ESSAYS ON MACRO-FINANCE AND INNOVATION By Mehmet Furkan Karaca A DISSERTATION Submitted to Michigan State University in partial fulfillment of the requirements for the degree of Economics—Doctor of Philosophy 2023 ABSTRACT This dissertation consists of two chapters of my work on macro-finance and innovation. In particular, it studies the impact of the dynamic process of credit reallocation on aggregate innovative activities. The first chapter introduces the main focus of my dissertation. In addition, it reviews the literature and discusses the contribution of this dissertation. The next chapter builds a model to draw out theoretical predictions. In the model economy, borrowing firms choose whether to innovate or retain a mature technology, while lenders decide their allocation of credit. The credit market is characterized with financial and matching frictions and investigates the consequences of lenders’ credit reallocation decisions on borrowers’ innovation choices. We posit that the innovation process is time consuming (e.g. due to the length of R&D projects). The different amount of time needed for production with the new and old technology exposes lenders to a liquidity risk. The analysis shows that lenders tend to reallocate credit when they face liquidity risks. We show that an intensification of the credit reallocation process improves the matching between lenders and innovative firms but, overall, it disrupts innovation activities. The final chapter empirically investigates the impact of credit reallocation on innovation and tests the predictions from the model. We use a novel data set on bank balance sheets and the number of patents in Italian (a bank-centered country) local markets (provinces) during a period of great economic growth and tighter banking regulation. We construct measures of credit reallocation following the established literature on job reallocation and examine their effect on innovation. To address the concerns about the endogeneity of credit reallocation in the provinces, we exploit indicators of the geographical diversity of the 1936 Italian Banking regulation. We then estimate a two-stage model that in the first stage projects the rate of credit reallocation in a province onto an indicator of tightness of the banking regulation in the province and in the second stage projects the measure of innovation (the number of patents) onto the value of credit reallocation in the province defined by the tightness of local banking regulation. Consistent with the predictions of the model, we find that an increase in credit reallocation depresses innovative activity while aggregate credit growth helps to expand it. Furthermore, we show that our results are robust across empirical specifications, and carry through when controlling for a broad battery of province characteristics or altering the estimation period. To my family and Serap, my better half. iv ACKNOWLEDGMENTS I am deeply indebted to many others who helped me throughout graduate school. I will never be able to fully express my gratitude to those who helped me complete my studies. I’d like to start by thanking my advisor and mentor, Raoul Minetti. I have continuously benefited from his insightful comments, endless support, and valuable advice. He always provided the motivation I needed to move forward and taught me how to think as an inde- pendent researcher. I will be eternally grateful for your help. My deep appreciation goes to Luis Araujo for generously sharing his knowledge and offering his support and guidance. My sincere gratitude goes to Qingqing Cao for her extensive support and extremely valuable suggestions. Luis and Qingqing greatly helped me improve myself and my research. I would also like to thank my external committee member Andrei Simonov for providing help in time of need and taking the place of my former committee member Zsuzsanna Fluck who sadly passed away and will be dearly missed. Further, I would like to thank Pierluigi Murro for his support on my research and job market paper. His generous guidance and suggestions on utilizing data from Italy immensely helped my research. I have been truly fortunate to be surrounded by amazing people during my time at MSU. Of these I’d like to thank Yogeshwar Bharat, David Hong, Giacomo Romanini, and Nick Rowe for their friendship and extensive discussions about research over the years. I would like to thank my friends for their continuous support. Last but not least, I would like to thank my parents, Gülay and Yılmaz, for their nev- erending support throughout my life. Without their help, I would not be able to pursue my dreams. I further thank my brothers, Erkam and Hakan for being a constant source of support in my life. Finally, I would like to thank my love, Serap, for her unconditional support and tenacious faith in me. v TABLE OF CONTENTS CHAPTER 1 INTRODUCTION . . . . . . . . . . . . . . . . . . . . . . . . . . . 1 1.1 Prior Literature . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4 CHAPTER 2 A THEORETICAL MODEL OF CREDIT REALLOCATION AND INNOVATION . . . . . . . . . . . . . . . . . . . . . . . . . 7 2.1 Agents, Goods, and Technology . . . . . . . . . . . . . . . . . . . . . . . 7 2.2 Credit . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 8 2.3 Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 20 CHAPTER 3 MEASURING THE EFFECT OF CREDIT REALLOCATION ON INNOVATION . . . . . . . . . . . . . . . . . . . . . . . . . . 22 3.1 Data and Methodology . . . . . . . . . . . . . . . . . . . . . . . . . . . . 22 3.2 The Empirical Model . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 30 3.3 Main Estimation Results . . . . . . . . . . . . . . . . . . . . . . . . . . . 33 3.4 Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 36 BIBLIOGRAPHY . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 38 APPENDIX A APPENDIX FOR A THEORETICAL MODEL OF CREDIT REALLOCATION AND INNOVATION . . . . . . . . . . . . . . 42 APPENDIX B APPENDIX FOR MEASURING THE EFFECT OF CREDIT REALLOCATION ON INNOVATION . . . . . . . . . . . . . . . 46 vi CHAPTER 1 INTRODUCTION The study of the allocation of resources in an economy often focuses on the distribution of labor and physical capital across firms. There is growing evidence that the reallocation of jobs (Davis and Haltiwanger (1992) and Davis et al. (1996)) and physical capital (Eisfeldt and Rampini (2006) and Eisfeldt and Shi (2018)) play a crucial role in economic growth. In contrast with the rich evidence on the importance of financial aggregates in boosting economic growth, the reallocation of financial resources is so far inadequately examined. Moreover, the interaction between the reallocation of financial resources and aggregate eco- nomic activity is under-explored. In particular, we know very little about the relationship between the reallocation of financial resources and innovation activities. The literature on finance and innovation provides evidence that well-functioning finan- cial markets can boost technological change. In addition, particularly in economies with underdeveloped stock and bond markets, banks have been shown to play a critical role in financing firms’ innovation. The allocation of bank credit can thus substantially impact in- novative activities due to differences in firms’ access to credit. Hence, credit reallocation can be an important channel through which aggregate shocks can affect innovation activities and ultimately, influence real economic activity. In light of these considerations, several questions arise. How does credit reallocation across firms affect firms’ innovation activities? Does a more intense credit reallocation foster innovation or, rather hinder innovation due potential financial instability? This dissertation consisting two chapters takes a step towards addressing these questions. First, we employ a model to investigate the consequences of lenders’ credit reallocation deci- sions on borrowers’ innovation choice. In the model economy, borrowing firms choose whether to innovate or retain a mature technology, while lenders decide their allocation of credit. We model the credit market as a decentralized one, characterized by matching frictions between borrowers and lenders. We posit that the innovation process is time consuming (e.g. due 1 to the length of R&D projects) and thus it takes more time to produce with the new tech- nology compared with the readily available (old) technology. Lenders and borrowers sign debt contracts promising a repayment to the lender in the event of production success. The different amount of time needed for production with the new and old technology exposes lenders to a liquidity risk (caused by a financial shock). Therefore, lenders have an incentive to terminate their lending agreements early (if they are lending to an innovating firm) and reenter the credit market to find a more profitable borrower (credit reallocation). Our anal- ysis shows that lenders tend to reallocate credit when they face liquidity risks. We obtain that in a region of parameter space, our economy exhibits multiple equilibria: the amount of innovative firms in the economy affects the credit reallocation decision of lenders, and in turn, lenders’ credit reallocation choices influence borrowers’ innovation choices. Model calibration reveals that overall an increase in the intensity of credit reallocation (as driven by easing of the credit matching process) disrupts innovative activities. Second, we test the predictions from the model using granular data from the Italian local markets. To investigate the effect of credit reallocation on innovation, we need an environ- ment in which firms heavily depend on bank financing. We also need a time period during which local markets experience a significant heterogeneity in the intensity of innovation ac- tivities. With these goals in mind, we pick Italy (a bank-centered country) and the boom years post Second World War (1950-1963) as the object of our investigation.1 We follow Herrera et al. (2011) to measure province-level credit reallocation. Their work on credit flow measures utilizes the methodology for measuring job flows developed by Davis and Haltiwanger (1992) and Davis et al. (1996). The measurement of credit reallocation within provinces relies on detailed bank balance sheets data. We use bank-level balance sheet data from the Historical Archive of Credit in Italy (ASCI) following Natoli et al. (2016). The data cover yearly balance sheets of nearly 600 banks for the time period under 1 The financial system can be characterized as bank-dependent since the stock market in Italy does not play a crucial role in financing firms’ activities. The exact time span is determined by the bank balance sheet data. 2 our analysis. We measure innovation using the number of patents in each province for each year.2 We complement our main data with information on province characteristics (such as financial development, education etc.) that might affect innovation activities. This data is manually extracted from historical censuses held in 1951, 1961, and 1971 using scanned census documents. To assuage concerns about the endogeneity of credit reallocation in the provinces, we exploit indicators of the provincial tightness of the 1936 Italian Banking regulation. Guiso et al. (2004a) and Guiso et al. (2004b) demonstrate that the banking regulation put in effect in 1936 created substantial heterogeneity in the degree of dynamism of provincial credit markets. Provincies where the regulation was tighter experience lower flows of entry, exit, and reallocation across banks than provinces with less tighter regulation. We then estimate a two-stage model that in the first stage projects the rate of credit reallocation in a province onto an indicator of tightness of the banking regulation in the province and in the second stage projects the measure of innovation (the number of patents) onto the value of credit reallocation in the province defined by the tightness of local banking regulation. The results reveal that the number of patents decreases as credit reallocation intensifies while higher credit growth increases the number of patents. Hence, in line with the model credit reallocation turns out to have a negative impact on the number of patents. The effects are sizable. A one percentage point increase in credit reallocation leads to a 9.8 percent decline in the number of patents. On the other hand, a one percentage point increase in credit growth causes a 1 percent increase in the number of patents. Our results are robust across empirical specifications, and carry through when controlling for a broad battery of province characteristics or altering the estimation period. The rest of the dissertation is organized as follows. Section 1.1 summarizes the related literature. The second chapter explains the model to study the effect of credit reallocation on firms’ innovation. The third chapter provides the result of the empirical investigation. 