The Oxford Martin Programme on Technological and Economic Change Disclaimer: This is a working paper, and represents research in progress. This paper represents the opinions of the authors and does not represent the position of the Oxford Martin School or other institutions or individuals. For more information on the Oxford Martin Programme on Technological and Economic Change, please visit: https://www.oxfordmartin.ox.ac.uk/technological-economic-change/ Do Court Delays Distort Capital Formation? Pantelis Koutroumpis and Farshad R. Ravasan May 2020 OMPTEC Working Paper No. 2020-5
Transcript
The Oxford Martin Programme on
Technological and Economic Change
Disclaimer: This is a working paper, and represents research in progress. This paper represents the opinions of the authors and does not represent the position of the Oxford Martin School or other institutions or individuals. For more information on the Oxford Martin Programme on Technological and Economic Change, please visit: https://www.oxfordmartin.ox.ac.uk/technological-economic-change/
Weak enforcement of financial contracts often distorts the incentive to investand delays efforts to catch up with the technology frontier. We study how longerbankruptcy trials affect the cost, structure and allocation of corporate capital acrossdifferent regions in Italy. We take advantage of an exogenous change in local courtefficiency caused by the reorganization of judicial districts under the legislativereform of 2012. Using an instrumental variable strategy, we find that the change inlength of bankruptcy proceedings had a strong impact on firm-level outcomes. Ourestimates show that the interquantile reduction in the length of bankruptcy trialslowers the marginal cost of capital by 11.5% and increases the firm’s capital stockand capital intensity of production by 9.7% to 11.5%. Significantly, our results showthat poor enforcement of financial contract lead firms to under-allocate capital intheir intangible assets. Moreover the effects are stronger in sectors that dependmore on external finance and firms with high leverage.
∗ University of Oxford, Oxford Martin school, 34 Broad Street, Oxford, OX1 3BD, UnitedKingdom. E-mail: [email protected]† Corresponding author. University of Oxford, Oxford Martin school, 34 Broad Street,
The economic performance of Southern European countries has been atthe centre of political and economic debate. Since the mid-1990s, theseeconomies have fallen behind and diverged from the global productivityfrontier. The economic slowdown in this region has often been linked tofinancial frictions and malfunctioning capital markets (Cette et al. (2016);Gopinath et al. (2017)). Economists and policy makers have often surmisedthat these frictions might stem from the poor quality of legal institutionsthat hinder creditor rights in bankruptcy, leading to a lower recovery ratefor loans and discouraging lending. (La Porta et al. (1997, 1998, 2000, 2008);Kumar et al. (1999); Demirgüç-Kunt and Maksimovic (1998); Djankov et al.(2007, 2008)).
The origins of these frictions partly stem from the adoption of the Frenchcivil law in this group of countries; including Italy, Spain, Portugal andGreece. The reason for this is that the civil laws originating in France pro-vide lower levels of legal protection to creditors compared to English com-mon law, the German and Scandinavian civil law (Djankov et al. (2003)). Toovercome these issues and improve their performance, Southern Europeancountries have implemented major legal reforms to update their bankruptcycodes1. Nevertheless, the recovery rate of loans remained low, due to thelack of efficient enforcement. For instance, the typical bankruptcy proce-dure takes 7.25 years in Italy whereas lenders expect to be repaid after only1.5 year in the U.S. 2. However, the bankruptcy trial length is not constantwithin countries exhibiting significant variations 3. This suggests that theorganization and bureaucracy of the sub-national legal systems plays aneminent role in defining the efficiency of debt contract enforcement in eachregion.
In this paper we look into the duration of bankruptcy cases across differ-ent Italian judicial districts and how they shape firms’ financial decisionsand outcomes. In this regard, our paper contributes on two grounds. First,
1For instance, the 2005–2006 Italian bankruptcy reforms introduced legal outlets, sim-ilar to US chapter 11, which made the renegotiation of credit contracts easier. Further-more, it significantly accelerated firms’ liquidation procedures Rodano et al. (2016). Newbankruptcy laws were also introduced in Spain in 2004 and Greece in 2007
2Both calculation for U.S. and Italy refer to 2012. Estimate for U.S. comes from Ponticelliand Alencar (2016). The estimate for Italy is based on authors’ calculation
3See Giacomelli and Menon (2016) for Italy and Fabbri (2010) for Spain
2
using a natural experiment, we aim to estimate the casual link betweencourt delays in bankruptcy trials and firm’s (under-) allocation of capital.This approach improves the identification strategy commonly used by theliterature and relies on the spatial variation of court efficiency. The latter islimited for two reasons: first, the spatial variation of judicial efficiency is of-ten highly correlated to other institutional factors even at the micro regionallevel. Second, the spacial relation between firm performance and local courtenforcement could be affected by firms’ decision to locate their HQ in spe-cific districts. To address these issues, we exploit the reorganization of 494 judicial districts in Italy. The legal reform assigned firms to the newlycreated courts, leading them to face a significant and unexpected changein the average trial length of bankruptcy proceedings compared to the pre-implementation period. Using this exogenous shock as an instrument, weestimate the elasticity of firms’ outcome to the trial length in bankruptcycases.
Second, we analyze the effect of court delays on firms spending acrossvarious types of assets. During the last two decades, the corporate assetstructure has seen a drastic shift from physical to intangible capital in aneffort to adjust to a high technological change environment. Hall (2001);Borisova and Brown (2013); Corrado et al. (2005, 2009); Corrado and Hul-ten (2010). The intangible investment in R&D activities , workforce train-ing, firm-specific know-how, new software, databases, copyrights, designs,trademarks, organisation and distribution networks are increasingly becom-ing key contributors in a firm’s ability to catch up with the expanding pro-ductivity frontier Demmou et al. (2019). This paper highlights both thequantitative and qualitative impact of the court delays in firms’ (under-)allocation of capital. We show that the longer trial lengths causes firms toboth accumulate lower levels of fixed capital but also to disproportionatelyunder-allocate capital in intangible assets. This effect could have played acrucial role in the productivity slowdown of the activities of new economies.
In this paper, we study the impact of the reorganization of judicial dis-tricts undertaken by the Italian government as part of a wider nationalagenda aimed to reduce public spending and create larger and more special-ized courts. The aim of this reform was not originally linked to addressing
4The reform covered 49 out of 165 districts.
3
the performance of credit markets or economic development but mainly tolimit public spending. In particular, the legal decree n. 155 of 7 September2012, enacted in September 2013, led to the reorganization of 49 out of 165judicial districts in Italy. This reform suppressed 26 courts, their districtsabsorbed by 23 other districts to create a new landscape of judicial districts.Starting from September 2013, firms were located in the reformed areas re-assigned to the newly created courts. The average trial length of the newcourts was often substantially different from that of their predecessors. Thekey reason being that the merger happened across districts with very dif-ferent pre-implementation levels of judicial efficiency. The trial lengths inthe new courts were close to the average trial length of the merged districts.The reform thereby drastically increased the trial length in more efficientdistricts and reduced it in less efficient ones. This heterogeneous impact ofthe reform provides an almost ideal setting to estimate the causal effect oflaw enforcement.
To measure the impact of the reform we need to estimate the difference intrial lengths between the new court and that of the merged districts. How-ever, the comparison is not straightforward. Ideally we would prefer tocompare the trial length in the new and old districts over the same year.This would guarantee that our comparison is not affected by year specificshocks in trial lengths. Nevertheless, the old courts ceased to exist after2013 and the new courts were only created afterwards. This implies thatwe don’t have comparable data on trial lengths for the same years both forold and new courts. To overcome this problem, we impute the hypotheti-cal length of proceedings in the newly created district according to the caseflow data. We then infer the impact of the reform by comparing the hypo-thetical length of trial case in the new and old courts, during 2011 and 2012,before the enactment of the reform.
Next, we use this estimated impact of the reform to instrument the differ-ence in the length of bankruptcy proceedings before and after the reform.We show that these estimated values are, indeed, highly and positively cor-related to actual changes in bankruptcy proceedings that occurred after thereform. Moreover, we show that these counterfactual changes are uncorre-lated with firm-level outcomes prior to the reform. Furthermore, we repeatthese tests in non-reformed districts using placebo simulations and confirmthat there is no significant change before or after the reform. This helps
4
ensure that the validity of our instrument satisfies both the relevance andexclusion conditions.
We combine the judicial data with a large and nationally representativedataset of manufacturing and service firms covering the period 2008-20175. This includes information on both financial and production decisions andcomes from Aida, which is a subset of the ORBIS - Europe firm-level dataset,covering Italian enterprises and is provided by the Bureau van Dijk. Usingthe instrumentation strategy described above, we provide casual estimatesof the delays in bankruptcy proceedings in local courts delays on firms’ out-comes.
The first firm-level outcome that we look into is the marginal return prod-uct of capital (MRPK), which is widely used to infer the firm-specific addi-tional cost of raising capital linked to financial frictions (Hsieh and Klenow(2009); Gopinath et al. (2017); Larrain and Stumpner (2017)). We comparethe temporal changes in MRPK for the firms affected by the reorganization,and focus on those that belong in the same two digit sectors but are lo-cated in different judicial districts which are therefore exposed to differentchanges after the reform. Our results indicate that the weak enforcement offinancial contracts increases a firm’s marginal cost of capital by almost 12%.Next, we look into the effects of the reform on the stock of capital using thesame identification strategy. Our results show that improvements in judi-cial efficiency can increase the stock of capital by 11.5%. In a similar way, wefind that capital intensity increased, with shorter bankruptcy proceedings,by 9.7% in the regions where judicial efficiency improved.
Next, we expand our analysis to look into the impact of the reform on thestructure of firms’ capital in terms of tangible and physical assets. Invest-ments in intangible capital, including the corporate spending in R&D andworkforce training, are particularly relevant and have increased sharplyover the past three decades, even surpassing the tangible investment Hall(2001); Borisova and Brown (2013); Corrado and Hulten (2010). In fact, thereis an emerging growth nexus that links productivity to the investment inintangible assets Corrado et al. (2005, 2009); Corrado and Hulten (2010).Although intangible assets represent a key feature in the transformation
5We only include non financial and non farm firms
5
of modern corporations in light of technological changes, almost a quar-ter of Italian firms in our sample report no investment in intangible capi-tal. Using the aforementioned instrumental variable strategy, we find thatthe share of intangible capital increased for firms located in districts withshorter bankruptcy proceedings by 2.1%.
