WPS6591 Policy Research Working Paper 6591 Institutional and Structural Determinants of Investment Worldwide Jamus Jerome Lim The World Bank Development Economics Prospects Group September 2013 Policy Research Working Paper 6591 Abstract This paper considers institutional and structural factors with latter displaying more stability in the sign and associated with investment activity in a panel of up to significance of its coefficient. Indeed, when endogeneity 129 developed and developing countries. It introduces concerns are addressed more explicitly using external these factors to a standard neoclassical investment instruments, and both interactions and subsamples are function for open economies, and find that financial considered, institutional quality tends to survive as the development and institutional quality are reasonably causal determinant of investment. robust determinants of cross-country capital formation, 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://econ.worldbank.org. The author may be contacted at jlim@ worldbank.org. 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 Institutional and Structural Determinants of Investment Worldwide Jamus Jerome Lim∗ Keywords: Investment, ï¬?nancial development, institutional quality JEL Classification: E22, E02, O16 Sector Board: EPOL ∗ The author thanks Sergio Kurlat for analytical inputs for, and Maurizio Bussolo for comments on, an earlier draft of the paper. This work served as a technical background paper for the policy-oriented discussions in Chapter 1 of World Bank (2013). Financial support for this paper from the RSB Research Grant P131352 “Structural Determinants of Aggregate Investment Behaviorâ€? is gratefully acknowledged. The ï¬?ndings, interpretations, and conclusions expressed in this article are entirely those of the authors. They do not necessarily represent the views of the World Bank, its Executive Directors, or the countries they represent. 1 1 Introduction The cross-country variation in investment activity is truly remarkable. For the 30-year period between 1980 and 2010, the rate of gross ï¬?xed capital formation worldwide ranged from 6 to 77 percent of production, a variance more than two times that of economic growth. Much of this variability stems from developing countries, which also exhibit a far greater diversity in terms of political-economic structure and institutions. However, since most empirical studies of aggregate investment tend to focus on a relatively small set of (mostly) developed countries (Byrne & Davis 2005; Davis 2010; Oliner, Rudebusch & Sichel 1995) and a well-deï¬?ned set of theories (Chirinko 1993; Ferderer 1993; Kopcke & Brauman 2001), they gloss over such structural and institutional detail, since the environments faced in those instances are reasonably similar. This is not the case when attempting to explain a broader cross-section of countries, which can differ along economic, legal, and political dimensions. Consequently, failure to take into account structural differences that exist in the cross-country data risks missing an important part of the explanation for variations in international investment patterns. Among the existing literature where a more general mix of economies is considered, the trend has been a focus on purely economic factors of a more cyclical nature, such as the real en 2003), ï¬?scal and monetary policy (Greene & Villanueva 1991), and capital exchange rate (Serv´ inflows (Wai & Wong 1982). The main shortcoming of such approaches is that they may fail to capture important discontinuities that may arise from longer-run changes in structural factors. A small number of papers do systematically examine the important role that institutional and structural factors play; however, most content themselves with the introduction of one or two such variables, such as the level of ï¬?nancial development (Benhabib & Spiegel 2000; Levine 2005; Love & Zicchino 2006) and structure (Ndikumana 2005), institutional quality (Campos & Nugent 2003; Mauro 1995; Morrissey & Udomkerdmongkol 2012) and structure Dawson (1998), alez and the business environment (Bartelsman, Haltiwanger & Scarpetta 2010; Utrero-Gonz´ 2007). When addressed in isolation, however, it is difficult to place the importance of different structural variables in context. Although there may be objection to the wholesale incorporation of such structural and insti- tutional measures as atheoretical, this is only the case when such determinants are understood narrowly. Many structural determinants are in fact implied by pure investment theory. For ex- ample, the user cost of capital in a standard neoclassical model (Jorgenson 1963) may differ by country due to differences in tax structure (Hall & Jorgenson 1967). Alternatively, adjustment costs in either a Tobin’s Q (Hayashi 1982; Tobin 1969) or (S, s)-type (Caballero & Engel 1999) setting may diverge between countries due to differences in the transactions costs related to the respective institutional frameworks. Modern theoretical models that incorporate frictions that arise from capital market imper- om & Tirole 1997) or uncertainty (Caballero & Pindyck 1996; Lucas & Prescott fections (Holmstr¨ 1971) also implicitly point to the need to account for structural and institutional factors, since such frictions suggest that, inter alia, a country’s ï¬?nancial structures and sophistication or 2 political-institutional risks may in fact matter for investment. More generally, the (at least partial) irreversibility of investment means that price (interest rate) signals alone may be in- sufficient to generate observed levels of investment activity (Dixit & Pindyck 1994), implying a need to pay greater attention to structural-institutional detail. Recent work seeking to explain differences in cross-country investment patterns (Caselli & en & Ventura 2005)—which stress the Feyrer 2007; Hsieh & Klenow 2007; Kraay, Loayza, Serv´ importance of uninsurable idiosyncratic investment risk—also support the notion that structural and institutional distinctions may be key frictions that prevent returns to capital, and hence investment, from normalizing across countries. Our work thus suggests that such distortions to the marginal product of capital may in fact derive, at least in part, from an economy’s economic structure or its institutions. Finally, the vast body of work examines the puzzle of high saving retention coefficients (Feldstein & Horioka 1980) in cross-country analyses of investment point, at least implicitly, to the need to account for endogeneity due to omitted variables, of which structural factors are key. While there have been subsequent theoretical (Bai & Zhang 2010; Kraay & Ventura 2000) and empirical (Byrne, Fazio & Fiess 2009; Hon 2012) attempts to either reconcile or reject the notion that a high correlation between investment and saving necessarily implies home bias in investment activity, the underlying misspeciï¬?cation concern underscored by this strand of literature strongly suggests that institutional and structural variation between countries should be properly accounted for in cross-country studies of capital formation. In this paper, we seek to empirically identify and estimate the relative importance of the structural and institutional determinants that may be associated with cross-country patterns of aggregate investment. Using a standard neoclassical model as our theoretical launching point, we systematically introduce various families of structural and institutional determinants. Our estimation methodology relies on dynamic panel estimation via GMM (Arellano & Bover 1995; Blundell & Bond 1998), which allows us to capture potential partial adjustment effects, as well as some (weak) control of potential endogeneity. Our main contribution is thus the simultaneous evaluation of a host of institutional and structural variables, with the goal of identifying key determinants of investment worldwide. We obtain two key ï¬?ndings. First, across a range of speciï¬?cations and alternative measures, ï¬?nancial development and institutional quality are reasonably robust determinants of invest- a-vis the latter, institutional ment. While the former typically enters with a larger magnitude vis-´ quality displays both a more stable coefficient and consistent statistical signiï¬?cance. Second, and related to the ï¬?rst, when potential endogeneity concerns are addressed more explicitly using external instruments, ï¬?nancial development drops out of statistical signiï¬?cance entirely, suggesting that—to the extent that the external instruments are reliable—institutional quality is less likely to be contaminated by reverse causality concerns, at least insofar as investment activity is concerned. The rest of the paper is organized as follows. The following section outlines the main data 3 sources and deï¬?nitions (Section 2.1), along with empirical methodology (Section 2.2). Section 3 discusses both the benchmark results as well as the robustness of these results to alternative spec- iï¬?cations and measurements (Section 3.3) and more stringent endogeneity testing (Section 3.4). The section also attempts to tease out the manner by which interaction effects (Section 3.5) and subsamples (Section 3.6) be driving the key ï¬?ndings. A ï¬?nal section concludes with some reflections on policy implications. 