Policy Research Working Paper 10811 The Role of Financial (Mis)allocation on Real (Mis)allocation Firm-level Evidence for European Countries Ana P. Cusolito Roberto N. Fattal-Jaef Davide S. Mare Akshat V. Singh Development Research Group Global Indicators Group & Europe and Central Asia Region June 2024 Policy Research Working Paper 10811 Abstract This paper leverages the novel methodology by Whited and 75 percent of these gains across countries result from better Zhao (2021) to identify financial distortions and applies it to access to finance, with the remainder from optimizing the a sample of 24 European countries. The analyses reveal that debt-to-equity ratio. The paper also quantifies the link less developed economies face more severe financial misal- between financial misallocation and real-input allocative location. Distortions in the allocation of financial resources inefficiency, showing that reducing financial misallocation raise the relative cost of finance for younger, smaller, and from the median to the 25th percentile of the cross-indus- more productive firms. Counterfactual analysis indicates try distribution can increase aggregate productivity by an that alleviating financial distortions could boost aggregate average of 5.2 percent. The effect is larger, at 6.4 percent, productivity by approximately 30-70 percent. On average, for industries heavily reliant on external finance. This paper is a product of the Development Research Group, the Global Indicators Group, Development Economics and the Office of the Chief Economist, Europe and Central Asia Region. It is part of a larger effort by the World Bank to provide open access to its research and make a contribution to development policy discussions around the world. Policy Research Working Papers are also posted on the Web at http://www.worldbank.org/prwp. The authors may be contacted at rfattaljaef@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 The Role of Financial (Mis)allocation on Real (Mis)allocation: Firm-level Evidence for European Countries Ana P. Cusolitoa , Roberto N. Fattal-Jaefa,∗, Davide S. Marea,b , Akshat V. Singhc a World Bank, Washington, D.C., US b University of Edinburgh, Edinburgh, UK c University of Oxford, Oxford, UK JEL classification : D24; D61; O47 Keywords: total factor productivity, financial misallocation, economic growth, financial frictions ∗ Corresponding Author Email addresses: acusolito@worldbank.org (Ana P. Cusolito), rfattaljaef@worldbank.org (Roberto N. Fattal-Jaef), dmare@worldbank.org (Davide S. Mare), akshat.singh@economics.ox.ac.uk (Akshat V. Singh) 1. Introduction The study of financial market imperfections and their overall impact on economic development is a major focus in macroeconomics research. Despite the extensive literature on this topic, the question of how significant financial frictions are and their implications for overall income and productivity remains unanswered. This paper aims to fill this gap and offer a comprehensive characterization of the topic across a diverse set of countries. We employ the novel framework developed by Whited and Zhao (2021) to quantify the effect of financial distortions on firm and aggregate productivity. In doing so, we work with a sample of 24 European economies for which we have firm-level balance sheet information for the period 2010-2016. The objective of this paper is threefold. First, we aim to provide a comprehen- sive understanding of the differences in finance misallocation across countries and its impact on overall productivity, considering the varying effects on firms of different sizes, ages and physical productivity. Second, we intend to examine the relationship between the measured misallocation of financial liabilities, as determined by Whited and Zhao (2021), and the misallocation of physical inputs, as computed by Hsieh and Klenow (2009). Specifically, we aim to understand how much of the observed variation in the marginal revenue products of physical inputs can be attributed to variation in the marginal returns to debt and equity across firms within specific industries. Third, we quantify the overall effect on total factor productivity of removing such financial inefficiencies. Our first quantitative result indicates a strong negative correlation between finan- cial misallocation and economic development. The aggregate productivity gains from efficiently reallocating finance across firms are more than twice as high in countries with the lowest per capita income in the sample relative to the richest counterparts. At a more disaggregated level and exploiting a rich set of country, sector, and time- fixed effects, we identify a substantial discount in the shadow cost of finance for larger and older firms. A 10% increase in the size of firms measured by a firm’s total assets is associated with a 5.4% reduction in the shadow cost of finance, whereas a 1-year 2 gap in a firm’s age corresponds to a reduction in the order of 0.6%. While this result has been documented earlier in the literature in different contexts and methodologies, we arrive at this conclusion from the direct observation of the distribution of financial liabilities interpreted through the lens of an efficient theory of finance allocation. Our second contribution is to document a strong productivity dependence on the distribution of financial distortions. Estimating regression coefficients between the logarithm of distortions against the logarithm of firm-level productivity within each country, we find that the relative cost of financing is higher for less productive firms, resulting in a misallocation of financial resources away from its more productive uses. Moreover, we find this elasticity to be decreasing with economic development, a feature that is reassuring of the subdued misallocation expected of economies with more developed capital markets. The productivity dependence of financial distortion echoes the pattern in the real misallocation literature, where productivity-dependent real distortions are a pervasive feature of the data. Our third contribution concerns quantifying the relationship between financial mis- allocation and real-input allocative inefficiency. Using the setting developed in Rajan and Zingales (1998), we classify industries into high and low dependence on external sources of finance. We then regress the standard deviation of the marginal revenue products of real inputs (i.e., real inputs misallocation) within a country, industry, and year against the standard deviation of the marginal return to financial liabilities (i.e., financial misallocation) and its interaction with the external finance indicator. We find that a decrease in financial misallocation from the median to the 25th percentile of the cross-industry distribution induces an improvement in the allocation efficiency of real inputs that, in turn, increases aggregate productivity by 5.2% on average and by 6.4% in industries with higher dependence on external finance. The theoretical framework providing the benchmark of an efficient allocation of real and financial resources is drawn directly from Hsieh and Klenow (2009) and Whited and Zhao (2021). At the framework’s core is the notion that, unless frictions and distortions are in place, the maximization of aggregate output requires that real 3 resources and financial liabilities are distributed across firms to equalize the marginal returns to additional units of real and financial inputs.1 A more fundamental connec- tion between the two methodologies is the assumption that the mechanism through which financial misallocation generates losses in aggregate productivity is by causing a misallocation of real inputs. Despite this strong link, the quantitative strength of this mechanism is not obvious since firms may appeal to internal funding resources to circumvent financial frictions and mitigate their disruptive effect.2 Therefore, the quantitative effect of finance misallocation on the misallocation of real resources is ultimately an empirical question we address in this paper. Our empirical strategy to conduct such an assessment involves estimating the rela- tionship between two model-based summary measures of misallocation: the standard deviation of the logarithm of firms’ marginal revenue product of real resources rela- tive to the industry mean, derived using the Hsieh and Klenow (2009) methodology; and the standard deviation of the logarithm of firms’ marginal revenue products of financial resources relative to the industry mean, yielded by the application of the methodology in Whited and Zhao (2021). Besides controlling for time, industry, and country fixed effects, we strengthen the identification by classifying industries into high or low external finance dependence, based on the external finance dependence of the industry being above or below the median across U.S. sectors and estimating the differential impact of finance misallocation on the highly external-finance depen- dent industries. We find a statistically significant decline in real-input misallocation resulting from reducing the dispersion of marginal returns to financial resources from the median to the 25th percentile value. The regression described above allows us to predict the reduction in the real-input misallocation associated with a particular improvement in financial misallocation but 1 A set of assumptions must be adopted to ensure this is indeed the efficient prescription. Among the most important ones are the imposition of a common technology across firms within a narrow industry and the frictionless reallocation of resources. 2 This insight plays a key role in the quantitative papers in the macro-development literature assessing the aggregate effects from financial frictions, such as Buera et al. (2011), Midrigan and Xu (2014), and Moll (2014). We discuss the literature in greater detail below. 