Furthermore, we show that the link between the quality of law enforce-ment and firms’ outcomes is driven, at least partly, by the financial con-straint channel. We test whether the court delays have a stronger impacton more financially vulnerable sectors and firms. Following the approachby Rajan and Zingales (1998); Braun (2005); Manova (2013), we find that thequality of financial contract enforcement has no effect on the size of fixedcapital in sectors relying more on internal finance. On the contrary, a largeelasticity of capital to court delay is observed among sectors that dependmore on external finance. We find this differential impact across sectors isstronger in a sub-sample of intangible capital intensive industries.
Moreover, we study the impact of court delays on highly levered firms.The high levels of corporate debt can distort a firm’s incentive to investsince it can reduce the profitability of the investment for shareholders My-ers (1977) and make future financing more costly Almeida and Campello(2007). This implies that firms with high levels of debt could face an under-investment problem that impede their future growth (Lang et al. (1996)).This situation is often referred as debt overhang in corporate finance litera-ture first coined by Myers (1977). However, the extent of debt overhang ismore severe when cyclical or institutional factors worsen the credit marketconditions (Lamont (1995); Occhino and Pescatori (2015); Chen and Manso(2017); Kalemli-Ozcan et al. (2018)). Our results suggest that court delaysindeed, substantially intensifies the debt overhang problem. We find thatcourt delays led firms with high leverage to reduce their capital to a greaterextent. They would also be more likely to under-allocate capital in theirintangible assets.
Departing from the effects of the reform on capital inputs, we now turnto that of financial contract enforcement on firm employment. The mecha-nism that moderates the effect on employment is two-fold. First, the lowermarginal cost of capital allows firms to "scale up" their production. In this
6
sense, judicial improvements would have a positive impact on firm employ-ment. However, the lower marginal cost of capital encourages firms to ad-just their input shares and focus on capital intensive process. This "substi-tutional" impact would then decrease the employment share of firms.
We find no impact between changes in law enforcement and the levelof employment. However, once we look into the growth rates of employ-ment, we find a strongly positive association between shorter bankruptcyproceedings and employment growth. This implied contradiction betweenthe static and dynamic effects is linked to the difference between the shortand long-run impacts (Kosova (2010); Lu et al. (2017)). This in turn suggeststhat the short term impact of the reform led to an increase in capital inten-sity which practically offset the scaling up mechanism. In the long run, thesubstitutional impact recedes and the scaling-up force dominates. The samestatic and dynamic impact is evident when looking at the wage bills. How-ever, the impact of the enforcement of the law is three times larger on thewage bill growth rate compared to the employment growth rate. This find-ing suggests that the shorter trial lengths also had a positive long-run effecton the average wage per employee. This in turn implies that improved lawenforcement has a positive long-run effect on the employment size and onthe quality of the workforce.
We then study the average effect of the reform from a more limited policy-maker’s perspective. The reorganization of judicial districts correspondedwith national objectives, such as spending cuts in the public budget. Thegovernment’s view of the reform was that creating larger courts would pro-vide an opportunity for judges’ specialization, and would increase judicialefficiency. In practice however, the reform led to unintended heterogeneouslocal changes. Some judicial districts gained more efficiency out of the re-form and others lost. This differential impact of the reform has allowedus to study the real effect of law enforcement. However, the question re-mains: what was the overall effect of the reform on firms’ outcomes? Ourresults show a positive but not significant effect across reformed districts.Furthermore, we find that reformed districts in the Italian South, on aver-age, gained substantially from the reform. However, there was a negativebut not significant impact of the reform in the North.
7
The remainder of this paper proceeds as follows: Section 2 reviews therelevant literature and Section 3 describes our estimation strategy coveringthe institutional details of the reform, data and empirical estimation. Section4 presents the results and Section 5 concludes.
II. LAW ENFORCEMENT AND FINANCIAL FRICTIONS
The quality of legal institutions has been found to positively correlatewith financial development and economic performance across countries (La Porta et al. (1997, 1998, 2000, 2008); Kumar et al. (1999); Giannetti (2003);Lerner and Schoar (2005); Qian and Strahan (2007); BAE and Goyal (2009);Haselmann et al. (2009)). Since judicial quality at the national level is alsocorrelated with other aspects of the institutional environment, isolating thecasual impact of the legal institutions in a cross-country setup can oftenprove challenging. (Acemoglu (2005); Acemoglu and Johnson (2005); Laevenand Woodruff (2007)). This leads the researcher to face two distinct issues:the identification and description of the mechanisms that moderate the linkbetween judicial efficiency and firm level outcomes. We look into these as-pects separately in this section.
A fundamental aspect of these analyses is that they build on metrics thatrely on the spatial variation of judicial efficiency within countries to de-sign a more reliable empirical identification and better control for potentialomitted variables (Brown et al. (2016); Ponticelli and Alencar (2016); Pezone(2018)). This process helps unbundle the national and sub-national effectsthat mediate the mechanism of judicial quality on firm-level outcomes. Toachieve this, researchers tried to distinguish between the legal content andlaw enforcement, in an attempt to exploit the regional variations of judi-cial efficiency. While the former is usually common across all regions in acountry, the latter can vary significantly due to sub-national organisation,the characteristics and workload of the local courts.
Building on this rationale, several studies look into regional variationsin local court enforcement for several countries including Mexico Laevenand Woodruff (2007) ,Italy (Jappelli et al. (2005); Giacomelli and Menon(2016); Pezone (2018)), Spain Fabbri (2010) and Brazil (Ponticelli and Alen-car (2016)). These studies use the local variation in the length of civil pro-ceedings to study the impact of judicial efficiency on firms’ employment size
8
(Jappelli et al. (2005); Laeven and Woodruff (2007); Giacomelli and Menon(2016); Ponticelli and Alencar (2016); Pezone (2018)) , firms’ capital (Fab-bri (2010)) , firms’ investment and capital intensity (Ponticelli and Alen-car (2016)) and bank loans (Jappelli et al. (2005); Schiantarelli et al. (2016)).These approaches meet the first part of the challenges described (data col-lection) but still need to address identification and mechanism design. Wenow turn to identification issues.
In order to draw a causal link within a specific country, some studies haverelied on regional historical differences to instrument the variation in thelength of civil proceedings. For instance, Laeven and Woodruff (2007) ex-ploit the systematic correlation between the quality of the legal system andthe prevalence of the indigenous population in year 1900 across differentstates in Mexico. The key hypothesis has been that in those places wherethe share of the indigenous population was higher, European settlers weremore likely to adopt institutions designed to exploit local labour. This inturn, had a protracted impact, hindering the quality of institutions in otherareas where indigenous populations had a more pronounced presence. Theproblem with this approach is that these historical factors are not them-selves exogenous to the regional judicial quality, as the characteristics of the"control-group" could also be correlated with confounding sub-national fac-tors. This results in a breach of the exclusion criterion and fails to effectivelyisolate the exogenous variation in the quality of law enforcement.
More recently, researchers used robust statistical methods to alleviate theseconcerns. This includes discontinuity designs that bind municipalities inpairs that are located across the judicial districts’ borders Giacomelli andMenon (2016); Brown et al. (2016); Ponticelli and Alencar (2016). While thisapproach has merit, its outcomes are still susceptible to firm behavior, in-cluding sorting their options about locating their HQ on either side of theborder. This in turn suggests that a reverse effect might be in place, as thequality of local law enforcement may drive firms’ decision to locate theirheadquarters in one side or the other of the judicial borders Pezone (2018).
The common caveat in both of these approaches is that they rely only onspacial variation. In this paper we use the reorganization of Italian courtdistricts as a natural experiment that enables us to exploit both the tempo-ral and the spatial variation in law enforcement. In this regard, our identi-
9
fication is very close to Pezone (2018) that exploits this event to study theimpact of average civil trial length on firms’ employment size.
Besides the identification constraints, the mechanism that connects judi-cial efficiency and firm outcomes is yet another aspect on which we focus.Throughout the literature, the poor quality of legal institutions has oftenbeen described as a key contributor that increases financial frictions and re-sults in weak contract enforcement. There is ample work across countriesthat studies the effects of legal institutions on the financial system and theirinfluence in capital allocation across the economy. Lerner and Schoar (2005);Qian and Strahan (2007); BAE and Goyal (2009); Haselmann et al. (2009).
The low recovery rate of loans is the main mechanism through whicha slow or weak contract enforcement process can generate frictions in localcredit markets. There are two separate forces that operate in tandem to formthis mechanism: the first, which we call "supply side frictions" signals tocreditors that their ability to recover their claims from distressed firms willbe limited (Qian and Strahan (2007); BAE and Goyal (2009); Haselmann et al.(2009)). The second, which we call "demand side frictions", is that weakenforcement can also affect the behavior of otherwise solvent borrowers,increasing the likelihood of "strategic defaults" (Jappelli et al. (2005); Guisoet al. (2013); Schiantarelli et al. (2016)).
These two forces, isolated or combined, affect credit supply, as they dis-courage lending and increase the risks of external financing, other thingsbeing equal. In spite of the overarching mechanism that moderates the flowof credit, the outcomes of this process do not affect every firm or sectorin the same way. Building on the insights of Rajan and Zingales (1998),the literature has often focused on the differential effect of law enforce-ment across industries to show that the link between institutional qualityand real outcomes works through the credit supply channel. To further dis-tinguish the industry’s sensitivity on credit supply conditions, researchersRajan and Zingales (1998) often relied on proxies that measure the exter-nal financial dependence of each industry’s investments . This proxy hasbeen used widely, in particular, among the very recent works that study thefirm level impact of law driven finance Ponticelli and Alencar (2016); Brownet al. (2016); Larrain and Stumpner (2017); Pezone (2018). Consistent withthis literature, our results show that the lax law enforcement hits sectors that
10
depend more on external funds disproportionately. We find this differentialimpact is stronger among the subset of sectors that use more intangible in-tensive assets in their production.