2 Data and methodology 2.1 Data sources and deï¬?nitions The dataset for the investment regressions is an unbalanced country-level panel, covering up to 129 economies over 5-year periods between 1980–2009.1 Variables for the benchmark regressions were sourced from the World Bank’s World Development Indicators (WDI) as well as Financial u¸ Development and Structure (Beck, Demirg¨ c-Kunt & Levine 2000) databases, the International Country Risk Guide (ICRG), and Chinn & Ito (2008). Additional variables included in the robustness tests were drawn from the World Bank’s Global Economic Monitor (GEM) and Doing Business databases, Beck, Clarke, Groff, Keefer & Walsh (2001), and Laeven & Valencia (2012). Full details of variable sources, deï¬?nitions, and other summary statistics are given in Ap- pendices A.1, A.3, and A.4. Two important statistical features are worth noting. First, the standard deviation in the institutional and structural variables, while small relative to the level of investment, are nevertheless larger than most of the economic controls, which supports the notion that variations in the former may be important for better understanding cross-country investment patterns. Second, the correlation among the distinct families of institutional and structural variables considered is actually fairly small; the highest correlation is between institutional quality and ï¬?nancial development (Ï? = 0.56), and even then the relationship is not particularly strong. This suggests that the various variables of interest are sufficiently distinct—statistically speaking—to warrant their inclusion as independent variables. Given the centrality of structural factors in this paper, we briefly discuss here the deï¬?nitions for the main institutional and structural variables of interest, along with the motivation behind their selection. To accommodate the host of variables that we consider, we organize them into various classes of determinants, as suggested by theory. One important factor we consider is the level of maturity of the ï¬?nancial sector as well as its its structure, which are measured respectively by domestic credit to the private sector (as a share of GDP) and the ratio of the total value traded in the stock market to domestic credit. Constraints arising from limited access to ï¬?nance have the potential to adversely affect 1 In the preferred benchmark speciï¬?cations, however, the sample coverage is 105 economies. These are listed in Appendix Table A.2. 4 investment activity (Schiantarelli 1996), and even the organizational form of corporate ï¬?nancing may impact the ease of investment by ï¬?rms (Dailami 1992). Another important factor is related to quality of institutional mechanisms such as contract enforcement and property rights, both of which can influence aggregate investment through either altering incentives for new investment (Besley 1995), or by increasing the sensitivity of investment to technological shocks at the macroeconomic level (Cooley, Marimon & Quadrini 2004). Even the overall structure of institutions may play a role in encouraging or discouraging investment, through the manner by which they may seek to resolve commitment problems (Gehlbach & Keefer 2011). We proxy for institutional quality by averaging indices of corruption and rule of law, while institutional structure is captured measure of democratic accountability. The overall business environment may also matter, especially as embodied by investor pro- tections (Shleifer & Wolfenzon 2002) or the nature of corporate taxation (Devereux 1996; Hall & Jorgenson 1967). While at ï¬?rst glance there may appear to be some overlap in such mea- sures with the overall institutional environment, business and regulatory factors typically affect investment more directly, and should be treated as distinct from the institutional setting that governs interactions between political-economic actors. Our gauge of the business environment is an index that approximates the strength of investor protection—selected in particular be- cause its reflects the investment-related aspects of business regulation—while the tax structure is represented by the highest marginal corporate tax rate. 2.2 Empirical methodology We motivate the empirical work to follow with a very simple theoretical speciï¬?cation of the (flexible) neoclassical model (Hall & Jorgenson 1967), where the optimal capital stock in country ∗ , is a function of production, Y , and the cost of capital, R , so that i at time t, Kit it it ∗ αYit Kit = σ , (1) Rit where α and σ are, respectively, the output and substitution elasticities of capital. To obtain investment, substitute the optimal capital stock with the equation of motion of capital Ki,t+1 = (1 − δ ) Kit + Iit , and applying the result that, in the steady state, the growth rate of capital is the growth rate of output (so that Ki,t+1 = (1 + µit ) Kit , where µ is the GDP growth rate), yields an estimable empirical speciï¬?cation it = β + yit + git − σrit , (2) where β ≡ ln α and git ≡ ln (µit + δ ) is the (depreciation-adjusted) growth rate, and lowercase letters indicate the logarithm of the respective uppercase variables. For the empirical speci- ï¬?cation that follows, we relax the parameter restriction of unity for the coefficient on growth 5 and output, and include additional economic variables Xit related to the open economy, and institutional and structural variables that may affect investment, Zit : iit = β + φii,t−1 + ψyit + γgit − σrit + Φ Xit + Γ Zit + it , (3) where it is a disturbance term. (3) further includes the lagged dependent variable ii,t−1 , to allow for partial adjustment in ï¬?xed capital formation. The econometric analysis of (3) is performed with system GMM (Arellano & Bover 1995; Blundell & Bond 1998), which is well-suited for this application since estimates both accounts for between and within variation in the data—especially important since structural and institutional variations may be more substantial across countries, rather than within a country alone—along with some (weak) control of endogeneity in the regressors. Moreover, system GMM resolves problems that may arise from Nickell (1981) bias due to the inclusion of the lagged dependent variable, which is especially important since aggregate investment is a persistent series (Bond, Hoeffler & Temple 2001). There are also additional efficiency gains that accrue to system GMM, which is important given the relatively small size of the cross-section. In all the speciï¬?cations that follow, output, growth, and the real interest rate are treated as endogenous, and entered into the (orthogonalized) instrument matrix with two lags and deeper, while lagged investment, trade openness, and ï¬?nancial openness are treated as predetermined and entered with one or more lags. The institutional and structural variables are instrumented with their lagged values. The instrument set is then collapsed to limit instrument proliferation (Roodman 2009), and all standard errors are corrected to account for heteroskedasticity and arbitrary patterns of autocorrelation within countries. 3 Results 3.1 Illustrative relationships In order to establish an initial grasp on how structural factors may be related to investment, we plot the ï¬?xed investment rate against each of the structural variables of interest. This is shown in Figure 1. Several features are worth noting. First, there appear to be signiï¬?cant bivariate relation- ships for a number of the structural variables of interest, notably for ï¬?nancial development, institutional quality, the business environment, and the tax environment. Since these are bi- variate relationships, however, it is premature to claim that these factors will all survive in a more systematic empirical treatment. Second, where applicable, the expected impact of these variables accord with a priori in- tuition. For example, higher levels of institutional quality correspond with higher rates of investment, while higher tax rates imply the opposite. With regard to ï¬?nancial and institu- tional structure—where there may be no deï¬?nitive theoretical hypothesis—the small positive 6 7 Log of fixed investment rate (% GDP) Log of fixed investment rate (% GDP) Financial development Institutional quality .1 .2 .3 .4 .5 .1 .2 .3 .4 .5 0 0 0 .5 1 1.5 0 .5 1 1.5 2 Log of financial development (% GDP) Log of institutional quality (1−6) Log of fixed investment rate (% GDP) Log of fixed investment rate (% GDP) Financial structure Institutional stucture .1 .2 .3 .4 .5 .1 .2 .3 .4 .5 0 0 0 .5 1 1.5 2 0 .5 1 1.5 2 Log of financial structure Log of institutional structure (0−6) Log of fixed investment rate (% GDP) Log of fixed investment rate (% GDP) Business environment Tax environment .1 .2 .3 .4 .5 .1 .2 .3 .4 .5 0 0 .5 1 1.5 2 2.5 2.5 3 3.5 4 Log of business environment (1−12) Log of tax environment (%) Source: Author’s calculations Figure 1: Scatterplots of ï¬?xed investment rate (as a percentage of GDP) to structural variables of interest, unbalanced 5-year average panel, 1980–2009. slopes appear to suggest that more market-based ï¬?nancial systems and more democratic sys- tems are more likely to be associated with greater investment (although the relationships are weak and unlikely to be signiï¬?cant). Finally, it is also worth noting that data limitations mean that the graphs are not all represented by the same sample. This is especially the case for ï¬?nancial structure and the tax environment, where the samples appear to be especially small. Such sample limitations may limit our ability to make strong inferences with the cross-country panel (an issue that we revisit in the more formal analysis that follows). 