4 it is silent about the aggregate productivity gains associated with such reduction. To make this extra step, we leverage the richness of our data to estimate the effect of an industry’s real misallocation on the aggregate productivity gains associated with its reversal. Equipped with this estimate, we then find that reducing finance misalloca- tion from the median to the 25th percentile of the cross-industry distribution yields an aggregate productivity gain of 5.2%, accruing through an improvement in the al- locative efficiency of real inputs. The aggregate gains are 1.2 percentage points higher in industries with high external finance dependence. Interestingly, these results fall within the range of estimates reported in the macro-development literature, appealing to radically different quantification strategies. Several studies attempt to analyze the link between financial frictions and produc- tivity growth. Finance is needed to buy production inputs (capital, labor, research, and development) and affects firm entry and exit, all of which influence a firm’s productivity. The presence of financial frictions may generate misallocation of re- sources, lowering firm productivity (Buera et al., 2011; Midrigan and Xu, 2014; and Moll, 2014). Frictions can also generate misallocation across sectors because they increase barriers to entry in more productive sectors. Furthermore, capital misallo- cation can derive from credit constraints (Banerjee and Duflo, 2005) and inefficient contract monitoring and enforcement (Cole et al., 2016). These studies quantify the aggregate effect of financial frictions, pursuing an approach that, unlike us, dispenses from the direct observation of the distribution of financial liabilities. They postulate a specific financial friction, typically a collateral constraint, and calibrate it to match measures of aggregate credit to GDP or moments from the distribution of firm dy- namics across age and size. Despite the methodological differences, we find aggregate productivity gains from alleviating financial frictions within the same range identi- fied in these macro-development studies. We interpret the similarity of the results as giving reassurance to the validity of our empirical approach. The objective of our analysis is not to pinpoint the source of financial misallo- cation, but rather to elucidate how the sub-optimal allocation of financial liabilities 5 can lead to real-input allocative inefficiency. One critical channel through which this occurs is weaknesses in the banking sector. Banks with capital buffers only marginally above the regulatory minimum may engage in “zombie lending”, continuing to finance weak or insolvent borrowers. Evidence from Japan and European banks in the after- math of the global financial crisis shows that following the bursting of an asset price bubble and widespread losses in the financial sector, banks with lower capital buffers were more likely to provide frequent rounds of loan restructuring – also known as “evergreening” (Peek and Rosengren, 2005 and Giannetti and Simonov, 2013) and to reduce lending to weak borrowers significantly less than to stronger ones (Acharya ¨ et al., 2018; Acharya et al., 2019; Schivardi et al., 2022; and Ozlem Dursun-de Neef and Schandlbauer, 2021). When banks’ capital gets locked up in troubled sectors and companies, this may prevent some second-round business failures, but it also diverts funds away from more productive sectors of the economy. Inefficient firms could thus have a dominant impact on the functioning of input and output markets, generating lower economic output, investment, and employment (Caballero et al., 2008). Another important channel works through the structure and maturity of a firm liabilities. De- parting from the Modigliani and Miller (1958) framework, empirical studies point to a reduction on long-term productive investment due, for example, to excessive reliance on short-term debt (Vermoesen et al., 2013). Sudden credit rationing may also affect disproportionately firms relying on debt financing affecting productivity through a reduction in investment in innovation.3 We contribute to this literature by employ- ing an identification strategy that allows us to identify the effect of finance on real misallocation. We show that mitigating capital market imperfections can lead to an average aggregate productivity gain of 5.2%. Furthemore, these productivity gains are even more pronounced in industries that are heavily dependent on external finance (an additional 1.2 percentage point gain). The remainder of the paper is organized as follows. In section 2 we introduce and 3 For example, Granja and Moreira (2021) show that in the aftermath of the global financial crisis, disruption in the credit market caused a decrease in product innovation. 6 summarize the data, and present the theoretical framework underlying the identifica- tion of real and financial misallocation. In section 3 we discuss the aggregate results and then in section 4 we present the heterogeneous effects based on firm age and size. Section 5 we present and implement our empirical strategy to connect financial with real misallocation. In section 6 we conclude. 2. Empirical approach In this section, we summarize the model and the data. We also provide some stylized facts about our sample. 2.1. Model Assessing the misallocation of financial and real resources requires an explicit no- tion of efficiency against which to compare the observed distribution of these inputs. To this end, we follow closely Hsieh and Klenow (2009)’s methodology for the identi- fication of real-input distortions and Whited and Zhao (2021)’s methodology for the case of financial inputs. Given that both types of misallocation feature prominently in our analysis, we introduce the methodologies in parallel. We assume there is a single final good Y produced under perfect competition combining output from all industries, Ys , under a Cobb-Douglas technology: S S Y = Ysθs with θs = 1 (1) s=1 s=1 We consider each industry s to be populated by a large number of monopolistically competitive firms (Ms ). Each sector’s output Ys is a constant elasticity of substitution (CES) aggregate of differentiated varieties, given by: σ Ms σ −1 σ −1 Ys = Ysv σ (2) v =1 7 where Ysv is the quantity produced of variety v in sector s and σ is the elasticity of substitution. The differentiated varieties, in turn, are produced by combining physical capital and labor input in a Cobb-Douglas production function with sector-specific factor shares: − αs αs Ysi = Asi L1 si Ksi (3) The physical productivity of the firm i, also referred to as T F P Q, is denoted with Asi . Notice that the capital and labor factor shares are assumed to be industry-specific. Firms need to issue debt and raise equity to finance the acquisition of the physical capital, the labor input, and the series of expenses that go into the determination of its T F P Q. Rather than imposing a specific theory for why debt and equity are not perfectly substitute and for how the total amount of financing is distributed into its various applications, we postulate a direct mapping from financial liabilities into real value added that captures these unmodeled elements in reduced form. The imperfect substitutability between debt and equity is reflected in a constant elasticity of substitution specification, which we estimate from the data. The distribution of finance into capital, labor, and innovation is subsumed in a finance-based measure of productivity which we label Total Finance Benefit (T F B ), which we will also back-out from the data. Formally: γs γs −1 γs −1 γs −1 γs γs Fsi = Zsi αs Dsi + (1 − αs ) Esi (4) where Zsi denotes the Total Finance Benefit, γ is the industry-specific elasticity of substitution between debt and equity, and αs is the industry-specific weight of debt in real value added. Notice that, for the sake of differentiating notation with re- spect to the real-input representation, the real value added here is denoted with Fsi . 8 Empirically, however, we shall extract information about real output from the same observable in the data, the value added of the firm. In terms of the optimal determination of the capital and labor inputs, we assume these are chosen every period, taking the capital rental rate and the wage rate in factor markets as given. To capture frictions and policies in these markets, we intro- duce wedges that distort the aggregate scale of the firm and the relative price between capital and labor. These are the output wedge τysi and the capital wedge τksi . Im- portantly, the wedges are assumed to be idiosyncratic to the firm, capturing the idea that the frictions and policies may exert a heterogeneous impact on the firms’ input choices. Given the monopolistically-competitive behavior of the variety producers, each firm maximizes πsi = (1 − τY si )Psi Ysi − wLsi − (1 + τKsi )RKsi (5) s.t. −1 1 Psi = Ysi σ Ps Ysσ (6) where equation 6 is the demand for variety i in sector s and where the firm’s output is given by equation 3. Solving the optimization problem yields −1 Aσ si (1 − τYsi ) σ Lsi ∝ (7) (1 + τKsi )α s (σ − 1) Ksi αs w 1 = (8) Lsi (1 − αs ) R (1 + τksi ) Equations 7 and 8 shows the direction in which the wedges distort the decisions away from the efficient level. Under no distortions, firm size is determined by the firm’s T F P Q, Asi , and capital-labor ratios are equalized within industries. With distortions, both properties break down. An implication of the optimality conditions in the model is that the revenue pro- 9 ductivity of the firm, T F P R, represents a summary statistic of the mix of capital and output wedges. Through a simple rearrangement of terms, it can be shown that Psi Ysi revenue productivity, defined as T F P Rsi = L1 −αs αs , Ksi becomes proportional to the si ratio of distortions in the following fashion: (1 + τksi )α s T F P Rreal,si ∝ (9) (1 − τYsi ) This representation of T F P R turns out to be very useful for the characterization of the misallocation in an economy. Since, in the efficient allocation with no distortions, the T F P R must be equalized across firms, any dispersion in revenue productivity is a sign of misallocation. Furthermore, the level of a given firm’s T F P R reveals information on the direction in which the distortions are affecting the firm relative to the average in its industry. A high T F P R is indicative of an inefficiently low level of labor and capital flowing to the firm, whereas the opposite is true if T F P R is lower than the average. One can arrive at an equivalent characterization of the optimal levels of debt and equity in a firm as a function of prices and distortions in capital markets. In this case, the profit maximization problem confronted by the firm is: πsi = Psi Fsi − r(1 + τDsi )Dsi − (1 + τEsi )λEsi (10) s.t. −1 1 Psi = Fsi σ Ps Ysσ (11) where r and λ are the prices of debt and equity, and τDsi and τEsi are the distortions in each market. Static optimization yields the following optimality conditions: 10 (σ − 1) Psi Fsi αs 1 (γs −1) = r(1 + τDsi ) (12) σ γs γs αs Dsi + (1 − αs )Dsi Esi (σ − 1) Psi Fsi αs (γs −1) 1 = λ(1 + τEsi ) (13) σ γs γs (1 − αs )Esi + αs Dsi Esi As it was the case when choosing real inputs, profit maximization requires that the marginal revenue products of equity and debt are equalized to their marginal costs. Under no distortions, our assumption of price-taking in capital markets would require that these marginal