To highlight the credit supply channel, our paper is also related to thestrand of literature that studies the marginal return product of capital (MRPK)to measure the frictions in the capital markets (Hsieh and Klenow (2009);Midrigan and Xu (2014); Larrain and Stumpner (2017); Gopinath et al. (2017)).We show that the improvement in law enforcement decreases the MRPKwhile it has no significant effect on the marginal return product of labour(MRPL). This provides further evidence that the reform primarily impactedfirms through changing the conditions in the capital markets and not inlabour markets.
Our results are also connected to the literature on the impact of finan-cial frictions on the labour market. This literature studies the impact ofthe variety of financial shocks including the Great Recession Chodorow-Reich (2014); Bentolila et al. (2018), the interstate branching deregulation ofRiegle-Neal Act Beck et al. (2010); Benmelech et al. (2011) the loan supplyshock transmitted by Japanese banks to markets in the United States Ben-melech et al. (2011); collateral constraint due to the lower real estate pricesChaney et al. (2012); Bahaj et al. (2019); collateral constraint driven by a le-gal reform in Sweden Cerqueiro et al. (ming) or the cyclical credit crunchMoscarini and Postel-Vinay (2012) among others. Our paper is closely re-lated to this strand of research that focuses on the financial frictions drivenby weak contract enforcement. These studies show that the poor qualityof law enforcement led firms to cut their employment Kumar et al. (1999);Laeven and Woodruff (2007); Giacomelli and Menon (2016); Pezone (2018).
Moreover we study whether court delays distort capital formation viathe debt overhang channel. In this regard our paper is broadly related tothe literature that studies the relationship between financial frictions anddebt hangover risk. The seminal work of Myers (1977) showed how exist-ing corporate debt could lead to an agency problem and sub-optimal in-vestment decisions. The debt overhang risk arises when a firm is highlyleveraged and faces the risk of default. In this case, firms would anticipatethe marginal benefit of any new investment will be absorbed by its creditorsin the event of default. Thus the investment decision not only depends on
11
the cash flows from investment, but also on the transfers between differentstakeholders. The probability of default acts as a net transfer from a firm’sshareholders to the existing debt holders. This leads firms to discount fur-ther the marginal profit of new investment, which in turn discourages themto invest. An alternative channel that could induce the debt overhang risk,is that high leverage increases the cost of future financing.
Lamont (1995) shows that the effect of debt overhang varies with creditmarket conditions. Debt overhang is binding when firms face more severefinancial frictions. Using a calibrated model, Occhino and Pescatori (2015)shows that the debt overhang effect on investment is higher during reces-sions when default risk is higher. Kalemli-Ozcan et al. (2018) provide firmlevel evidence that the credit crunch following the 2008 Global Crisis ledto an under-investment problem in South European countries via the debtoverhang channel. Our paper adds to this literature, showing that courtdelays increase the debt overhang cost. Considering the large North-Southdivide in the efficiency of court enforcement and corporate debt, our papersheds light on a new channel that might contribute to the divergence of theEuropean periphery countries.
Our paper contributes to this literature on two grounds. Firstly, these pa-pers, with the exception of Pezone (2018), use the spatial variation of lawenforcement that is defined by historical factors or long-existing adminis-trative borders. This implies that their estimates only capture the long-runor equilibrium impact. Our quasi experimental design, contrarily, enablesus to also capture the short-run or transitory impact of improvements incontract enforcement. Secondly, our results reveal that the long run positiveimpact on firm’s wage-bill is much stronger than employment size. Thiseffect, which is consistent with Kumar et al. (1999), implies that better lawenforcement not only enables firms to scale up their employment, but alsoto create organizational changes within the firms Kumar et al. (1999). Thelarger firms hire high-quality workers Kremer (1993) or more employees ata higher level in the hierarchy Rosen (1982).
Finally, one of our main results contributes to the rising literature on theimpact of financial friction on firm’s under-allocation of capital in intangi-ble assets. Studying the impact of the Great Recession, Duval et al. (2019)show that financial frictions could be an important factor in explaining the
12
low share of firms’ intangible assets. In fact, they show that intangible cor-porate spending is a lot more likely to be affected by financial frictions com-pared to investments in physical assets. As firms adjust their investmentdepending on their expected risk and return for various types of assets 6, aweak contract enforcement environment can decrease the perceived returnof intangible assets. This is because it is harder to use intangible assets ascollateral, owing to the difficulty of external valuation of patents and tacitworkforce training and skills. Given the complexity of property right cases,it also becomes even harder to secure a return for intangible assets Claessensand Laeven (2003); Braun (2005). Therefore, the weak law enforcement candistort the optimal allocation of spending on intangible assets.
The earlier works by Claessens and Laeven (2003); Demirgüç-Kunt andMaksimovic (1998) provide cross-country evidence that highlights the asso-ciation between a better legal system and the higher share of spending inintangible assets. However, the cross sectional nature of these works makesit hard to establish the causal link. Using a quasi experimental instrument,our paper is the first to provide the micro level evidence and casual links tosupport this hypothesis. We find that the longer bankruptcy trials have ledfirms to under-allocate capital in intangible assets.
III. ESTIMATION STRATEGY
III.A Reorganization of court districts in Italy
Italy adopted the French civil law that has several institutional features thatare particularly significant to our study. Firstly, the content of these lawsover credit relations are the same nationwide, but enforcement varies sub-stantially across local court districts. Secondly, financial disputes are as-signed to the local courts on a territorial basis. Thereby, the variation inlocal court efficiency leads to a variation in financial contract enforceabilityfor local firms Giacomelli and Menon (2016).
Figure 1 points to these drastic differences in judicial efficiency at the localcourt level, ranging between two to more than ten years. Although there isa distinct pattern of improving court enforcement as we move to the north
6See (Matsuyama (2007))
13
of the country, there are often substantial differences between two neigh-bouring districts.
Before 2013, the territorial organization of the Italian judicial system wasbased on 165 court jurisdiction areas. This territorial distribution of courtswas mainly determined by historical factors and largely resembled the oneshaped in 1865, immediately after the unification of Italy. This, in turn, wasbased on the judicial systems of the previous states.
Until 2013, no existing court had ever been removed Giacomelli and Menon(2016). The major reform of the court reorganization occurred on the 7th ofSeptember 2012 and became effective on the 13th of September 2013. Thereform reorganised 49 out of 165 pre-reform judicial tribunes. The reorgani-zation of the courts led to the suppression of 26 courts that merged with theadjacent district of a surviving court.
The main aim of the reform was to reduce costs in the judicial system.Although, no judge or employee was fired during the reform, the reorgani-zation led to massive running cost savings in court facilities, including theuse of large, historical buildings. However, the reorganization of the courtswas based on some administrative criteria that are not related to the initialefficiency of the courts. For example, the reform suppressed the courts thatwere not located in provincial capitals. This kept the selection of surpassedcourts exogenous in the sense of judicial efficiency.
Before the reform, the Italian Minister for Justice highlighted this pointby mentioning that the reorganization of the courts, aimed to update theadministrative boundaries maintained in the country since the unification,when "people were moving with horse-drawn carriages, not high-speedtrains" 7. Nevertheless, there were still some exceptions made against thesuppression of several southern local courts in districts with a high pres-ence of organised crime Pezone (2018).
III.A.1 Impact of the reform
Figure 2 highlights the impact of the reform in Piemonte region. There were12 judicial districts before the reorganisation, in 2013. After the reform,
7From Pezone (2018) that cites this from Accorpamenti di tribunali e procure. Severino:Una riforma epocale. corriere.it, July 6, 2012
14
seven courts were suppressed and joined with their neighbouring districts.For instance, in the Cuneo province, the two districts Saluzzo and Mon-doví were surpassed and joined to their neighbouring district, the provincialcapital Cuneo. Alba, another pre-reform district in the province of Cuneo,was also suppressed and absorbed by the neighbouring district Asti in theprovince of Alessandria.
Although the merged districts were nearby and often located in the sameprovince, they exhibited substantial differences in the length of their bankruptcyproceedings. For instance, the average length of bankruptcy proceedingsduring 2011 and 2012 before enactment of the reform, was much shorterin Saluzzo (1527 days) compared to Cuneo (2596 days) and Mondoví (2770days). However, after the September 2013, the firms that were located inthese three districts were assigned to the same new court that was createdout of their merger.
To study the impact of the reform we need to compare the trial lengthbetween the new court and these three pre-reform districts. However, thecomparison is not straightforward. Ideally we want to compare the triallength in the new and old districts in the same year. This would guaranteethat our comparison is not affected by year specific shocks in trial lengths.Nevertheless, the old courts existed until 2013 and the new courts werecreated afterwards. This implies we don’t have comparable data on triallengths for the same year for old and new courts.
To overcome this problem, we impute the hypothetical length of proceed-ings in the newly created district during 2011 and 2012. We estimate thetrial length in the new court according to the aggregate number of pending,initiated and resolved bankruptcy cases in the three districts during 2011and 2012. We follow the approach, initially used by Pezone (2018) whichwill be explained in section III.C.
Using this estimate we can quantify the impact of the reform in thesethree districts by comparing the hypothetical trial length of the new courtwith the actual trial length in these districts between the 2011 and 2012.
Our calculation indicates that the hypothetical length of proceedings inthe newly created court, would be on average 2248 days during 2011 and
15
2012. This number lies in the range between the shortest trial period inSaluzzo (1527 days) and the longest proceeding length in Mondoví (2770days). This suggests that the merger had a very different impact on the firmsin these three districts. Our estimation implies that the reorganization ofcourt districts increased the trial length by 49.2% for firms that are located inthe district of Saluzzo whilst reducing trial lengths in Cuneo and Mondovíby 12.9% and 15.1% respectively.