3.2 Benchmark results Our benchmark results for (3) are reported in Table 1. Across all speciï¬?cations, the included variables are jointly signiï¬?cant (as measured by the Wald χ2 test), and the exogeneity of the instrument set is veriï¬?ed by the insigniï¬?cant Hansen J statistics. The z statistic for the Arellano- Bond AR(2) tests do indicate that autocorrelation may be an issue for the ï¬?rst two speciï¬?cations; however, these two are offered more as baselines, and hence their potential misspeciï¬?cation is less of a concern. Column (B1) is a minimal speciï¬?cation—corresponding to (2)—while column (B2) allows for open-economy effects by introducing two medium-term determinants of external accounts on, Chong & Loayza 2002; Chinn & Prasad 2003) are included: trade openness and (Calder´ ï¬?nancial openness. The coefficients on these economic determinants are consistent with a priori expectations on their sign: economic size and growth are both positively correlated with the level of ï¬?xed investment, and the series displays a fair degree of persistence. The cost of capital— as proxied by the real interest rate—is statistically insigniï¬?cant, a result consistent with the broader literature, which has struggled to establish a strong empirical relationship between the two variables (Caballero 1999).2 Interestingly, the coefficient on ï¬?nancial openness is negative and signiï¬?cant. This effect is nontrivial: a ten percent increase in ï¬?nancial openness—an decrease in restrictions on capital flows roughly comparable to moving from, say, that of Egypt to that of Singapore (for the year 2009)—could trigger a decrease in investment by between one and two percent. This implies that, ceteris paribus, more ï¬?nancially open economies tend to experience lower levels of investment; this would be the case if foreign direct investment (FDI) flows not only substitute but displace domestic flows more than one-for-one. Although somewhat surprising, this would be the case if FDI were more productive than domestic investment, and the relatively weak contribution of FDI to new domestic investment and growth is a result that has some limited org & Greenaway 2004; Narula support in the empirical literature (Agosin & Machado 2005; G¨ & Driffield 2012). 2 Indeed, this has generally been the case even when more precise measures of the cost of capital (which account additional complications such as the corporate tax rate and investment tax credits) and more sophisticated econometric techniques, including the exploitation of natural experiments. 8 Table 1: Benchmark regressions for ï¬?xed investment, unbalanced 5-year average panel, 1980–2009† B1 B2 B3 B4 B5 B6 B7 B8 Lagged investment 0.463 0.608 0.373 0.466 0.471 0.475 0.359 0.458 (0.18)∗∗∗ (0.11)∗∗∗ (0.23) (0.17)∗∗∗ (0.16)∗∗∗ (0.18)∗∗∗ (0.10)∗∗∗ (0.08)∗∗∗ Output 0.583 0.393 0.614 0.518 0.509 0.495 0.663 0.536 (0.21)∗∗∗ (0.12)∗∗∗ (0.22)∗∗∗ (0.19)∗∗∗ (0.17)∗∗∗ (0.19)∗∗∗ (0.12)∗∗∗ (0.09)∗∗∗ Output growth 0.594 1.071 1.488 1.430 1.395 1.423 1.249 1.241 (0.24)∗∗ (0.25)∗∗∗ (0.38)∗∗∗ (0.26)∗∗∗ (0.26)∗∗∗ (0.30)∗∗∗ (0.23)∗∗∗ (0.29)∗∗∗ Cost of capital 0.419 -0.895 -0.201 1.241 1.168 1.251 0.246 0.662 (1.98) (1.04) (1.08) (1.67) (1.54) (1.49) (1.23) (1.13) Trade openness -0.166 -0.046 -0.079 -0.017 -0.105 0.098 0.028 (0.25) (0.27) (0.26) (0.23) (0.23) (0.14) (0.11) Financial openness -0.141 -0.140 -0.185 -0.169 -0.184 -0.116 -0.108 (0.06)∗∗ (0.08)∗ (0.08)∗∗ (0.07)∗∗ (0.07)∗∗ (0.08) (0.05)∗∗ Financial 0.275 0.273 0.270 0.329 0.007 0.048 development (0.15)∗ (0.13)∗∗ (0.13)∗∗ (0.13)∗∗∗ (0.07) (0.08) Institutional 0.138 0.140 0.149 0.158 0.159 9 quality (0.08)∗ (0.08)∗ (0.09)∗ (0.09)∗ (0.08)∗∗ Business -0.015 -0.130 -0.123 -0.023 environment (0.10) (0.18) (0.12) (0.12) Institutional 0.101 0.112 0.053 structure (0.08) (0.08) (0.08) Tax environment -0.045 (0.04) Financial 0.007 structure (0.10) Wald χ2 1,858∗∗∗ 3,287∗∗∗ 10,798∗∗∗ 8,977∗∗∗ 9,642∗∗∗ 11,448∗∗∗ 20,917∗∗∗ 18,830∗∗∗ Hansen J 19.663 35.256 34.427 29.065 29.541 30.135 30.029 40.157 AR(2) z -1.936∗ -2.364∗∗ -1.588 -1.570 -1.547 -1.446 0.000 -0.763 Instruments 24 39 40 39 41 43 46 44 N (countries) 483 (129) 467 (125) 364 (123) 333 (105) 333 (105) 333 (105) 138 (79) 191 (81) † All variables are in log form. Heteroskedasticity and autocorrelation-robust standard errors are reported in parentheses. Period ï¬?xed effects and a constant term were included in the regressions, but not reported. ∗ indicates signiï¬?cance at 10 percent level, ∗∗ indicates signiï¬?cance at 5 percent level, and ∗∗∗ indicates signiï¬?cance at 1 percent level. Columns (B3)–(B8) incrementally introduce structural and institutional controls: ï¬?nancial development, institutional quality, the business environment, institutional structure, the tax environment, and ï¬?nancial structure. Due to data limitations, the ï¬?nal two speciï¬?cations are added independently (as evident, the sample size drops dramatically as a result of their inclusion). These two variables are, in any case, insigniï¬?cant; we henceforth proceed with speciï¬?cation (B6) as our preferred benchmark speciï¬?cation. Across these different speciï¬?cations, institutional quality typically enters with a statistically signiï¬?cant coefficient (although often only at the 10 percent level). The coefficient is bound by [0.136, 0.158], which, while small, is nonetheless economically relevant: a ten percent increase in institutional quality could translate into an increase of investment by 1.6 percent. This would be equivalent to an improvement from 2009 levels in Ukraine to that of Italy, or around the improvement in Chile’s institutional quality between 1996 and 2009, the period where it transitioned away from the military junta under Augusto Pinochet. It is interesting to contrast the positive and signiï¬?cant coefficient on the institutional quality variable against that of the business environment variable, which is insigniï¬?cant. Given the speciï¬?city of the latter variable for investment activity, this result suggests that the importance of the rule of law goes beyond the manner by which institutions foster investment; rather, a strong institutional framework likely affords broad-based economic opportunity and fosters competition dynamics, which in turn leads to economywide incentives toward greater levels of investment. This result provides an alternative view of the institutions that are central to glu & Johnson 2005), who argue that property rights investment activity, in contrast to (Acemo˘ institutions dominate contracting institutions in the determination of investment.3 The magnitude of the positive coefficient on ï¬?nancial development—which averages 0.20 across the six speciï¬?cations in which it is included—is also economically relevant, and around twice that of institutional quality in most speciï¬?cations (although in the limited subsample of the ï¬?nal two speciï¬?cations, the coefficient drops out of statistical signiï¬?cance). Given the sharp contraction in the size of the sample resulting from the inclusion of institutional structure or the tax environment, it is difficult to draw strong conclusions regarding the robustness of the statistical signiï¬?cance of ï¬?nancial development; however, we revisit the issue in the following subsections. 3.3 Robustness of the benchmark In this section we consider the robustness of the benchmark results—as embodied by speciï¬?ca- tion (B6)—to alternative measures of our variables of interest. Our choices of these alternative measures for the institutional and structural variables were predicated by the desire to offer a 3 Acemo˘ glu & Johnson (2005) favor legal measures—such as the extent of formalism and procedural complexity and depth—as measures of contractual institutions, while they treat protection against expropriation as a property rights institution. We believe that all these measures are more reflective of the commercial and business climate, whereas the broader institutional environment, as measured by the rule of law and corruption, represents a more distinctive alternative determinant of investment activity. 10 variant to the conceptualization of the variable in the benchmark, rather than simply an alter- native measure. Nevertheless, we recognize that different data sources may result in changes to the potential accuracy, reliability, and coverage of the variable in question. Accordingly, we considered several alternative sources for the variables in our benchmark (as before, detailed deï¬?nitions are provided in Annex Table A.1). In columns (R1) and (R2) of Table 2, we consider two alternative deï¬?nitions of our de- pendent variable. (R1) uses the ï¬?xed investment rate (the ï¬?xed capital formation share of GDP), while (R2) employs gross investment (inclusive of inventory accumulation). Although the coefficients are not directly comparable, the qualitative messages are similar; notably, that ï¬?nancial development and institutional quality are important structural determinants, and the magnitude of the coefficient on the former is larger than that on the latter.4 Somewhat interestingly, the coefficient on business environment enters with a negative sign (and is marginally signiï¬?cant) in speciï¬?cation (R2). While counterintuitive at ï¬?rst, a careful perusal of the underlying data is illuminative: many economies with strong investor protection scores tend to be relatively less developed. This result may be rationalized by the acceding to the possibility that when investor protection clauses are in conflict with the more general sense of the rule of law (captured by institutional quality), investors may regard de jure laws as a negative signal and reduce their investment activity, resulting in a negative relationship. Columns (R3) and (R4) introduce two alternatives to the baseline speciï¬?cation for the economic controls. The ï¬?rst of these imposes the constraint, suggested by (2), where the coefficient on growth and output are held at unity. The second substitutes the real interest rate measure of the cost of capital with an alternative computed from the differential between the domestic interest rate and an exchange rate-adjusted risk-free interest rate (an interest rate “arbitrageâ€? measure); this alternative is to allow for the possibility that the real interest rate operates at the margin relative to a global risk-free rate.5 Both changes have little impact on the main results, although predictably the coefficient on the alternative cost of capital measure is much smaller (although still statistically insigniï¬?cant). The robustness of the two key structural variables of interest is considered in columns (R5)– (R7). Speciï¬?cation (R5) utilizes an alternative deï¬?nition of ï¬?nancial development, domestic credit by banks, which excludes nonbank sources of credit. Since investment ï¬?nancing in many developing economies are typically obtained from bank lending, using this alternative measure provides a better sense of the importance of ï¬?nancial development via the pure credit channel, as opposed to the possibility that the presence of deep capital markets may also play some role (which introduces elements of ï¬?nancial structure). Speciï¬?cations (R6) and (R7) decompose the institutional quality variable into, respectively, its rule of law and corruption subindices. Doing so renders the coefficient on the rule of law 4 The coefficient on institutional quality in (R2), while statistically insigniï¬?cant, is approaching signiï¬?cance (p = 0.20), and the sign remains unchanged. 