returns are equalized across firms. Therefore, any dispersion in marginal returns would once again constitute evidence of misallocation, in this case of financial liabilities The CES structure of finance-based real value added precludes a transparent char- acterization of T F P R as a function of distortions as it was possible for real inputs. For this reason, we define the finance-based marginal returns (T F P Rf in ) as the following weighted average of the debt and equity distortions: Dsi Esi T F P Rf in,si = (1 + τDsi ) + (1 + τEsi ) (14) Dsi + Esi Dsi + Esi In the empirical and quantitative analysis that follows, we characterize the loga- T F P Rreal,si rithm of the demeaned values of the firm-level marginal returns, log T F P Rreal,s and T F P Rf in,si log T F P Rf in,s . By demeaning the marginal returns with the industry average we are acknowledging that the only type of misallocation we are capturing is within a sector, being silent about any misallocation of real and financial inputs across indus- tries. Lastly, following the literature, we conduct the within-sector average at the lowest level of aggregation allowed for by the data. The last piece in the characterization of the equilibrium that feeds directly into the empirical analysis relates to the aggregation of firm-level outcomes. Since we are ulti- mately interested in the aggregate productivity gains that are reaped from efficiently 11 reallocating real and financial inputs from the observed to the undistorted allocation, we must therefore characterize the aggregate output of an industry under no distor- tions. The real-input-based aggregation in the undistorted economy is simplified by the Cobb-Douglas nature of the production function and boils down to the following expression: 1 Ms σ −1 σ −1 T F P Rreal,s T F Preal,s = Asi (15) i=1 T F P Rreal,si where Asi , the firm-level T F P Q, can be backed out from the observation of the firm’s value-added, the real inputs, and the CES structure for the demand system as: σ (Psi Ysi ) σ−1 Asi ∝ −αs αs (16) (wL)1 si Ksi Notice that the aggregate productivity under the efficient allocation can be easily computed from equation 15 recalling that in such allocation, T F P R is equalized across firms, so that equation 15 becomes: 1 Ms σ −1 T F P real,s = (Asi )σ−1 i=1 The aggregate productivity gain from reversing all the real-input misallocation in a given industry, then, is given by the ratio of the observed and the efficient aggregate productivity. The CES structure of the function mapping the financial liabilities into real value added does not allow for a simple characterization of the aggregate total benefit as a function of demeaned marginal returns. Therefore, we must construct it for the undistorted and the observed allocations separately. Solving a benevolent social plan- ner’s problem of maximizing aggregate real value added subject to a given aggregate 12 amount of debt and equity in the industry yields the following solution to the optimal debt and equity allocations: σ −1 Zsi Dsi = D σ −1 s (17) ΣM i=1 Zsi s σ −1 Zsi Esi = Ms σ−1 Es (18) Σi=1 Zsi where Ds and Es stand for the aggregate debt and equity holdings allocated to industry s. Both expressions show the well-established result that, in an undistorted allocation, the more productive firms are assigned higher amounts of debt and equity, limited by the degree of substitutability between product varieties in the industry. Given the definitions of efficient debt and equity holdings, the efficient real value added at the firm level is given by: γs γs −1 γs −1 γs −1 γs γs Fsi = Zsi αs Dsi + (1 − αs ) Esi (19) which can be obtained by plugging in the efficient debt and equity levels derived in equations 17 and 18 and appealing to the finance-based measure of the firm’s T F P Q, Zsi , which is given by: σ (Psi Ysi ) σ−1 Zsi ∝ γs (20) γs −1 γs −1 γs −1 γs γs αs Dsi + (1 − αs ) Esi The aggregate real value added of an industry under the efficient allocation is simply σ σ −1 σ −1 Ms Ys = i=1 Fsi σ . As was the case for the real-input allocative gains, in the quantitative analysis we shall focus on the aggregate gains from resolving financial misallocation as given by the ratio between the aggregate real value added in the undistorted economy and the one observed in the data. 13 2.2. Data and stylized facts We use firm-level information from Bureau van Dijk’s Orbis global database, the largest cross-country database containing information on firms’ financial statements, production activity, and firm ownership (Kalemli-Ozcan et al., 2019). We consider all industrial sectors, including financial companies and services (see table A.1 in Appendix A). We observe European firms over a full economic cycle between 2010 and 2016. To clean up the data for our intended purposes, we follow Cusolito and Didier (2020) and select firms for which we have information on the key variables in our study, namely total sales, compensation of employees, interest payments, taxes, paid-in shareholder equity, total debt and date of establishment. Our final sample comprises information on approximately 7.8 million observations across 24 European countries. Table 1 reports the number of countries and firms in our sample. Of the 24 countries, according to the 2021 World Bank classification of countries into income groups, 18 are high-income countries and 6 are upper-middle-income countries. Italy is the country with the largest number of firms (25% of the sample) and Luxembourg the lowest (0.04% of the sample). On average, more than 60% of total assets are financed through debt. Liabilities to total assets range from 41% (Ukraine) to 74% (Italy). The ratio of value-added (VA) to total assets is on average around 0.9, varying between a minimum of 43% (Bosnia and Herzegovina) to a maximum of 143% (Bulgaria). Table 2 summarizes the cost of financing through debt of firms classified according to their size. In all years, smaller firms face a higher cost of financing through debt. Moreover, not only does the average cost of finance appear to be higher for smaller firms, but also there is a higher dispersion, suggesting potentially greater misalloca- tion. Table 3 shows similar patterns but with some distinctions. The cost of equity financing is still greater for smaller firms but variability within size clusters appears to be elevated also for larger firms. Tables 4 and 5 summarize the cost of debt and equity by age group. We do not observe a clear pattern in the cost of funding among age groups. In general, firms 14 Table 1: Summary Statistics Country name Firms Liabilities to Assets VA to Assets Austria 10,014 0.62 0.95 Belgium 72,547 0.60 0.75 Bosnia and Herzegovina 28,352 0.51 0.43 Bulgaria 351,913 0.44 1.43 Croatia 182,335 0.59 0.70 Czech Republic 364,312 0.50 0.82 Estonia 92,486 0.44 1.03 Finland 131,778 0.58 1.36 France 831,697 0.61 1.21 Germany 116,200 0.63 1.09 Hungary 62,543 0.53 0.88 Italy 1,935,265 0.74 0.79 Luxembourg 2,806 0.58 0.86 Montenegro 7,500 0.48 0.66 North Macedonia 67,900 0.42 1.33 Norway 86,216 0.64 1.38 Poland 50,766 0.50 0.90 Portugal 551,947 0.61 0.71 Romania 627,369 0.55 0.91 Serbia 149,100 0.54 0.80 Slovak Republic 232,676 0.59 0.95 Slovenia 144,325 0.54 1.11 Spain 1,579,529 0.58 0.76 Ukraine 78,658 0.41 1.32 Total 7,758,234 0.61 0.91 older than 10 years appear to have the lowest mean cost but not necessarily the lowest variability in the cost of funding. For the other age groups, there is not a monotonic relationship among the groups, both in terms of the mean cost of funding and the within-group variability of the cost of debt or equity. 3. Aggregate Results Equipped with a theoretical framework and a suitable global database, we begin the presentation of the aggregate results obtained using our empirical framework. At the aggregate level, the headline object of interest is the aggregate productivity gain that would be obtained from the reversal of the misallocation of financial liabilities. Following Whited and Zhao (2021), the thought experiment underlying this result is the computation of the aggregate T F P under the hypothetical efficient allocation 15 Table 2: Cost of Debt (by Firm Size) 0-5 5-15 15-30 30-50 50-70 70-85 85-95 95-100 Mean S.D. Mean S.D. Mean S.D. Mean S.D. Mean S.D. Mean S.D. Mean S.D. Mean S.D. 2010 1.20 3.25 0.78 1.35 0.61 2.29 0.49 0.90 0.41 0.81 0.35 0.74 0.30 0.38 0.23 0.32 2011 1.27 15.97 0.80 2.67 0.61 1.41 0.49 1.51 0.40 1.25 0.34 0.64 0.29 0.47 0.23 0.29 2012 1.13 4.72 0.79 5.51 0.60 2.14 0.47 1.10 0.39 0.77 0.34 0.90 0.29 0.46 0.24 0.36 2013 1.19 9.84 0.79 2.73 0.61 1.98 0.49 2.79 0.40 1.31 0.35 0.82 0.31 0.97 0.26 2.54 2014 1.17 7.28 0.80 3.12 0.61 1.41 0.49 1.22 0.41 1.12 0.36 0.71 0.31 0.59 0.25 0.34 2015 1.18 3.18 0.83 1.73 0.67 4.26 0.54 2.02 0.45 2.56 0.37 0.72 0.35 8.56 0.26 0.64 2016 1.40 13.41 0.94 13.83 0.69 2.69 0.55 1.92 0.44 0.96 0.38 0.66 0.34 1.84 0.27 0.48 Note: This table reports the mean and standard deviation in each year of the cost of debt of firms classified according to their size, from the smallest size (up to the 5th percentile) to the largest size (95th percentile to 100th percentile). Table 3: Cost of Equity (by Firm Size) 0-5 5-15 15-30 30-50 50-70 70-85 85-95 95-100 Mean S.D. Mean S.D. Mean S.D. Mean S.D. Mean S.D. Mean S.D. Mean S.D. Mean S.D. 2010 1.28 3.24 0.98 2.86 0.84 6.93 0.69 4.17 0.57 2.54 0.54 19.58 0.42 4.04 0.41 10.68 2011 1.25 3.94 1.00 3.38 0.87 3.76 0.76 5.52 0.68 14.81 0.53 3.58 0.43 2.31 0.44 10.85 2012 1.31 5.21 1.03 3.34 0.90 5.95 0.76 4.48 0.71 23.97 0.54 4.34 0.44 2.21 0.37 5.34 2013 1.30 3.54 1.11 6.51 0.91 6.97 0.76 3.07 0.67 14.36 0.53 2.38 0.44 2.04 0.40 5.07 2014 1.40 7.40 1.10 4.94 0.91 2.93 0.78 5.02 0.67 6.43 0.55 3.27 0.46 2.71 0.42 7.92 2015 1.46 4.28 1.16 4.88 1.01 12.26 0.83 9.80 0.66 3.09 0.64 36.06 0.67 55.72 0.40 7.41 2016 1.43 3.22 1.17 4.53 0.96 5.03 0.80 4.13 0.66 3.96 0.56 5.59 0.44 2.44 0.37 2.15 Note: This table reports the mean and standard deviation in each year of the cost of equity of firms classified according to their size, from the smallest size (up to the 5th percentile) to the largest size (95th percentile to 100th percentile). Table 4: Cost of Debt (by Firm Age) 0-2 2-5 5-10 10+ Mean S.D. Mean S.D. Mean S.D. Mean S.D. 2010 0.44 0.59 0.53 0.68 0.58 1.29 0.48 1.38 2011 0.50 0.79 0.71 23.44 0.59 3.23 0.48 1.50 2012 0.60 2.60 0.59 9.06 0.60 2.90 0.46 1.13 2013 0.60 1.23 0.61 1.61 0.63 5.56 0.47 1.85 2014 0.58 1.85 0.63 2.28 0.60 3.05 0.47 1.91 2015 0.55 1.23 0.68 1.62 0.62 3.44 0.50 3.97 2016 0.73 1.68 0.69 8.07 0.62 2.45 0.50 5.11 Note: This table reports the mean and standard deviation in each year of the cost of debt of firms classified according their age, from the youngest (up to 2 years) to the oldest firms (greater than 10 years). 