2 Patent data comes from the Italian Patent Office (IPO) and the European Patent Office’s (EPO) PATSTAT database which includes the international patents. Please see Bianchi and Giorcelli (2020) for the details of patent data. 3 In chapter 3, Section 3.1 describes the data, the credit reallocation measures, and summary statistics. Section 3.2 provides details about the estimation process, while Section 3.3 present our main empirical results. Details on the data, proofs and additional robustness tests are relegated to the Appendices. 1.1 Prior Literature This dissertation is related to three strands of literature. The first studies how finan- cial markets affect the allocation and reallocation of physical capital and financial resources. Eisfeldt and Rampini (2006) and Chen and Song (2013) investigate the impact of financial frictions on the allocation and reallocation of physical capital across firms. Galindo et al. (2007) study the effect of financial shocks on the allocation of physical investment. Eisfeldt and Shi (2018)3 argue that the empirical literature on capital reallocation demonstrates two main results. First, capital reallocation occurs during either high productivity or high equity market valuation periods. Second, capital flows from less productive firms to more produc- tive ones. Firm-level differences in technology, risk characteristics, and uncertainty cause a dispersion in firm-level productivity. As a result, firms can choose between the reallocation of existing capital and the production of new capital to use as a form of investment. Lanteri (2018) provides a microfoundation for the interplay between new and used capital and shows that used capital prices are more volatile and procyclical than prices of new capital. The reallocation of financial resources remains overlooked in the literature. The credit reallocation happens in two analogous ways. Either a firm increases the number of credit relationships to multiple banks (reallocation across banks) or a bank expands its lending to multiple firms (reallocation across firms). Theoretical evidence provides results stemming from different frictions. The negative effect of credit reallocation is that having multiple creditors reduces the available funding, but relationship lending increases available funds to a firm due the frictions in the credit market (Petersen and Rajan (1994)). On the contrary, expanding credit relationships to multiple banks decreases the probability of failure for a 3 See for more details about capital reallocation literature. 4 project due to monitoring effect (Detragiache et al. (2000)). Furthermore, lending com- petition forces lenders to reallocate credit toward more captured borrowers due to higher expected profit (Dell’Ariccia and Marquez (2004)). Banks can also play a ‘Schumpeterian role’ in the economy. Credit rellocation can be interpreted as a creative destruction process. Banks reallocate credit from firms with poor prospects to expanding and successful firms (Keuschnigg and Kogler (2020)). There is growing empirical evidence on the dynamics of credit reallocation. Dell’Ariccia and Garibaldi (2005) provide evidence that inter-bank loan reallocation is intense using data from U.S. Banks’ Call Report Files. Chang et al. (2010) find that there is no correlation be- tween credit reallocation and regional economic growth in China from 1991 to 2005. Herrera et al. (2011) lay out stylized facts on credit reallocation across U.S. businesses. Credit real- location is slightly procyclical, substantially volatile and intense. Additionally, it is mainly across firms in similar industries, geography and size. Hyun (2016) finds that debt financing of large firms is mostly affected by a national factor, while, in contrary, a regional factor plays a crucial role for small firms in Korea from 1984 to 2013. Hyun and Minetti (2019) reveal that credit reallocation across Korean firms intensifies and become more procyclical after the 1997 crisis. They conclude that intensified credit reallocation enhances firm efficiency. De Jonghe et al. (2020) show that banks reallocate credit toward low-risk firms, to sectors where they have more specialization, or to sectors in which their market share is high after a negative funding shock. The second strand of literature studies the effect of finance on technological change and innovation. Hall and Lerner (2010), Brown et al. (2012), and Kerr and Nanda (2015)4 con- clude that well-functioning financial markets can boost technological innovation. Caballero and Hammour (1994) study how innovative production can replace old technologies during a recession. Caballero and Hammour (2005) find that credit frictions can cause an excess destruction of production units during a recession. The recovery in the aftermath of the 4 See for a very detailed review of literature on financing innovation. 5 recession is through an attenuation in the product destruction rate rather than a rise in the creation rate. Garcia-Macia (2017) investigates the effects of a crisis on the investment deci- sion on tangible and intangible capital by heterogeneous firms. In case of a default, lenders cannot collect intangible capital easily due to its structure. This feature of intangible capital increases the borrowing cost. As a result, firms reduce their investment in intangible capital due to higher financing costs amplifying the negative effects of the crisis. Furthermore, as in Wang (2017) show that firms with initially high knowledge capital tend to save more to increase their financial assets which they can pledge as collateral. In addition, firms can also increase their investment in pledgeable assets to protect themselves from the negative effects of financial crisis. Araujo et al. (2019) investigate the effect of a credit crunch on a firm’s technology choice. Collateral-poor firms loose access to credit due to a contraction in collateral value. On the contrary, collateral-rich firms gain easy access to credit market which fosters innovation. Entry to credit market can play an important role as in Malamud and Zucchi (2019). Costly access to external financing discourages innovative firms’ entry disrupting creative destruction. Lastly, this dissertation is also related to the literature on how banking regulations impact growth and innovation. Differences in local financial development can substantially impact lending practices and growth (Jayaratne and Strahan (1996), Guiso et al. (2004a), Dehejia and Lleras-Muney (2007)). Additionally, a frequent result in the literature is that banking regulations may hinder innovation due to the long-term nature of the innovation process. 6 CHAPTER 2 A THEORETICAL MODEL OF CREDIT REALLOCATION AND INNOVATION This chapter describes a general equilibrium model of the credit market where borrowers can retain a mature technology or adopt a new technology. We then explore the impact of lenders’ credit reallocation decisions on borrowers’ technology choice. Our objective is to analyze how credit reallocation affects the innovation process represented by the operation of a new technology. 2.1 Agents, Goods, and Technology Consider a three-period economy (t = 1, 2, 3). There is a final good and distinct, indivisi- ble assets (machines) that produce the final good. The population consists of a continuum of risk neutral agents who derive utility from their period 3 consumption of final good. There are two groups of agents each of measure one: unskilled (u) and skilled (s) agents with different initial endowments. Unskilled agents start with one machine while skilled agents have no initial endowment. Besides storage, there are two technologies available for production: new and old tech- nology. The new technology represents the innovation process. We assume that only skilled agents can produce with the new technology. The two groups of agents differ in their produc- tivity. The probability of success for the skilled agents, λs , is higher than for the unskilled agents, λu (λs > λu ), and these probabilities are independent of the technology. For simplicity, machines cannot be used again once the production process ends. The new technology takes more time to yield production than the old technology. In particu- lar, production with the old technology takes one period, while production with the new technology takes two periods. Returns differ between the new and the old technology. Innovation offers a productivity edge, yielding a higher amount of final good: the old technology yields an output ys of final good while the new technology returns ys (1 + γ) final good. Furthermore, if an unskilled 7 agent does not transfer her machine and successfully implements the old technology, the machine yields a lower return yu . In other words, skilled agents who choose to produce with the new technology obtain the highest amount of output and skilled agents who choose to produce with the old technology obtain a higher amount of output than unskilled agents producing with the old technology (ys (1 + γ) > ys > yu ).1 2.2 Credit Skilled and unskilled agents can effectively act as borrowers and lenders in the economy, respectively. If an unskilled agent lends a machine to a skilled agent but production fails, the machine is returned to the lender. If the machine was used in the old technology, a salvage value of a can be recovered by the lender. The salvage value from a failed new technology production is instead normalized to 0 for simplicity. Intuitively, machines accompanying new technologies are likely to be firm specific and illiquid. This makes it hard for lenders to liquidate them compared to machines used in old technologies. Given the long-term nature of the new techology, it is more exposed to interim liquidity shocks. In the event of such an unfavorable shock, an unskilled agent who lent to an inno- vating skilled agent can end the credit relationship and reallocate the machine to avoid a continuation cost σ. 2.2.1 Market Frictions We follow Kiyotaki and Wright (1993) to introduce market frictions: the inability to match an agent who wants to transfer her machine to another agent who wants to use it. We capture this friction by introducing an exogenous parameter x that captures the level of specilization in the economy. Particularly, x denotes the proportion of machines that can be used by skilled agents and the proportion of skilled agents who can use a specific machine. In addition, unskilled and skilled agents meet in pairs under a uniform random matching 1 This assumption will make the trade of machines meaningful. It also follows the typical assumption from the search literature that an agent is not satisfied with her endowment, which motivates trade; see. e.g., Kiyotaki and Wright (1993). 8 technology. There are two markets where agents meet. The first one opens in period 1. Unskilled agents (the lenders) lend their machines to skilled agents (the borrowers). We call this market the credit market. The second market opens in period 2 after the lenders observe borrowers’ technology choice and any unfavorable shocks. The specification of the production technology provides a rationale for credit reallocation in period 2. We call this market the reallocation market. If a borrower chooses the new technology and an interim negative shock occurs, the lender have the opportunity either to break the credit relationship and reallocate their machines or to continue with the innovation process facing the continuation cost σ. If the lenders reallocate their machines, the borrowers can only produce with old technology which takes one period to complete. In particular, the lenders are informed about the features of the innovation process and can realize a liquidity shock or an exogenous shock which hinders the innovation process in period 2. Therefore, a positive measure of lenders have the incentive to reallocate their machines in order to make themselves prone to these shocks. The existence of the reallocation market is ensured with this rationale. If no shock is realized in period 2, the reallocation market is not formed and innovation process continues without any interruption. 