 1412 1412 1412 1178 R-squared 0.870 0.868 0.870 0.872 0.875 0.870 0.868 0.879 0.872 0.872 0.873 0.871 0.871 0.880 0.874 0.870 0.870 0.870 0.890 This table reports the results of the first stage IV estimation of Table 2. Robust standard errors are in parentheses. All regressions include the constant, not reported in the table. * p < 0.1, ** p < 0.05, *** p < 0.01 IDOS Discussion Paper 3/2025 13 5.2 Alternative subsamples and measures Next, we explore the sensitivity of our main results using alternative subsamples. We start by re-estimating our baseline specification, excluding hyperinflation periods, defined in Lin and Ye (2009) as country-year observations with inflation rates of 40% or more.9 The underlying intuition is that, since such high inflation rates would reflect significant economic imbalances, this could affect private sector investment independently of macroprudential policies. Along the same lines, we exclude from the sample the years 2008–2009, given the economic imbalances resulting from the global financial crisis. Third, our sample includes 14 fragile states, classified by the IMF as countries with strong economic, institutional and structural vulnerabilities hindering their economic development. Given that these cases exhibit significantly different characteristics from the rest of the sample, it is worth considering whether their inclusion in the study affects the main results. Hence, for robustness, we exclude them from the main sample. Fourth, we explore the sensitivity of our results to outliers, by excluding country-year observations with values above the 95th percentile of the sample, for the variable of interest and the dependent variable respectively. Fifth, we re-estimate the baseline model using exclusively countries that have implemented at least one macroprudential instrument during the study period. In other words, we exclude from the main sample the eight countries for which the macroprudential policy index is zero over the entire study period thus focusing solely on within-country variation.10 In all cases (Columns [1]-[6] of Table A1 in the appendix), the results are very similar to our baseline results, indicating that outliers or specific subsamples do not drive our results. Furthermore, the Kleibergen-Paap F-statistics and the results of the first-stage equation support the relevance and the instrument in all cases. In Column [7] of Table A1, we further consider an alternative measure of the instrumental variable, weighting the instrumental variable by annual GDP to assign greater influence to regional neighbours with larger economic size. Next, the macroprudential policy data used so far is drawn from the Cerutti et al. (2017) database, which covers 162 advanced and developing countries from 2000 to 2017. The integrated Macroprudential Policy (iMaPP) database, published by the International Monetary Fund (Alam et al., 2019), also provides a summary measure of macroprudential actions for a panel of 135 countries over 1990–2021. The main advantage of the Cerutti et al. index is that it covers a larger sample of countries, enabling greater international comparability, albeit over a relatively shorter period than the IMF database. The iMaPP dataset includes 17 main categories of macroprudential tools, classifying them into no action (0), tightening (+1), or loosening (-1). Based on this database, we conduct additional robustness checks (Table A1), considering the iMaPP dataset. In Column [8], we use the sum of the 17 instruments. In Column [9], we follow Sever and Yücel (2022) and compute tightening episodes through a dummy variable equal to 1 if the number of tightening episodes across months is greater than the number of easing episodes in a given year, and 0 otherwise. In both cases, the coefficient of the variable of interest is negative and significant, although the magnitude of the coefficient increases slightly compared to that of the main model. In Appendix A, we perform additional robustness tests, based on the two-step System-GMM method and using variables aggregated into non-overlapping three-year average to reduce stationarity issues, respectively. The results remain stable. 