Table 2 presents the hypothetical average trial length for all 23 new courtsbetween 2011 and 2012 (before the reform). Comparing these with the actualtrial lengths for 49 pre-reform districts, we find a similar pattern as we do inour example from the Cuneo districts. The districts that merged with moreefficient courts benefited from the reform. However, the reorganization in-creased the trial length in districts that merged with courts with relativelylonger trial periods.
Furthermore, it is important to note that the reorganization of courts re-sulted in increasing public concern and discontent. The complaints weremore evident in districts where judicial efficiency dropped post- reform.For instance, in 2017, in a petition to the Minister for Justice, the mayorsof the district of Crema complained that their territory was “gravely andunjustly penalized by the closure of the Court of Crema in 2013, which wasamong the best in the country in terms of efficiency. In another example,the mayor of the city of Bossano del Grappa publicly complained about thesuppression of the district and merging to the neighbouring Vicenza court.He wrote in a public letter in 2012, to the prime minister, Minster for Justiceand the Home Secretary that the length of civil trials is much longer in theVicenza court compared to the court of Bossano del Grappa. He stressedthat the merger could "gravely penalize a community and an economic areaof enormous dimension". 8
III.A.2 Estimating the reform impact using the hypothetical trial lengthof new courts before the reform
In this section we explain how we impute the hypothetical pre-reformtrial length of new districts that were created after the reform. First, it is
8These two examples come from Pezone (2018). He cites the Crema petition that isavailable at www.cremaonline.it and the letter form mayor of Bossano del Grappa that isavailable at www.bassanonet.it
16
worth noting that, in Italy, there is no information on the exact length ofbankruptcy proceedings at the court level. Nevertheless, we could estimatethem through the available case flow data as follows.
Dt =Pt + Pt+1
Et + Ft× 365 (1)
where Dt is trial length to resolve matters regarding the bankruptcy pro-ceedings in year t, Pt are pending cases at the beginning of the year t, Pt+1
are pending cases at the end of the year t, Ft are new cases filed duringyear t and Et are cases ending with a judicial decisions or withdrawn by theparties during the year t.
We apply the same formula adopted by the Italian Ministry of Justiceand the Italian National Institute of Statistics (Istat) to calculate and reportthe trial length in different districts. This approach is also widely used inthe academic works to calculate the average length of civil proceedings asa sound measure of judicial inefficiency Jappelli et al. (2005); Schiantarelliet al. (2016); Giacomelli and Menon (2016); Pezone (2018). Furthermore, us-ing the rarely available data from actual length of civil proceedings in 2016,Pezone (2018) shows that above index is remarkably reliable estimate of theactual proceedings.
Here, we use the same concept to estimate the hypothetical pre-reformduration of court proceedings in the new districts. If the new court wouldhave existed before the reform, its pending, new and solved cases wouldbe the sum of the all pending, new and solved cases in the merged districts.Thus we could calculate the trial length of the new court j at time t that werecreated by the merger of the district of i ε 1,..,N with other N-1 courts at timeT > t 9.
DNewjt =
∑Ni=1 Pti + ∑N
i=1 Pt+1i
∑Ni=1 Eti + ∑N
i=1 Fti
∗ 365 (2)
Then the hypothetical impact of the reform on district i at time t < T canbe calculated as follows
∆Re f ormijt = log(DNew
jt )− log(Dpre−re f ormit ) (3)
The DNewjt and Dpre−re f orm
it can, respectively, be driven from equation (3) and
9This approach proposed by Pezone (2018) to calculate the impact of the reform
17
(1).
III.B Data and definition of variables
III.B.1 Judicial data
Data on the judicial system comes from Italian ministry of justice and in-cludes information for all Italian civil courts between 2008 and 2017. Pub-licly available data, between 2014 to 2017, reports only the length of civilproceedings and we don’t have access to the case-flow data. However, thedata for pending, incoming and resolved cases are available between 2007and 2013. That makes it possible to calculate the hypothetical length ofbankruptcy proceedings for the new courts before the reform in 2013.
Data for civil cases includes the bankruptcy cases as well as labour dis-putes, property and torts. Fortunately, we have disaggregated data for thesesubjects. That makes our analysis different from Giacomelli and Menon(2016); Pezone (2018) that use the average length of civil proceedings dueto lack of more detailed data. Following Schiantarelli et al. (2016); Pon-ticelli and Alencar (2016), we focus on the length of bankruptcy proceed-ing to measure the quality of financial contract enforcement. It is impor-tant to note that bankruptcy proceedings in our data does not include thevoluntary bankruptcy cases that the borrower files himself. As voluntarybankruptcy cases tend to be much shorter, this increases the average lengthof bankruptcy trials than we observe in the data.
To combine our judicial and firm level data, we assign each firm to a ju-dicial jurisdiction. To this end, we use the judicial court where the firmis located. This approach is similar to that used by Giacomelli and Menon(2016); Schiantarelli et al. (2016); Pezone (2018) to study the real impact of ju-dicial inefficiency in Italian civil courts. This choice is motivated by the factthat post-enforcement, however, this requires several steps. First, lendersneed an injunction from the court typically located in the province of itsheadquarters. Having gotten an injunction, to take possession of collateralthe lender must then adjudicate before the court in the location of the collat-eral that often coincides with the firm’s location Schiantarelli et al. (2016).
18
III.B.2 Firm level data
Our firm level data comes from AIDA, the subset of ORBIS - Europe, cover-ing both listed and unlisted Italian enterprises provided by Bureau van Dijk.The database includes information on firms’ value added, output, capitalstock and employment. It also includes data on firms’ balance sheets andincome statements including debt, asset , capital depreciation and share oftangible vs intangible fixed capital.
We focus on the period of 2008 to 2017, where we have a decent availabil-ity of main variable of interest. The number of observations shrinks drasti-cally before 2008 and data are sparse before 2005. In Italy, unlike US ,there isa strict rules for all firms including being unlisted to disclose their balancesheet to the Italian Chamber of Commerce, this being the main source forAIDA Pezone (2018). As a result, our data has a nationally representativecoverage of Italian firms in the period of our study.
Similar to Duval et al. (2019), we study firms in non-farm and non-financialbusinesses. We follow Kalemli-Ozcan et al. (2015) to clean data from dupli-cates and inconsistent entries. In case of multi-plant firms, we keep theheadquarters and remove all other plants from the same establishment. Wealso restrict our sample to the firms that have data entries before 2012 andafter 2013. This excludes new firms that entered the market after the reformpassed into law in 2012. This could eliminate the potential sorting effect,as the introduction of the reform could have led new entries to strategicallylocate themselves in districts that expected to gain more from the reform.
III.B.3 Firm level measure of financial friction
When firms face idiosyncratic frictions in access to external funds, it im-pacts on their return of capital. There is an extensive literature 10 , thatstudies these financial frictions by examining the differences between thefirm’s marginal revenue product of capital (MRPK) within similar sectors11.
We follow closely Gopinath et al. (2017); Larrain and Stumpner (2017) toconstruct the measure for the marginal revenue product of capital (MRPK)
10See for instance Midrigan and Xu (2014); Gopinath et al. (2017); Larrain and Stumpner(2017)
11often within sectors classified at two digits industry identification
19
at firm level based on the ORBIS data set. Both papers use the frameworkand set of assumptions that were developed by Hsieh and Klenow (2009).First, we briefly explain the conceptual framework behind our measure.Then we discuss how we construct it based on our firm level data.
We consider an industry at time t that is populated by a large number Nstof monopolistic competitive firms. We define industries in the data by theirtwo-digit industry classification. Total industry output is given by a CESproduction function as follows
Yst =
[Nst
∑i=1
Dist(yist)ε−1
ε
] ε1−ε
(4)
where yist denotes firm i’s real output, Dist denotes a demand shifter forfirm i’s variety, and ε denotes the elasticity of substitution between varieties.We denote by pist the price of variety i and by Pst the price of industry outputYst. Firms face an isoelastic demand for their output given by
yist = (pist
Pst)−ε(Dist)
εYst. (5)
This leads all firms in sector s to impose a constant and similar markup µs =ε
1−ε over their marginal cost when they choose their price, capital, and laborto maximize their profits. Moreover, we consider that firms are exposed tothe different degree of frictions k in capital markets12. Denoting the inversedemand function by p(yist) we could write down the profit maximizationproblem of firm i in sector s in time t:
Then we could derive the variation in marginal cost of capital for eachfirm within the sector and across the time by tracking firm’s log( pistyist
kist)
log(MCKist) = log(pistyist
kist) + Cs (8)
12In our case reflects the court delays
20
We measure firm nominal value added, pistyist, as real gross output (op-erating revenue) deducted by cost of materials. We measure real output,yist,as nominal value added divided by an output price deflator. Given that wedo not observe prices at the firm level, we use gross output price deflatorsfrom EU KLEMS at the two-digit industry level. We measure the capitalstock, kist, with the book value of fixed assets and deflate this value withthe price of investment goods. In fixed assets we include both tangible andintangible fixed assets. We use a two-digit industry specific deflator basedon the investment good prices for Italy. For our price deflators, for bothconsumption and investment goods, we choose 2010 as our reference year.
III.C Estimation specification
Equation 10 shows our benchmark model to estimate the impact of financialcontract enforcement on firms’ outcome.
y f spit = βlog(Dit) + δ f + δs,t + δp,t + ε f spit (9)
Where f, s, p, i and t denote the firm, two digit industry, province, pre-reform judicial districts and year respectively; y f spit measures the perfor-mance of firm f of industry s at time t that is located in province p and pre-reform judicial district i. In our estimation y f spit includes firms’ outcomessuch as the logarithms of stock of capital, capital per employee, MRPK andshare of intangible in fixed capital. The specification contains firm fixed ef-fect δ f to control for time invariant characteristics of the firms, sector-yearfixed effect δs,t to control for sectoral shocks and and province-year fixedeffects δp,t to control for time varying regional characteristics.