2.2.2 Contractual Structure Suppose that an unskilled agent (the lender) transfer her machine to a skilled agent (the borrower) in period 1 when they meet in a pair in the credit market. The two agents sign a contract to formalize transactions between the lenders and the borrowers. We only consider debt contracts to focus on credit reallocation. A debt contract specifies the upfront payment to the lender, property rights of the machine in case of failure, and the repayment in case of successful production. The debt contract requires no upfront payment to the unskilled agent, gives full property rights to the unskilled agent in the event of failure, and promises a repayment to the unskilled agent in the event of success. In other words, the unskilled agent lends her machine to the skilled one in exchange for future payments. This 9 agreement can be interpreted as a debt contract. In addition, we describe the repayments in terms of final goods. The repayment of new technology (yd (1 + γ)) is higher than the old technology (yd ) if the production is successful. The amount of repayment is lower than the return of successful production (ys (1 + γ) > yd (1 + γ) and ys > yd ). The unskilled agent (lender) has the option to break the credit relationship if the skilled agent (borrower) innovates (produces with new technology). The costly action that the unskilled agent needs to perform provides a rationale for credit reallocation. Contractually, a lender cannot be enforced to commit to continue the credit relationship that facilitates the innovation because the state of nature (or an unexpected unfavorable shock) cannot be contracted on as in Aghion and Bolton (1992) and Diamond and Rajan (2001). 2.2.3 Summary The timing and mechanisms of the model can be summarized in the following way. Fig- ure 2.1 illustrates the timing of events. Period 1. In the credit market, unskilled (lenders) and skilled (borrowers) agents meet in pairs under a uniform random matching technology. An unskilled agent can transfer the machine when she meets a skilled agent whose specialization matches with the type of her machine. In this case, they sign a debt contract. Then, the skilled agents choose whether to innovate (new technology) or not (old technology). Period 2. Skilled and unskilled agents producing with old technology are either successful or failed. Skilled agents incur a maintenance cost to prevent further depreciation of the ma- chine before it is returned to the unskilled agents in the event of failure. Skilled and unskilled agents who successfully produce with old technology have access to a storage technology that is used to preserve the payoffs for one period until the end (period 3). Machines that fail cannot be reused. The unskilled agents get a salvage value for the failed machines and use the storage technology to preserve the salvage value for one period until the end (period 3). The unskilled agents who lend their machines to innovating skilled agents observe the skilled agents’ technology choice. In addition, they realize if there is an unfavorable shock (a 10 liquidity shock etc.) that can hinder the innovation process. If no shock hits the economy, innovation process continues without interruption and the outcome is observed in period 3. If an unfavorable shock is realized, the unskilled agents can either continue with the innovation process facing an effort (continuation) cost or reallocate their machines to another skilled agent who can only produce with old technology. After the unskilled agents’ decision is observed the reallocation market is formed in period 2. The participants in this market are the unskilled agents who take back their machines and skilled agents whose production process with new technology is interrupted. The machines recalled early or before production process ends can be used by other agents. Agents meet in pairs under a uniform random matching technology in the reallocation market. All skilled agents are matched with an unskilled agent in the reallocation market because the amount of skilled and unskilled agents are equal and all machines can be used by a skilled agent. Agents can only produce with old technology after the reallocation market since there is only one period remaining. The outcome is realized in period 3. Period 3. Agents producing with old technology after the reallocation market and new technology are either successful or failed. In the event of failure, the unskilled agents do not get the salvage value as the machines fully depreciate in this period. Agents consume. 11 Period 1 Period 2 Period 3 ) sful (λ s Succes No shock (1 − β) No interruption e Shock (β) Lenders continue tinu Failed Con (1 − η) with cost σ (1 − λ Skilled s) Agents α) g y( sful (λ s ) olo Rea lloc Succes chn ate Shock (β) Te Reallocation w x (Matched Ne Lenders reallocate (η) Market Borrowers produce Skilled Agents) with old technology Failed Credit (1 − λ s) Market Ol d Te ch no λ s) log ed (1 − y (1 Fail − Unskilled α) Agents Suc 1 − x (Unmatched ce ssfu l (λ s) Use storage tech- Unskilled Agents) nology for 1 period Storage technology pays constant gross interest rate r λ u) ed (1 − Fail Suc cess ful ( λu ) Figure 2.1 Flow of Agents 12 2.2.4 Equilibrium To solve for the equilibrium we proceed in steps. First, we start with the reallocation market in period 2. Second, we analyze the credit market and then, finally, we characterize the equilibrium. 2.2.4.1 Reallocation Market Credit is reallocated in this market in period 2. Unskilled agents who reallocate their machines become sellers in this market. Skilled agents seek to obtain a machine to produce with old technology for the last period. The amount of credit reallocation is equal to Ω = xαβη (2.1) since we assume a unit continuum of both types of agents. It is the amount of machines that are reallocated by the unskilled agents (η) if an unfavorable shock is realized (β) after observing the skilled agents’ technology choice (α) among all matched skilled agents (x). The matching technology in the credit market (x) represents economy-wide matching technology. There are equal measure of agents in the credit market and, consequently, the measure of skilled and unskilled agents are the same in the reallocation market due to symmetry assumption. Additionally, all machines can be used by a skilled agent. Thus, x captures the matching in the reallocation market. Now, we can derive the value functions of the skilled and unskilled agents in the credit reallocation market as follows Vsn,R = xλs (ys − yd ) (2.2) Vun,R = xλs yd + (1 − x)λu yu (2.3) where Vsn,R is the value function of a skilled agent and Vun,R is the value function of an unskilled agent in the credit reallocation market. 13 2.2.4.2 Credit Market We now analyze the credit market, the choice between reallocation and continuation for the unskilled agents (lenders), and the technology choice of skilled agents (borrowers) at period 1. Denote by Ws the value function of a skilled agent and Wu the value function of an unskilled agent at the beginning of period 1 net endowments of final good. The value functions are n o Wu = Vo + x(1 − α)Vuo + xα(1 − β)Vun,C + xαβ max Vun,C − σ, Vun,R (2.4) n o Ws = x max Vsn , Vso (2.5) where for the unskilled agents, Vo is the value of the outside option (being unmatched), Vuo is the value of lending their machine to a skilled agent producing with old technology, Vun,C represents the continuation value of producing with new technology, Vun,R displays the value when they reallocate their machines. For the skilled agents, Vsn is the value of producing with new technology and Vso is the value of producing with old technology. The expected payoff of an unskilled agent, Wu , is the sum of the expected payoffs from the outside option (unmatched and producing on their own), lending to a skilled agent producing with old technology, lending to a skilled agent producing with new technology without an unfavorable shock to the economy, and the decision between continuation and reallocation in the event of an unfavorable shock is realized. Similarly, the expected payoff of a skilled agent depends on the choice between new and old technology. Next, we define the value functions mentioned above. Firstly, consider the value functions of the unskilled agent h i Vo = (1 + r) λu yu + (1 − λu )a (2.6) h i o Vu = (1 + r) λs yd + (1 − λs )a (2.7) Vun,C = λs (1 + γ)yd (2.8) 14 and, secondly, consider the value functions of the skilled agent Vso = (1 + r)λs (ys − yd ) (2.9) h i n n,C n,R n,C Vs = (1 − β)Vs + β ηVs + (1 − η)Vs (2.10) Vsn,C = λs (1 + γ)(ys − yd ) (2.11) where Vsn,R and Vun,R are as defined in the previous section. The next lemma formalizes the conditions for a credit relationship. Under the following conditions the unskilled agents prefer meeting skilled agents in the credit market. Lemma 1 In period 1, an unskilled agent will always prefer lending her machine to a skilled agent than producing on her own if (i) yu < a < yd and η 1+r (ii) σ < (1 − x)λu yu − . 1 − η β(1 − η) Proof. See the Appendix A.1. 2.2.4.3 Unskilled Agents’ Choice The expected payoff from continuation is higher than the expected payoff from realloca- tion as long as Vun,C − σ > Vun,R . Lemma 2 outlines the unskilled agents’ decision. Lemma 2 Suppose that an unskilled and a skilled agent meet at period 1. Then, at period 2, conditional on skilled agents’ technology choice, the unskilled agent will continue with the innovation process if and only if σ − (1 − x)[λs yd − λu yu ] γ> (2.12) λs y d and, they believe that a positive measure of the unskilled agents continues with the innova- tion process. Otherwise, the unskilled agents will reallocate their machines (disrupting the innovation process). Proof. See the Appendix A.1. 15 2.2.4.4 Skilled Agents’ Choice The expected payoff of new technology is higher than the expected payoff from old tech- nology as long as Vsn > Vso − µ. Lemma 3 characterizes the skilled agents’ technology choice. Lemma 3 Suppose that an unskilled and a skilled agent meet at period 1. Then, the skilled agents will innovate if and only if r + (1 − x)βη γ> (2.13) 1 − βη and, they believe that a positive measure of the skilled agents choose to innovate. Otherwise, the skilled agents will choose to produce with old technology. Substituting for the unskilled agents’ decision, η, one of the following cases is realized: (i) If γ < r, all of the skilled agents will only choose to produce with old technology. r + (1 − x)β (ii) If r ≤ γ ≤ , all of the skilled agents will choose either old technology or 1−β new technology, and there exists a γ ′ where the skilled agents will be indifferent between old and new technology. r + (1 − x)β (iii) If γ > , all of the skilled agents will only choose to produce with new 1−β technology. Proof. See the Appendix A.1. 2.2.4.5 Disruptive Credit Reallocation Credit reallocation negatively impacts innovation process. Combining the conditions from Lemmas 1-3, Lemma 4 characterizes the the conditions that facilitate the hindering effect of credit reallocation. Lemma 4 Credit reallocation disrupts the innovation process if r + (1 − x)βη σ (1 − x)[λs yd − λu yu ] <γ≤ − (2.14) 1 − βη λs yd λs yd Proof. See the Appendix A.1. 