9 The results remain robust when considering alternative thresholds, such as inflation rates above 50%, 70%, 90%, or 100%. These results are not reported but are available on request. 10 These countries include: Burkina Faso, Guyana, Madagascar, Mali, Niger, Senegal, Togo, and Venezuela. IDOS Discussion Paper 3/2025 14 6 Heterogeneity This section conducts a series of heterogeneity analyses, first distinguishing the effect of macroprudential policies between borrower-targeted and financial-institution-targeted instruments.11 We could expect a higher impact of policies targeting borrowers on investment, as they have the most restrictive effects on credit and financial inclusion (see our discussion in Section 2). However, as highlighted by Cerutti et al. (2017), even tools targeted at financial institutions tend to lower credit growth, notably in emerging and developing markets – driven by tools such as dynamic provisioning, leverage ratios, counter-cyclical requirements, tax measures, interconnection and concentration limits. Indeed, although these instruments target financial institutions to mitigate systemic risk, they largely influence intermediary tools to regulate credit. Along the same lines, using data on 900,000 firms from 48 countries from 2003–2011, Ayyagari et al. (2018) find that young firms have lower investment and sales growth after implementation of both borrower-targeted and financial institution-targeted policies. The results reported in Columns [1] and [2] of Table 4 show that both instruments targeting borrowers and financial institutions significantly reduce investment, almost to roughly the same scale. Next, we examine the role of several macroeconomic, institutional, and structural factors. More precisely, we consider our main model (Equation 1) and augment it with several interactive terms. First, we interact the macroprudential policy index with the business cycle, approximated by annual GDP growth and the output gap, respectively.12 Since credit tightening can be more pronounced during economic downturns (Lown & Morgan, 2006), one may expect the negative effect of macroprudential policies on investment to be less pronounced during the expansion phase of the business cycle. The results reported in Columns [3] and [4] seem to confirm our hypothesis. Second, we consider financial and monetary factors, namely: the level of financial development (proxied by domestic credit to the private sector), financial openness and the exchange rate regime.13 The potential effect of macroprudential policies on private sector investment with regard to the level of financial development is not so clear-cut. As greater financial development comes with greater economic development, and probably with better institutional frameworks, it can be argued that financially more developed countries are more likely to strengthen their macroprudential policies effectively. On the other hand, following the perspective of Cerutti et al. (2017) perspective, we can consider that a more developed financial system also implies greater sophistication, making the application of macroprudential policies more complex, which can weaken their effectiveness. Regarding financial openness, we can expect borrowers in more open economies to successfully circumvent macroprudential policies, by finding ways to access other sources of financing, such as non-bank or cross-border banking activities. In this case, the effect of macroprudential policies on private sector investment would be more limited, given the potential substitution between domestic credit and other sources of financing. Regarding the exchange rate regime, Cerutti et al. (2017) note that it is more 11 Borrower-targeted instruments include loan-to-value ratios and debt-to-income ratios. Financialinstitution-targeted instruments include dynamic loan-loss provisioning; countercyclical capital buffer requirement; leverage ratio; capital surcharges on systemically important financial institutions; limits on interbank exposures; concentration limits; limits on foreign currency loans; reserve requirement ratios, limits of domestic currency loans; and levy/tax on financial institutions. 