Dit is our variable of interest. It denotes the length of bankruptcy proceed-ing in days in judicial district i at time t 13. The trial length has been usedwidely in the literature to capture the efficiency of contract enforcement14.In our specification, we use the logarithm of trial lengths and firms’ perfor-mance variables; with the exception of the share of tangible assets in fixedcapital. This helps us to correct the right skewness of our dependent and in-
13Dit shows the trial length for pre-reform district i before the reform in 2013 and the newdistrict j after 2013. However, since the new districts uniquely is defined by pre-reformdistricts we drop the index j from our equations
14See Fabbri (2010); Schiantarelli et al. (2016); Ponticelli and Alencar (2016); Giacomelliand Menon (2016); Pezone (2018)
21
dependent variable. Furthermore, this logarithmic transformation enablesus to report our coefficient of interest β in elasticises of firm outcome to thebankruptcy trial length.
To obtain an unbiased estimate of β we need to make sure our variable ofinterest is not correlated with the error term ε f spit. There are many potentialconfounding factors that could violate this condition. For instance, districtswith poor judicial efficiency could have less effective institutions in general,which binds the capacity of local firms and markets.
To address this identification problem, we use exogenous variations intrial lengths across districts caused by the reorganization of local courts in2013. To this end we use the following first stage IV estimation:
log(Dit) = β′Impact of the reformi × Post-reformt + δ f + δs,t + δp,t + ε f spit
(10)Where Post-reform is a dummy that takes one for all in the years after 2012(t≥2013). Impact of the reform is our indicator of reform impact that we con-struct in section . It estimates the impact of the reorganization on the re-formed districts and takes zero for all in districts that were not impacted bythe reform.
Significantly, since our instrument varies at pre-reform judicial districts,we cluster the error terms at the pre-reform districts in all our estimations.This follows the Bertrand et al. (2004) method to address the potential serialcorrelation and heteroskedasticity issue.
III.D Validity of the instrument and exclusion conditions
For our identification strategy to be valid, our instrument needs to capturethe exogenous variation in trial length. This implies that our instrumentshould not be correlated to the error terms ε f spit in equation 10. This exclu-sion condition can be denoted as follows:
cov(Impact of the reformi × Post-reformt, ε f spit|X f spit) = 0
Where X f spit is the vector of control variables that we used in equation 10.This exclusion condition could be violated under two circumstances. Firstcase is that Impact of the reformi and error terms in equation 10 are correlated.
22
This suggest that the selection of the reformed districts and the way theyhave been impacted by the reorganization was not random.
To test this possibility, we examine the dynamics of the relation betweenfirms’ outcome and Impact of the reformi. We trace a year-by-year effect dur-ing the 10-years window between 2008 to 2017. We run the following re-gression that includes the same control variables in equation 10.
log(y f spit) = Impact of the reformi×2017
∑t′=2008
βt′×1(t = t′))+ δ f + δs,t + δp,t + ε′f spit
(11)We could reject the possibility of correlation between Impact of the reformi
and error terms in equation 10 if there is no association between the Impactof the reformi and firms outcome prior to the reform in 2013. This impliesthat βt′ = 0 for all t′ < 2013.
The second case is that Post-reformt and error terms in equation 10 arecorrelated. This suggests that 2013, the timing of the reform, was not co-incidental. To absorb the the general year specific shocks we include thesector-year and province-year fixed effects. However, our results still couldbe contaminated by other shocks that occurred during same time span whenthe reorganisation of the courts was executed. The main concern is that theless efficient courts would respond differently to the other existing shocks.This could contaminate our results as the predicted impact of the reformdepends on the ex-ante quality of the courts.
To reject this hypothesis, we simulate a placebo effect of the reform indistricts that are not covered by the reform. We randomly select a 23 po-tential merger between neighbouring judicial districts from provinces thatare not impacted by the reform. We calculate the potential impact of thesehypothetical mergers in 2013. We then regress the stock of capital on theimpact of this hypothetical reform in our baseline reduced form regressionwith the same range of control variables. We repeat these placebo exercises3000 times.
Finally, it is important to note that our instrument belongs to the familyof difference in difference based instruments that have been used recently
23
in studies such as Lu et al. (2017); Pezone (2018); Malgouyres et al. (2019). Itis constructed based on the event that impacted some districts while othersremain unaffected 15. Our two exclusion conditions reflect the exogeneityof the timing and impact of the event 16. These are the two important con-ditions that needed to be held for the validity of this group of instruments.
IV. RESULTS
IV.A Main Results
We begin by reporting our baseline results for the impact of court delaysin bankruptcy proceedings on firm outcome. Table 3 reports the estima-tion for our baseline instrumental variable regression. The first and sec-ond stage estimations are respectively reported in panel B and A in table 3.In addition to firm fixed effect, all regressions include the sector-year andprovince-year fixed effect to absorb the sectoral or local specific shocks. Fol-lowing Bertrand et al. (2004), the standard errors clustered at the pre-reformjudicial districts in order to address the potential serial correlation and het-eroskedasticity.
Panel B, indicates that our instrument, Predicted impact of the reform × Postreform (year2013), has a positive and statistically significant effect on theLog(Length of proceedings). This confirms that the reform triggered changesin the length of bankruptcy proceedings in the post reform period and there-fore satisfies the relevance condition. The main issue with the relevancecondition is the weak instrument problem. To address this concern, we fur-ther report the Kleibergen-Paap F-test that is from the magnitude of 33 inour main regressions.
With respect to our central research question, panel A indicates that, afterbeing instrumented, the longer length of bankruptcy proceedings cast a neg-ative impact on firm outcomes.These results further confirm the findings in
15However, ours differs from normal difference in difference instruments when treat-ment variable is a dummy that takes 1 if the observation belongs to the treatment group. Inour case, the treatment variable is continuous, reflecting the heterogeneous impact of thereform
16In the normal difference in difference based instrument the impact of the reform isbinary. Thus the exogeneity of the impact of the event implies random selection to thetreatment group
24
the literature (e.g. Kumar et al. (1999); Laeven and Woodruff (2007); Fabbri(2010); Giacomelli and Menon (2016); Brown et al. (2016); Pezone (2018)).
Column (1), in panel A, indicates that 10% rise in the length of bankruptcyproceedings increases the marginal return production of capital by 1.63%.The size of MRPK used in the literature to capture the size of the firm-specific financial friction (i.e., Larrain and Stumpner (2017) ). Our resulttherefore implies the 10% increase in the length of bankruptcy proceedingsincrease the firms’ marginal cost of raising capital by 1.63%. This result isin line with the estimates of Fabbri (2010) that found 10% increases in theaverage length of proceedings decrease the cost of bank finance by 3.62%.
The estimate, in column (2) of panel A, shows that the 10% longer bankruptcyproceedings leads firms to reduce their capital stock by , the same magni-tude, 1.63%. column (3) of panel A, demonstrates that 10% longer bankruptcytrials reduce the capital per employee by 1.38%.This suggests that the weakenforcement of financial contracts could distort the optimal choice of inputsin the production process and thus reduce firm level productivity.
Column (4) of panel A, shows the results for the impact of law enforce-ment on firms’ intangible capital stock. Intangible capital includes the cor-porate spending on RD and workforce training. The estimate in column (4)indicates that the 10% increase in the length of proceedings reduce the shareof intangible capital by 0.3-percentage point. Considering that spending onintangible assets is typically much lower than on physical assets, the esti-mate is also economically significant. The result suggests that, facing finan-cial friction caused by court delays, firms under-allocate spending in theirintangible assets. This is consistent with the recent findings of Duval et al.(2019) that show the financial frictions, caused by the Great Depression, ledfirms to cut their investment in intangible more so than on physical capital.Duval et al. (2019) use the share of debt maturing in 2008 to measure thefinancial vulnerability of the firms. They found 10% increase in debt matur-ing in 2008 is associated with the 0.58-percentage-point decline in the shareof intangible assets when firms hit by financial crisis.
We further report the reduced form estimation when we regress the firm’sstock of capital directly on our instrument along with the same set of con-trols. In column (5) of panel C. The estimate is negative and strongly signif-
25
icant which is consistent with the result in column (2) and further confirmsthe relevance condition. Column (6), in panel C, reports the estimate forOLS regression. The estimated coefficient is negative, but much smaller, andnot statistically significant. This suggests that naive OLS estimation couldunderestimate the impact of the length of bankruptcy proceedings due tothe confounding variables (Lu et al. (2017)).
IV.A.1 Economic magnitude
To highlight the economic magnitude of this result, we perform a hypothet-ical experiment. Moving from the areas in the bottom quartile of judicialefficiency to the top quartile, according to the average length of bankruptcyproceedings between 2008 and 2012, decreases(increases) the marginal cost(size) of capital by 11.47 % for firms within the similar sectors17. This alsoboosts the intensity of capital in firm production process by 9.71 %. Finally,the share of intangible capital will expand by 2.11 percentage-point. Whilethe effect may not seem large, one should bear in mind that many Italianfirms have no or a very small stock of intangible capital in their assets. Forinstance, the bottom quartile and median level of the share of intangiblecapital in Italy were respectively 0.16% and 5.18% between 2008 and 2012.
IV.B Robustness checks
IV.B.1 Dynamic effects and exclusion condition
Next, we examine the dynamics of the relationship between the changesin the length of bankruptcy proceedings induced by the reform on firms’outcomes. We estimate the elasticity of a firm’s capital size and share of in-tangibles to the predicted impact of the reform for each year in the periodof five years before and after the law was enacted. Figure 3 shows these es-timated effects. It illustrates that the impact of the reform does not predatethe reform. This confirms the validity of our instrumentation strategy toexploit the reform as a natural experiment. This suggest that we could usethe impact of the reform as an exogenous source of variation in the lengthof bankruptcy proceedings to study the impact of financial contract enforce-
17The bottom and top quartile of Log(Length of proceedings) between 2008 and 2012respectively are 7.7129 and 8.4167. This implies that interquartile difference is 0.7038. Theimpact of interquartile improvement on capital stock in percentage can be obtained by(8.4167− 7.7129)× 0.163× 100 = 11.47194. Where 0.163 is the Elasticity of capital to courtdelay that we measured in column 2 of table 3.