16 2.2.4.6 Equilibrium Characterization Now, we can present the equilibrium. There are two choices in the model: the tech- nology choice of skilled agents and the unskilled agents’ decision between reallocation and continuation. Let M denote the set of period 1 meetings between skilled and unskilled agents in the credit market at period 1. Define Ss as the choice of the skilled agent and Su as the choice of the unskilled agent. Consider a generic point i in this set and let Ss × Su be the profile of actions. We have Ss = {N, O} and Su = {C, R} where N and O represent the choice of new and old technology, respectively, for the skilled agent and C and R represent the choice of continuation and reallocation, respectively, for the unskilled agent. Now, define C(i, ss , su , v) as the outcome of ith meeting, where the skilled agent chooses ss , the unskilled agent chooses su and v = (α, η) is the distribution of skilled agents between two technologies and the distribution of unskilled agents between continuation and reallocation. Definition 1 A Nash equlibrium is a pair (C, v) such that: (i) In any meeting i ∈ M , agents’ choice c(i, ss , su , v) maximizes surplus. (ii) The aggregation of agents’ choices across meetings generates a distribution of skilled agents between two technologies and a distribution of unskilled agents between contin- uation and reallocation. Proposition 1 combines the results of Lemmas 1-4. It characterizes the distribution of skilled agents between two technologies, and, the distribution of unskilled agents between the choice of continuation and reallocation if the borrower chooses to produce with new technology in the event of an unfavorable shock to the economy. Proposition 1 (Distribution of Agents) Suppose that an unskilled agent (lender) and a skilled agent (borrower) meet at period 1. Regardless of the unskilled agents decision, (i) (no innovation) the skilled agents will not innovate if γ < r, and (ii) (innovation) the 17 r+(1−x)β r+(1−x)β skilled agents will innovate if γ > 1−β . If γ < γ ′ in the interval r ≤ γ ≤ 1−β , the skilled agents will not innovate regardless of the unskilled agents decision. Assuming γ > γ ′ , (i) (no disruption to innovation process) the skilled agents will innovate and the σ−(1−x)[λs yd −λu yu ] r+(1−x)β unskilled agents will continue if r ≤ λs yd <γ ≤ 1−β , and (ii) (disruption to innovation process) the skilled agents will innovate but the unskilled agents will reallocate σ−(1−x)[λs yd −λu yu ] r+(1−x)β their machines if r ≤ γ ≤ λs yd ≤ 1−β . Proof. See the Appendix A.1. Figure 2.2 displays the proposition. The intervals from the proposition presented on the figure. To summarize, the skilled agents will not innovate if γ < r and γ < γ ′ regardless of the unskilled agents’ choice. Hence, we get α = 0. The skilled agent will innovate if γ > γ ′ . r+(1−x)β The intervals matter for this case. If γ > 1−β the skilled agents will innovate regardless of the unskilled agents’ choice. We get α = 1. The disruptive effect of credit reallocation r+(1−x)β appears in the interval r ≤ γ ≤ 1−β . The innovation process will not be interrupted, if σ−(1−x)[λs yd −λu yu ] r+(1−x)β σ−(1−x)[λs yd −λu yu ] r+(1−x)β γ′ ≤ λs yd <γ ≤ 1−β or λs yd ≤ γ′ < γ ≤ 1−β . Again, we have α = 1. On the other hand, the unskilled agents’ decision will interrupt the innovation σ−(1−x)[λs yd −λu yu ] r+(1−x)β process if γ ′ < γ ≤ λs yd ≤ 1−β . Thus, we start with α = 1, but end up with α = 0. The choice of reallocating credit harms the innovation process as depicted in Figure 2.2. In conclusion, the analysis predicts that, as long as the productivity edge provided by innovation is not too small, agents will innovate. However, if the productivity edge is not big enough, the lenders will reallocate credit and disrupt the innovation process. 2.2.5 A Numerical Example In this section, we develop some numerical experiments that help further grasp the intu- ition behind the results of the model. Table 2.1 outlines the exercise. We will consider two cases in two different scenarios: high and low local financial development in an economic boom period or an economic downturn period. An economic boom period represents lower interest rate for the storage technology, the effort cost of lenders to continue will be lower, 18 Innovate (α) 1 0 r γ′ σ−(1−x)[λs yd −λu yu ] r+(1−x)β 1 Productivity λs yd 1−β edge (γ) Figure 2.2 Relationship between reallocation and innovation and the probability of a shock is lower. An economic downturn period indicates that the storage technology pays a higher interest rate, the continuation cost of lenders will be higher, and a shock is more likely to occur. We define local financial development depending on x which shows matching efficiency in the credit market. We fix technology parameters. The probability of success for the skilled agents is λs = 0.8. The probability of success for the unskilled agents is λu = 0.6. The return amount production yields is ys = 1.8 for the skilled agents and yu = 1.3 for the unskilled agents. The repayment in the event of successful production is yd = 1.5. The unskilled agents receives the salvage value a = 1.4 in period 2 if noninnovating skilled agents fail to produce. We calculate the thresholds for the productivity edge given the parameters. Column 4 of Table 2.1 presents the thresholds for the productivity edge. Given in Proposition 1, γ represent the lower bound of the interval below which the skilled agent will never innovate, γ is the upper bound after which the skilled agents will innovate regardless of the unskilled agents decision. γ̂ shows the cutoff point for the unskilled agents. Below γ̂ the unskilled agents will reallocate credit and above they will continue with the innovation process. Col- umn 5 displays the intervals for the skilled agents to innovate and the threshold for the unskilled agents’ decision. 19 Table 2.1 Value of Parameters Parameters Parameters Parameters Thresholds Cases Result (Technology) (Economy) (Credit Market) Panel A: Economic boom λs = 0.8 r = 0.1 High Local Financial Development λu = 0.6 β = 0.1 x = 0.8 γ = 0.1 (i)γ < γ < γ̂ Disruptive credit reallocation ys = 1.8 σ = 0.5 γ = 0.13 (ii)γ̂ < γ No interruption to innovation process yu = 1.3 γ̂ = 0.35 yd = 1.5 Low Local Financial Development a = 1.4 x = 0.2 γ = 0.1 (i)γ < γ < γ̂ Disruptive credit reallocation γ = 0.2 (ii)γ̂ < γ < γ No interruption to innovation process γ̂ = 0.14 Panel B: Economic downturn r = 0.3 High Local Financial Development β = 0.6 x = 0.8 γ = 0.3 (i)γ < γ < γ̂ Disruptive credit reallocation σ=1 γ = 1.05 (ii)γ < γ̂ < γ No interruption to innovation process γ̂ = 0.76 Low Local Financial Development x = 0.2 γ = 0.3 (i)γ < γ < γ̂ Disruptive credit reallocation γ = 1.95 (ii)γ < γ̂ < γ No interruption to innovation process γ̂ = 0.55 In an economic boom scenario, it is more likely to have higher productivity edge provided by innovation and better economic conditions. The lower bound of unconditional innovation choice of the skilled agents γ and the continuation threshold for the unskilled agents γ̂ are lower in the economic boom environment. The lenders will be more likely to continue with the innovation process. On the contrary, the productivity edge will be lower in an economic downturn and the lenders will be more likely to reallocate credit in the economic downturn environment. Furthermore, local financial development matters for the decisions of agents. In highly developed local financial markets, the lenders are more likely to reallocate and interrupt the innovation process. 2.3 Conclusion In this chapter, we study the impact of credit reallocation on innovation building a general equilibrium model. We employ a model to investigate the consequences of lenders’ credit re- allocation decisions on borrowers’ innovation choice. We show that lenders tend to reallocate credit when they face liquidity risks. The amount of innovative firms in the economy affects the credit reallocation decision of lenders, and in turn, lenders’ credit reallocation choices 20 influence borrowers’ innovation choices. A simple model calibration reveals that overall, an increase in the intensity of credit reallocation (as driven by easing of the credit matching process) disrupts innovative activities. 21 CHAPTER 3 MEASURING THE EFFECT OF CREDIT REALLOCATION ON INNOVATION This chapter test the predictions from the model in the previous chapter in an empirical setting. We investigate the impact of credit reallocation on innovation. First, we start with explaining data and methodology. Then, we provide the estimation results. 3.1 Data and Methodology In this section, we describe the data and the methodology for credit reallocation measures. We collect patent counts, bank-level credit flows, and province characteristics from census data. The bank loan data covers the period between 1890 and 1973 and the patent data is from 1950 to 2010 with a gap between 1963 and 1968. Thus, the final data set comprises the period between 1950 and 1963. However, including data from 1968 to 1973 does not change the results.1 We perform our investigation at the province level. A province is a unit of analysis very similar to a county in the US. In Italy, the relevant local market in banking is the province according to the Italian Antitrust authority. Additionally, the Bank of Italy used the same rule to define a local market concerning opening new branches and extending credit outside of a bank’s location. Therefore, we collect data at the province level. 3.1.1 Institutional Background In this paper, we study the effect of credit reallocation on innovation and technological change. To achieve that we require an environment in which firms heavily depend on bank financing and a time during which a great expansion in innovative activities was experienced. Picking Italy as the subject of the investigation provides abundant advantages in focusing on bank financing. The financial system can be characterized as bank-dependent since the stock market in Italy does not play a crucial role in financing firms’ activities. Hence, Italy provides a very useful environment to isolate the role of banks, in particular credit reallocation, in 1 The inclusion of extra data and results from this exercise is discussed in Section B.3. 22 fostering innovation. Overall, Italy traditionally has a financial system dominated by credit institutions (De Bo- nis et al. (2012)). The ratio of loans to deposits rose above one during the economic boom from 1958 to 1963. Bank loans and deposits reached 75% of GDP, and the total factor productivity growth was particularly exceptional from the 1950s to the mid-1970s, the so- called “Italian economic miracle” period.2 Thus, focusing on Italy during this period is very informative in terms of investigating the role of credit reallocation in boosting innovation. 3.1.2 Patent Data In the literature, R&D spending and patent counts are commonly used as two main measures of innovation. Even though each has advantages and disadvantages, we choose to use patent counts because R&D spending cannot tell us whether the innovation process is successful.3 We start with European Patent Office’s (EPO) PATSTAT database. However, the miss- ing information (i.e. location) on patent applications seriously affected the data collecting process. We use a matched patent count data set to overcome the issue. The data set matches the names on patents with individuals and location. Then, to refine and improve the matching the data set uses work histories provided by Italy’s Social Security Admin- istration. In addition, observations are manually checked and confirmed for the matched names on patents to increase precision.4 As a result, the data set has more accurate infor- mation and more complete picture at the province level. It includes patent data using the Italian Patent Office (IPO) between 1950 and 2010, and the international patents included in the European Patent Office’s (EPO) PATSTAT database. The data set provides number of patents at each province in Italy during the given time period. The patent data is able to distinguish between the assignees and the inventors of a patent. An assignee can be a firm or an individual who holds the intellectual property rights over 2 See Malanima and Zamagni (2010), De Bonis et al. (2012) and Nuvolari and Vasta (2015) for more details. 3 See Hall (2011) for a more detailed discussion. 4 Please see Bianchi and Giorcelli (2020) for a detailed discussion of patent count data set. 23 the patented invention. Hence, patent counts only for assignees can disrupt the geographical variety. For example, a large firm headquartered in province A may be the patent’s assignee, while the inventor of this patent works in a plant of the large firm in province B. In this example, the patent would be counted in province A if we use patent assignee and province B if we use the patent’s inventor. The separation between the assignees and the inventors provides a better way to capture the effect of credit reallocation on innovation. 