12 We compute the output gap by extracting potential output from observed real GDP, using the Hodrick– Prescott filter. 13 We have also considered potential heterogeneity with regard to the central bank interest rate, since monetary policy decisions have significant demand effects as well. The interactive term does not suggest any heterogeneity between macroprudential policies and the central bank interest rate. However, these results should be interpreted with caution, as monetary policy decisions may themselves be strongly endogenous to macroprudential policies (see Kim & Mehrotra, 2018 for a comprehensive discussion). IDOS Discussion Paper 3/2025 15 challenging for economies to control overall credit in more flexible exchange rate regimes, particularly given the impact of exchange rate appreciations or depreciation on capital movements. This suggests that the effect of macroprudential policies may be more limited in flexible exchange rate regimes. The results reported in Columns [5]-[7] reveal that the negative effect of macroprudential policies on investment is less pronounced in countries with more developed financial systems and flexible exchange rate regimes. However, no heterogeneity seems to emerge with regard to financial openness. Third, Column [8] examines potential heterogeneity in macroprudential policies according to the size of the informal sector, with the idea that policies targeting borrowers could lead them to shift towards informal financial services as an alternative form of financing.14 The results seem to corroborate this hypothesis. In Columns [9]-[11], we cross the macroprudential policy index with per capita income – using a dummy variable based on deviations from the sample mean – and the quality of institutions – proxied by corruption control and the level of democracy, respectively. We find strong evidence that the adverse effect of macroprudential policies on investment is mitigated in economically and institutionally more developed countries. These results can be aligned with our findings and discussion on the degree of financial development. In the last column, we differentiate the effect based on the periods before and after the 2008–2009 global financial crisis, in the idea that the impact of macroprudential policies may have been more pronounced after the crisis, as these policies were significantly intensified during that time. However, one may equally expect the effect to be less pronounced after the crisis, as the post-crisis intensification of macroprudential policies may also be associated with greater complexity in the application of macroprudential tools and the financial environment, thus hindering the effective and rigorous implementation of these tools. The results appear to align with the second hypothesis, indicating a diminished impact of macroprudential policies in the post-crisis period. Finally, Table A4 (in the appendix), which reports the results of the first-stage equation, does not invalidate the relevance of the instrument, regardless of the heterogeneity analysed. 14 The informal sectoral is from the World Bank’s Prospects Group (Elgin et al., 2021) which measures the informal economic activity using Multiple Indicators Multiple Causes (MIMIC) model-based estimates of informal output. IDOS Discussion Paper 3/2025 16 Table 4: The effect of macroprudential policies on private domestic investment: heterogeneity (1) (2) (3) (4) (5) (6) (7) (8) (9) (10) (11) (12) BorrowerTargeted Instruments -1.464** (0.695) Financial InstitutionTargeted Instruments -1.977*** (0.654) MPI -1.406*** -1.750*** -3.246*** -1.495*** -4.554*** -2.610*** -2.813*** -3.235*** -6.271*** -2.599*** (0.434) (0.544) (1.048) (0.471) (1.518) (0.778) (0.833) (1.204) (2.232) (0.837) MPI x Annual GDP growth 0.065*** (0.018) MPI x Output gap 0.661*** (0.184) MPI x Financial development 0.032*** (0.009) MPI x Capital openness -0.002 (0.076) MPI x Exchange rate regime 0.397*** (0.137) MPI x Informal sector 0.060*** (0.019) MPI x High income 1.883*** (0.521) MPI x Democracy 0.557** (0.221) MPI x Corruption control 2.167*** (0.777) MPI x Post 2008-09 crisis 1.018*** (0.344) Observations 1412 1412 1412 