5 The reason why this measure is not favored for the baseline, however, is that there remain signiï¬?cant frictions to cross-border capital flows, so that domestic investors do not typically have ready access to global capital markets. 11 Table 2: Robustness regressions for ï¬?xed investment, unbalanced 5-year average panel, 1980–2009† R1 R2 R3 R4 R5 R6 R7 R8 R9 R10 R11 R12 Lagged investment/ 0.404 0.429 0.473 0.438 0.578 0.528 0.503 0.315 0.445 0.667 0.476 0.498 share (0.16)∗∗ (0.09)∗∗∗ (0.18)∗∗∗ (0.18)∗∗ (0.18)∗∗∗ (0.23)∗∗ (0.19)∗∗∗ (0.12)∗∗∗ (0.17)∗∗∗ (0.09)∗∗∗ (0.15)∗∗∗ (0.18)∗∗∗ Output 0.202 1.334 1.422 1.464 1.565 1.129 1.474 1.636 1.376 1.535 1.643 1.419 (0.04)∗∗∗ (0.22)∗∗∗ (0.30)∗∗∗ (0.27)∗∗∗ (0.33)∗∗∗ (0.34)∗∗∗ (0.32)∗∗∗ (0.24)∗∗∗ (0.25)∗∗∗ (0.34)∗∗∗ (0.34)∗∗∗ (0.28)∗∗∗ Output growth 0.507 0.498 0.543 0.402 0.431 0.470 0.711 0.543 0.343 0.162 0.472 (0.08)∗∗∗ (0.19)∗∗∗ (0.18)∗∗∗ (0.18)∗∗ (0.22)∗∗ (0.19)∗∗ (0.13)∗∗∗ (0.18)∗∗∗ (0.10)∗∗∗ (0.30) (0.18)∗∗∗ Cost of capital 0.210 0.422 1.248 0.870 1.399 1.152 1.994 1.538 -1.850 2.537 1.361 (0.17) (0.76) (1.53) (1.45) (1.55) (1.55) (1.38) (1.76) (1.37) (1.06)∗∗ (1.61) Cost of capital 0.003 alt. (0.01) Trade openness 0.005 0.085 -0.101 0.001 -0.021 -0.020 -0.052 0.058 0.021 -0.425 -0.155 -0.120 (0.03) (0.13) (0.23) (0.21) (0.19) (0.20) (0.22) (0.11) (0.27) (0.15)∗∗∗ (0.20) (0.23) Financial openness -0.035 -0.048 -0.183 -0.202 -0.181 -0.176 -0.192 -0.108 -0.173 -0.178 -0.188 -0.185 (0.01)∗∗∗ (0.05) (0.07)∗∗ (0.07)∗∗∗ (0.08)∗∗ (0.07)∗∗ (0.07)∗∗∗ (0.07)∗ (0.07)∗∗ (0.08)∗∗ (0.09)∗∗ (0.07)∗∗ Financial 0.046 0.238 0.327 0.257 0.305 0.337 0.063 0.246 -0.031 0.366 0.332 development (0.02)∗∗∗ (0.11)∗∗ (0.14)∗∗ (0.12)∗∗ (0.15)∗∗ (0.14)∗∗ (0.10) (0.13)∗ (0.11) (0.15)∗∗ (0.13)∗∗ Financial 0.257 development, alt. (0.17) Institutional 0.026 0.086 0.149 0.053 0.170 0.185 0.142 0.254 0.178 0.147 quality (0.01)∗ (0.06) (0.09) (0.10) (0.09)∗ (0.08)∗∗ (0.08)∗ (0.12)∗∗ (0.10)∗ (0.09)∗ Rule of law 0.182 (0.10)∗ Corruption 0.060 (0.07) Business -0.033 -0.152 -0.129 0.105 -0.117 -0.194 -0.124 -0.094 -0.009 -0.224 -0.139 environment (0.03) (0.08)∗ (0.18) (0.18) (0.18) (0.21) (0.20) (0.10) (0.11) (0.20) (0.17) Business -0.014 environment, alt. (0.02) Institutional 0.008 0.032 0.097 0.091 0.128 0.130 0.140 0.065 0.068 0.115 0.104 structure (0.02) (0.05) (0.11) (0.09) (0.11) (0.12) (0.11) (0.05) (0.09) (0.10) (0.11) Institutional -0.003 structure, alt. (0.02) Financial -0.050 structure, alt. (0.05) Capital stock 0.327 (0.23) Financial crisis -0.075 (0.15) Wald χ2 131∗∗∗ 8,638∗∗∗ 11,449∗∗∗ 14,417∗∗∗ 19,590∗∗∗ 10,673∗∗∗ 11,240∗∗∗ 8,955∗∗∗ 9,806∗∗∗ 9,179∗∗∗ 10,976∗∗∗ 10,721∗∗∗ Hansen J 23.814 38.293 30.135 28.374 35.864 30.080 28.900 31.060 28.159 21.872 25.591 30.596 AR(2) z -1.763∗ -1.205 -1.446 -1.282 -1.528 -1.344 -1.668 0.108 -1.465 0.000 -1.173 -1.435 Instruments 39 50 43 43 46 41 41 45 42 38 46 44 N (countries) 333 (105) 340 (107) 333 (105) 324 (105) 334 (105) 333 (105) 333 (105) 234 (82) 330 (104) 174 (97) 317 (102) 333 (105) † All variables, except the ï¬?nancial crisis dummy, are in log form. Heteroskedasticity and autocorrelation-robust standard errors are reported in parentheses. Period ï¬?xed effects and a constant term were included in the regressions, but not reported. ∗ indicates signiï¬?cance at 10 percent level, ∗∗ indicates signiï¬?cance at 5 percent level, and ∗∗∗ indicates signiï¬?cance at 1 percent level. remains signiï¬?cant, while that on corruption is insigniï¬?cant. This implies that the results may be driven more by cross-country variations in property rights and the rule of law, as opposed to the pervasiveness of corruption.6 In columns (R8)–(R10) we consider alternative measures of the other structural variables. (R8) substitutes the ï¬?nancial structure variable with the ratio of stock market capitalization to domestic credit, which better approximates the influence of ï¬?nancial structure size as distinct from ï¬?nancial structure activity (Levine 2002). Nevertheless, using this alternative measures makes little different to the coefficient, which remains insigniï¬?cant. We conclude that, in con- trast to ï¬?nancial development, ï¬?nancial structure appears to exert no independent effect on investment, a ï¬?nding that echoes that of Ndikumana (2005). Column (R9) offers an alternative measure of the structure of political institutions, a con- centration index of the relative size of parties in parliament. This measure may offer a stronger sense of the level of political competition, as opposed to an index of democratic accountability alone. Finally, column (R10) replaces the business environment variable with an index of the extent of commercial contract enforcement. The main results in Table 1 are largely undisturbed by these three alternative measures, although we note that the coefficient on institutional quality tends to retain its statistical signiï¬?cance (and increase its magnitude) relative to the benchmark. Finally, Table 2 also considers the robustness of the benchmark results to the inclusion of several additional covariates. Column (R11) adds the capital stock, depreciated at a constant 5 percent.7 In the ï¬?nal column (R12), we introduce an additional indicator variable for ï¬?nancial crises. We deï¬?ne ï¬?nancial crises as the coincidence of banking and currency crises. In contrast to, say, a currency crisis—which may only result in nominal dislocations—such “twin crisesâ€? typically exact a large output cost (Hutchison & Noy 2005), and hence are likely to be especially devastating for investment. The coefficients on these variables are of the expected sign, but are statistically indistinguishable from zero, and the other results are qualitatively unaltered by the inclusion of these additional factors. 3.4 Possible channels of endogeneity In this subsection we consider the issue of endogeneity in the two structural variables of interest—ï¬?nancial development and institutional quality—more seriously. In particular, we exploit two external instruments for institutional quality and ï¬?nancial development that have opez-de Silanes, Shleifer been commonly used in the existing literature: legal origin (La Porta, L´ & Vishny 1998) (for ï¬?nancial development) and fraction of population speaking European lan- 6 The correlation on the two is Ï? = 0.57, which is certainly high but not excessively so. Indeed, replications of the benchmark regressions in Table 1 using only the rule of law variable generally result in more statistically signiï¬?cant coefficients for institutional quality (these are available from the author on request). We have retained the aggregate measure in the benchmark as we regard an aggregated measure as a more complete representation of institutional quality, rather than a measure of rule of law alone. 7 Using an alternative depreciation method, such as hyperbolic discounting, does not markedly change the results. 13 guages (Hall & Jones 1999) (for institutional quality),8 and embed them in the system GMM framework as additional exogenous instruments. Table 3: Regressions for ï¬?xed investment with exogenous instruments, unbalanced 5-year average panel, 1980–2009† E1 E5 E2 E6 E3 E7 Financial -0.297 -0.131 -0.042 0.292 -0.202 -0.185 development (0.26) (0.39) (0.27) (0.15)∗∗ (0.23) (0.20) Institutional 0.461 0.453 0.368 0.256 0.266 0.201 quality (0.19)∗∗ (0.23)∗∗ (0.20)∗ (0.33) (0.13)∗∗ (0.08)∗∗ Economic controls Yes Yes Yes Yes Yes Yes Structural controls No Yes No Yes No Yes Wald χ2 17,411∗∗∗ 5,766∗∗∗ 18,052∗∗∗ 9,832∗∗∗ 8,422∗∗∗ 16,822∗∗∗ Hansen J 38.280 33.215 35.482 28.811 36.237 39.364 AR(2) z -1.446 -1.339 -1.286 -1.457 -1.536 -1.481 Instruments 42 41 41 41 41 43 External? Both Both IQ only IQ only FD only FD only N (countries) 408 (106) 337 (105) 403 (106) 333 (105) 337 (105) 337 (105) † All variables are in log form. Heteroskedasticity and autocorrelation-robust standard errors are reported in parentheses. All variables are in log form. Heteroskedasticity and autocorrelation-robust standard errors are reported in parentheses. Period ï¬?xed effects, a constant term, and economic (all speciï¬?cations) and additional structural controls (E2, E4, E6) were included in the regressions, but not reported. IQ = institutional quality, FD = ï¬?nancial development. ∗ indicates signiï¬?cance at 10 percent level, ∗∗ indicates signiï¬?cance at 5 percent level, and ∗∗∗ indicates signiï¬?cance at 1 percent level. The results can be found in Table 3, both without (columns (E1)–(E3)) and with (columns (E5)–(E7)) additional institutional and structural controls (so that they are analogous to to speciï¬?cations (B4) and (B6), respectively). To better understand the sensitivity of the results to the use of internal instruments, the ï¬?rst two columns (E1/E5)9 include both external instru- ments, while the next two (E2/E6) take the (external) institutional quality instrument seriously by using only the language share instruments alongside lagged ï¬?nancial development (as inter- nal instruments) in the exogenous instrument matrix. The ï¬?nal two columns (E3/E7) treat the external ï¬?nancial development instrument seriously by using only legal origins alongside lagged institutional quality in the exogenous instrument matrix. Taken together, these results convey a consistent message: Conditional on the external in- struments being valid, institutional quality is more likely to have a causal impact on investment, as opposed to ï¬?nancial development. Institutional quality retains its positive and signiï¬?cant coefficient in virtually all speciï¬?cations, while ï¬?nancial development is only signiï¬?cant in one speciï¬?cation (E6), which relies on the internal instruments for ï¬?nancial development. Although the relatively weak result for ï¬?nancial development does not necessarily negate the possibility 8 glu, Johnson An alternative (and somewhat popular) instrument for institutions is settler mortality (Acemo˘ & Robinson 2001). For the sake of parsimony, we report results using this instrument—which are similar to the language share instrument—in the annex. 