3.1.3 Banking Data Following Herrera et al. (2011), we use bank-level loan data to measure credit flows. For the same time period it is almost impossible to find firm-level debt structures in Italy. The banking data clearly represents the banking system with detailed balance sheet items. This feature makes it very well suited for analyzing credit flows. We use bank-level balance sheet data from Historical Archive of Credit in Italy (ASCI) following Natoli et al. (2016). ASCI provides data for nearly 2,600 banks for the time between 1890 and 1973. The data includes yearly balance sheet of banks and there are more than 41,000 balance sheets in the data set. Bank balance sheet data collection is built on Bank of Italy’s earlier work. Due to confidentiality of bank supervision statistics, the data ends in 1973. Under our analysis, we use the yearly balance sheets of nearly 600 banks for the time period. There are 14 types of assets and 9 types of liabilities included in the data set. The important feature of the data set is that the main balance sheet items are comparable over time since the construction is done with a uniform balance sheet structure.5 The data set has information on each bank’s province and region. Thus, we can create aggregate measures at province level. We obtain total loans summing short-term and long- term loans from balance sheet. We use total loans to calculate credit reallocation measures. 3.1.4 Province Characteristics We use province characteristics as controls in our analysis. We collect data from histori- cal censuses held in 1951, 1961, and 1971. The main problem is that the data is not digitally 5 More details about the data set can be found in Natoli et al. (2016). 24 available. Only scanned census documents are accessible at the Italian National Institute of Statistics’ (ISTAT) website.6 We manually extracted data for province characteristics using scanned census documents. Particularly, we use general population censuses (“Censimento Generale Della Popolazione") and industry and commerce censuses (“Censimento Generale Dell’Industria E Del Commercio") to obtain province characteristics. Using general sum- mary data (“Dati Generali Riassuntivi") from censuses, we can extract a good amount of useful data at province level. We acquire population and education related characteristics from general population censuses. We use share of active population as an indicator of labor force participation and share of higher education degrees as an indicator for level of education at a province. We obtain economic province characteristics from industry and commerce censuses. We use share of individual firms as an indicator for economic development7 . We also get number of firms, workers, and bank branches from industry and commerce censuses. We add number of workers per firm and number of bank branches per firm to control for economic and financial characteristics of provinces. We calculate credit market concentration as a simple Herfindahl–Hirschman Index (HHI) using the bank level loan data for each province. Lastly, we measure productivity as to- tal value added for each firm. Although, the results are robust to different definitions of productivity. 3.1.5 Measurement Issues The measurement of credit flows using bank loans has an important caveat. Bank loans do not account for inflation making it hard to measure the real exposure of patenting activities to banks. We deflate the original bank loan data using an implicit GDP deflator to overcome this issue. Additionally, we deflate province characteristics if necessary, in particular we deflate total value added. We acknowledge that using non-deflated (nominal) credit flows 6 ISTAT catalog can be accessed at ebiblio.istat.it. 7 Guiso et al. (2004a) show that individuals are more likely to start a business in more developed regions in Italy. 25 might have important insights. However, the results are all in real terms. 3.1.6 Credit Reallocation Measures To obtain credit reallocation measures we closely follow Herrera et al. (2011). Their work on credit flow measures utilizes the methodology for measuring job flows developed by Davis and Haltiwanger (1992) and Davis et al. (1996). We use credit and loan interchangeably. Let us define cbt as the average of the loans of a bank b at time t − 1 and at time t. Then, we define Cst as the average of loans for a set s of banks where the set is a province. We define time t loan growth rate of a bank, gbt , taking the first difference of its loans divided by cbt . Now, given a set s of banks, we can define credit creation and credit destruction to establish credit reallocation measures. We calculate credit creation at time t, P OSst , as the weighted sum of the loan growth rates of banks with rising loans or newborn banks. Similarly, we calculate credit destruction at time t, N EGst , as the weighted sum of the absolute values of the loan growth rates of banks with shrinking loans or dying banks. Then, for both measures, we weight the loan growth rate of a bank b with the ratio cbt /Cst . We obtain the following measures   X cbt P OSst = gbt (3.1) b∈st Cst gbt >0   X cbt N EGst = |gbt | (3.2) b∈st Cst gbt <0 where st is the set of banks at time t. Finally, we can define credit reallocation, SU Mst , as the sum of credit creation and credit destruction, SU Mst = P OSst + N EGst . In addition, we can define the net credit growth as N ETst = P OSst − N EGst , and the excess credit reallocation as the reallocation in excess of the minimum required to accommodate the net credit change, EXCst = SU Mst − |N ETst |. 26 3.1.7 Choosing a Credit Reallocation Measure The intensity of credit reallocation is important in its own right and its movement along- side the economic activity. However, our goal is to understand whether credit reallocation influences economic activity, in our case innovation. Therefore, an important task is to decide which credit reallocation measure provides better information on credit markets. To deal with this task, we discuss some features and properties of each credit reallocation measure in this section. Davis (1998) argues that using gross job reallocation, the sum of job creation and job destruction, as the main indicator of reallocation intensity is harmless enough in many contexts. However, he concludes that gross job reallocation becomes a questionable measure of reallocation in a time-series context. Instead, he offers excess job reallocation as a robust measure of reallocation. We adopt the same approach and disregard gross credit reallocation in our analysis. We explain the main problem of gross credit reallocation using its definitions. We define gross credit reallocation in two different ways. First, gross credit reallocation increases with simultaneous credit creation and destruction, SU Mst = P OSst + N EGst . Second, gross credit reallocation also rises with a change in the absolute value of the net credit growth, SU Mst = EXCst + |N ETst | where N ETst = P OSst − N EGst . Thus, using gross credit reallocation makes it hard to compare two provinces in our case. A simple example given for gross job reallocation helps to better understand. An economy with a 5% credit creation rate and 0% credit destruction rate has 5% gross credit reallocation rate, while an economy with 0% credit creation and destruction rates has 0% gross credit reallocation rate. However, we cannot say that the first economy has more reallocation activity than the second economy. Because both economies have 0% excess credit reallocation and we define excess credit reallocation as the part of gross credit reallocation over and above the amount required to accommodate the net credit growth. Hence, it is a better measure of simultaneous credit creation and destruction. 27 Overall, we choose excess credit reallocation as our main measure of credit reallocation and use net credit growth as an indicator of development in credit markets. 3.1.8 Properties of Credit Reallocation The intensity of credit reallocation can help shed light on the impact of reallocation on innovation. In particular, examining the dynamic behavior of credit reallocation and differences at the province level can be informative about how credit reallocation affects innovation and what factors can play a key role. This section investigates properties of credit reallocation across provinces from 1950 to 1963.8 Figure B.1 and Table B.1 display how credit reallocation measures change from 1950 to 1963 compared to real GDP growth. We take the average of credit reallocation measures for each province in a given year. In the early 1950s, gross credit reallocation and real GDP growth move in the opposite directions. On the other hand, credit destruction, con- sequently excess credit reallocation, moves hand in hand with the real GDP growth in the early 1950s. Gross credit reallocation and net credit growth declined in the early 1950s and then increased towards the mid-1950s. However, they gradually decreased until late 1950s. Until this point, we can say that gross credit reallocation and net credit growth demonstrate an opposite movement compared to real GDP growth. This negative relationship reverses after 1958. Starting in the 1960s, gross credit reallocation and net credit growth started to follow a more similar pattern with real GDP growth. Lastly, credit destruction and excess credit reallocation stay relatively low during sample period. Overall, credit creation, gross credit reallocation, and net credit growth closely follow each other over time, while credit destruction and excess credit reallocation display a similar movement. These results are not unexpected considering that the time coincides with the greatest development of the Italian economy. Also, we work with bank loans instead of firm debts and we expect banks to increase the amount of loans during an economic expansion period. Figure B.2 presents the relationship between innovation and credit reallocation from 1950 8 Please see Data Appendix (Section B.2) for inclusion of additional data. 28 to 1963. Again we take the average of credit reallocation measures and number of patents for each province for a given year. Patents increase towards the end of 1950s after a slight decline in the early 1950s. This period coincides with the Italian economic boom. However, after this prosperous period, there is a large decline in the number of patents in the early 1960s. Nuvolari and Vasta (2015) argue that scientific activities prevail patenting during this period. Next, we try to explore more how innovation and credit reallocation are related at the province level. We examine how provinces are distributed using number of patents and credit reallocation measures. We present the results of this exercise in Figure B.3, Figure B.4, Figure B.5, Figure B.6, and Figure B.7. We take the average of number of patents and credit reallocation measures for the whole sample period to draw the scatter plots. First thing to notice is that Milan, Rome, Turin, Florence, Bologna, and Genoa are the provinces with the highest average number of patents. This result is expected, particularly for Milan. Bianchi and Giorcelli (2020) show that 12.7% of patents granted between 1968 and 2010 were assigned to an individual or a firm located in Milan. However, credit reallocation is not amongst the highest for these five provinces. Smaller provinces have higher credit reallocation compared to larger provinces. We see a similar picture for credit creation and credit destruction. Hence, this exercise suggest a negative relationship between innovation and credit reallocation. 