26
ment on the firm’s capital size and structure. As a result, this reassures usthat our instrument satisfies the exclusion condition.
IV.B.2 Placebo test
We further conduct placebo tests to assure that our results are not drivenby the other shocks that occurred simultaneously with the reorganisation ofthe courts. The key concern is that the less efficient courts would responddifferently to the other existing shocks. This could contaminate our resultsas the predicted impact of the reform depends on the ex-ante quality of thecourts. To reject this hypothesis, we simulate a placebo effect following amethod used by Lu et al. (2017); Pezone (2018); Malgouyres et al. (2019).
We randomly select a 23 potential merger between neighbouring judicialdistricts from provinces that are not impacted by the reform. We calculatethe potential impact of these hypothetical mergers in 2013. We then regressthe stock of capital on the impact of this hypothetical reform in our baselinereduced form regression with the same range of control variables. We repeatthis placebo exercise for 3000 different random selections.
Figure 4 shows the cumulative distribution function of the estimated co-efficients.The distribution of these estimates are centered around zero withthe mean value of 0.0041 and standard deviation of 0.0194. The real coef-ficient lies outside the 99% of the 3000 placebo estimates. These results arevery strong and show that our estimates are not driven by other simultane-ous shocks unrelated to the reform.
IV.C Economic channel
Next we examine the mechanism that moderates the estimated effects. Wetest whether the contract enforcement affects firms’ capital formation throughthe credit channel. Following the standard approach in the literature, we ex-amine whether the impact of law enforcement was different for sectors andfirms that are more financially vulnerable. At the sectoral level we constructfinancial vulnerability according to the external finance dependence and as-set intangibility. These two factors make the performance of some sectorsmore sensitive to financial development and access to credit (Rajan and Zin-gales (1998); Braun (2005); Almeida and Campello (2007); Manova (2013)).At the firm level, we focus on the role of debt overhang and ex-ante balance
27
sheet vulnerability. Lamont (1995); Hennessy (2004); Kalemli-Ozcan et al.(2018); Duval et al. (2019).
IV.C.1 Sectoral financial vulnerability
Following the approach in Rajan and Zingales (1998), we construct a sec-toral proxy for external finance dependence. The index measures the dif-ference between investment and cash flow as a share of the investment. Inthat sense, the lower value of the index. the higher the dependence on ex-ternal funding. Since capital expenditures are not directly reported in ourdata (Aida), we follow the method used by Acharya et al. (2018) and Pe-zone (2018) to calculate investment. To this end, we compute investment asfixed capital minus lagged fixed capital plus depreciation. We set all neg-ative values to zero. Since investment appears in the denominator of ourproxy, the proxy is calculated only for firms with strictly positive levels ofinvestment. We calculate the average of this index for each firm between2008 and 2012. Using the average index of each firm, we find the median ofeach two digit industries. Then, we use this sectoral index to divide sectorsinto two groups. The sectors that are more dependent on external financehave a lower index than median. Similarly, they depend on internal financeif their index is higher than the median.
Column (1), in panel A of table 4, shows that the length of bankruptcyproceedings does not have any impact on the firms that depend more oninternal finance. The estimated coefficient is almost zero. However, firmsare very sensitive to the length of proceedings when they operate in sectorsthat are dependent on external finance. The estimate suggest that a 10%increase in the length of bankruptcy proceedings led firms to reduce theirstock of capital by 3.11%. The magnitude is almost twice as high comparedto our baseline estimation 1.63%. However, column (1), panel B of table 4,shows that there is no statistically significant differences between sectors, inthe elasticity of the share of intangible capital.
In a cross-country study, Claessens and Laeven (2003) show that thereis positive association between law enforcement for property rights andgrowth. This association is stronger in sectors with a higher share of in-tangible assets. This implies the intangible intensive sectors should be morevulnerable to weak law enforcement. Nevertheless, some of the recent firm-
28
level empirical evidence shows that improvements in law enforcement tendto benefit more the tangible intensive sectors (Braun (2005); Berkowitz et al.(2015); Pezone (2018)).
One explanation for these contradictory results lies in the fact that in-tangible intensive sectors might rely more on internal rather than externalfunds. Thus, in this paper, we study the the spread of law enforcement ef-fect between sectors, that depend on internal and external funding. Thenwe compare the size of this spread in the subsample of tangible and intan-gible intensive sectors. Similar to the external funding index, we calculatethe average share of intangible capital for each firm between 2008 and 2012.Using the average index of each firm, we compute the median of each two-digit industry. We then use this sectoral index to divide sectors into twogroups. The intangible intensive sectors that have a share of intangible cap-ital above the sample median value and intangible sectors that have a shareof intangible capital below the median.
Figure 5, panel a, b and c show that quality of law enforcement has noeffect on size of capital in sectors with internal funding across tangible andintangible intensive sectors. However, the weak financial contract enforce-ment exhibits a strong negative impact on externally funded sectors. Thisgap, between internally and externally funded sectors, is more substantialand statistically significant amongst intangible intensive sectors than in tan-gible intensive sectors. Columns (2) and (3) of panel A, in table 4 showthat a 10% increase in the length of proceedings led firms to reduce the sizeof their capital by 1.80 % , which is not statistically significant, in the sub-sample of tangible intensive sectors. However, a 10% increase in lengthof proceedings generates a sharp and significant decline in firm capital by5.69% among intangible intensive sectors. Nevertheless, the results in panelB of table 4 indicate that the impact of court delays on share of intangiblecapital is the same across sectors with high and low dependence on exter-nal finance. Furthermore, column 2 and 3, show the similar pattern acrosssub-sample of tangible and intangible intensive sectors.
IV.C.2 Corporate debt overhang
Following Kalemli-Ozcan et al. (2018), we construct the index for the debtoverhang using the ratio of total debt to total assets as a proxy for firm in-
29
debtedness. We measure total debt as the sum of long term debt and currentliabilities. Then we divide firms in tertiles according to their level of debtoverhang in 2012 right before the reform occurred. Following the commonpractice in this literature, we classify bottom, second and top tertiles as thelow, medium and highly levered firms.
Columns 1 to 3 in table 5 show that the relationship between court en-forcement and capital size is much stronger among highly leveraged firms.The impact of court delays increases by more than twice when we movefrom the sub-sample of firms with low to high leverage. Column 4 indicatesthe difference between estimated coefficients, -0.195 is statistically signifi-cant at 5%. Columns 4 to 7 point to a similar pattern in the impact of courtdelay on capital structure. The coefficient is as large as five times in the sub-sample of firms with high leverage compared to the low leveraged firms.Column 8 reports the difference between the estimated coefficient in thesub-sample of firms in the top and bottom tertiles. The difference is sub-stantial and significant at 5%. Figure 6 visualizes the results of table 5. Itshows that corporate investment and capital formation in highly leveragedfirms are much more sensitive to the court delays.
IV.D Financial contract enforcement and firm’s employment
There is a growing interest in the literature that looks into the impact oflaw enforcement on firms’ employment size Kumar et al. (1999); Laevenand Woodruff (2007); Giacomelli and Menon (2016); Pezone (2018). In thissection we study the consequences of better financial contract enforcementon a firm’s employment size. We argue that there could be two opposingimpacts. First, the lower marginal cost of capital lets firms increase theircapital and "scale up" their production. Thus, it will boost the employmentsize of firms. Second, the lower marginal cost of capital will encourage firmsto use a more capital intensive process. This "substitutional" impact willthen decrease a firm’s employment share.
Column (1) , in table 6, shows there is no impact due to changes in lawenforcement and the level of employment size and cost. However, the es-timate in column (3) demonstrates that there is a negative and statisticallysignificant impact of the weak financial contract on employment growth. A10% increase in the length of bankruptcy proceedings reduces the employ-
30
ment growth rate by 3%. The literature interprets this dispersion betweenstatic and dynamic effect as a difference between the short term and long-run impact Kosova (2010); Lu et al. (2017). These results suggest that theshort term impact of an increase in capital intensity is offset by the positiveimpact of scaling up. Nevertheless, in the long run, the positive impact offinancial contract enforcement on the firm’s employment size dominates.
Furthermore, the estimates that are reported in columns (2) and (4) showthe same static and dynamic impact on a firm’s wage bill. Column (4) indi-cates that a 10% increase in the length of bankruptcy proceedings reduces afirm’s wage-bill growth rate by 10.5%. Comparing the coefficients in column(2) and (4) highlights the magnitude of the impact on the firms’ wage-bill asalmost three times larger. This suggests that law enforcement has a posi-tive long run effect on both employment size and on the average wage peremployees.
IV.E Average impact of the reform
In this section, we discuss the average effect of the reform. The reorganiza-tion of judicial districts followed the national objectives to introduce spend-ing cuts in the public sector budget. The government supported that thereform would create larger courts and provide the opportunity for judges’specialization, thereby increasing judicial efficiency. However, the reformcreated unintended heterogeneous local changes. Some judicial districtsgained more efficiency out of the reform and others lost. So far, we ex-ploited this differential impact of the reform to study the real effect of lawenforcement. In this section we test whether the reform had, on average, apositive impact on firms in the reformed areas. Panel A,B and C , in Table 7,illustrates that there is an overall positive but not significant impact of thereform. However, Southern districts, on average, gained substantially fromthe reform. Average capital stock increased by 7% in the reformed areascompared to non reformed areas in the same two digit sectors. The firmsalso increased capital intensity by 6.7% post the reform. On other hand, wefind a negative but not significant impact of the reform in Northern districts.
31
V. CONCLUSION
In this paper, we looked into the casual link between court delays in bankruptcyproceedings and firm’s under-allocation of capital. There are both shortand long term impacts of these delays. The direct effects lead to increas-ing financial frictions which is indicated by the changes in firms’ marginalreturn productivity of capital (MRPK). This affects both the size and thestructure of firms’ capital, limiting firms’ ability to accumulate capital andunder-allocate resources to riskier forms of capital such as intangible assets.