16 Table 5: Cost of Equity (by Firm Age) 0-2 2-5 5-10 10+ Mean S.D. Mean S.D. Mean S.D. Mean S.D. 2010 0.99 3.04 1.24 3.69 1.08 5.51 0.59 9.44 2011 1.52 10.13 1.41 5.84 1.08 5.62 0.62 8.40 2012 1.48 7.95 1.62 46.17 1.11 6.77 0.60 7.97 2013 1.50 5.61 1.51 12.13 1.03 4.40 0.60 7.81 2014 1.54 4.02 1.49 10.40 0.98 4.69 0.59 3.78 2015 1.46 9.45 1.39 5.22 0.99 13.33 0.65 27.88 2016 1.70 8.72 1.25 5.59 0.92 5.74 0.58 3.18 Note: This table reports the mean and standard deviation in each year of the cost of equity of firms classified according their age, from the youngest (up to 2 years) to the oldest (greater than 10 years) firms. and compare it with the T F P under the observed distribution of debt and equity. In Figure 1, we plot the aggregate gains against the log of GDP per-capita, which we take as proxy for the degree of financial development across countries. Figure 1 shows a strong negative relationship between finance misallocation and economic development. At the bottom of the income distribution, the conditional T F P gains of efficiently reallocating financial liabilities (dotted line in 1) reach ap- proximately 75%, almost doubling the conditional gains to be reaped by the richer countries. As a benchmark, Whited and Zhao (2021) find gains in the order of 11% to 12% in the U.S. and 70% to 80% in China. These numbers reinforce the plausibility of our findings and imply that most of the countries in our sample fall between the polar cases of China and that of the U.S. The correlation between firms’ productivity and financial distortions has an im- portant bearing on the aggregate effects of financial distortions. Are financial frictions relatively more stringent for productive firms? There are various mechanisms through which distortions in credit markets may favor less productive firms at the expense of more productive ones. Government policy often subsidizes credit to micro and small enterprises, which, in turn, are relatively less productive than their larger counter- parts. Moreover, capable yet young entrepreneurs may also confront higher costs of credit due to limited commitment and informational asymmetries. Such policy and 17 Figure 1: Productivity Gains from Reversing Finance Misallocation Note: The figure shows the counterfactual aggregate T F P gain that each country would enjoy if finance misallocation was reversed. The gains are computed based on the methodology described in section 2.1. The GDP per capita is based on the Penn World Tables Database. market-driven frictions also apply to the allocation of physical inputs, where the real misallocation literature established a strong productivity dependence on output and factor market distortions. Accounting for such features of the data proved important for helping rationalize cross-country differences in firm-level outcomes, such as the size distribution and the life cycle dynamics of firms.4 Following the procedure applied in the real misallocation literature, we assess the productivity dependence of financial distortions by estimating the elasticity between financial distortions and firms’ physical productivity. Defined in equation 20, the T F P Q of the firm measures its efficiency in turning a given amount of debt and equity into value-added. The productivity elasticity of financial distortions, then, can be backed out by estimating the following regression: log (T F P Rf ini ,c,s,t /T F P Rf ins ,t ) = βc,t log (Zi,c,s,t /Zs,t ) + i,c,s,t (21) where log (T F P Rf ini ,c,s,t /T F P Rf ins ,t ) denotes a firm’s idiosyncratic financial cost in 4 See, for instance, Hsieh and Klenow (2014) and Fattal-Jaef (2022). 18 a given period and country relative to its industry’s average cost. log (Zi,c,s,t /Zs,t ) stands for the firm’s physical productivity relative to its industry’s average. A positive sign of the regression coefficient would capture the situation where financial frictions misallocate resources away from firms with above-average physical productivity to firms with below-average physical productivity. We estimate this regression for each country and year separately and report the average of regression coefficients of each country against the log of real GDP per capita in Figure 2. Figure 2: Productivity Dependence of Financial Distortions Note: The figure shows the elasticity between the log of T F P Rsi and log of Asi of each country and the log of real GDP per capita. Each dot represents each country’s average regression coefficient resulting from estimating equation 21. The dotted line provides a general tendency in the association between financial distortions and the log real GDP per capita. Two salient features emerge from Figure 2. Firstly, all countries exhibit a positive elasticity between financial distortions. Secondly, such elasticity is lower in more developed economies, suggesting these countries’ more developed financial systems allow for a more efficient allocation of financial resources. We now focus on assessing the degree of correlation between financial and real measures of misallocation. The theory posits that productivity losses from finance misallocation operate from an induced misallocation of physical capital, labor, and 19 material. Therefore, assessing the extent to which this mechanism is indeed taking place in the data is of great value for the validity of the methodology. Our multi- country database enables us to perform such an assessment. To this end, then, we apply the real misallocation methodology based on Hsieh and Klenow (2009) to our data and investigate the relationship between the productivity gains accruing from the reversal of the real misallocation vis-a-vis the gains resulting from undoing financial misallocation. At this stage, we present a simple correlation between the two measures and later pursue a richer empirical strategy controlling for other confounding factors. The relationship between the two outcomes is plotted in Figure 3. It shows that the aggregate gains from reversing financial and real misallocation are strongly correlated, offering some preliminary support to the mechanisms of the finance misallocation theory. Figure 3: Finance and Real Misallocation Note: The figure shows the counterfactual aggregate T F P gain that each country would enjoy if finance misallocation was reversed (X-axis, Reallocation Gains F inance computed following Whited and Zhao (2021)), against the counterfactual productivity gain that would be obtained from reversing real-input misallocation (Y axis, Reallocation Gains Real, computed following Hsieh and Klenow (2009)). To provide an intuition of what may cause financial misallocation, we plot the country’s measure of financial misallocation against a measure of financial develop- ment (Figure 4). Figure 4 shows that when financial markets and institutions are 20 more developed, reallocation gains from lowering financial frictions are lower. This suggestive relationship may provide a mechanism through which financial misalloca- tion relates to real misallocation, in line with findings in the literature that frictions associated with lower financial development can distort the entry decisions of high- growth potential firms (Aghion et al., 2007, and Ardagna and Lusardi, 2010). Figure 4: Finance Misallocation and Financial Development Note: The figure plots finance misallocation (Y axis, Reallocation Gains F inance, computed fol- lowing Whited and Zhao (2021)), against a measure of financial development (X axis, F inancial Development as computed in Sahay et al. (2015)) . To close this section, we provide a decomposition of the contribution of the ag- gregate level vis-a-vis the composition of the distribution of financial liabilities across firms in explaining the overall degree of finance misallocation. We do this by compar- ing the baseline results under our estimates of the elasticity of substitution between debt and equity across sectors against those emerging from the imposition of perfect substitution between financial liabilities. Since, in the latter case, the ratio of debt and equity is irrelevant for real value added, the extent to which this specification captures the productivity gains from the baseline represents a measure of the contribution of the levels of finance in reaping these gains. Figure 5 shows that aggregate productivity gains accrue mostly from achieving the efficient level of finance across firms rather than from attaining the efficient com- position of liabilities. Keeping debt-to-equity ratios fixed, reallocating funds from 21 Figure 5: The Role of Levels of Finance versus Composition Note: The figure shows the counterfactual aggregate T F P gain that each country would enjoy if finance misallocation was reversed under the baseline estimation of the elasticity of substitution between debt and equity and under the alternative scenario of perfect substitutability. The gains under perfect substitution are depicted in green, and the difference between the baseline and the perfect substitution, which measure the contribution of the debt-to-equity ratios, is depicted in orange. the low to the high marginal return firms would reap in all countries more than 60% of the total gains in aggregate productivity. This share is remarkably robust across our sample of countries and is only slightly below the 79% to 83% share reported by Whited and Zhao (2021) for China. In short, this section reviewed some of the most salient findings derived from a novel methodology to measure finance misallocation in a broad cross-section of coun- tries. The main conclusion is that the methodology delivers a systematic relationship between the level of economic development and the degree of finance misallocation. We also find suggestive evidence that, as presumed by the methodology, there could be a causal relationship between financial and real-input misallocation, a conjecture that we shall revisit under a richer empirical strategy in the coming sections. Lastly, our examination of the contribution of the levels vis-a-vis the composition of finance in a cross-country setting validates the finding that is documented for China: it is the level rather than the composition that accounts for the largest share of the measured 22 financial misallocation. 