3.1.9 Summary Statistics for Province Characteristics The relationship between credit reallocation and macroeconomic variables can offer im- portant insights about what possible factors can play a key role between the credit market and the aggregate economy. This section studies province characteristics that can offer some insights on the impact of reallocation on innovation. We present Table B.2 to examine province characteristics in our analysis. We take the average of all considered variables for all provinces at a given year. Data collected from censuses are presented only at the year the census held. First, the number of patents follows 29 a path similar to an inverted-U shape between 1950 and 1963.9 We measure productivity as the total value added per firm in a province. Productivity gradually decreases until 1961 and starts to increase after. The evidence suggests that inno- vation and productivity follow a similar path over time. The number of banks is stable over time moving around 4 banks on average in each province, while number of bank branches on average increases substantially over time. There are 96 branches on average in each province in 1951, while the number of bank branches reaches 118 on average in 1961. Additionally, credit market in Italy is highly concentrated between 1950 and 1963. Average number of workers for each firm increases from 3.64 in 1951 to 3.99 in 1961, while the share of active population decreases from 46.2% in 1951 to 40.4% in 1961. Italy’s great economic development period pays out as share of higher education degrees increases from 3.8% in 1951 to 4.95% in 1961. 3.2 The Empirical Model In this section, we describe the empirical strategy in detail. Our goal is to identify the effect of credit reallocation on innovation. However, we suspect that credit reallocation can be endogenous to financial development. For instance, highly developed regions in terms of economic and financial output may also have the most financially developed banking systems. We present Figure B.8 to display the regional differences. We take the average number of patents, net credit growth, and excess credit reallocation for each region for the entire period. Panel (a) shows that the average number of patents is higher in highly developed regions. Net credit growth is higher in less developed (mostly southern) regions (Panel (b)). Finally, excess credit reallocation, our main credit reallocation measure, is higher in highly developed regions but two less developed regions have the highest credit reallocation rates (Panel (c)). Moreover, unobserved factors that affect economic and financial activity may be corre- lated with credit reallocation. This relationship may cause a bias in our results. Therefore, we must use exogenous factors of financial development to instrument our credit reallocation 9 Please see Data Appendix (Section B.2) for inclusion of additional data. 30 measures. Considering these endogeneity issues, our empirical strategy is estimating a two- stage model that in the first stage projects the rate of credit reallocation in a province onto an indicator of local financial development and in the second stage projects a the measure of innovation (number of patents) onto value of credit reallocation in the province defined by local financial development. We first discuss our instruments and their validity. Then, we lay out the empirical model employed for estimation. 3.2.1 Instruments Banking regulations play an important role in shaping the financial system in Italy. There is considerable diversity in the banking development due to regulatory reforms in the bank- ing system. Particularly, the banking regulation in effect from 1936 to the end of the 1980s is the source of a large fraction of diversity observed in Italian banking development. Guiso et al. (2004a) and Guiso et al. (2004b) discuss in great detail that the banking regulation put in effect in 1936 creates a partly exogenous geographical diversity in banking development, which might be informative in isolating the effect of bank financing on real outcomes. There- fore, it might help identify the effect of credit reallocation on innovation. Additionally, the banking sector structure allows us to safely rely on the geographical diversity in the banking sector to examine the impact of credit reallocation on innovation. The Italian Government enacted the banking legislation of 1936 in response to the 1930–1931 banking crisis. The government introduced strict market entry regulations to preserve the banking system from instability. Four categories for credit institutions were established: national, cooperative, local commercial, and savings banks. The regulation required all banks to shut down branches located outside its geographical boundaries deter- mined by the legislation. Furthermore, national banks were allowed to open new branches in the main cities. Cooperative and local commercial banks were allowed to open new branches within the boundaries of the province where they were located in 1936. On the other hand, savings banks were allowed to expand within the boundaries of the region where they were 31 located in 1936. Finally, the Bank of Italy was designated as the sole authority enabling banks to extend credit outside their geographical boundaries determined by the legislation. The banking regulation passed in 1936 remained in effect until 1985. Guiso et al. (2004a) argue that this regulation significantly hampered the growth of the financial system. Furthermore, they document that banks in these four categories experience substantially different growth paths. Considering this fact, they show that these differences in growth can explain the regional variation in credit supply after 60 years. They select the number of total branches (per million inhabitants) in a region in 1936, the fraction of branches owned by local versus national banks, the number of savings banks (per million inhabitants), and the number of cooperative banks (per million inhabitants) to instrument financial development. They find that these candidate variables can explain 72% of the cross-sectional variation in the supply of credit in the 1990s. We also perform a similar exercise to find our instruments. We choose the number of savings banks in 1936 (per 100,000 inhabitants) to instrument credit reallocation. The main reason for selecting the number of savings banks is that they are the only category of banks allowed to extend credit outside of the province where they were located. In addition, we choose the inverse of credit market concentration in 1936 to instrument net credit growth. We measure credit market concentration with a Herfindahl–Hirschman Index (HHI) of bank loans. The inverse of credit market concentration provides the effective number of banks in the credit market making it a good candidate instrument for net credit growth. Additionally, the first-stage regression results (See Table B.13 and Table B.14) confirm that these two variables are correlated with the variables of interest, namely excess credit reallocation and net credit growth. Finally, Guiso et al. (2004a) discuss in great detail how and why these instruments are uncorrelated with the error term. They extensively argue that the structure of local credit markets in 1936 was not the outcome of characteristics of the region or forced by the legislation. On the contrary, the structure of the credit markets was random and mostly the outcome of politics. 32 3.2.2 Fixed Effects Model The empirical fixed effects model can be expressed as follows P atentit = αt + βi + γCreditit + εit (the second stage) (3.3) Creditit = κt + ηi + F inDevit + νit (the first stage) (3.4) where P atentit is the average number of patents (per 1,000 firms) in province i in year t, αt is a time fixed effect that captures nation-wide shocks to economic activity in year t, βi is a regional fixed effect10 that measures the component of economic activity specific to the region of province i (reflecting time-invariant unexplained factors that differ across regions), Creditit is the rate of credit reallocation or credit growth in province i in year t, and εit is the residual. In addition to time and regional fixed effects, Equation 3.3 (the first stage) includes F inDevit , local financial development indicators as instrumental variables to account for different development levels. We expect that local financial development indicators are correlated with credit reallocation but they affect economic activity only through the credit market. We use Savingsi , the number of savings banks in province i in 1936 (per 100,000 inhabitants), to instrument credit reallocation. Then, we use IHHIi the inverse of credit market concentration in province i in 1936, to instrument net credit growth. 3.3 Main Estimation Results In this section we present our main results. Empirical evidence suggests a negative relationship between credit reallocation and innovation (See Figure B.4). Thus, our primary goal is to further investigate whether credit reallocation negatively impacts innovation. In addition, we examine the relationship between net credit growth and innovation. Because the time coincides with Italy’s great economic boom, the economic growth would reflect on net credit growth. Thus, we try to see the impact of economic advancement on innovation. We use excess credit reallocation, which nets out the minimum reallocation needed to 10 Since we instrument credit reallocation with 1936 local financial development indicators, we can only exploit cross sectional variation at province level. Hence, we need to drop province fixed effects and add regional fixed effects. 33 accommodate net credit growth, as our main indicator of reallocation intensity to examine the impact of credit reallocation on innovation. We also use net credit growth to reestimate the model to shed light on the economic growth and innovation mechanisms. Table B.3 and Table B.4 reports coefficient estimates from estimation and associated heteroskedasticity- robust standard errors in parentheses. We start by discussing the baseline estimates (Column 1 in Table B.3 and Table B.4). The estimates reveal that the number of patents decreases as credit reallocation increases, while, expanding credit helps increase the number of patents. Hence, unsurprisingly, we confirm that credit reallocation harms innovation measured by the number of patents. A one percentage point increase in excess credit reallocation leads to a 9.8 percent decline in the number of patents. On the other hand, a one percentage point increase in net credit growth causes a 1 percent increase in the number of patents. We support the fact that great economic development boosts innovation. First-stage regression results reveal that an increase in the number of savings banks rises excess credit reallocation. This result is expected because savings banks were the only category of banks allowed to expand within the boundaries of the region where they were located in 1936. Moreover, a rise in the effective number of banks (inverse of credit market concentration) causes an increase in net credit growth. Next, we investigate whether credit reallocation and net credit growth in previous periods impact innovation. We present the results of this exercise in Columns 3 and 4 in Table B.3 and Table B.4. With the inclusion of lags of excess credit reallocation, the coefficient on contemporary credit reallocation does not substantially change. However, the sign of lags of credit reallocation is the opposite of current credit reallocation, although the coefficients are not statistically significant. Net credit growth experiences a similar sign reversal with the only difference that the first lag of net credit growth is statistically significant. The sign reversal for lags is not entirely surprising. Italy experienced a miraculous economic development during the 1950s, followed by a slowdown in economic growth and innovative 34 activities. Nuvolari and Vasta (2015) claim that scientific activities prevail patenting between 1960 and 1970. Hence, this fact explains the sign reversal for net credit growth and its significance. We show that an increase in net credit growth in the previous period leads to a decrease in the number of patents in the current period. Furthermore, we control for province characteristics in addition to the baseline estimates. Columns 2, 5 and 6 in Table B.3 and Table B.4 present the results of our experiments with different specifications. We pick province characteristics to account for various aspects of development in a province. First, we start by controlling for the share of the active population, the number of bank branches per firm, the share of individual firms, the share of higher education degrees, and productivity measured as total value added per firm in the estimation for excess credit reallocation (Column 2). We do not use the number of banks instead of the number of bank branches. We think that bank branches may better capture unobserved effects of credit reallocation considering savings banks can expand outside of the province but within the region where they are located. The results reveal that controlling for province characteristics decreases the magnitude of coefficient estimates for excess credit reallocation compared to the baseline estimates. Still, the direction of impact remains the same. The same exercise results differently for net credit growth. The magnitude and sign of coefficient estimates for net credit growth do not substantially change compared to the baseline estimates with the inclusion of province characteristics as control variables. Additionally, human capital (share of higher education degrees) and approximate labor participation (share of the active population) have a positive and statistically significant effect on innovation. This result is expected considering the evidence provided in the lit- erature. These results are the same in both estimations. On the other hand, the share of individual firms, an indicator of economic development, negatively impacts the number of patents in both estimations. The number of bank branches negatively affects innovation in the estimation with excess credit reallocation, while it positively impacts innovation in the estimation with net credit growth, but the coefficient estimate is statistically insignificant. 