Our findings highlight channels through which financial frictions in SouthEuropean countries affect firms’ input allocation. The key source of finan-cial frictions is related to the low recovery rate of secured loans in this regionwhich discourages lending. To address these implications, governmentsacross the region have implemented several legal reforms to limit the ex-tend of these frictions through nation-wide legislation aimed to improvethe creditor’s rights in cases of bankruptcy. These interventions had lim-ited impact, as the recovery rate remained low. The key contribution of thispaper is to highlight the significant impact of local civil court inefficienciesfor judicial enforcement. As policy-makers often suggest that the economicslowdown in southern Europe stems from the poor institutional quality ofcontract enforcement, we aimed to provide more robust empirical evidenceand a clear mechanism to explain this conjecture.
32
Figure 1
Average length of bankruptcy proceedings across Italian civil courts
Length of bankruptcy proceedings (in years)>108-106-84-6<4No data
Figure indicates the average length of proceedings (in years) during the 2011and 2012 before the enactment of the reform in 2013. The red line indicatesthe area covered by the reform.
33
Figure 2
Impact of the reform in the northern region of Piemonte
Figure shows the impact of the reform in the northern region of Piemonte.The variation in greyscale indicates the average length of bankruptcy pro-ceedings between 2011 and 2012 before enactment of the reform. The thickgreen line indicates the border of the new judicial districts generated by themerging of smaller districts. The lines within the encircled green zones in-dicate the border of judicial districts before the reorganization. The number,below the name of each prereform judicial districts, indicates the predictedpercent change in length of bankruptcy proceedings due to merging.
34
Figure 3
Response to the court delays induced by the reform
-.04
-.02
0.0
2.0
4El
astic
ity o
f tan
gibl
e ca
pita
l sha
re to
cou
rt de
lay
-.2-.1
0.1
.2El
astic
ity o
f cap
ital t
o co
urt d
elay
2008 2009 2010 2011 2012 2013 2014 2015 2016 2017
Log Capital Share of tangible capital
Figure shows that the court delays induced by the reform reduce the sizeof firms’ capital and lead them to under-allocate capital in their intangibleassets. We use the impact of the reform to instrument trial length in ourbenchmark IV regression. The impact of the reform measures the differ-ence between actual trial length of reformed districts and hypothetical triallength of new districts between 2011 and 2012, before the enactment of thereform. The year-by-year court delay elasticity is captured by the coeffi-cients βt′ in following reduced form regression
log(y f spit) = Impact of the reformi×2017
∑t′=2008
βt′×1(t = t′))+ δ f + δs,t + δp,t + ε′f spit
Where f, s, p, i and t denote the firm, two digit industry, province, pre-reform judicial districts and year respectively and y refers to firm’s capitalsize and share of tangible capital.
35
Figure 4
Placebo test
Figure plots the empirical cumulative distribution function of estimatedplacebo coefficients from 3000 regressions. Each time, we simulate a placeboreform in 2013 that creates 23 new courts out of the merger between neigh-bouring districts that are randomly selected. The draws exclude areas thatwere covered by the reform. The solid red line corresponds to the actualestimate for the elasticity of the capital in the reduced form specification intable 3 column (5). It lies outside of the 95% and 99% confidence intervalthat are delineated respectively by the doted and dashed lines in gray.
36
Figure 5: Response to Court delays: Internal vs external finance dependence
Elasticity of capital stock to court delay
(a)Tangible intensive
-1-.8
-.6-.4
-.20
.2.4
.6.8
1El
astic
ity o
f cap
ital t
o co
urt d
elay
Inte
rnal
Fun
d
Exte
rnal
vs
Inte
rnal
Fun
d
(b)All sectors
-1-.8
-.6-.4
-.20
.2.4
.6.8
1El
astic
ity o
f cap
ital t
o co
urt d
elay
Inte
rnal
Fun
d
Exte
rnal
vs
Inte
rnal
Fun
d
(c)Intangible intensive
-1-.8
-.6-.4
-.20
.2.4
.6.8
1El
astic
ity o
f cap
ital t
o co
urt d
elay
Inte
rnal
Fun
d
Exte
rnal
vs
Inte
rnal
Fun
d
Elasticity of intangible capital share to court delay
(d)Tangible intensive
-.2-.1
0.1
.2El
astic
ity o
f int
angi
ble
capi
tal s
hare
to c
ourt
dela
y
Inte
rnal
Fun
d
Exte
rnal
vs
Inte
rnal
Fun
d
(e)All sectors
-.2-.1
0.1
.2El
astic
ity o
f int
angi
ble
capi
tal s
hare
to c
ourt
dela
y
Inte
rnal
Fun
d
Exte
rnal
vs
Inte
rnal
Fun
d
(f)Intangible intensive
-.2-.1
0.1
.2El
astic
ity o
f int
angi
ble
capi
tal s
hare
to c
ourt
dela
y
Inte
rnal
Fun
d
Exte
rnal
vs
Inte
rnal
Fun
d
Figure depicts the differential impact of court delays on sectors that dependon internal finance(sectors that has lower than median dependency on ex-ternal finance) and external finance (above median dependency on exter-nal finance). It depicts he estimates for three different samples: All sectors,sectors with low share of intangible capital and sectors with high share ofintangible capital. The figure visualize the results of the table 4.
Figure 6
Court delays and firm’s debt overhang
-.6-.4
-.20
.2.4
.6El
astic
ity o
f cap
ital t
o co
urt d
elay
LowLeverage
MediumLeverage
HighLeverage
(a) Log(Capital)-.0
8-.0
40
.04
.08
Elas
ticity
of i
ntan
gibl
e ca
pita
l sha
re to
cou
rt de
lay
LowLeverage
MediumLeverage
HighLeverage
(b) Share of intangible capital
Figure depicts the differential impact of court delays across distribution ofleverage ratio . It depicts he estimates for subsamples of firms in top middleand bottom tertiles of leverage ratio in 2012 before the reform took place.The figure visualize the results of the table 5.
Share of intangible capital 2,200,585 0.187 0.271 0.000 1.000
Panel B : Court level variables
Log(Length of proceedings) 2,190,843 7.814 0.377 4.875 10.728
Log(Length of proceedings) in reformed area 312,556 7.812 0.402 6.553 9.575
Log(Length of proceedings) in non-reformed area 1,878,287 7.814 0.373 4.875 10.728
Reformed 2,190,987 0.143 0.350 0.000 1.000
Impact of the reform 2,190,987 0.000 0.043 -0.696 1.215
Impact of the reform × Post reform (year≥2013) 1219201 0.001 0.058 -0.696 1.215
Note: log(MRPK) is the logarithm of marginal return product of capital, log(Capital) is thelogarithm of capital stock deflated at 2010 prices of two digit industry, Log(Capital Inten-sity) is the capital per employee. "Log(Length of proceedings) "is a continuous variable thatestimates the logarithm of bankruptcy proceedings in days . Reformed is a dummy vari-able that takes 1 if firm located in the reformed judicial districts and zero otherwise.Finally,"Impact of the reform" estimates the the change in bankruptcy trial length that caused bythe reform t. It is calculated as follows
Impact of the reform = log(Length of proceedings)2012,2011Old district(Actual)
− log(Length of proceedings)2012,2011New district(Hypothetical)
39
Table 2
The reformed judicial districts: Actual length of proceedings in preformedcourts and hypothetical length of proceedings in new courts.
Region Province Pre reformjudicial district
Pos treformjudicial district
2011-2012ActualPrereform courts
2011-2012HypotheticalNew courts
Nord Ovest
Liguria Genova Chiavari Genova 2159 1825Liguria Genova Genova Genova 1787 1825Liguria Imperia Imperia Imperia 5707 4986Liguria Imperia Sanremo Imperia 4782 4986Lombardia Cremona Crema Cremona 2138 2555Lombardia Cremona Cremona Cremona 2767 2555Lombardia Pavia Pavia Pavia 2096 2388Lombardia Pavia Vigevano Pavia 2132 2388Lombardia Pavia Voghera Pavia 3486 2388Piemonte ssandria Acqui Terme Alessandria 2476 2417Piemonte ssandria Alessandria Alessandria 2877 2417Piemonte ssandria Tortona Alessandria 1373 2417Piemonte Asti Asti Asti 1802 1763Piemonte Cuneo Alba Asti 1736 1763Piemonte Cuneo Cuneo Cuneo 2581 2247Piemonte Cuneo Mondovi’ Cuneo 2646 2247Piemonte Cuneo Saluzzo Cuneo 1506 2247Piemonte Torino Pinerolo Torino 1641 1876Piemonte Torino Torino Torino 1892 1876Piemonte ssandria Casale Monferrato Vercelli 1732 2813Piemonte Vercelli Vercelli Vercelli 4906 2813
Note: OLS in all columns in all panels. The dependent variable, log(MRPK) , in column (1),
is a continuous variable that estimates the logarithm of marginal return product of capital
at each year for each firm. The dependent variable, log(Capital), in columns (2), is a con-
tinuous variable that estimates the logarithm of capital stock deflated at 2010 prices of two
digit industry at each year for each firm. The dependent variable,Log(Capital Intensity), in
column (3) measures the capital per employee. The dependent variable, share of intangible
capital , in column (4), is a continuous variable that estimates the share of intangible asset
in total fixed capital at each year for each firm. Reformed area is a dummy variable that
takes the value of one for all firms that are located in the districts that covered by the reform
and zero otherwise.
45
REFERENCES
Acemoglu, D. (2005). Constitutions, politics, and economics: A review essayon persson and tabellini’s the economic effects of constitutions. Journal ofEconomic Literature 43(4), 1025–1048.
Acemoglu, D. and S. Johnson (2005). Unbundling institutions. Journal ofpolitical Economy 113(5), 949–995.
Acharya, V. V., T. Eisert, C. Eufinger, and C. Hirsch (2018). Real effects ofthe sovereign debt crisis in europe: Evidence from syndicated loans. TheReview of Financial Studies 31(8), 2855–2896.
Almeida, H. and M. Campello (2007). Financial constraints, asset tangibility,and corporate investment. The Review of Financial Studies 20(5), 1429–1460.