9 The nonconsecutive numbering of the columns is to allow correspondence with the full results, which are provided in the annex. 14 that it could still be an important structural determinant of ï¬?xed investment activity—there are potential issues with the quality of legal origin as an instrument, after all (Kraay 2012)— we are nevertheless led to the conclusion that institutional quality is more likely to exert an unequivocal causal effect on investment. 3.5 Interactions between development and structure In this subsection we explore the interaction effects of ï¬?nancial development and institutional quality—which we regard as development measures—with that of structure measures corre- sponding to each. In particular, we interact our measure of ï¬?nancial development with that of ï¬?nancial structure, and institutional quality with that of institutional structure. In doing so, we hope to obtain further insight on the conditions in which our key variables of interest may or may not be operative. These results are summarized in Table 4. We consider interaction effects pertaining to ï¬?nancial development and structure (I1)–(I3), and institutional quality and structure (I4)–(I5). In an analogous fashion to Table 3, we report the results with only economic controls (I1/I4), and with both economic and structural controls (I2–I3/I5) (for reasons documented in Section 3.2, including ï¬?nancial structure severely decreases the sample size; we therefore allow for either the exclusion or inclusion of this control to ensure that sample choice is not driving our results). We consider these effects in turn. Insofar as institutional quality is concerned, the effect of institutional quality does appear to be conditioned by structure; the coefficient on the interac- tion term is signiï¬?cant across all three speciï¬?cations (I1)–(I3). This suggests that, conditional on the quality of institutions, the degree of democratic development in an economy (recall, our benchmark institutional structure variable is an index of democratic accountability) raises the level of investment; this contrasts to the unconditioned effect of institutional structure being in- signiï¬?cant (Table 1 and Table 2).10 The important conditioning effect required by institutional quality for institutional structure to play a role serves as an important caveat to more straight- forward claims that merely improving democratic accountability and voice will necessarily lead to improved economic performance (in this case, investment).11 Note that, while the coefficient on institutional quality is now negative, the total effect— which requires that we add this coefficient to the product of institutional structure and the coefficient on the interaction term—is likely to be positive for the majority of observations. For example, for the fullest speciï¬?cation (I3), the sample mean of institutional quality and structure are 1.54 and 1.70, respectively, which yields the partial derivative of −1.40 + 0.87 (1.70) = 0.08. Furthermore, when taken in tandem with the negative (and signiï¬?cant in 2 of the 3 speciï¬?cations) coefficient on institutional structure, the combination points to why including 10 It is useful to recall, as noted in Table A.4, that these two variables are actually fairly distinct, with the correlation between them (in our sample) being 0.45. 11 Another way to frame this point is that inclusive political institutions (Acemo˘ glu & Robinson 2012) re- quire not only that such institutions encourage broad-based participation from economic agents, but that this participation be premised on rules of the game that are supportive of economic activity. 15 Table 4: Regressions for ï¬?xed investment with interaction terms (variables of interest), unbalanced 5-year average panel, 1980–2009† I1 I2 I3 I4 I5 Financial 0.017 0.028 0.034 0.102 0.097 development (0.11) (0.11) (0.12) (0.16) (0.15) Financial 0.052 0.166 0.084 structure (0.14) (0.24) (0.23) Fin. dev. × -0.153 -0.057 ï¬?n. struc. (0.22) (0.21) Institutional -1.158 -1.152 -1.396 0.162 0.136 quality (0.60)∗ (0.60)∗ (0.91)∗ (0.08)∗∗ (0.10) Institutional -0.934 -0.935 -1.221 0.098 structure (0.48)∗ (0.47)∗∗ (0.75) (0.09) Inst. qual. × 0.692 0.697 0.869 inst. struc. (0.34)∗∗ (0.33)∗∗ (0.51)∗ Economic controls Yes Yes Yes Yes Yes Structural controls No Partial Full No Yes Wald χ2 18,421∗∗∗ 18,225∗∗∗ 9,250∗∗∗ 10,117∗∗∗ 13,206∗∗∗ Hansen J 29.435 28.634 30.755 28.874 29.156 AR(2) z -1.927∗ -1.853∗ -1.003 0.165 0.105 Instruments 45 46 47 41 44 N (countries) 321 (105) 321 (105) 229 (82) 236 (82) 236 (82) † All variables are in log form. Heteroskedasticity and autocorrelation-robust standard errors are reported in parentheses. Period ï¬?xed effects, a constant term, and economic (all speciï¬?- cations) and additional structural controls (I2–I4) were included in the regressions, but not reported. ∗ indicates signiï¬?cance at 10 percent level, ∗∗ indicates signiï¬?cance at 5 percent level, and ∗∗∗ indicates signiï¬?cance at 1 percent level. institutional structure alone (without an interaction term) may yield a coefficient statistically indistinguishable from zero, as the two cancel out. For ï¬?nancial development, including an interaction term with ï¬?nancial structure leads to the coefficient on all three being statistically insigniï¬?cant. This echoes the result in column (B8) of Table 1, and could be due to a more restrictive sample being employed when ï¬?nancial structure is included. However, another reason can be surmised by examining the coefficient on the interaction term: since it is negative (and relatively large), allowing for interaction effects likely means that the negative conditioning effect of ï¬?nancial structure on development may potentially give rise to a statistically insigniï¬?cant coefficient on the independent term. Finally, we should also note that, across all speciï¬?cations, institutional quality tends to be statistically signiï¬?cant,12 but not so for ï¬?nancial development. While we hesitate to rule out ï¬?nancial development altogether due to the more restrictive sample in most of the speciï¬?cations in Table 4, it is nonetheless the case that—as it was in Table 3—the signiï¬?cant impact of ï¬?nancial development on investment is a somewhat more fragile result. 12 Even for column (I5), where the coefficient on institutional quality is insigniï¬?cant, p = 0.158. 16 3.6 Subsample analysis In this subsection we probe further into when ï¬?nancial development and institutional quality may matter by splitting the main sample into distinct subsamples, chosen to potentially offer additional insight into the circumstances under which these variables are operative. The ï¬?rst column (S1) of Table 5 presents results for a subsample comprising industrialized economies, as captured by membership in the Organisation for Economic Cooperation and Development (OECD) or its status as a Newly Industrialized Economy (NIE),13 using our preferred speciï¬?cation (B6) that includes both structural and economic controls. Column (S2) reports results from the mutually exclusive (from S1) subsample of non-industrialized economies. For the ï¬?nal two columns, we split the sample by the mean level of ï¬?nancial development and institutional quality, and report regressions using the above-average subsample for the former (S3) and latter (S4). Table 5: Regressions for ï¬?xed investment on selected subsam- ples, unbalanced 5-year average panel, 1980–2009† S1 S2 S3 S4 Financial 0.192 -0.025 0.117 0.154 development (0.10)∗∗ (0.14) (0.11) (0.09)∗ Institutional -0.279 0.315 0.223 0.139 quality (0.18) (0.19)∗ (0.10)∗∗ (0.17) Economic controls Yes Yes Yes Yes Structural controls Yes Yes Yes Yes Wald χ2 33,096∗∗∗ 3,454∗∗∗ 18,604∗∗∗ 51,763∗∗∗ Hansen J 16.147 26.271 33.258 29.902 AR(2) z 0.004 -1.191 -0.834 0.287 Instruments 42 42 43 48 N (countries) 104 (32) 220 (73) 144 (51) 177 (68) Subsample? Ind. Non-ind. High FD High IQ † All variables are in log form. Heteroskedasticity and autocorrelation-robust standard errors are reported in parentheses. Period ï¬?xed effects and a con- stant term were included in the regressions, but not reported. FD = ï¬?nancial development, IQ = institutional quality. ∗ indicates signiï¬?cance at 10 percent level, ∗∗ indicates signiï¬?cance at 5 percent level, and ∗∗∗ indicates signiï¬?cance at 1 percent level. The results in Table 5 offers further hints as to what drives our main results. Consider, ï¬?rst, the results from the industrialized/non-industrialized subsamples. It is clear that, for industrial- ized economies, ï¬?nancial development is far more important in stimulating investment activity, whereas institutional quality is more central for investment in non-industrialized ones.14 This result suggests that—in non-industrialized economies where the strength of institutions is typi- 13 Deï¬?ned to include Hong Kong SAR, South Korea, Singapore, and Taiwan; in our dataset, this only expands the OECD subsample to include Hong Kong and Singapore, since Taiwan is not in our data, and South Korea is in any case a member of the OECD. 14 This ï¬?nding survives in a pure-OECD/non-OECD subsample as well; these results are available on request. 