1412 1222 1350 1330 1373 1412 1412 1412 1412 R-squared 0.811 0.774 0.796 0.785 0.782 0.783 0.726 0.778 0.752 0.747 0.697 0.745 KleibergenPaap LM stat (p-value) 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 KleibergenPaap F-stat 123.55 44.67 67.46 49.39 18.40 53.55 26.07 69.23 28.51 26.90 15.29 27.99 Notes: In Columns [3]-[11] vector X variables in isolation (without interaction with macroprudential policies) and controls are included but not reported for the sake of space. Robust standard errors are in parentheses. All regressions include the constant, not reported in the table. * p < 0.1, ** p < 0.05, *** p < 0.01 IDOS Discussion Paper 3/2025 17 7 Mechanisms So far, we have mainly supported our potential transmission channels through theoretical discussions. In this section, we attempt to examine them empirically, drawing on existing work and the literature dealing with causal mediation analysis (for instance, see Acemoglu et al., 2019; Apeti & Edoh, 2023; Bambe, 2023; Bambe et al., 2024; Imai et al., 2010). First, we estimate the effect of macroprudential policies on the potential channel, accounting for key potential determinants of the channel. The findings from the first three columns of Table 5 indicate that macroprudential policies significantly reduce credit growth, financial inclusion and the probability of banking crises, highlighting the importance of these factors as potential transmission channels.15 Second, in columns [4]-[6] we re-estimate the effect of macroprudential policies on private domestic investment (Equation 1), including the potential channel among the vector of controls. The results in Column [4] are particularly noteworthy: when credit growth is included, the effect of macroprudential policies diminishes and becomes statistically insignificant. Moreover, the coefficient for credit growth is positive and significant, indicating that the latter is a key transmission channel. In Column [5], accounting for financial inclusion slightly reduces the effect of macroprudential policies compared to the coefficient of the baseline model (Table 2, Column [1]). Still, the coefficient of the financial inclusion index is positive and significant, suggesting that the latter is also an important channel. In the last column, although the results indicate that banking crises reduce investment, the inclusion of this variable does not significantly alter the coefficient of macroprudential policies, whose effect remains close to that of the baseline model. In other words, while financial stability – particularly banking crises – may be a channel through which macroprudential policies can promote domestic investment, the negative effect of these policies on investment, via the reduction in credit supply and financial inclusion, appears to outweigh their potentially beneficial impact via financial stability. In short, these results seem to support our theoretical hypotheses, providing evidence that the reduction in credit supply and financial inclusion resulting from macroprudential policies are relevant channels through which these policies affect private sector investment in developing countries – though credit growth seems to be the most dominant channel. 15 We follow previous studies (e.g. see Bozkurt et al., 2018; Ozili, 2022) and compute a composite index to capture financial inclusion. We consider two dimensions of financial inclusion, using data from the Financial Access Survey (IMF). The access dimension includes the number of commercial bank branches per 100,000 adults and the number of deposit accounts with commercial banks per 1,000 adults. The usage dimension includes outstanding loans from commercial banks (percent of GDP). We further consider the availability dimension, including the number of ATMs per 1,000 km2 and 100,000 adults. The index is computed following Anderson (2008), i.e., using generalised least squares estimators that account for variables with missing data, giving them less weight. Since the determinants of the channels considered may differ from those of private investment, in Columns [1] and [2] (Table 5) we draw on the literature on the determinants of financial inclusion (for instance, see Bozkurt et al., 2018) and consider the following control variables: per capita GDP, the level of education, the quality of institutions (proxied by government durability), the employment rate, the size of the informal sector, and financial sector reforms. In Column [3], we consider the following determinants of banking crises: inflation, lagged GDP growth, financial development, trade and financial globalisation. IDOS Discussion Paper 3/2025 24 Nickell, S. (1974). On the role of expectations in the pure theory of investment. The Review of Economic Studies, 41(1), 1–19. Nickell, S. (1981). Biases in dynamic models with fixed effects. Econometrica: Journal of the Econometric Society, 1417–1426. Osinski, J., Seal, K., & Hoogduin, M. L. (2013). Macroprudential and microprudential policies: toward cohabitation. International Monetary Fund. Ozili, P. K. (2022). Financial inclusion and sustainable development: an empirical association. Journal of Money and Business, 2(2), 186–198. Persson, T., & Tabellini, G. (2009). Democratic capital: The nexus of political and economic change. American Economic Journal: Macroeconomics, 1(2), 88–126. Raksmey, U., Lin, C.-Y., & Kakinaka, M. (2022). Macroprudential regulation and financial inclusion: Any difference between developed and developing countries? Research in International Business and Finance, 63, 101759. Reinhart, C. (2012). The return of financial repression (CEPR Discussion Paper No. DP8947). CEPR. Reinhart, C. M., & Reinhart, V. R. (2015). Financial crises, development, and growth: a long-term perspective. The World Bank Economic Review, 29(suppl_1), S53–S76. Reinhart, C. M., & Sbrancia, M. B. (2015). The liquidation of government debt. Economic Policy, 30(82), 291–333. Richter, B., Schularick, M., & Shim, I. (2019). The costs of macroprudential policy. Journal of International Economics, 118, 263–282. Rivera-Batiz, L. A., & Romer, P. M. (1991). Economic integration and endogenous growth. The Quarterly Journal of Economics, 106(2), 531–555. Rubio, M., & Carrasco-Gallego, J. A. (2016). The new financial regulation in Basel III and monetary policy: A macroprudential approach. Journal of Financial Stability, 26, 294–305. Sachs, J. D., & Warner, A. M. (2001). The curse of natural resources. European economic review, 45(46), 827–838. Schularick, M., & Taylor, A. M. (2012). Credit booms gone bust: monetary policy, leverage cycles, and financial crises, 1870–2008. American Economic Review, 102(2), 1029–1061. Sever, C., & Yücel, E. (2022). The effects of elections on macroprudential policy. Journal of Comparative Economics, 50(2), 507–533. Shin, H. S. (2013). Adapting macro prudential approaches to emerging and developing economies. Dealing with the challenges of macro financial linkages in emerging markets, 17–55. Shipan, C. R., & Volden, C. (2008). The mechanisms of policy diffusion. American Journal of Political Science, 52(4), 840–857. Staiger, D. O., & Stock, J. H. (1997). Instrumental variables regression with weak instruments. Econometrica, 65(3), 1997, 557–86. JSTOR, 65(3), 557–86. Straub, L., & Ulbricht, R. (2024). Endogenous uncertainty and credit crunches. Review of Economic Studies, 91(5), 3085–3115. Teixeira, A., & Venter, Z. (2023). Macroprudential policy and aggregate demand. International Journal of Central Banking, 19(4), 1–40. Teorell, J., Dahlberg, S., Holmberg, S., Rothstein, B., Khomenko, A., & Svensson, R. (2016). The quality of government standard dataset, version Jan16. University of Gothenburg: The Quality of Government Institute. http://www.qog.pol.gu.se doi:10.18157/QoGStdJan16 Tillmann, P. (2015). Estimating the effects of macroprudential policy shocks: A qual var approach. Economics Letters, 135, 1–4. Van der Ghote, A. (2021). Interactions and coordination between monetary and macroprudential policies. American Economic Journal: Macroeconomics, 13(1), 1–34. IDOS Discussion Paper 3/2025 25 Appendix A: Further robustness GMM estimates Without valid external instruments, the literature sometimes relies on alternative econometric strategies, such as the Generalised Method of Moments (GMM), to mitigate endogeneity issues. In addition to correcting