4. Heterogeneous Effects: The Role of Age and Size A long-lasting hypothesis in the macro-finance literature is that the size and the age of the firm constitute important determinants of a firm’s access to finance (Beck et al., 2008). Larger firms are better able to pledge collateral in contexts where the relevant financial frictions is a limited commitment problem. Similarly, moral hazard may be attenuated when the borrower is a large firm. A similar reasoning applies to older firms where the accumulation of capital over time and the availability of information reduce financial frictions. In this section, we leverage the richness of the data to explore these hypotheses. We begin by exploring the differences in the average cost of finance across firms of different sizes. Within an industry in a particular country, we classify firms into small and large based on their asset holdings being above or below the median in the sector-country. Then, we take the model-implied average cost of finance among small and large firms and average it across countries to construct a global measure of the average cost of finance by firm size within each sector. The results are plotted in Figure 6. Figure 6 reveals a clear pattern: in all industries, the average cost of finance is higher for smaller than for larger firms. A higher cost of finance in our model is reflective of a high marginal return to additional units of finance and is thus suggestive of a scarcity of funds relative to the optimal level. Therefore, our results dictate that in response to a hypothetical reform that liberalizes capital and credit markets, we should expect finance to flow from the larger to the smaller firms. Indeed, this prediction is in accordance with the findings in the recent literature studying the firm-level response to financial liberalization episodes,5 although in these papers the inference is based on the observation of the reallocation of real inputs. 5 See, for instance, the work of Larrain and Stumpner (2017), and Bau and Matray (2020). 23 Figure 6: Heterogeneous Costs of Finance: The Role of Size Note: The figure shows the model-based average cost of finance among small (less than the median size of the relevant sector in each country) and large (more than the median size of the relevant sector in each country) firms within each sector classified according to NACE 1. We turn now to exploring differences in the average cost of finance across firms of different ages. We classify firms within an industry into young and old based on them being less than or greater than five years old (if exactly five years old, the firm is considered young). We then compute the average cost of finance across young and old firms within an industry in a country and the average across countries. Figure 7 also shows a common pattern of heterogeneity across firms: young firms are systematically confronted with higher shadow costs of finance than older firms. As in the case of size, the model-based nature of our measure of the cost of finance is reflective of a higher marginal return to finance among younger firms, and hence funding would be reallocated to these firms in the event of a removal of financial frictions. The same literature tracking real-input reallocation in response to capital market liberalizations also confirms this prediction. While figures 6 and 7 are suggestive of a negative relationship between firm size, firm age, and the average cost of finance, we validate these relationships and assess 24 Figure 7: Heterogeneous Costs of Finance: The Role of Age Note: The figure shows the model-based average cost of finance among old (more than five years) and young (less of or equal to five years) firms within each sector classified according to NACE 1. their quantitative magnitude in a controlled regression setting. To this end, then, we estimate the following equation: log (T F P Ri,c,s,t /T F P Rs,t ) = β1 log (Assetsi,c,s,t ) + β2 Agei,c,s,t +β3 log (T F P Qi,c,s,t /T F P Qs,t ) + αi + αt + αc ∗ αt + αs ∗ αt + i,s,t (22) where subscripts i, c, t, and s denote the firm, country, year, and industry, respec- tively; log (T F P Ri,c,s,t /T F P Rs,t ) denotes the model-based measure of a firm’s cost of finance relative to the average cost in the corresponding industry in a country, and log (T F P Qi,c,s,t /T F P Qs,t ) stands for the firm’s financed-based T F P Q relative to the average in the industry and coountry, as defined in equation 20. We use a diverse set of fixed effects to control for both firm-specific and systematic factors that may affect a firm’s cost of finance. In particular, we intend to capture the effect of firm-specific risk premiums on the firms’ average cost of finance, a factor that would manifest as a 25 financial distortion in the model though it is a feature that is also part of the social planner’s assessment of the economic environment (David et al., 2020). In addition to isolating the effect of a firm’s age and size on its average cost of finance from other factors, the regression specifications also seek to assess pro- ductivity dependence in the distribution of financial wedges. This is motivated by the real misallocation literature, where it is a well-established feature of resource misallocation that distortions tend to tax productive firms more heavily than less productive ones. As shown in the literature, idiosyncratic distortions carry larger aggregate effects when increasing with the firms’ physical productivity.6 By including log (T F P Qi,s,t /T F P Qs,t ), in contemporaneous or lagged fashion, we seek to establish if such a relationship manifests as well in the context of financial distortions. Table 6: Cost of Finance: The Role of Firm Age and Size (1) (2) (3) (4) Age of the firm 0.0114∗∗∗ 0.0097∗∗∗ -0.0057∗∗ -0.0242∗∗∗ (0.0033) (0.0033) (0.0027) (0.0083) lnasset -0.5315∗∗∗ -0.5384∗∗∗ -0.5460∗∗∗ (0.0002) (0.0002) (0.0002) log TFPQ 0.3970∗∗∗ 0.3995∗∗∗ 0.4160∗∗∗ (0.0001) (0.0001) (0.0001) lnassett−1 -0.1715∗∗∗ (0.0009) log TFPQt−1 0.0680∗∗∗ (0.0004) Observations 7758234 7758234 7758234 5263441 Time fixed effects Y N N N Firm fixed effects Y Y Y Y Industry-time fixed effects N Y Y Y Country-time fixed effects N N Y Y Note: Standard errors in parentheses. This table presents the results obtained using equation 22. ∗ p < 0.1, ∗∗ p < 0.05, ∗∗∗ p < 0.01. 6 See, for instance, Hsieh and Klenow (2014) and Bento and Restuccia (2017). 26 Table 6 confirms that older and larger firms confront a relatively lower cost of fi- nancing relative to their younger and smaller peers in the same industry and country. Recall that the cost of finance is measured according to the theory and reflects the cost of acquiring an extra unit of debt and equity that rationalizes marginal returns exhibited by the firm in the data. Using the most conservative estimates reported in Column 3 Table 6, a 10% increase in the size of a firm is associated with a 5.4% reduc- tion in the shadow cost of finance, whereas a 1-year increase in firm age corresponds to a reduction in the order of 0.6%. Column 4 shows that these results are robust to using lagged measures of firm size and TFPQ. The productivity dependence of idiosyncratic financial distortions shows promi- nently in table 6. Comparable to estimates in the real misallocation literature, fi- nancial frictions are significantly and strongly increasing in the underlying level of physical productivity of the firm. Even when projecting distortions on lagged values of physical productivity and controlling for a wider range of firm and other types of fixed effects, the positive relationship between productivity and financial distortions is weakened but remains statistically significant. To summarize, in this section, we provided empirical validation for the widely be- lieved and convincingly proven idea that firm age and size are important determinants of the ease of access to financing opportunities by firms. We investigated this hypoth- esis in the context of a novel inference strategy of financial misallocation that relies on the direct observation of the distribution of financial liabilities across firms, applied to a rich firm-level database with rich cross-country and time-series coverage. We found that larger and older firms do face significantly lower costs of external financing. We also show evidence that the most productive firms are more financially distorted than they should optimally be. 5. Establishing the Link between Financial and Real (Mis)allocation We now turn to address the second contribution of the paper, namely, answering the question we postulated in the introduction: how much of the observed misalloca- 27 tion of real inputs can be attributed to a misallocation of finance? As stated earlier in the text, the core of the measurement of finance misallocation in our paper lies in the assumption that disruptions in the allocation of finance across firms cause a mis- allocation of real inputs, which in turn translates into productive inefficiency. While the aggregate evidence presented in section 3 is suggestive of such a relationship being at play, in this section we propose a richer empirical strategy aimed at identifying a causal link between finance and real misallocation. Our strategy involves estimating the relationship between two model-based sum- mary measures of misallocation: the standard deviation of the logarithm of the de- meaned marginal revenue product of real inputs within an industry, yielded by the application of Hsieh and Klenow (2009)’s methodology, and the standard deviation of the logarithm of the demeaned marginal revenue products of financial resources, yielded by the application of the methodology in Whited and Zhao (2021). Finan- cial market imperfections may affect industries in a different fashion, hence the first channel to be explored is whether industries with higher measured finance misallo- cation are associated with a higher real misallocation, controlling for country, year, and industry fixed effects. Moreover, to get one step closer to identifying a causal relationship, we follow Rajan and Zingales (1998) in splitting industries into high or low external finance dependence, based on a classification applied to industries in the United States. Equipped with this taxonomy, we then also consider the interaction between the high/low external finance indicator and the finance misallocation of an industry as another explanatory variable for the degree of real misallocation. The idea here is that industries that rely more heavily on external borrowing should exhibit a higher degree of real misallocation, for any given degree of finance misallocation. Formally, our estimating equation is given by: sd(log T F P R real)s,c,t = αs + γc + τt + β1 sd(log T F P R f in)s,c,t + β2 sd(log T F P R f in)s,c,t × 1{EF Ds,U SA } + s,c,t (23) 28 where s denotes industry, c denotes country, t denotes time, EF D denotes external finance dependence of an industry, and sd(log T F P R)s,c,t is the standard deviation of log(TFPR/T F P R) within an industry-country. real and f in