17 cally weak—it is institutional quality that binds as a constraint on higher levels of investment, whereas ï¬?nancial depth is more central in industrialized economies. The results from the high ï¬?nancial development/institutional quality subsamples further indicate that the influence of each on investment may well be nonlinear: At above-average levels of each respective variable, their effects flatten out, so that—while they retain their positive coefficients—their magnitudes are smaller, so they are no longer statistical signiï¬?cance (although the effect of the other corresponding variable remains at least marginally signiï¬?cant). Importantly, there is limited overlap between the two high subsamples: 34 economies appear in the high institutional quality subsample that do not appear in the high ï¬?nancial development subsample, and conversely, 18 countries appear in the high ï¬?nancial development subsample but not the high institutional quality one. Nor do these countries appear to be mainly high-income or developing. The implication of this fairly large non-overlap, then, is that the nonlinearity result does not seem to be driven by a small set of countries, but is reflective of a more systematic difference between economies that demonstrate high levels of either ï¬?nancial development or institutional quality. 4 Conclusion In this paper, we have sought to empirically examine the manner by which structural and in- stitutional factors contribute to cross-country variation in investment activity. We obtain two main ï¬?ndings. First, we ï¬?nd that ï¬?nancial development and institutional quality are reason- ably robust determinants of investment, even after controlling for a host of additional candidate structural variables and economic controls, alternative measures of investment and other struc- tural variables, additional confounding variables. Second, while these results are likely to be robust to weak endogeneity concerns, using external instruments leads to the conclusion that institutional quality is likely to be less sensitive to reverse causality concerns. Our ï¬?ndings offer a nice complement to the existing literature on the role of ï¬?nancial devel- opment and institutions in economic growth. But in contrast to that voluminous literature, we are able to establish the contribution of these variables on a speciï¬?c channel for growth—capital accumulation—and to demonstrate that the dominance of institutional quality in influencing economic performance (Rodrik, Subramanian & Trebbi 2004), while not ruling out the im- portant role that ï¬?nancial development can play, in contrast to other structural determinants. Future research that seeks to model the key dynamics of investment can thus beneï¬?t from a more intentional modeling of these two factors, in particular the manner by which the two may interact to influence capital accumulation decisions. The results in this paper point to the fact that a favorable investment climate is characterized not only by traditional policy areas that can foster private sector investment—such as a stable macroeconomic and regulatory regime, and tax credits favoring business investment—but also by the broader institutional environment in which ï¬?rms operate, which includes secure property 18 rights and stable rule of law, and by the governance framework, such as adequate control of corruption. In an analogous fashion, policy that seeks to enhance investment ï¬?nancing should probably focus on improving the level of development of the ï¬?nancial sector, as opposed to narrowly-conceived investment credits and incentives. Such well-functioning ï¬?nancial systems are more likely to ensure a superior mobilization and corresponding allocation of saving toward the most productive investment opportunities, and raise the level of investment in the economy overall. References glu, K. Daron & Simon Johnson (2005). “Unbundling Institutionsâ€?. Journal of Political Economy 113(5) Acemo˘ (October): 949–995 glu, K. Daron, Simon Johnson & James A. Robinson (2001). “The Colonial Origins of Comparative Acemo˘ Development: An Empirical Investigationâ€?. American Economic Review 91(5) (December): 1369–1401 glu, K. Daron & James A. Robinson (2012). Why Naitons Fail: The Origins of Power, Prosperity, and Acemo˘ Poverty. New York, NY: Crown Business Agosin, Manuel R. & Roberto Machado (2005). “Foreign Investment in Developing Countries: Does it Crowd in Domestic Investment?â€? Oxford Development Studies 33(2) (Spring): 149–162 Arellano, Manuel & Olympia Bover (1995). “Another Look at the Instrumental Variable Estimation of Error- Components Modelsâ€?. Journal of Econometrics 68(1) (July): 29–51 Bai, Yan & Jing Zhang (2010). “Solving the Feldstein-Horioka Puzzle With Financial Frictionsâ€?. Econometrica 78(2) (March): 603–632 Bartelsman, Eric J., John C. Haltiwanger & Stefano Scarpetta (2010). “Cross-Country and Within-Country Differences in the Business Climateâ€?. International Journal of Industrial Organization 28(4) (July): 368–371 Beck, Thorsten, George R.G. Clarke, Alberto Groff, Philip E. Keefer & Patrick Walsh (2001). “New Tools in Comparative Political Economy: The Database of Political Institutionsâ€?. World Bank Economic Review 15(1) (January): 165–176 Beck, Thorsten, Aslı Demirg¨ c-Kunt & Ross Levine (2000). “A New Database on the Structure and Development u¸ of the Financial Sectorâ€?. World Bank Economic Review 14(3) (September): 597–605 Benhabib, Jess & Mark M. Spiegel (2000). “The Role of Financial Development in Growth and Investmentâ€?. Journal of Economic Growth 5(4) (December): 341–360 Besley, Timothy J. (1995). “Property Rights and Investment Incentives: Theory and Evidence from Ghanaâ€?. Journal of Political Economy 103(5) (October): 903–937 Blundell, Richard & Stephen R. Bond (1998). “Initial Conditions and Moment Restrictions in Dynamic Panel Data Modelsâ€?. Journal of Econometrics 87(1) (August): 115–143 Bond, Stephen R., Anke Hoeffler & Jonathan R.W. Temple (2001). “GMM Estimation of Empirical Growth Modelsâ€?. Discussion Paper 3048, London, England: Centre for Economic Policy Research 19 Byrne, Joseph P. & E. Philip Davis (2005). “The Impact of Short- and Long-run Exchange Rate Uncertainty on Investment: A Panel Study of Industrial Countriesâ€?. Oxford Bulletin of Economics and Statistics 67(3) (June): 307–329 Byrne, Joseph P., Giorgio Fazio & Norbert M. Fiess (2009). “The Global Side of the Investment-Saving Puzzleâ€?. Journal of Money, Credit and Banking 41(5) (August): 1033–1040 Caballero, Ricardo J. (1999). “Aggregate Investmentâ€?. In John B. Taylor & Michael D. Woodford (editors), Handbook of Macroeconomics, volume 1 of B, pp. 813–861. Amsterdam, the Netherlands: Elsevier Caballero, Ricardo J. & Eduardo M.R.A. Engel (1999). “Explaining Investment Dynamics in U.S. Manufacturing: A Generalized (S, s) Approachâ€?. Econometrica 67(4) (July): 783–826 Caballero, Ricardo J. & Robert S. Pindyck (1996). “Uncertainty, Investment, and Industry Evolutionâ€?. Interna- tional Economic Review 37(3) (August): 641–662 on, C´ Calder´ esar A., Alberto E. Chong & Norman V. Loayza (2002). “Determinants of Current Account Deï¬?cits in Developing Countriesâ€?. B.E. Journal of Macroeconomics 2(1) (March): Article 2 Campos, Nauro F. & Jeffrey B. Nugent (2003). “Aggregate Investment and Political Instability: An Econometric Investigationâ€?. Economica 70(279) (August): 533–549 Caselli, Francesco & James D. Feyrer (2007). “The Marginal Product of Capitalâ€?. Quarterly Journal of Economics 122(2) (May): 535–568 Chinn, Menzie D. & Hiro Ito (2008). “A New Measure of Financial Opennessâ€?. Journal of Comparative Policy Analysis 10(3) (September): 309–322 Chinn, Menzie D. & Eswar S. Prasad (2003). “Medium-Term Determinants of Current Accounts in Industrial and Developing Countries: An Empirical Explorationâ€?. Journal of International Economics 59(1) (January): 47–76 Chirinko, Robert S. (1993). “Business Fixed Investment Spending: Modeling Strategies, Empirical Results, and Policy Implicationsâ€?. Journal of Economic Literature 31(4) (December): 1875–1911 Cooley, Thomas F., Ramon Marimon & Vincenzo Quadrini (2004). “Aggregate Consequences of Limited Contract Enforceabilityâ€?. Journal of Political Economy 112(4) (August): 817–847 Dailami, Mansoor (1992). “Optimal Corporate Debt Financing and Real Investment Decisions under Controlled Banking Systemsâ€?. In Klaus P. Fischer & George J. Papaioannou (editors), Business Finance in Less Developed Capital Markets, pp. 229–249. Westport, CT: Greenwood Press Davis, E. Philip (2010). “New International Evidence on Asset-Price Effects on Investment, and a Survey for Consumptionâ€?. OECD Journal: Economic Studies 2010(1) (December): 1–50 Dawson, John W. (1998). “Institutions, Investment, and Growth: New Cross-Country and Panel Data Evidenceâ€?. Economic Inquiry 36(4) (October): 603–619 Devereux, Michael P. (1996). “Investment, Saving, and Taxation in an Open Economyâ€?. Oxford Review of Economic Policy 12(2) (Summer): 90–108 Dixit, Avinash K. & Robert S. Pindyck (1994). Investment under Uncertainty. Princeton, NJ: Princeton Uni- versity Press 20 Feldstein, Martin S. & Charles Y. Horioka (1980). “Domestic Saving and International Capital Flowsâ€?. Economic Journal 90(358) (June): 314–329 Ferderer, J. Peter (1993). “The Impact of Uncertainty on Aggregate Investment Spending: An Empirical Anal- ysisâ€?. Journal of Money, Credit and Banking 25(1) (February): 30–48 Gehlbach, Scott & Philip E. Keefer (2011). “Investment Without Democracy: Ruling-Party Institutionalization and Credible Commitment in Autocraciesâ€?. Journal of Comparative Economics 39(2) (June): 123–139 org, Holger & David Greenaway (2004). “Much Ado About Nothing? Do Domestic Firms Really Beneï¬?t from G¨ Foreign Direct Investment?