35 Lastly, we combine province characteristics with the lags of excess credit reallocation and net credit growth. Columns 5 and 6 present the results in Table B.3 and Table B.4. The direction of the effect stays the same for both excess credit reallocation and net credit growth. However, the magnitude of coefficient estimates for excess credit reallocation compared to the baseline estimates decreases in this case. At the same time, they stay around the same for coefficient estimates for net credit growth. Additionally, we perform robustness checks to examine the strength of our instruments. We leave the robustness of results to the north-south divide, an alternative specification and inclusion of additional data from 1968 to 1973 to Section B.3. 3.4 Conclusion In this chapter, we empirically study the impact of credit reallocation on innovation. Combining a theoretical and an empirical approach, we test the predictions from the model using granular data from the Italian local markets. We use bank-level loan data to calculate credit reallocation and patent count data to measure innovation. Focusing on Italy provides a very informative environment to isolate the effect of banks, in particular credit reallocation, on innovation. We also use the fact that the sample time period coincides with the so-called “Italian economic miracle" period and tighter banking regulations. In addition, we suspect that highly developed regions in terms of economic output may also have the most financially developed banking systems. Hence, we estimate a two-stage model using instruments from the banking regulations. We exploit the banking legislation in 1936 to pick our instruments. The banking legislation in 1936 creates a partly exogenous geographical diversity in banking system that lasts without substantial changes until 1985. Our results reveal a negative relationship between credit reallocation and innovation. We find that an intensification in credit reallocation disrupts firms’ innovation through a decline in the number of patents. On the other hand, a rise in net credit growth boosts innovation by increasing the number of patents. Our results carry through when we control for various 36 province characteristics. We find that human capital and labor force participation indicators positively affect innovation, while financial and economic development indicators depress in- novation. We further show that our results are robust to weak instruments, across alternative empirical specifications, and altering the estimation period via inclusion of additional data. 37 BIBLIOGRAPHY Aghion, P. and Bolton, P. (1992). An incomplete contracts approach to financial contracting. Review of Economic Studies, 59(3):473–494. Anderson, T. W. and Rubin, H. (1949). Estimation of the parameters of a single equation in a complete system of stochastic equations. Annals of Mathematical Statistics, 20(1):46–63. Andrews, I., Stock, J., and Sun, L. (2019). Weak instruments in IV regression: Theory and practice. Annual Review of Economics. Araujo, L., Cao, Q., Minetti, R., and Murro, P. (2019). Credit crunches, asset prices and technological change. Review of Economic Dynamics, 32:153–179. Bianchi, N. and Giorcelli, M. (2020). Scientific education and innovation: From technical diplomas to university stem degrees. Journal of the European Economic Association, 18(5):2608–2646. Brown, J. R., Martinsson, G., and Petersen, B. C. (2012). Do financing constraints matter for R&D? European Economic Review, 56(8):1512–1529. Caballero, R. J. and Hammour, M. L. (1994). The cleansing effect of recessions. American Economic Review, 84(5):1350–1368. Caballero, R. J. and Hammour, M. L. (2005). The cost of recessions revisited: A reverse- liquidationist view. Review of Economic Studies, 72(2):313–341. Chang, P. C., Jia, C., and Wang, Z. (2010). Bank fund reallocation and economic growth: Evidence from China. Journal of Banking and Finance, 34(11):2753–2766. Chen, K. and Song, Z. (2013). Financial frictions on capital allocation: A transmission mechanism of TFP fluctuations. Journal of Monetary Economics, 60(6):683–703. Davis, S. (1998). Discussion: Job reallocation and the business cycle: New facts for an old debate. In Conference Series-Federal Reserve Bank of Boston, volume 42, pages 349–358. Davis, S. J. and Haltiwanger, J. (1992). Gross job creation, gross job destruction, and employment reallocation. Quarterly Journal of Economics, 107(3):819–863. Davis, S. J., Haltiwanger, J. C., and Schuh, S. (1996). Job Creation and Destruction. The MIT Press. De Bonis, R., Farabullini, F., Rocchelli, M., Salvio, A., and Silvestrini, A. (2012). A quanti- tative look at the Italian banking system: Evidence from a new dataset since 1861. Bank of Italy, Economic History Working Papers, (26). 38 De Jonghe, O., Dewachter, H., Mulier, K., Ongena, S., and Schepens, G. (2020). Some borrowers are more equal than others: Bank funding shocks and credit reallocation. Review of Finance, 24(1):1–43. de Rassenfosse, G., Kozak, J., and Seliger, F. (2019). Geocoding of worldwide patent data. Scientific Data, 6(260). Dehejia, R. and Lleras-Muney, A. (2007). Financial development and pathways of growth: State branching and deposit insurance laws in the united states, 1900-1940. Journal of Law and Economics, 50(2):239–272. Dell’Ariccia, G. and Garibaldi, P. (2005). Gross credit flows. Review of Economic Studies, 72(3):665–685. Dell’Ariccia, G. and Marquez, R. (2004). Information and bank credit allocation. Journal of Financial Economics, 72(1):185–214. Detragiache, E., Garella, P., and Guiso, L. (2000). Multiple versus single banking relation- ships: Theory and evidence. The Journal of Finance, 55(3):1133–1161. Diamond, D. W. and Rajan, R. G. (2001). Liquidity risk, liquidity creation, and financial fragility: A theory of banking. Journal of Political Economy, 109(2):287–327. Eisfeldt, A. L. and Rampini, A. A. (2006). Capital reallocation and liquidity. Journal of Monetary Economics, 53(3):369–399. Eisfeldt, A. L. and Shi, Y. (2018). Capital reallocation. Annual Review of Financial Eco- nomics, 10(1):361–386. Galindo, A., Schiantarelli, F., and Weiss, A. (2007). Does financial liberalization improve the allocation of investment?: Micro-evidence from developing countries. Journal of De- velopment Economics, 83(2):562–587. Garcia-Macia, D. (2017). The financing of ideas and the great deviation. Working Paper No. 17/176, International Monetary Fund. Guiso, L., Sapienza, P., and Zingales, L. (2004a). Does local financial development matter? Quarterly Journal of Economics, 119(3):929–969. Guiso, L., Sapienza, P., and Zingales, L. (2004b). The role of social capital in financial development. American Economic Review, 94(3):526–556. Hall, B. H. (2011). Innovation and productivity. Working Paper Series, National Bureau of Economic Research, (17178). 39 Hall, B. H. and Lerner, J. (2010). The financing of R&D and innovation. In Hall, B. and Rosenberg, N., editors, Handbook of the Economics of Innovation, volume 1, pages 609–639. Elsevier, Amsterdam. Herrera, A. M., Kolar, M., and Minetti, R. (2011). Credit reallocation. Journal of Monetary Economics, 58(6-8):551–563. Hyun, J. (2016). Financial crises and the evolution of credit reallocation: Evidence from Korea. Economic Modelling, 56:25–34. Hyun, J. and Minetti, R. (2019). Credit reallocation, deleveraging, and financial crises. Journal of Money, Credit and Banking, 51(7):1889–1921. Jayaratne, J. and Strahan, P. E. (1996). The finance-growth nexus: Evidence from bank branch deregulation. Quarterly Journal of Economics, 111(3):639–670. Keane, M. P. and Neal, T. (2021). A practical guide to weak instruments. UNSW Economics Working Paper, 2021-05a. Kerr, W. R. and Nanda, R. (2015). Financing innovation. Annual Review of Financial Economics, 7:445–462. Keuschnigg, C. and Kogler, M. (2020). The Schumpeterian role of banks: Credit reallocation and capital structure. European Economic Review, 121:103349. Kiyotaki, N. and Wright, R. (1993). A search-theoretic approach to monetary economics. American Economic Review, pages 63–77. Lanteri, A. (2018). The market for used capital: Endogenous irreversibility and reallocation over the business cycle. American Economic Review, 108(9):2383–2419. Lotti, F. and Marin, G. (2013). Matching of PATSTAT applications to AIDA firms: Discus- sion of the methodology and results. Bank of Italy, Occasional Papers, (166). Malamud, S. and Zucchi, F. (2019). Liquidity, innovation, and endogenous growth. Journal of Financial Economics, 132(2):519–541. Malanima, P. and Zamagni, V. (2010). 150 years of the Italian economy, 1861-2010. Journal of Modern Italian Studies, 15(1):1–20. Natoli, S., Piselli, P., Triglia, I., and Vercelli, F. (2016). Historical archive of credit in Italy. Bank of Italy, Economic History Working Papers, (36). Nuvolari, A. and Vasta, M. (2015). The ghost in the attic? the Italian national innovation system in historical perspective, 1861-2011. Enterprise & Society, 16(2):270–290. 40 Olea, J. L. M. and Pflueger, C. (2013). A robust test for weak instruments. Journal of Business & Economic Statistics, 31(3):358–369. Petersen, M. A. and Rajan, R. G. (1994). The benefits of lending relationships: Evidence from small business data. The Journal of Finance, 49(1):3–37. Staiger, D. and Stock, J. H. (1997). Instrumental variables regression with weak instruments. Econometrica, pages 557–586. Stock, J. and Yogo, M. (2005). Testing for weak instruments in linear IV regression. Identi- fication and Inference for Econometric Models: Essays in Honor of Thomas Rothenberg, page 80. Thoma, G., Torrisi, S., Gambardella, A., Guellec, D., Hall, B. H., and Harhoff, D. (2010). Harmonizing and combining large datasets - an application to firm-level patent and ac- counting data. Working Paper Series, National Bureau of Economic Research, (15851). Wang, Y. (2017). Debt market friction, firm-specific knowledge capital accumulation and macroeconomic implications. Review of Economic Dynamics, 26:19–39. 