denotes real and fi- nancial distortions, respectively. To attenuate concerns about reverse causality and endogeneity, we consider specifications of the estimating equation where the explana- tory variable of interest, the standard deviation of the log of financial distortions, is lagged to the previous period’s value. The external finance dependence within an industry is computed as an average of the following firm-level indicator: (T otalCapitalExpenditureis − T otalCashF lowis ) EF Dis = (24) T otalCapitalExpenditureis That is, following Rajan and Zingales (1998), we define the external finance depen- dence of the firm i in sector s as the fraction of its capital expenditures that could not be financed with internal sources of funding. The total cash flow, in turn, is computed as an aggregate of total funds from operations, plus increases in accounts payable, de- creases in inventories, and decreases in receivables. Importantly, these cash flows are independent of the internal equity of the firm, which we treat as a source of external finance in the measurement of finance misallocation. The sector-wide external finance is the average across firms in the industry. As said, for identification, we apply this indicator to the U.S. economy based on Compustat data. Finally, we classify an in- dustry as highly dependent on external finance if its indicator is above the median across all industries. Table 7 presents the results from estimating equation 23. We report a positive and statistically significant coefficient for the effect of finance-based on real-based T F P R dispersion. Moreover, we also document a positive and statistically significant coefficient on the interaction between financial misallocation and the high external finance dependence indicator. In industries with a higher reliance on external credit, 29 Table 7: The link between financial and real misallocation (1) (2) sd(log TFPR fin) 0.2737∗∗∗ (0.0101) sd(log TFPR fin) × 1{EF Ds,U SA } 0.0336∗∗ (0.0134) sd(log TFPR fin)t−1 0.1674∗∗∗ (0.0101) sd(log TFPR fin)t−1 × 1{EF Ds,U SA } 0.0405∗∗∗ (0.0138) Observations 50,064 42,100 Country fixed effects Y Y Time fixed effects Y Y Industry fixed effects Y Y Note: Standard errors in parentheses. This table presents the results obtained using equation 23. ∗ p < 0.1, ∗∗ p < 0.05, ∗∗∗ p < 0.01 the misallocation of real resources that are created by the misallocation of finance is about 4 percentage points higher than the average. Our goal next is to provide an economic interpretation of the regression coefficients reported in table 7. What are the T F P consequences of the real misallocation that results from financial frictions? The question cannot be answered directly from the estimates in the table. The missing piece is a conversion factor that translates changes in the standard deviation of real marginal returns into changes in aggregate produc- tivity. Such a conversion factor can be gauged by running the following regression: T F P gains,c,t = αs + γc + τt + θSDT F P Rreal ,s,c,t + s,c,t (25) which projects a given industry-country’s T F P gain from reversing misallocation, T F P gains,c,t , against the given industry-country’s degree of real-input misalloca- tion as captured by the standard deviation of the marginal return to real factors, SDT F P Rreal ,s,c,t . 30 Table 8: T F P Effects of Dispersion in Marginal Returns to Real Inputs 100*{(YEf f ective /YActual )-1} sd(log TFPR real) 192.7697∗∗∗ (9.5454) Observations 60,815 Country fixed effects Y Time fixed effects Y Industry fixed effects Y Note: Standard errors in parentheses. ∗ p < 0.1, ∗∗ p < 0.05, ∗∗∗ p < 0.01 As expected, table 8 reports a strong positive coefficient for the relationship be- tween dispersion in marginal returns and counterfactual gains in aggregate produc- tivity. For instance, reducing SDT F P Rreal from its median value (0.559) to the 25th percentile (0.399) would imply an aggregate productivity gain of 31%.7 We are now ready to answer the question of what is the economic significance of the effect of finance misallocation on real misallocation.8 ˆ1 × ∆sd(log T F P R f in)s,c,t + ∆T F P gains,c,t = [β ˆ2 × ∆sd(log T F P R f in)s,c,t × 1{EF Ds,U SA }] × θ β ˆ (26) Consider a reduction in the dispersion of marginal returns to financial inputs from the median to the 25th percentile of the cross-industry distribution of standard de- viations, i.e. a reduction of 0.160.9 According to our estimates in table 7, such a decline reduces the standard deviation of marginal returns to real inputs by 0.027 in the average industry, and by 0.033 in industries with high external finance depen- 7 The aggregate productivity gain is computed as follows: (0.559-0.399)*192.7697 = 30.84%. 8 As a reminder, every time we refer to financial misallocation, we mean the within-industry- country dispersion in the log of marginal returns to financial liabilities, demeaned by the industry- country average, i.e. log(T F P Rf in /T F P Rf in ). Likewise, with real-input misallocation we mean the dispersion in log(T F P Rreal /T F P Rreal ). 9 The reduction in the dispersion of marginal returns is computed as follows: 0.559-0.399 = 0.160. 31 dence.10 Then, feeding these effects on the coefficient mapping real misallocation to aggregate productivity (table 8), we find that the proposed improvement in finance misallocation ends up increasing aggregate productivity by 5.2% on average and by 6.4% in industries with high external finance dependence.11 To put the quantitative implications of our identification strategy in perspective, we contrast our results with those found in the macro-development literature. In Buera et al. (2011), Midrigan and Xu (2014), and Moll (2014), salient examples in the literature, the quantification is based on the postulation of specific financial friction (a collateral constraint) in the context of a general equilibrium model of firm dynamics and the quantification of the productivity gains resulting from the alleviation of the friction. Our results fall in the ballpark of their findings. Midrigan and Xu (2014), for instance, document productivity losses in the order of 5% to 10% arising from the misallocation of labor and capital across firms within a sector, including the 5.2% that we estimate in our study.12 We interpret the proximity of our results to those emerging from alternative quantification strategies as providing reassurance to the validity of our empirical specification. In short, in this section we constructed a bridge between the finance misalloca- tion methodology proposed in Whited and Zhao (2021) and the real misallocation counterpart pioneered in Hsieh and Klenow (2009). We proposed an identification strategy to tease out the association of finance with real misallocation and ultimately aggregate productivity. We found that a plausible improvement in the degree of cap- 10 The reduction in the standard deviation of marginal returns to real inputs in the average in- dustry is computed using equation 26 as follows: 0.160*0.1674 = 0.0268. That is the computed change in the finance misallocation measure (0.160) multiplied by the estimated β ˆ1 from equation 23 (0.1674; estimate in Table 7, Column 2). Similarly, the value for industries with high external finance dependence is computed using the coefficients β ˆ2 estimated through equation 23: ˆ1 and β (0.160*0.1674)+(0.160*0.0405)=0.0333. 11 0.0268*192.7697 = 5.2; 0.0333*192.7697 = 6.4. 12 Midrigan and Xu (2014) and Buera et al. (2011) find larger productivity gains from reversing financial frictions when accounting for compounding forces, such as the effect of these frictions on limiting the adoption of better and more modern technologies. However, since our methodology is designed to capture the allocative channel only, it is not comparable to the magnitudes generated by these magnifying mechanisms. 32 ital market imperfections that brings finance misallocation down from the median to the 25th percentile of the cross-industry distribution of financial misallocation would trigger aggregate productivity gain of 5.2% on average, and an extra 1.2% gain in industries with a higher dependence on external borrowing. We found reassurance for the plausibility of these findings in that they fall in the ballpark of the gains found in macro-development studies under very different quantification strategies. 6. Conclusion Financial frictions play an important role in determining total factor productivity. In this study, we advance the extant literature by providing new evidence on the global properties of financial misallocation and its ability to account for the misallocation of real resources that is observed in the data. In light of a promising new methodology to measure finance misallocation and its narrow application to few countries, an important goal of the paper is to assess whether such methodology would deliver reasonable implications for the distribution of finance misallocation across countries with different degrees of financial develop- ment. Such an assessment is challenged by the data requirements: one must be able to count with firm-level databases offering a wide international and sectoral coverage, as well as rich information on the firm’s balance sheet. In this paper, we took this challenge confronting the new methodology with the ORBIS database, arguably the only suitable database for the task. We confirm that the novel measure of finance misallocation proposed in Whited and Zhao (2021) varies systematically with a country’s economic development in the direction that one would expect: finance misallocation is twice as severe in the poorest than in the richest countries in our sample. We also validate a number of well-rooted conjectures in the literature concerning the distribution of financing costs across firms of different ages and sizes. Based on the theoretical implication that younger and smaller firms are more likely to be financially constrained, we show that this is actually the case under our measure of theory-based 33 financing costs: small and young firms confront a significantly higher cost of accessing external finance. Perhaps the most novel of our contributions was the ability to establish an em- pirical connection between the novel measure of finance misallocation and the most standard measure of real-input misallocation that has been widely adopted in the literature since the work of Hsieh and Klenow (2009). Establishing such a connection is essential for the validity of one of the central assumptions behind the approach to measuring financial misallocation: that by disrupting the costs of accessing debt and equity markets, firms are in effect unable to acquire the efficient levels of productive factors. We close this gap in this paper and show that there is a strong positive cor- relation between financial misallocation in an industry and its degree of misallocation of real resources. Our estimates yield significant productivity gains from reducing fi- nance misallocation. Mediated by its implied improvement in real allocative efficiency, we find that reducing finance misallocation generates aggregate gains in Total Factor Productivity in the order of 5% to 6%, in the ballpark of what has been found earlier in the literature. While the paper builds upon a novel approach to measuring financial misallocation, it remains silent about the potential causes of such misallocation. In future research, we expect to leverage the rich time and cross-sectional dimensions of our database to identify plausible exogenous natural experiments that can provide us with a clean identification of the effect of particular types of financial reform or friction on financial misallocation and, ultimately, aggregate productivity. Acknowledgements The views in this paper are those of the authors and do not necessarily represent those of the World Bank, their Executive Directors, or the countries they represent. We thank Tatiana Didier, Leonardo Iacovone, Ha Nguyen, Jean Pesme and the par- ticipants to a World Bank internal workshop for constructive comments. 