â€? World Bank Research Observer 19(2) (September): 171–197 Greene, Joshua E. & Delano S. Villanueva (1991). “Private Investment in Developing Countries: An Empirical Analysisâ€?. IMF Staff Papers 38(1) (March): 33–58 Hall, Robert E. & Charles I. Jones (1999). “Why Do Some Countries Produce So Much More Output Per Worker Than Others?â€? Quarterly Journal of Economics 114(1) (February): 83–116 Hall, Robert E. & Dale W. Jorgenson (1967). “Tax Policy and Investment Behaviorâ€?. American Economic Review 57(3) (June): 391–414 Hayashi, Fumio (1982). “Tobin’s Marginal q and Average q: A Neoclassical Interpretationâ€?. Econometrica 50(1) (January): 147–159 om, Bengt R. & Jean Tirole (1997). “Financial Intermediation, Loanable Funds, and the Real Sectorâ€?. Holmstr¨ Quarterly Journal of Economics 112(3) (August): 663–691 Hon, Chu. Kam (2012). “The Feldstein-Horioka Puzzle and Spurious Ratio Correlationâ€?. Journal of International Money and Finance 31(2) (March): 292–309 Hsieh, Chang-Tai & Peter J. Klenow (2007). “Relative Prices and Relative Prosperityâ€?. American Economic Review 97(3) (June): 562–585 Hutchison, Michael M. & Ilan Noy (2005). “How Bad Are Twins? Output Costs of Currency and Banking Crisesâ€?. Journal of Money, Credit, and Banking 37(4) (August): 725–752 Jorgenson, Dale W. (1963). “Capital Theory and Investment Behaviorâ€?. American Economic Review 53(2) (May): 247–259 Kopcke, Richard W. & Richard S. Brauman (2001). “The Performance of Traditional Macroeconomic Models of Businesses’ Investment Spendingâ€?. New England Economic Review 1(2) (Summer): 3–39 Kraay, Aart C. (2012). “Instrumental Variables Regressions with Uncertain Exclusion Restrictions: A Bayesian Approachâ€?. Journal of Applied Econometrics 27(1) (January/February): 108–128 en & Jaume Ventura (2005). “Country Portfoliosâ€?. Journal of the Kraay, Aart C., Norman V. Loayza, Luis Serv´ European Economic Association 3(4) (June): 914–945 Kraay, Aart C. & Jaume Ventura (2000). “Current Accounts in Debtor And Creditor Countriesâ€?. Quarterly Journal of Economics 115(4) (November): 1137–1166 opez-de Silanes, Andrei Shleifer & Robert W. Vishny (1998). “Law and Financeâ€?. La Porta, Rafael, Florencio L´ Journal of Political Economy 106(6) (December): 1113–1155 21 an Valencia (2012). “Systemic Banking Crises Database: An Updateâ€?. Working Paper Laeven, Luc & Fabi´ WP/12/163, Washington, DC: International Monetary Fund Levine, Ross (2002). “Bank-Based or Market-Based Financial Systems: Which Is Better?â€? Journal of Financial Intermediation 11(4) (October): 398–428 Levine, Ross (2005). “Finance and Growth: Theory and Evidenceâ€?. In Philippe M. Aghion & Steven N. Durlauf (editors), Handbook of Economic Growth, volume 1A, pp. 865–934. Amsterdam, the Netherlands: Elsevier Love, Inessa & Lea Zicchino (2006). “Financial Development and Dynamics Investment: Evidence from Panel VARâ€?. Quarterly Review of Economics and Finance 46(2) (May): 190–210 Lucas, Robert E., Jr & Edward C. Prescott (1971). “Investment Under Uncertaintyâ€?. Econometrica 39(5) (September): 659–681 Mauro, Paolo (1995). “Corruption and Growthâ€?. Quarterly Journal of Economics 110(3) (August): 681–712 Morrissey, Oliver & Manop Udomkerdmongkol (2012). “Governance, Private Investment and Foreign Direct Investment in Developing Countriesâ€?. World Development 40(3) (March): 437–445 Narula, Rajneesh & Nigel L. Driffield (2012). “Does FDI Cause Development? The Ambiguity of the Evidence and Why it Mattersâ€?. European Journal of Development Research 24(1) (February): 1–7 Ndikumana, L´ eonce (2005). “Financial Development, Financial Structure, and Domestic Investment: Interna- tional Evidenceâ€?. Journal of International Money and Finance 24(4) (June): 651–673 Nickell, Stephen J. (1981). “Biases in Dynamic Models with Fixed Effectsâ€?. Econometrica 49(6) (November): 1417–1426 Oliner, Stephen D., Glenn D. Rudebusch & Daniel E. Sichel (1995). “New and Old Models of Business Investment: A Comparison of Forecasting Performanceâ€?. Journal of Money, Credit and Banking 27(3) (August): 806–826 Rodrik, Dani, Arvind Subramanian & Francesco Trebbi (2004). “Institutions Rule: The Primacy of Institutions over Geography and Integration in Economic Developmentâ€?. Journal of Economic Growth 9(2) (June): 131– 165 Roodman, David M. (2009). “A Note on the Theme of Too Many Instrumentsâ€?. Oxford Bulletin of Economics and Statistics 71(1) (February): 135–158 Schiantarelli, Fabio (1996). “Financial Constraints and Investment: Methodological Issues and International Evidenceâ€?. Oxford Review of Economic Policy 12(2) (Summer): 70–89 en, Luis (2003). “Real-Exchange-Rate Uncertainty and Private Investment in LDCsâ€?. Review of Economics Serv´ and Statistics 85(1) (February): 212–218 Shleifer, Andrei & Daniel Wolfenzon (2002). “Investor Protection and Equity Marketsâ€?. Journal of Financial Economics 66(1) (October): 3–27 Tobin, James (1969). “A General Equilibrium Approach To Monetary Theoryâ€?. Journal of Money, Credit and Banking 1(1) (February): 15–29 Utrero-Gonz´alez, Natalia M. (2007). “Banking Regulation, Institutional Framework and Capital Structure: International Evidence from Industry Dataâ€?. Quarterly Review of Economics and Finance 47(4) (September): 481–506 22 Wai, U Tun & Chorng-huey Wong (1982). “Determinants of Private Investment in Developing Countriesâ€?. Journal of Development Studies 19(1) (Winter): 19–36 World Bank (2013). Global Development Horizons 2013: Capital for the Future. Washington, DC: The World Bank 23 Technical Appendix Data description This subsection reports basic features related to the data, for the main variables of interest. This includes detailed sources and deï¬?nitions (Table A.1), countries included in the sample (Table A.2), standard summary statistics (Table A.3), and the corresponding correlation matrix (Table A.4). Detailed robustness regression results This subsection reports the full results of the regressions for ï¬?xed investment with exogenous instruments, with (Table A.6) and without (Table A.5) additional institutional and structural variables, analogous to speciï¬?cations (B4) and (B6), respectively. The speciï¬?cations below rely on exogenous instrument sets that vary from the benchmark according to: (A.E1) and (A.E5) utilize the Hall & Jones (1999) language share and La Porta et al. (1998) legal origin instruments; (A.E2) and (A.E6) utilize only language shares, with lagged domestic credit included as internal exogenous instruments; (A.E3) and (A.E7) utilize only legal origin, with lagged institutional glu quality included as internal exogenous instruments; and (A.E4) and (A.E8) utilize the Acemo˘ et al. (2001) settler mortality and legal origin instruments. We also report full results for the regressions with interaction terms; these are likewise reported with and without additional institutional and structural controls (Table A.7). (A.I1) includes only economic controls for regressions that include an interaction term for institutional quality and structure, while (A.I2) includes additional structural controls, with the exception of ï¬?nancial structure (since this reduced the sample signiï¬?cantly). To ensure that the results were not dependent on the expanded sample, (A.I3) includes ï¬?nancial structure in the set of structural controls. (A.I4) and (A.I5) repeat the exercise with the interaction between ï¬?nancial development and structure, both without and with additional structural controls, respectively. Finally, we report full results for regressions on subsamples, all with additional institutional and structural controls, analogous to speciï¬?cation (B6) (Table A.8). (A.S1) is for a subsample comprised of only economies in the OECD or are NIEs, while (A.S2) is for the mutually exclusive subsample of non-OECD/NIE economies. (A.S3) and (A.S4) are, respectively, subsamples where economies possess levels of ï¬?nancial development and institutional quality higher than their respective sample means. 24 Table A.1: Sources and deï¬?nitions for main variables of interest Variable Deï¬?nition Source Economic variables Fixed investment Gross ï¬?xed capital formation in constant 2000 U.S. dollars WDI† Output Gross domestic product (GDP) in constant 2000 U.S. dollars WDI Output growth Growth in real output‡ WDI Cost of capital Real interest rate (lending rate adjusted for inflation) WDI Trade openness Imports plus exports divided by GDP WDI Financial openness Index of restrictions on capital account openness Chinn & Ito (2008) Structural and institutional variables Financial development Domestic credit to private sector WDI Financial structure Stock market value traded divided by domestic credit Beck et al. (2000) Institutional quality Simple average of rule of law and control of corruption indices ICRG† Institutional structure Index of democratic accountability ICRG Business environment Index of strength of investment protection ICRG Tax structure Highest marginal corporate tax rate WDI 25 Alternative measures Gross investment Gross capital formation in constant 2000 U.S. dollars WDI Investment rate Gross ï¬?xed capital formation as share of GDP WDI Alternative cost of capital Difference in domestic and exchange-adjusted risk-free rate* Bloomberg, WDI Alternative ï¬?nancial development Domestic credit by banking sector WDI Alternative ï¬?nancial structure Stock market capitalization divided by domestic credit Beck et al. (2000) Alternative institutional structure Herï¬?ndahl index of government parties Beck et al. (2001) Alternative business environment Commercial contract enforcement index Doing Business Capital stock Stocks of capital in constant 2000 U.S. dollars§ WDI and GEM† Financial crisis Indicator variable for occurrence of ï¬?nancial crisis¶ Laeven & Valencia (2012) † WDI = World Development Indicators, ICRG = International Country Risk Guide, GEM = Global Economic Monitor. ICRG indicators are measured such that higher values indicate lower risk (better outcomes). ‡ Since the depreciation rate is (assumed) constant across countries, the difference between adjusting output growth for depreciation, as implied by the theoretical model, is trivial. * Computed as the U.S. real interest rate, multiplied by the change in the exchange rate. § Computed using the perpetual inventory method, with an assumed constant depreciation rate of 5 percent. Countries with insufficient data in the constant investment series were backcasted using a regression of the investment deflator on the GDP deflator and available investment data. ¶ Financial crisis deï¬?ned as the coincident occurrence of a banking and currency crisis within the 5-year period. 