for endogeneity bias using internal instruments, the GMM method also allows correcting for Nickell bias (Nickell, 1981), which is common in dynamic panel models. Therefore, for robustness, we rely on the two-step System-GMM method of Blundell and Bond (1998) which combines lagged differences and levels of explanatory variables as instruments, thus improving estimation efficiency. The GMM estimates are reported in Column [1] of Table A3. The new coefficient of the variable of interest remains comparable to that obtained from the main model, supporting our main conclusions. Three-year window The stationarity of our variables is an important consideration, as it plays a key role in ensuring the reliability of our results. Hence, to address potential stationarity issues, we draw on previous studies (e.g. see De Haan & Sturm, 2017 and Apeti et al., 2025) and re-estimate our baseline model using variables aggregated into non-overlapping three-year averages. The results are reported in the last column of Table A3 and align with our initial conclusions. Furthermore, the magnitude of the new coefficients is highly consistent with those of the main model, indicating that non-stationarity is unlikely to introduce bias into our main estimates. The results of the firststage equation (not reported, but available on request) also support the validity of the instrumental variable. IDOS Discussion Paper 3/2025 26 Table A1: Macroprudential policies (MPI) and private domestic investment: alternative subsamples and measures (1) (2) (3) (4) (5) (6) (7) (8) (9) invest invest invest invest invest invest invest invest invest MPI -1.535*** -1.467*** - 1.576*** -2.477*** -2.019*** -1.378*** -0.918* (0.464) (0.437) (0.473) (0.622) (0.563) (0.509) (0.540) Log. Employment 11.057 *** 12.354 *** 9.816 *** 11.915 *** 10.978 *** 11.315 *** 9.693 *** 8.753 *** 10.129 *** (2.455) (2.524) (2.348) (2.689) (2.254) (2.317) (2.154) (2.027) (2.216) Trade globalisation -0.016 -0.005 -0.016 0.003 0.003 -0.011 -0.003 -0.003 0.001 (0.017) (0.019) (0.017) (0.018) (0.017) (0.018) (0.016) (0.019) (0.019) Corruption control 0.398 ** 0.368 * 0.718 *** 0.631 *** 0.689 *** 0.505 ** 0.454 *** 0.375 ** 0.325 (0.187) (0.208) (0.198) (0.193) (0.188) (0.196) (0.174) (0.189) (0.202) Log. Government durability -0.305*** -0.253* -0.306*** -0.127 -0.265** -0.240** -0.290*** -0.076 -0.082 (0.110) (0.135) (0.113) (0.127) (0.115) (0.117) (0.103) (0.110) (0.112) Log. GDP per capita 6.077 *** 4.942 *** 5.744 *** 5.986 *** 3.405 *** 6.022 *** 5.446 *** 2.343 ** 2.197 ** (1.267) (1.429) (1.211) (1.088) (1.031) (1.202) (1.131) (1.101) (1.106) MPI (Alam et al., 2019) -5.760* (3.409) MPI Tightening -3.045 *** (0.845) Observations 1356 1181 1244 1367 1357 1293 1412 1174 1174 R-squared 0.778 0.764 0.775 0.753 0.757 0.781 0.803 0.805 0.781 Kleibergen-Paap LM stat (p-value) 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 Kleibergen-Paap Fstat 54.99 61.32 53.10 35.95 36.76 45.24 35.38 31.57 89.80 This table reports estimates of the effect of macroprudential policies (MPI) on private domestic investment, using alternative subsamples and measures. In all cases, the instrumental variables (IV) is the average macroprudential policy index in regional countries. In Columns [1]-[3] we re-estimate our baseline specification, excluding hyperinflation periods; fragile states; and the 2008–2009 global final crisis, respectively. Columns [4] and [5] exclude outliers, i.e., country-year observations with values above the 95th percentile of the sample, for the variable of interest and the dependent variable respectively. Column [6] uses exclusively countries that have implemented at least one macroprudential instrument during the study period. In Column [7], we weight the instrumental variable by annual GDP to assign greater influence to regional neighbours with larger economic size. Column [8] considers the sum of the 17 macroprudential tools from the integrated Macroprudential Policy (iMaPP) database (Alam et al., 2019). Column [9] uses tightening episodes through a dummy variable equal to 1 if the number of tightening episodes across months is greater than the number of easing episodes in a