34 Appendix A. Supplementary tables Table A.1: Classification of firms into industries Code Value Description Industry 01 Agricultural Production - Crops A. Agriculture, Forestry, & Fishing 02 Agricultural Production - Livestock and Animal Specialties A. Agriculture, Forestry, & Fishing 07 Agricultural Services A. Agriculture, Forestry, & Fishing 08 Forestry A. Agriculture, Forestry, & Fishing 09 Fishing, Hunting and Trapping A. Agriculture, Forestry, & Fishing 10 Metal Mining B. Mining 12 Coal Mining B. Mining 13 Oil and Gas Extraction B. Mining 14 Mining and Quarrying of Nonmetallic Minerals, Except Fuels B. Mining 15 Construction - General Contractors & Operative Builders C. Construction 16 Heamy Construction, Except Building Construction, Contractor C. Construction 17 Construction - Special Trade Contractors C. Construction 20 Food and Kindred Products D. Manufacturing 21 Tobacco Products D. Manufacturing 22 Textile Mill Products D. Manufacturing 23 Apparel, Finished Products from Fabrics & Similar Materials D. Manufacturing 24 Lumber and Wood Products, Except Furniture D. Manufacturing 25 Furniture and Fixtures D. Manufacturing 26 Paper and Allied Products D. Manufacturing 27 Printing, Publishing and Allied Industries D. Manufacturing 28 Chemicals and Allied Products D. Manufacturing 29 Petroleum Refining and Related Industries D. Manufacturing 30 Rubber and Miscellaneous Plastic Products D. Manufacturing 31 Leather and Leather Products D. Manufacturing 32 Stone, Clay, Glass, and Concrete Products D. Manufacturing 33 Primary Metal Industries D. Manufacturing 34 Fabricated Metal Products D. Manufacturing 35 Industrial and Commercial Machinery and Computer Equipment D. Manufacturing 36 Electronic & Other Electrical Equipment & Components D. Manufacturing 37 Transportation Equipment D. Manufacturing 38 Measuring, Photographic, Medical, & Optical Goods, & Clocks D. Manufacturing 39 Miscellaneous Manufacturing Industries D. Manufacturing 40 Railroad Transportation E. Transportation & Public Utilities 41 Local & Suburban Transit & Interurban Highway Transportation E. Transportation & Public Utilities 42 Motor Freight Transportation E. Transportation & Public Utilities 43 United States Postal Service E. Transportation & Public Utilities 44 Water Transportation E. Transportation & Public Utilities 45 Transportation by Air E. Transportation & Public Utilities 46 Pipelines, Except Natural Gas E. Transportation & Public Utilities 47 Transportation Services E. Transportation & Public Utilities 48 Communications E. Transportation & Public Utilities 49 Electric, Gas and Sanitary Services E. Transportation & Public Utilities 50 Wholesale Trade - Durable Goods F. Wholesale Trade 51 Wholesale Trade - Nondurable Goods F. Wholesale Trade 52 Building Materials, Hardware, Garden Supplies & Mobile Homes G. Retail Trade 53 General Merchandise Stores G. Retail Trade 54 Food Stores G. Retail Trade 55 Automotive Dealers and Gasoline Service Stations G. Retail Trade 56 Apparel and Accessory Stores G. Retail Trade 57 Home Furniture, Furnishings and Equipment Stores G. Retail Trade 58 Eating and Drinking Places G. Retail Trade 59 Miscellaneous Retail G. Retail Trade 60 Depository Institutions H. Finance, Insurance, & Real Estate Continued on next page 35 Table A.1 – continued from previous page Code Value Description Industry 61 Nondepository Credit Institutions H. Finance, Insurance, & Real Estate 62 Security & Commodity Brokers, Dealers, Exchanges & Services H. Finance, Insurance, & Real Estate 63 Insurance Carriers H. Finance, Insurance, & Real Estate 64 Insurance Agents, Brokers and Service H. Finance, Insurance, & Real Estate 65 Real Estate H. Finance, Insurance, & Real Estate 67 Holding and Other Investment Offices H. Finance, Insurance, & Real Estate 70 Hotels, Rooming Houses, Camps, and Other Lodging Places I. Services 72 Personal Services I. Services 73 Business Services I. Services 75 Automotive Repair, Services and Parking I. Services 76 Miscellaneous Repair Services I. Services 78 Motion Pictures I. Services 79 Amusement and Recreation Services I. Services 80 Health Services I. Services 81 Legal Services I. Services 82 Educational Services I. Services 83 Social Services I. Services 84 Museums, Art Galleries and Botanical and Zoological Gardens I. Services 86 Membership Organizations I. Services 87 Engineering, Accounting, Research, and Management Services I. Services 88 Private Households I. Services 89 Services, Not Elsewhere Classified I. Services 91 Executive, Legislative & General Government, Except Finance J. Public Administration 92 Justice, Public Order and Safety J. Public Administration 93 Public Finance, Taxation and Monetary Policy J. Public Administration 94 Administration of Human Resource Programs J. Public Administration 95 Administration of Environmental Quality and Housing Programs J. Public Administration 96 Administration of Economic Programs J. Public Administration 97 National Security and International Affairs J. Public Administration 99 Nonclassifiable Establishments K. Nonclassifiable Establishments 36 Appendix B. Data Cleaning Following Bureau van Dijk (2011), Kalemli-Ozcan et al. (2015), Cusolito and Di- dier (2020), and Kalemli-Ozcan et al. (2023), we document the steps we apply to clean the financial information. 1. Fill time-invariant data gaps : for a given BvD.ID-year combination, with BvD.ID standing for firm unique identifier, replace missing highly-likely time-invariant information with information available for previous years (e.g., US SIC code, NAICS, NACE, NACE main sector, company name, city, region, postal code, legal form, incorporation date, thicker, isin). To perform this step, the team first worked with auxiliary raw tables, which collect legal and sectoral information of the firm, and collapsed the time-invariant variables at the BvD.ID level. 2. Harmonize timeframe : convert variable closedate from string to numeric format. Then create a new variable, name it year, and assign a year to the observation according to the following rule. If closing month corresponding to the obser- vation is June or any other month after June, then make Year take the year reported in closedate. Otherwise, make Year the year reported in closedate minus 1. 3. Drop duplicates : the raw database presents a large number of duplicates at the BvD.ID-year level. The team noticed that the information was the same, except in the SIC primary code variable. Thus, we collapsed all the SIC primary codes reported by the same BvD.ID-year in one variable, using semicolons to list all the SIC primary codes, and eliminated duplicates. 4. Drop firms with missing relevant information : drop all the firms with no in- formation for the following set of variables: US SIC code, NAICS, NACE core code, NACE main sector. 5. Drop observations with missing information for the currency code : eliminate observations with missing information for the currency code. 37 6. Drop observations with missing information for variable closedate : eliminate observations with missing information for the close date of the financial state- ment. 7. Drop observations with relevant missing information eliminate observations that at the BvD.ID-year level have missing information in all the following variables: operating revenue (turnover), sales, employment, total assets. 8. Drop duplicates and keep most updated information : keep observations with the most recent closing date if there are duplicates at the BvD.ID-year-first letter of consolidation code (e.g., C, U) level. 9. Drop duplicates and keep information from annual reports : keep observations with annual report in Use FillingType variable if there are still duplicates and keep the standardized information. Using annual reports (IFRS preferred, in- stead of local reports) guarantees standardization of reporting protocol at inter- national level. 10. Eliminate firms with noisy data : drop all the observations corresponding to a specific BvD.ID if any of the following variables has a negative value in a specific year – total fixed assets, tangible fixed assets, intangible fixed assets, other fixed assets, current assets, sales, and employment. 11. Deflate values : use country GDP deflators from the World Bank database to deflate nominal variables and set year 2005 as the base year.13 12. Harmonize currencies : convert values in local currency to USD dollars, using the average of the monthly exchange rate for year 2005. 