26 Table A.2: Sample of countries Albania Finland Netherlands Algeria France New Zealand Argentina Gabon Nicaragua Armenia Gambia Norway Australia Germany Pakistan Austria Greece Panama Azerbaijan Guatemala Papua N/ Guinea Bahamas Guinea Paraguay Bangladesh Honduras Peru Barbados* Hong Kong SAR Philippines Belarus Hungary Poland Belgium Iceland Portugal Belize* India Romania Benin* Indonesia Russia Bolivia Iran Senegal Bosnia & Herz.* Ireland Serbia* Botswana Israel Seychelles* Brazil Italy Singapore Brunei* Japan Slovak Rep. Bulgaria Jordan Slovenia Burkina Faso Kenya South Africa Cameroon Kyrgyz Rep.* South Korea Canada Lao PDR* Spain Cape Verde* Latvia Sri Lanka Cent. Afr. Rep.* Lebanon Swaziland* Chad* Lesotho* Sweden Chile Liberia Switzerland China Lithuania Syria Colombia Luxembourg* Tajikistan* Costa Rica Macao SAR* Tanzania Cote d’Ivoire Macedonia, FYR* Thailand Croatia Madagascar Togo Cyprus Malaysia Trin. & Tob. Czech Republic Maldives* Tunisia* Denmark Mali Uganda Djibouti* Malta Ukraine Dominica* Mauritania* United Kingdom Dominican Rep. Mauritius* United States Ecuador Mexico Uruguay Egypt Moldova Venezuela El Salvador Morocco Vietnam Estonia Mozambique Yemen Ethiopia Namibia Zambia * Countries that were excluded (due to data limitations) from the preferred benchmark speciï¬?cations (B4)–(B6) are denoted with an asterisk. 27 Table A.3: Summary statistics for main variables of interest Variable N Mean Std Dev Min Max Fixed investment 483 22.331 2.294 16.810 28.368 Output 483 23.916 2.229 19.319 30.066 Output growth 483 0.177 0.141 -0.691 0.865 Cost of capital 483 0.717 0.051 0.370 1.199 Trade openness 482 0.584 0.244 0.124 1.646 Financial openness 468 1.051 0.510 0.000 1.670 Financial development 482 0.370 0.262 0.016 1.223 Financial structure 323 0.241 0.279 0.000 1.256 Business environment 418 2.117 0.270 1.071 2.565 Tax environment 234 3.338 0.478 0.000 3.976 Institutional quality 418 1.490 0.275 0.405 1.946 Institutional structure 418 1.598 0.336 0.024 1.946 Table A.4: Correlation matrix for main variables of interest Fixed Output Output Cost of Trade Fin. Fin. Fin. Inv. Tax Inst. Inst. inv. growth capital open. open. dev. struc. climate env. quality struc. Fixed investment 1.000 Output 0.989 1.000 Output growth 0.056 0.007 1.000 Cost of capital -0.101 -0.091 -0.027 1.000 28 Trade openness -0.292 -0.326 0.132 -0.112 1.000 Financial openness 0.302 0.316 0.049 0.080 0.159 1.000 Financial development 0.585 0.575 0.053 -0.120 0.172 0.408 1.000 Financial structure 0.569 0.559 0.075 -0.185 0.037 0.233 0.424 1.000 Investment climate 0.303 0.290 0.245 0.006 0.319 0.478 0.478 0.343 1.000 Tax environment 0.265 0.285 -0.049 0.085 -0.254 0.138 -0.032 0.074 -0.199 1.000 Institutional quality 0.372 0.376 -0.052 -0.049 0.184 0.328 0.561 0.245 0.348 -0.021 1.000 Institutional structure 0.339 0.358 -0.109 0.013 0.005 0.417 0.379 0.081 0.384 -0.008 0.453 1.000 29 Table A.5: Regressions for ï¬?xed investment with exogenous instru- ments (economic controls only), unbalanced 5-year average panel, 1980–2009† A.E1 A.E2 A.E3 A.E4 Lagged investment 0.647 0.537 0.606 0.382 (0.13)∗∗ * (0.19)∗∗∗ (0.15)∗∗∗ (0.23)∗ Output 0.371 0.458 0.420 0.688 (0.13)∗∗∗ (0.18)∗∗ (0.16)∗∗∗ (0.22)∗∗∗ Output growth 1.233 1.134 1.495 1.253 (0.32)∗∗∗ (0.28)∗∗∗ (0.35)∗∗∗ (0.61)∗∗ Cost of capital 0.711 1.119 -0.588 -0.052 (1.35) (1.34) (0.89) (0.76) Trade openness -0.214 -0.091 -0.286 -0.020 (0.26) (0.23) (0.22) (0.38) Financial openness -0.194 -0.193 -0.140 -0.280 (0.07)∗∗∗ (0.07)∗∗∗ (0.07)∗∗ (0.08)∗∗∗ Financial -0.297 -0.042 -0.202 -0.452 development (0.26) (0.27) (0.23) (0.38) Institutional 0.461 0.368 0.266 0.660 quality (0.19)∗∗ (0.20)∗ (0.13)∗∗ (0.31)∗∗ Wald χ2 17,411∗∗∗ 18,052∗∗∗ 8,422∗∗∗ 2,299∗∗∗ Hansen J 38.280 35.482 36.237 22.966 AR(2) z -1.446 -1.286 -1.536 -1.332 Instruments 42 41 41 36 N (countries) 408 (106) 403 (106) 337 (105) 408 (106) † All variables are in log form. Heteroskedasticity and autocorrelation-robust stan- dard errors are reported in parentheses. Period ï¬?xed effects and a constant term were included in the regressions, but not reported. ∗ indicates signiï¬?cance at 10 per- cent level, ∗∗ indicates signiï¬?cance at 5 percent level, and ∗∗∗ indicates signiï¬?cance at 1 percent level. 30 Table A.6: Regressions for ï¬?xed investment with exogenous in- struments (economic and structural controls), unbalanced 5-year average panel, 1980–2009† A.E5 A.E6 A.E7 A.E8 Lagged investment 0.608 0.523 0.657 0.321 (0.22)∗∗∗ (0.23)∗∗ (0.08)∗∗∗ (0.26) Output 0.415 0.454 0.368 0.755 (0.23)∗ (0.24)∗ (0.10)∗∗∗ (0.26)∗∗∗ Output growth 1.415 1.513 1.395 1.053 (0.42)∗∗∗ (0.35)∗∗∗ (0.26)∗∗∗ (0.68) Cost of capital 0.423 1.573 0.824 0.848 (1.30) (1.48) (0.94) (2.14) Trade openness -0.157 -0.083 -0.024 0.319 (0.20) (0.21) (0.16) (0.52) Financial openness -0.140 -0.195 -0.187 -0.148 (0.08)∗ (0.08)∗∗ (0.06)∗∗∗ (0.10) Financial -0.131 0.292 -0.185 -0.133 development (0.39) (0.15)∗∗ (0.20) (0.42) Institutional 0.453 0.256 0.201 1.068 quality (0.23)∗∗ (0.33) (0.08)∗∗ (0.55)∗ Business -0.250 -0.268 0.031 -0.720 environment (0.34) (0.30) (0.17) (0.53) Institutional 0.028 0.118 0.081 -0.192 structure (0.14) (0.12) (0.09) (0.27) Wald χ2 5,766∗∗∗ 9,832∗∗∗ 16,822∗∗∗ 1,817∗∗∗ Hansen J 33.215 28.811 39.364 20.809 AR(2) z -1.339 -1.457 -1.481 -0.459 Instruments 41 41 43 37 N (countries) 337 (105) 333 (105) 337 (105) 408 (106) † All variables are in log form. Heteroskedasticity and autocorrelation-robust standard errors are reported in parentheses. Period ï¬?xed effects and a constant term were included in the regressions, but not reported. ∗ indicates signiï¬?cance at 10 percent level, ∗∗ indicates signiï¬?cance at 5 percent level, and ∗∗∗ indicates signiï¬?cance at 1 percent level. 31 Table A.7: Regressions for ï¬?xed investment with interaction terms, unbalanced 5-year average panel, 1980–2009† A.I1 A.I2 A.I3 A.I4 A.I5 Lagged investment 0.600 0.620 0.421 0.356 0.403 (0.11)∗∗∗ (0.11)∗∗∗ (0.08)∗∗∗ (0.09)∗∗∗ (0.10)∗∗∗ Output 0.414 0.393 0.572 0.629 0.592 (0.12)∗∗∗ (0.11)∗∗∗ (0.10)∗∗∗ (0.10)∗∗∗ (0.12)∗∗∗ Output growth 1.078 1.068 1.723 1.175 1.374 (0.29)∗∗∗ (0.29)∗∗∗ (0.30)∗∗∗ (0.23)∗∗∗ (0.33)∗∗∗ Cost of capital 0.768 1.095 0.306 1.278 1.555 (0.88) (0.87) (0.93) (1.29) (1.31) Trade openness 0.158 0.185 0.056 0.086 0.087 (0.16) (0.16) (0.13) (0.12) (0.11) Financial openness -0.156 -0.146 -0.075 -0.103 -0.123 (0.06)∗∗ (0.06)∗∗ (0.08) (0.06)* (0.07)∗ Financial 0.017 0.028 0.034 0.102 0.097 development (0.11) (0.11) (0.12) (0.16) (0.15) Financial 0.052 0.166 0.084 structure (0.14) (0.24) (0.23) Fin. dev. × -0.153 -0.057 ï¬?n. struc. (0.22) (0.21) Institutional -1.158 -1.152 -1.396 0.162 0.136 quality (0.60)∗ (0.60)∗ (0.91)∗ (0.08)∗∗ (0.10) Institutional -0.934 -0.935 -1.221 0.098 structure (0.48)∗ (0.47)∗∗ (0.75) (0.09) Inst. qual. × 0.692 0.697 0.869 inst. struc. (0.34)∗∗ (0.33)∗∗ (0.51)∗ Business -0.069 -0.055 -0.099 environment (0.16) (0.18) (0.20) Wald χ2 18,421∗∗∗ 18,225∗∗∗ 9,250∗∗∗ 10,117∗∗∗ 13,206∗∗∗ Hansen J 29.435 28.634 30.755 28.874 29.156 AR(2) z -1.927∗ -1.853∗ -1.003 0.165 0.105 Instruments 45 46 47 41 44 N (countries) 321 (105) 321 (105) 229 (82) 236 (82) 236 (82) † All variables are in log form. Heteroskedasticity and autocorrelation-robust standard errors are reported in parentheses. Period ï¬?xed effects and a constant term were in- cluded in the regressions, but not reported. ∗ indicates signiï¬?cance at 10 percent level, ∗∗ indicates signiï¬?cance at 5 percent level, and ∗∗∗ indicates signiï¬?cance at 1 percent level. 32 Table A.8: Regressions for ï¬?xed investment on selected subsam- ples, unbalanced 5-year average panel, 1980–2009† A.S1 A.S2 A.S3 A.S4 Lagged investment 0.634 0.591 0.471 0.698 (0.10)∗∗∗ (0.13)∗∗∗ (0.18)∗∗ (0.21)∗∗∗ Output 1.012 1.164 1.060 1.385 (0.39)∗∗∗ (0.39)∗∗∗ (0.39)∗∗∗ (0.32)∗∗∗ Output growth 0.318 0.437 0.501 0.266 (0.11)∗∗∗ (0.14)∗∗∗ (0.18)∗∗∗ (0.21) Cost of capital -3.686 1.995 -0.431 -0.439 (1.94)∗ (1.26) (0.84) (1.02) Trade openness -0.099 0.243 -0.094 -0.071 (0.16) (0.24) (0.11) (0.14) Financial openness -0.096 -0.174 -0.068 0.002 (0.08) (0.09)∗ (0.06) (0.05) Financial 0.192 -0.025 0.117 0.154 development (0.10)∗∗ (0.14) (0.11) (0.09)∗ Institutional -0.279 0.315 0.223 0.139 quality (0.18) (0.19)∗ (0.10)∗∗ (0.17) Business 0.641 -0.209 -0.085 -0.103 environment (0.26)∗∗ (0.19) (0.28) (0.14) Institutional -0.041 -0.058 -0.162 0.063 structure (0.14) (0.11) (0.13) (0.08) Wald χ2 33,096∗∗∗ 3,454∗∗∗ 18,604∗∗∗ 51,763∗∗∗ Hansen J 16.147 26.271 33.258 29.902 AR(2) z 0.004 -1.191 -0.834 0.287 Instruments 42 42 43 48 N (countries) 104 (32) 220 (73) 144 (51) 177 (68) Subsample? Ind. Non-ind. High FD High IQ † All variables are in log form. Heteroskedasticity and autocorrelation-robust standard errors are reported in parentheses. Period ï¬?xed effects and a con- stant term were included in the regressions, but not reported. FD = ï¬?nancial development, IQ = institutional quality. ∗ indicates signiï¬?cance at 10 percent level, ∗∗ indicates signiï¬?cance at 5 percent level, and ∗∗∗ indicates signiï¬?cance at 1 percent level.