BAE, K.-H. and V. K. Goyal (2009). Creditor rights, enforcement, and bankloans. The Journal of Finance 64(2), 823–860.
Bahaj, S., G. Pinter, A. Foulis, and P. Surico (2019). Employment and thecollateral channel of monetary policy.
Beck, T., R. Levine, and A. Levkov (2010). Big bad banks? the winnersand losers from bank deregulation in the united states. The Journal of Fi-nance 65(5), 1637–1667.
Benmelech, E., N. K. Bergman, and A. Seru (2011). Financing labor. Techni-cal report, National Bureau of Economic Research.
Bentolila, S., M. Jansen, and G. Jiménez (2018). When credit dries up: Joblosses in the great recession. Journal of the European Economic Associa-tion 16(3), 650–695.
Berkowitz, D., C. Lin, and Y. Ma (2015). Do property rights matter? evidencefrom a property law enactment. Journal of Financial Economics 116(3), 583–593.
Bertrand, M., E. Duflo, and S. Mullainathan (2004). How much should wetrust differences-in-differences estimates? The Quarterly journal of eco-nomics 119(1), 249–275.
46
Borisova, G. and J. R. Brown (2013). R&d sensitivity to asset sale proceeds:New evidence on financing constraints and intangible investment. Journalof Banking & Finance 37(1), 159–173.
Braun, M. (2005). Financial contractability and asset hardness. Available atSSRN 2522890.
Brown, J. R., J. A. Cookson, and R. Z. Heimer (2016). Law and finance mat-ter: Lessons from externally imposed courts. The Review of Financial Stud-ies 30(3), 1019–1051.
Cerqueiro, G., S. Ongena, and K. Roszbach (forthcoming). Collateral dam-age? on collateral, corporate financing and performance. Journal of Finan-cial Intermediation.
Cette, G., J. Fernald, and B. Mojon (2016). The pre-great recession slowdownin productivity. European Economic Review 88, 3–20.
Chaney, T., D. Sraer, and D. Thesmar (2012). The collateral channel: Howreal estate shocks affect corporate investment. American Economic Re-view 102(6), 2381–2409.
Chen, H. and G. Manso (2017). Macroeconomic risk and debt overhang.Review of Corporate Finance Studies 6(1), 1–38.
Chodorow-Reich, G. (2014). The employment effects of credit market dis-ruptions: Firm-level evidence from the 2008–9 financial crisis. The Quar-terly Journal of Economics 129(1), 1–59.
Claessens, S. and L. Laeven (2003). Financial development, property rights,and growth. The Journal of Finance 58(6), 2401–2436.
Corrado, C., C. Hulten, and D. Sichel (2005). Measuring capital and tech-nology: an expanded framework. In Measuring capital in the new economy,pp. 11–46. University of Chicago Press.
Corrado, C., C. Hulten, and D. Sichel (2009). Intangible capital and us eco-nomic growth. Review of income and wealth 55(3), 661–685.
Corrado, C. A. and C. R. Hulten (2010). How do you measure a" technolog-ical revolution"? American Economic Review 100(2), 99–104.
47
Demirgüç-Kunt, A. and V. Maksimovic (1998). Law, finance, and firmgrowth. The Journal of Finance 53(6), 2107–2137.
Demmou, L., I. Stefanescu, and A. Arquie (2019). Productivity growth andfinance: The role of intangible assets-a sector level analysis.
Djankov, S., O. Hart, C. McLiesh, and A. Shleifer (2008). Debt enforcementaround the world. Journal of political Economy 116(6), 1105–1149.
Djankov, S., R. La Porta, F. Lopez-de Silanes, and A. Shleifer (2003). Courts.The Quarterly Journal of Economics 118(2), 453–517.
Djankov, S., C. McLiesh, and A. Shleifer (2007). Private credit in 129 coun-tries. Journal of financial Economics 84(2), 299–329.
Duval, R., G. H. Hong, and Y. Timmer (2019). Financial frictions and thegreat productivity slowdown. The Review of Financial Studies 33(2), 475–503.
Fabbri, D. (2010). Law enforcement and firm financing: Theory and evi-dence. Journal of the European Economic Association 8(4), 776–816.
Giacomelli, S. and C. Menon (2016). Does weak contract enforcement af-fect firm size? evidence from the neighbour’s court. Journal of EconomicGeography 17(6), 1251–1282.
Giannetti, M. (2003). Do better institutions mitigate agency problems? evi-dence from corporate finance choices. Journal of Financial and QuantitativeAnalysis 38(1), 185–212.
Gopinath, G., S. Kalemli-Özcan, L. Karabarbounis, and C. Villegas-Sanchez(2017). Capital allocation and productivity in south europe. The QuarterlyJournal of Economics 132(4), 1915–1967.
Guiso, L., P. Sapienza, and L. Zingales (2013). The determinants of attitudestoward strategic default on mortgages. The Journal of Finance 68(4), 1473–1515.
Hall, R. E. (2001). The stock market and capital accumulation. AmericanEconomic Review 91(5), 1185–1202.
Haselmann, R., K. Pistor, and V. Vig (2009). How law affects lending. TheReview of Financial Studies 23(2), 549–580.
48
Hennessy, C. A. (2004). Tobin’s q, debt overhang, and investment. The Jour-nal of Finance 59(4), 1717–1742.
Hsieh, C.-T. and P. J. Klenow (2009). Misallocation and manufacturing TFPin China and India. The Quarterly journal of economics 124(4), 1403–1448.
Jappelli, T., M. Pagano, and M. Bianco (2005). Courts and banks: Effectsof judicial enforcement on credit markets. Journal of Money, Credit, andBanking 37(2), 223–244.
Kalemli-Ozcan, S., L. Laeven, and D. Moreno (2018). Debt overhang,rollover risk, and corporate investment: Evidence from the european cri-sis. Technical report, National Bureau of Economic Research.
Kalemli-Ozcan, S., B. Sorensen, C. Villegas-Sanchez, V. Volosovych, andS. Yesiltas (2015). How to construct nationally representative firm leveldata from the orbis global database. Technical report, National Bureau ofEconomic Research.
Kosova, R. (2010). Do foreign firms crowd out domestic firms? evidencefrom the czech republic. The Review of Economics and Statistics 92(4), 861–881.
Kremer, M. (1993). The o-ring theory of economic development. The Quar-terly Journal of Economics 108(3), 551–575.
Kumar, K. B., R. G. Rajan, and L. Zingales (1999). What determines firmsize? Technical report, National Bureau of Economic Research.
La Porta, R., F. Lopez-de Silanes, and A. Shleifer (2008). The economic con-sequences of legal origins. Journal of economic literature 46(2), 285–332.
La Porta, R., F. Lopez-de Silanes, A. Shleifer, and R. Vishny (2000). Investorprotection and corporate governance. Journal of financial economics 58(1-2),3–27.
La Porta, R., F. Lopez-de Silanes, A. Shleifer, and R. W. Vishny (1997). Legaldeterminants of external finance. The journal of finance 52(3), 1131–1150.
La Porta, R. L., F. Lopez-de Silanes, A. Shleifer, and R. W. Vishny (1998).Law and finance. Journal of political economy 106(6), 1113–1155.
49
Laeven, L. and C. Woodruff (2007). The quality of the legal system, firmownership, and firm size. The Review of Economics and Statistics, 601–614.
Lamont, O. (1995). Corporate-debt overhang and macroeconomic expecta-tions. The American Economic Review, 1106–1117.
Lang, L., E. Ofek, and R. Stulz (1996). Leverage, investment, and firmgrowth. Journal of financial Economics 40(1), 3–29.
Larrain, M. and S. Stumpner (2017). Capital account liberalization and ag-gregate productivity: The role of firm capital allocation. The Journal ofFinance 72(4), 1825–1858.
Lerner, J. and A. Schoar (2005). Does legal enforcement affect financial trans-actions? the contractual channel in private equity. The Quarterly Journal ofEconomics 120(1), 223–246.
Lu, Y., Z. Tao, and L. Zhu (2017). Identifying fdi spillovers. Journal of Inter-national Economics 107, 75–90.
Malgouyres, C., T. Mayer, and C. Mazet-Sonilhac (2019). Technology-induced trade shocks? evidence from broadband expansion in france.
Manova, K. (2013). Credit constraints, heterogeneous firms, and interna-tional trade. Review of Economic Studies 80(2), 711–744.
Matsuyama, K. (2007). Credit traps and credit cycles. American EconomicReview 97(1), 503–516.
Midrigan, V. and D. Y. Xu (2014). Finance and misallocation: Evidence fromplant-level data. American economic review 104(2), 422–58.
Moscarini, G. and F. Postel-Vinay (2012). The contribution of large and smallemployers to job creation in times of high and low unemployment. TheAmerican Economic Review 102(6), 2509–2539.
Myers, S. C. (1977). Determinants of corporate borrowing. Journal of financialeconomics 5(2), 147–175.
Occhino, F. and A. Pescatori (2015). Debt overhang in a business cyclemodel. European Economic Review 73, 58–84.
Pezone, V. (2018). The real effects of judicial enforcement.
50
Ponticelli, J. and L. S. Alencar (2016). Court enforcement, bank loans, andfirm investment: evidence from a bankruptcy reform in brazil. The Quar-terly Journal of Economics 131(3), 1365–1413.
Qian, J. and P. E. Strahan (2007). How laws and institutions shape financialcontracts: The case of bank loans. The Journal of Finance 62(6), 2803–2834.
Rajan, R. and L. Zingales (1998). Financial dependence and growth. Ameri-can Economic Review 88(3), 559–86.
Rodano, G., N. Serrano-Velarde, and E. Tarantino (2016). Bankruptcy lawand bank financing. Journal of Financial Economics 120(2), 363–382.
Rosen, S. (1982). Authority, control, and the distribution of earnings. TheBell Journal of Economics, 311–323.
Schiantarelli, F., M. Stacchini, and P. E. Strahan (2016). Bank quality, judicialefficiency and borrower runs: Loan repayment delays in Italy.