given year, and 0 otherwise. All regressions include the constant, not reported in the table. * p < 0.1, ** p < 0.05, *** p < 0.01 IDOS Discussion Paper 3/2025 27 Table A2: Alternative subsamples and measures: first stage results (1) (2) (3) (4) (5) (6) (7) (8) (9) pi Contiguity 0.842*** 0.910*** 0.833*** 0.629*** 0.646*** 0.754*** (0.109) (0.112) (0.109) (0.101) (0.102) (0.108) Log. Employment 1.168* 0.794 0.837 1.113* 0.978 0.217 0.934 -0.021 0.542* (0.685) (0.734) (0.682) (0.648) (0.622) (0.696) (0.672) (0.065) (0.289) Trade globalisation -0.004 -0.007 -0.005 -0.003 -0.002 -0.004 -0.003 0.001 0.003 (0.004) (0.005) (0.004) (0.004) (0.004) (0.005) (0.004) (0.001) (0.002) Corruption control 0.029 0.026 0.095** 0.062 0.078* 0.017 0.061 -0.004 -0.030 (0.042) (0.051) (0.043) (0.040) (0.042) (0.046) (0.042) (0.007) (0.027) Log. Government durability 0.006 -0.066 0.025 0.073** 0.063** -0.005 -0.001 -0.006** -0.014 (0.034) (0.044) (0.036) (0.029) (0.029) (0.038) (0.033) (0.003) (0.013) Log. GDP per capita 0.395 0.411 0.316 0.233 -0.135 0.179 0.727*** 0.013 0.006 (0.270) (0.301) (0.268) (0.214) (0.226) (0.267) (0.259) (0.024) (0.098) Weighted contiguity 0.482*** (0.078) Contiguity (Alam et al., 2019) 1.059*** (0.181) Contiguity MPI Tightening 1.023*** (0.104) Observations 1356 1181 1244 1367 1357 1293 1412 1174 1174 R-squared 0.870 0.865 0.865 0.848 0.881 0.859 0.866 0.598 0.586 This table reports the results of the first stage IV estimation of Table A1. Robust standard errors are in parentheses. All regressions include the constant, not reported in the table. * p < 0.1, ** p < 0.05, *** p < 0.01 IDOS Discussion Paper 3/2025 28 Table A3: Macroprudential policies and private domestic investment: System-GMM and IV-three-year window estimates (1) (2) System-GMM IV-three-year window Lag. Investment 0.676** (0.135) MPI -1.097* -1.744** (0.606) (0.823) Log. Employment 23.781* 8.877** (12.998) (4.015) Trade globalisation -0.143 -0.016 (0.093) (0.027) Corruption control 0.030 1.112** (0.751) (0.362) Log. Government durability -0.836 -0.395** (1.221) (0.201) Log. GDP per capita 7.381** 5.294** (3.113) (1.876) Observations 1412 499 R-squared 0.8018 The AR (1), AR(2), and Hansen test p-values reported in Column [1] are respectively 0.002, 0.794, and 0.730. We report 84 groups for 42 instruments in the first column. In Column [2], we use variables aggregated into non-overlapping three-year averages, to reduce stationarity issues. Robust standard errors are in parentheses. All regressions include the constant, not reported in the table. * p < 0.1, ** p < 0.05, *** p < 0.01 IDOS Discussion Paper 3/2025 29 Table A4: Heterogeneity: first stage results This table reports the results of the first stage IV estimation of Table 4. In Columns [3]-[12] vector X variables in isolation (without interaction with macroprudential policies) and the interactive terms with macroprudential polices are not reported for the sake of space. Robust standard errors are in parentheses. All regressions include the constant, not reported in the table. * p < 0.1, ** p < 0.05, *** p < 0.01 (1) (2) (3) (4) (5) (6) (7) (8) (9) (10) (11) (12) Contiguity 0.920*** (0.080) Log. Employment 0.245 0.599 0.837 0.608 1.024* 0.382 0.887** 0.980*** -1.104*** 0.215 -1.197*** -0.663 (0.275) (0.500) (0.597) (0.611) (0.572) (0.566) (0.430) (0.306) (0.408) (0.409) (0.240) (0.544) Trade globalisation -0.003 -0.002 -0.004 -0.005 0.008* -0.007* 0.001 -0.004* -0.012*** -0.013*** -0.002 -0.006* (0.002) (0.003) (0.004) (0.004) (0.004) (0.004) (0.002) (0.002) (0.003) (0.002) (0.002) (0.003) Corruption control 0.027 0.021 0.038 0.052 0.060 0.033 0.059** 0.063*** 0.030 -0.079*** -0.643*** -0.016 (0.019) (0.035) (0.039) (0.040) (0.039) (0.043) (0.025) (0.023) (0.036) (0.028) (0.045) (0.037) Log. Government durability 0.005 0.010 0.026 0.043 0.031 0.008 0.040*** 0.063*** -0.009 -0.036** 0.017 -0.028 (0.015) (0.025) (0.030) (0.030) (0.027) (0.038) (0.014) (0.014) (0.027) (0.018) (0.014) (0.025) Log. GDP per capita -0.125 0.432** -0.184 0.163 -0.276 0.468* -0.101 0.482* 0.430** 0.306 0.401*** 0.239 (0.098) (0.200) (0.235) (0.234) (0.190) (0.246) (0.168) (0.247) (0.180) (0.186) (0.103) (0.180) Contiguity 