13 https://data.worldbank.org/indicator/NY.GDP.DEFL.ZS. 38 Appendix C. Validation of Final Database We validate the representatives of the final database by calculating the ratio of the sum of employment and gross output in the database to their corresponding aggregates, in the same manner as Gopinath et al. (2017). Aggregates for employment and gross output are obtained from Eurostat’s Structural Business Statistics Database (SBS). Tables C.1 and C.2 show the coverage of our sample by country, separately for non-manufacturing and manufacturing sectors. Table C.1: Coverage of Final Database Relative to Eurostat (SBS) - Non-Manufacturing Country Employment Turnover 2010 2011 2012 2013 2014 2015 2016 2010 2011 2012 2013 2014 2015 2016 Austria 19% 21% 28% 32% 26% 31% 26% 16% 18% 34% 42% 36% 39% 36% Belgium 44% 41% 43% 45% 44% 44% 43% 55% 56% 55% 55% 55% 53% 52% Bosnia and Herzegovina 72% 34% 66% 65% 53% 60% 73% 40% 70% 63% 59% 62% Bulgaria 47% 58% 73% 70% 72% 76% 71% 53% 58% 61% 62% 62% 63% 62% 39 Croatia 53% 57% 59% 60% 60% 61% 63% 66% 67% 74% 76% 72% 72% 75% Czechia 64% 70% 68% 69% 66% 68% 63% 48% 48% 49% 51% 52% 54% 51% Estonia 41% 42% 42% 43% 46% 45% 47% 38% 39% 41% 39% 43% 42% 44% Finland 42% 45% 45% 44% 44% 41% 39% 52% 60% 56% 61% 60% 60% 57% France 25% 25% 24% 28% 32% 32% 29% 30% 31% 26% 34% 39% 37% 35% Germany 26% 28% 29% 29% 30% 28% 27% 47% 51% 49% 50% 49% 45% 45% Hungary 38% 39% 41% 43% 44% 39% 36% 73% 77% 82% 84% 85% 79% 72% Italy 39% 54% 55% 54% 56% 59% 56% 48% 58% 55% 54% 55% 55% 56% Luxembourg 39% 41% 41% 45% 125% 147% 128% 57% 64% 70% 69% 71% 89% 83% Montenegro North Macedonia 58% 63% 66% 67% 71% 77% 83% 87% 90% 84% Norway 10% 10% 9% 7% 7% 69% 69% 9% 9% 8% 7% 11% 59% 65% Poland 19% 18% 13% 7% 6% 4% 12% 21% 20% 14% 7% 7% 6% 15% Portugal 42% 46% 44% 44% 46% 46% 43% 50% 56% 55% 53% 57% 56% 51% Romania 43% 47% 49% 51% 54% 58% 58% 56% 63% 69% 71% 72% 81% 82% Serbia 59% 102% Slovak Republic 52% 60% 59% 65% 65% 59% 58% 54% 71% 68% 68% 66% 64% 61% Slovenia 64% 65% 69% 69% 68% 66% 67% 70% 76% 78% 79% 78% 76% 74% Spain 42% 44% 43% 44% 45% 45% 42% 50% 52% 52% 53% 53% 54% 49% Note: Blanks correspond to country-year pairs for which the Eurostat’s SBS Database does not report information. Montenegro and Ukraine are excluded from this table as the SBS Database does not include information for any of the years in our sample. Table C.2: Coverage of Final Database Relative to Eurostat (SBS) - Manufacturing Country Employment Turnover 2010 2011 2012 2013 2014 2015 2016 2010 2011 2012 2013 2014 2015 2016 Austria 14% 21% 34% 41% 44% 45% 47% 12% 17% 35% 51% 54% 58% 58% Belgium 63% 64% 64% 65% 67% 67% 66% 78% 76% 76% 72% 74% 79% 76% Bosnia and Herzegovina 64% 44% 67% 60% 62% 59% 62% 49% 68% 62% 66% 62% Bulgaria 54% 65% 73% 71% 70% 74% 70% 45% 50% 41% 40% 42% 45% 45% Croatia 61% 65% 65% 62% 65% 67% 65% 68% 76% 79% 75% 73% 71% 65% Czechia 74% 77% 76% 77% 75% 79% 72% 50% 52% 53% 55% 53% 62% 60% Estonia 39% 41% 41% 43% 44% 43% 44% 33% 33% 33% 35% 35% 36% 38% Finland 47% 52% 53% 54% 49% 48% 48% 45% 50% 52% 49% 46% 46% 49% France 30% 28% 23% 31% 37% 38% 40% 27% 25% 19% 30% 35% 37% 37% Germany 32% 34% 34% 34% 35% 34% 33% 46% 49% 48% 48% 50% 50% 49% Hungary 57% 61% 62% 62% 64% 63% 60% 75% 78% 79% 83% 89% 89% 90% Italy 47% 59% 59% 62% 63% 65% 64% 54% 61% 59% 62% 61% 61% 62% 40 Luxembourg 60% 67% 64% 62% 62% 55% 53% 80% 100% 96% 77% 77% 67% 62% North Macedonia 61% 65% 66% 60% 61% 56% 63% 47% 43% Norway 32% 36% 40% 28% 25% 68% 75% 31% 35% 31% 27% 26% 64% 73% Poland 26% 22% 16% 10% 7% 7% 16% 31% 25% 20% 12% 10% 10% 23% Portugal 61% 65% 66% 68% 68% 69% 68% 62% 64% 65% 67% 56% 59% 56% Romania 50% 53% 55% 57% 59% 60% 59% 45% 48% 57% 57% 60% 59% 58% Serbia 65% 87% Slovak Republic 61% 59% 63% 63% 61% 57% 52% 44% 48% 57% 59% 53% 43% 42% Slovenia 71% 74% 75% 71% 72% 74% 66% 71% 74% 75% 72% 70% 73% 65% Spain 52% 54% 54% 57% 59% 59% 56% 46% 46% 43% 49% 47% 49% 47% Note: Blanks correspond to country-year pairs for which the Eurostat’s SBS Database does not report information. Montenegro and Ukraine are excluded from this table as the SBS Database does not include information for any of the years in our sample. References Viral V Acharya, Tim Eisert, Christian Eufinger, and Christian Hirsch. Real Effects of the Sovereign Debt Crisis in Europe: Evidence from Syndicated Loans. The Review of Financial Studies, 31(8):2855–2896, 8 2018. ISSN 0893-9454. doi: 10.1093/rfs/ hhy045. URL https://academic.oup.com/rfs/article/31/8/2855/4982027. Viral V Acharya, Tim Eisert, Christian Eufinger, and Christian Hirsch. Whatever It Takes: The Real Effects of Unconventional Monetary Policy. The Review of Financial Studies, 32(9):3366–3411, 9 2019. ISSN 0893-9454. doi: 10.1093/rfs/ hhz005. URL https://academic.oup.com/rfs/article/32/9/3366/5298322. Philippe Aghion, Thibault Fally, and Stefano Scarpetta. Credit constraints as a barrier to the entry and post-entry growth of firms. Economic policy, 22(52):732–779, 2007. Silvia Ardagna and Annamaria Lusardi. Heterogeneity in the effect of regulation on entrepreneurship and entry size. Journal of the European Economic Association, 8 (2-3):594–605, 2010. Abhijit V. Banerjee and Esther Duflo. Growth Theory through the Lens of Develop- ment Economics. In Philippe Aghion and Steven N. Durlauf, editors, Handbook of Economic Growth, volume 1, chapter 7, pages 473–552. Elsevier B.V., 2005. ISBN 9780444520418. doi: 10.1016/S1574-0684(05)01007-5. Natalie Bau and Adrien Matray. Misallocation and capital market integration: Evi- dence from India. 2020. u¸ Thorsten Beck, Asli Demirg¨ c-Kunt, Luc Laeven, and Ross Levine. Finance, Firm Size, and Growth. Journal of Money, Credit & Banking, 40(7):1379–1405, 2008. Pedro Bento and Diego Restuccia. Misallocation, Establishment Size, and Produc- tivity. American Economic Journal: Macroeconomics, 9(3):267–303, 7 2017. ISSN 1945-7707. doi: 10.1257/mac.20150281. URL https://pubs.aeaweb.org/doi/10. 1257/mac.20150281. 41 Francisco J. Buera, Joseph P. Kaboski, and Yongseok Shin. Finance and development: A tale of two sectors. American Economic Review, 101(5):1964–2002, 2011. ISSN 00028282. doi: 10.1257/aer.101.5.1964. Bureau van Dijk. Orbis - User Guide, 2011. Ricardo J Caballero, Takeo Hoshi, and Anil K Kashyap. Zombie lending and depressed restructuring in Japan. American Economic Review, 98(5):1943–1977, 11 2008. ISSN 0002-8282. doi: 10.1257/aer.98.5.1943. URL https://pubs.aeaweb.org/ doi/10.1257/aer.98.5.1943. Harold L. Cole, Jeremy Greenwood, and Juan M. Sanchez. Why doesn’t technology flow from rich to poor countries? Econometrica, 84(4):1477–1521, 2016. ISSN 0012-9682. doi: 10.3982/ecta11150. Ana Paula Cusolito and Tatiana Didier. Methodology, sql, and stata codes to create a clean financial firm-level data using orbis raw files. Working paper, World Bank, Washington, DC, US, 2020. Joel Michael David, Lukas Schmid, and David Zeke. Risk-Adjusted Capital Allocation and Misallocation. 2020. URL https://www.ssrn.com/abstract=3753061. Roberto N. Fattal-Jaef. Entry barriers, idiosyncratic distortions, and the firm size distribution. American Economic Journal: Macroeconomics, 14(2):416–68, April 2022. doi: 10.1257/mac.20200234. URL https://www.aeaweb.org/articles?id= 10.1257/mac.20200234. Mariassunta Giannetti and Andrei Simonov. On the real effects of bank bailouts : Micro evidence from Japan. American Economic Journal: Macroeconomics, 5(1): 135–167, 2013. Gita Gopinath, Sebnem Kalemli-Ozcan, Loukas Karabarbounis, and Carolina Villegas-Sanchez. Capital allocation and productivity in south europe. Quarterly 42 Journal of Economics, 132(4):1915–1967, 2017. ISSN 15314650. doi: 10.1093/qje/ qjx024. Joao Granja and Sara Moreira. Product innovation and credit market disruptions. 2021. URL https://www.ssrn.com/abstract=3477726. Chang-Tai Hsieh and Peter J. Klenow. Misallocation and manufacturing TFP in China and India. The Quarterly Journal of Economics, 124(4):1403–1448, 2009. doi: 10.1093/qje/qjs020.Advance. Chang-Tai Hsieh and Peter J. Klenow. The life cycle of plants in India and Mexico. Quarterly Journal of Economics, 129(3):1057–1106, 2014. doi: 10.1093/qje/qjs020. Advance. Sebnem Kalemli-Ozcan, Bent Sorensen, Carolina Villegas-Sanchez, Vadym Volosovych, and Sevcan Yesiltas. How to Construct Nationally Representative Firm Level Data from the Orbis Global Database: New Facts and Aggregate Implica- tions. NBER Working Papers 21558, National Bureau of Economic Research, Sep 2015. URL https://ideas.repec.org/p/nbr/nberwo/21558.html. Sebnem Kalemli-Ozcan, Bent Sorensen, Carolina Villegas-Sanchez, Vadym Volosovych, and Sevcan Yesilitas. How to construct nationally representative firm level data from the orbis global database: New facts and aggregate implications. 2019. Sebnem Kalemli-Ozcan, Bent Sorensen, Carolina Villegas-Sanchez, Vadym Volosovych, and Sevcan Yesiltas. How to construct nationally representative firm level data from the Orbis global database: New facts on SMEs and aggregate im- plications. American Economic Journal: Macroeconomics, (Forthcoming):–, 2023. ISSN 1945-7707. Mauricio Larrain and Sebastian Stumpner. Capital Account Liberalization and Ag- gregate Productivity: The Role of Firm Capital Allocation. Journal of Finance, 72 (4):1825–1858, 2017. ISSN 15406261. doi: 10.1111/jofi.12497. 43 Virgiliu Midrigan and Daniel Yi Xu. Finance and misallocation: Evidence from plant- level data. American Economic Review, 104(2):422–458, 2 2014. ISSN 00028282. doi: 10.1257/aer.104.2.422. Franco Modigliani and Merton H. Miller. The cost of capital, corporation finance and the theory of investment. The American Economic Review, 48(3):261–297, 1958. URL https://www.jstor.org/stable/1809766. Benjamin Moll. Productivity losses from financial frictions: Can self-financing undo capital misallocation? American Economic Review, 104(10):3186–3221, 2014. ISSN 00028282. doi: 10.1257/aer.104.10.3186. ¨ H. Ozlem Dursun-de Neef and Alexander Schandlbauer. COVID-19 and lending re- sponses of European banks. Journal of Banking & Finance, 133:106236, 12 2021. ISSN 03784266. doi: 10.1016/j.jbankfin.2021.106236. URL https://linkinghub. elsevier.com/retrieve/pii/S0378426621001953. Joe Peek and Eric S. Rosengren. Unnatural selection: Perverse incentives and the misallocation of credit in Japan, 9 2005. ISSN 00028282. Raghuram G. Rajan and Luigi Zingales. Financial dependence and growth. American Economic Review, 88(3):559–586, 1998. URL https://www.jstor.org/stable/ 116849. ˇ ak, Papa N’Diaye, Adolfo Barajas, Ran Bi, Diana Ayala, Ratna Sahay, Martin Cih´ Yuan Gao, Annette Kyobe, Lam Nguyen, Christian Saborowski, Katsiaryna Sviry- dzenka, and Seyed Reza Yousefi. Rethinking Financial Deepening: Stability and Growth in Emerging Markets. Technical report, The International Monetary Fund, Washington, DC, USA, 2015. Fabiano Schivardi, Enrico Sette, and Guido Tabellini. Credit Misallocation During the European Financial Crisis. The Economic Journal, 132(641):391–423, 1 2022. ISSN 0013-0133. doi: 10.1093/ej/ueab039. URL https://academic.oup.com/ej/ article/132/641/391/6258430. 44 Veronique Vermoesen, Marc Deloof, and Eddy Laveren. Long-term debt ma- turity and financing constraints of SMEs during the Global Financial Cri- sis. Small Business Economics, 41(2):433–448, 8 2013. ISSN 0921-898X. doi: 10.1007/s11187-012-9435-y. URL http://link.springer.com/10.1007/ s11187-012-9435-y. Toni M. Whited and Jake Zhao. The Misallocation of Finance. The Journal of Finance, 76(5):2359–2407, 10 2021. ISSN 0022-1082. doi: 10.1111/jofi.13031. URL https://onlinelibrary.wiley.com/doi/10.1111/jofi.13031. 45