0.864*** (0.124) Contiguity 0.859*** 0.701*** 0.410*** 0.829*** 0.301*** 0.473*** 0.475*** 0.352*** 0.212*** 0.468*** (0.101) (0.096) (0.091) (0.109) (0.057) (0.055) (0.085) (0.065) (0.052) (0.085) Observations 1412 1412 1412 1412 1222 1350 1330 1373 1412 1412 1412 1412 R-squared 0.785 0.869 0.887 0.894 0.941 0.877 0.964 0.970 0.918 0.953 0.970 0.911 IDOS Discussion Paper 3/2025 30 Appendix B: Sample and descriptive statistics Table B1: Summary statistics of the baseline model variables Variables Obs. Mean Sd. Min. Max. MPI 1,566 2.044 1.869 0 10 Private domestic investment 1,555 13.003 6.083 0.036 33.490 Log. Employment 1,566 4.035 0.199 3.427 4.453 Trade globalisation 1,566 49.849 15.471 15.934 85.464 Corruption control 1,560 2.171 0.733 0.5 5 Log. Government durability 1,524 2.429 1.125 0 4.595 Log. GDP per capita 1,548 9.058 1.035 6.588 11.453 IDOS Discussion Paper 3/2025 31 Table B2: Sample Country Average MPI Country Average MPI Country Average MPI Albania 1.28 Jordan 2.83 Turkey 3.94 Algeria 1.94 Kazakhstan 1.44 Uganda 2.33 Angola 1 Kenya 0.61 Ukraine 2.89 Argentina 4.78 Kuwait 5.11 Uruguay 1.89 Armenia 3 Lebanon 3.83 Venezuela, RB 0 Azerbaijan 2.39 Liberia 1.67 Vietnam 1.83 Bahamas, The 2.56 Madagascar 0 Zambia 1 Bahrain 3 Malawi 1.78 Bangladesh 3.78 Malaysia 2 Belarus 1.5 Mali 0 Bolivia 2.61 Mexico 2.33 Botswana 1.11 Moldova 2.56 Brazil 4.28 Mongolia 2.89 Brunei Darussalam 1.94 Morocco 3 Bulgaria 2.39 Mozambique 2.94 Burkina Faso 0 Myanmar 0.11 Chile 6.5 Namibia 1.44 China 4.83 Nicaragua 0.11 Colombia 6.61 Niger 0 Congo, Dem Rep 1.67 Nigeria 1.17 Costa Rica 3 Oman 1.56 Cote d’Ivoire 0.28 Pakistan 7.5 Croatia 2 Panama 1.33 Dominican Republic 2.17 Paraguay 3.89 Ecuador 5.11 Peru 4.28 Egypt, Arab Rep 0.22 Philippines 2.61 El Salvador 1 Poland 1.89 Ethiopia 0.39 Romania 3.28 Gambia, The 1.78 Russian Federation 1.28 Ghana 2 Saudi Arabia 2.22 Guatemala 0.28 Senegal 0 Guinea-Bissau 0.06 Serbia 3.11 Guyana 0 Sierra Leone 0.67 Haiti 2.5 South Africa 0.94 Honduras 1.11 Sri Lanka 1.22 Hungary 1.78 Sudan 1.78 India 2.17 Tanzania 1 Indonesia 1.39 Thailand 1.78 Iran, Islamic Rep 0.44 Togo 0 Iraq 0.83 Tunisia 2.11 IDOS Discussion Paper 3/2025 32 Table B3: Sources of variables Variables Nature Sources 1. Main model variables Macroprudential Policy Index Scores ranging from 0 to 10 Cerutti et al. (2017) Private domestic investment Continuous International Monetary Fund (IMF)’s Investment and Capital Stock database Employment rate Continuous World Bank’s World Development Indicators (WDI) database Trade globalisation Index ranging from 0 to 100 KOF index (Dreher, 2006a; Gygli et al., 2019) Corruption control Index ranging from 0 to 6 International Country Risk Guide (ICRG) Government durability Continuous Polity IV 2. Additional variables Trade openness Continuous WDI Terms of trade Continuous WDI Capital openness Index ranging from -2 to 2 Chinn and Ito (2008) Inflation Continuous WDI Remittances Continuous WDI Exchange rate regime Dummy Authors, from Ilzetzki et al. (2019) Real effective exchange rate Continuous Darvas (2012) Public investment Continuous IMF’s Investment and Capital Stock database Natural resources Continuous WDI Property rights Index ranging from 0 to 1 V-DEM Political Pressures and Controls on Media Content Index ranging from 0 to 40 Quality of Government (Teorell et al., 2016) IMF programmes Dummy Dreher (2006b) Economic freedom Index ranging from 0 to 100 The Heritage Foundation Regional GDP growth Continuous Authors, from WDI Borrower-Targeted Instruments Scores ranging from 0 to 2 Cerutti et al. (2017) Financial Institution-Targeted Instruments Scores ranging from 0 to 8 Cerutti et al. (2017) Output gap Dummy Authors, using real GDP from WDI Financial openness Index ranging approximately from -2 to 2 Chinn and Ito (2008) Financial development Continuous WDI Informal sector Index ranging from 0 to 100 Elgin et al. (2021) Democracy Index ranging from 0 to 6 International Country Risk Guide (ICRG) Credit growth Continuous Authors, using data from WDI Financial inclusion Index ranging from 0 to 100 Authors, using data from the Financial Access Survey (IMF) Banking crises Dummy Laeven and Valencia (2020)