41 APPENDIX A APPENDIX FOR A THEORETICAL MODEL OF CREDIT REALLOCATION AND INNOVATION A.1 Proofs Proof of Lemma 1. An unskilled agent will always prefer to transfer her machine to a skilled agent if Vu > Vo . We can define the expected payoff of lending a machine Vu as the sum of the expected payoffs from lending to a skilled agent producing with old technology Vuo and new technology Vun . Next, we plug in the value functions. We obtain Vu > Vo (1 − α)Vuo + αVun > Vo n h io (1 − α)Vuo + α (1 − β)Vun,C + β ηVun,R + (1 − η)(Vun,C − σ) > Vo n o (1 − α)Vuo + α (1 − βη)Vun,C + βηVun,R − β(1 − η)σ > Vo n o α (1 − βη)(1 + γ)λs yd + βηxλs yd + βη(1 − x)λu yu − β(1 − η)σ + n h io h i (1 − α) (1 + r) λs yd + (1 − λs )a − (1 + r) λu yu + (1 − λu )a > 0 .. . nh i o α (1 − βη)(1 + γ) + βηx λs yd + βη(1 − x)λu yu − β(1 − η)σ − (1 + r) + | {z } | {z } >0 (ii) n o (1 + r) (1 − α)λs (yd − a) + λu (a − yu ) >0 | {z } (i) We derive two conditions and get (i) yu < a < yd η 1+r (ii) σ < (1 − x)λu yu 1−η − β(1−η) Proof of Lemma 2. The expected payoff from continuation is higher than the expected payoff from reallocation as long as Vun,C − σ > Vun,R . We substitute for the necessary value 42 functions. Then, we obtain Vun,C − σ > Vun,R λs (1 + γ)yd − σ > xλs yd + (1 − x)λu yu solving for γ, we get the condition needed for the unskilled agents to continue as following σ − (1 − x)[λs yd − λu yu ] γ> λs y d Proof of Lemma 3. The expected payoff from new technology is higher than the expected payoff from old technology as long as Vsn > Vso . We get the following h i (1 − β)Vsn,C n,R + β ηVs + (1 − η)Vs n,C > Vso (1 − βη)Vsn,C + βηVsn,R > Vso (1 − βη)λs (1 + γ)(ys − yd ) + βηxλs (ys − yd ) > (1 + r)λs (ys − yd ) .. . r + (1 − x)βη γ> 1 − βη After substituting for the necessary value functions, we find inequality (2.13) as the condition. Then, we substitute for the unskilled agents choice η to define intervals for the skilled agents decision. First, we set η = 0 so that all of the unskilled agents continues and the expected payoff from new technology is maximized. Inequality (2.13) becomes γ ≥ r. In this region of the parameter space, the skilled agents will choose to produce only with either new technology or old technology. However, if we have γ 1−β In this region of the parameter space, the expected payoff of producing with new technology is higher than the expected payoff of producing with old technology. Therefore, choosing new technology dominates choosing old technology for the skilled agents (case (iii)). Finally, we can consider the case r + (1 − x)β r≤γ≤ . 1−β In this region of the parameter space, there is a value γ ′ such that the skilled agents are indifferent choosing between new and old technology. For this value of γ, there exists a case in which some of the skilled agents will choose new technology and others will choose old technology. Proof of Lemma 4. We combine the conditions from Lemmas 1-3. First, we consider the condition for the skilled agents’ new technology decision. A skilled agent will innovate if r + (1 − x)βη γ> . 1 − βη Next, the unskilled agents lend their machines if η 1+r σ λu y u η 1+r σ < (1 − x)λu yu − =⇒ < (1 − x) − 1 − η β(1 − η) λs yd λs yd 1 − η λs yd β(1 − η) and they will reallocate their machines if σ (1 − x)[λs yd − λu yu ] γ≤ − . λs yd λs yd Thus, we can derive an interval in which the skilled agents choose new technology and innovate in period 1, while, the unskilled agents will reallocate their machines in period 2. If γ is in the following interval r + (1 − x)βη σ (1 − x)[λs yd − λu yu ] λu yu η 1+r <γ≤ − < (1 − x) − 1 − βη λs y d λs y d λs yd 1 − η λs yd β(1 − η) 44 the reallocation decision of the unskilled agents hinders the innovation process. Proof of Propsition 1. Using Lemmas 1-4, the proof of the proposition is immediate. Considering parts (i) and (iii) of Lemma 3, we show innovation and no innovation cases regardless of the unskilled agents’ decision. Then, using part (ii), we show that the skilled agents will not innovate if γ < γ ′ . Next, we combine part (ii) of Lemma 3 and and Lemma 2. The skilled agents will innovate if γ > γ ′ . Within the interval from part (ii) of Lemma 3, we plug in the cutoff point from Lemma 2. Hence, we get the results. 45 APPENDIX B APPENDIX FOR MEASURING THE EFFECT OF CREDIT REALLOCATION ON INNOVATION B.1 Tables and Figures This section includes the tables and figures mentioned in the main body of text. Figure B.1 Credit reallocation measures and the real GDP growth Figure B.2 Credit reallocation measures and the average number of patents per firm 46 Figure B.3 Distribution of provinces - Gross credit reallocation and the number of patents per 1,000 firms Figure B.4 Distribution of provinces - Net credit growth and the number of patents per 1,000 firms 47 Figure B.5 Distribution of provinces - Excess credit reallocation and the number of patents per 1,000 firms Figure B.6 Distribution of provinces - Credit creation and the number of patents per 1,000 firms 48 Figure B.7 Distribution of provinces - Credit destruction and the number of patents per 1,000 firms 49 (a) Number of Patents (b) Net Credit Growth (c) Excess Credit Reallocation Figure B.8 Regional overview of variables of interest Note: This figure plots the regional overview of three main variables of interest. Panel (a) displays the average number of patents for each region. The Northern regions have higher number of patents. Panel (b) presents the regional distribution of net credit growth. The Southern regions have higher net credit growth as expected. Panel (c) shows the regional differences in the excess credit reallocation measure. Overall, the Northern regions have higher levels of excess credit reallocation, but two of the Southern regions have the highest levels. 50 Table B.1 Summary statistics for credit reallocation measures The table reports the yearly averages of credit reallocation measures for each year over the sample period. The summary statistics refer to the 1950-1963 period. Credit flows are computed from the bank-level loan changes using the methodology described in the paper. Real GDP growth is added at the last column to make comparisons. Year Gross Net Excess Credit Credit Real Credit Credit Credit Creation Destruction GDP Reallocation Growth Reallocation Growth 1950 17.61% 11.44% 1.84% 14.53% 3.09% 8.41% 1951 13.73% 5.79% 1.65% 9.76% 3.97% 9.68% 1952 20.38% 19.68% 0.70% 20.03% 0.35% 4.75% 1953 16.73% 15.87% 0.45% 16.30% 0.43% 7.35% 1954 17.62% 16.55% 0.73% 17.08% 0.54% 3.80% 1955 14.86% 13.11% 0.72% 13.99% 0.88% 6.97% 1956 13.58% 12.44% 0.82% 13.01% 0.57% 4.97% 1957 12.68% 11.11% 0.84% 11.89% 0.78% 5.72% 1958 11.79% 9.52% 0.82% 10.65% 1.13% 5.94% 1959 14.72% 13.88% 0.38% 14.30% 0.42% 7.12% 1960 18.75% 17.77% 0.51% 18.26% 0.49% 7.71% 1961 13.62% 12.91% 0.41% 13.27% 0.36% 8.47% 1962 14.98% 13.91% 0.62% 14.45% 0.54% 6.98% 1963 11.96% 10.41% 1.08% 11.19% 0.77% 6.22% 51 Table B.2 Summary statistics for province characteristics The table reports the yearly averages of province characteristics over the sample period. The statistics are computed averaging across all of the provinces for each year in the sample period which refers to the 1950-1963 period. The number of patents are the total number of patents divided by the total number of firms. Productivity is measured as the total value added per firm. The number of banks is the total banks divided by number of provinces. Credit market concentration is measured by a Herfindahl Index of number of banks. Share of individual firms is the average of share of sole proprietary firms across all provinces. Share of higher education degrees represents the average share of population obtained higher education degrees and an indicator of human capital. Share of active population is the fraction of population actively working or searching for a job, an approximation for labor force participation. Year Number of Productivity Number of Credit Number of Number of Share of Share of Share of Patents (000 lire) Banks Market Workers Bank Individual Higher Active Concentration per Firm Branches Firms Education Population Degrees 1950 73.43 269.43 3.45 0.68 1951 68.53 248.85 4.26 0.62 3.64 96.33 91.37% 3.79% 46.24% 1952 71.49 240.87 4.50 0.61 1953 79.02 232.74 4.50 0.61 1954 82.72 225.95 4.51 0.61 1955 84.83 218.88 4.52 0.61 1956 82.10 210.35 4.50 0.61 1957 76.31 206.52 4.37 0.62 1958 72.88 202.40 3.34 0.68 1959 79.30 203.61 4.45 0.61 1960 71.60 200.43 4.45 0.61 1961 61.73 349.22 4.44 0.61 3.99 118.60 91.45% 4.94% 40.44% 1962 60.93 331.51 4.43 0.61 1963 31.15 305.50 4.43 0.61 52 Table B.3 The effect of credit reallocation on innovation The table reports regression coefficients for the impact of credit reallocation on innovation within provinces. The regressions are estimated by two-stage least squares to control for the endogeneity of credit flows. The dependent variable is the number of patents per firm in each province. Heteroskedasticity-robust standard errors are in parentheses. All regressions include region and year fixed effects. *, **, and *** denote statistical significance at the 10, 5 and 1% level, respectively. We use the number of savings banks in 1936 (per 100,000 inhabitants) to instrument excess credit reallocation. The main reason for selecting the number of savings banks is that they are the only category of banks allowed to extend credit outside of the province where they were located. Last row of the table reports the F-statistic for an F-test of joint significance of the instrument. (1) (2) (3) (4) (5) (6) VARIABLES Patents Patents Patents Patents Patents Patents per firms per firms per firms per firms per firms per firms Excess Credit Reallocation -0.098*** -0.027** -0.102*** -0.111*** -0.026** -0.027** (0.028) (0.012) (0.030) (0.033) (0.012) (0.013) Share of Active Population 0.000*** 0.000*** 0.000*** (0.000) (0.000) (0.000) No. of Bank Branches per Firms -0.136*** -0.136*** -0.135*** (0.040) (0.040) (0.040) Share of Individual Firms -0.035*** -0.035*** -0.035*** (0.004) (0.004) (0.004) Share of Higher Education Degrees 0.049*** 0.049*** 0.049*** (0.006) (0.006) (0.006) Productivity (Total Value Added per Firm) 0.000*** 0.000*** 0.000*** (0.000) (0.000) (0.000) Excess Credit Reallocation (First lag) 0.004 0.004 -0.000 -0.000 (0.004) (0.004) (0.002) (0.002) Excess Credit Reallocation (Second lag) 0.011 0.001 (0.007) (0.002) Observations 1,204 1,204 1,204 1,204 1,204 1,204 Region FE YES YES YES YES YES YES Year FE YES YES YES YES YES YES F-Stat 17.43 19.02 15.74 14.27 17.36 15.95 53 Table B.4 The effect of credit growth on innovation The table reports regression coefficients for the impact of credit growth on innovation within provinces. The regressions are estimated by two-stage least squares to control for the endogeneity of credit flows. The dependent variable is the number of patents per firm in each province. Heteroskedasticity-robust standard errors are in parentheses. All regressions include region and year fixed effects. *, **, and *** denote statistical significance at the 10, 5 and 1% level, respectively. We use the inverse of credit market concentration in 1936 to instrument net credit growth. We measure credit market concentration with a Herfindahl–Hirschman Index (HHI) of bank loans. The inverse of credit market concentration provides the effective number of banks in the credit market making it a good candidate instrument for net credit growth. Last row of the table reports the F-statistic for an F-test of joint significance of the instrument. (1) (2) (3) (4) (5) (6) VARIABLES Patents Patents Patents Patents Patents Patents per firms per firms per firms per firms per firms per firms Net Credit Growth 0.010** 0.014** 0.012* 0.012* 0.017** 0.017** (0.005) (0.006) (0.006) (0.006) (0.007) (0.007) Share of Active Population -0.001 -0.001 -0.001 (0.000) (0.000) (0.000) No. of Bank Branches per Firms -0.067 -0.070 -0.069 (0.068) (0.074) (0.074) Share of Individual Firms -0.037*** -0.036*** -0.036*** (0.005) (0.005) (0.005) Share of Higher Education Degrees 0.046*** 0.045*** 0.045*** (0.008) (0.009) (0.009) Productivity (Total Value Added per Firm) 0.000*** 0.000*** 0.000*** (0.000) (0.000) (0.000) Net Credit Growth (First lag) -0.003** -0.003** -0.003** -0.003** (0.001) (0.001) (0.001) (0.001) Net Credit Growth (Second lag) -0.000 0.000 (0.001) (0.001) Observations 1,162 1,162 1,162 1,162 1,162 1,162 Region FE YES YES YES YES YES YES Year FE YES YES YES YES YES YES F-Stat 7.631 9.369 5.853 5.898 7.182 7.345 54 B.2 Data Appendix In this section, we further describe the data sources and provide additional information and summary statistics. B.2.1 Banking Data Following Herrera et al. (2011), we use bank-level loan data to measure credit flows. For the same time period it is almost impossible to find firm-level debt structures in Italy. The banking data clearly represents the banking system with detailed balance sheet items. This feature makes it very well suited for analyzing credit flows. We use bank-level balance sheet data from Historical Archive of Credit in Italy (ASCI) following Natoli et al. (2016). ASCI provides data for nearly 2,600 banks for the time between 1890 and 1973. The data includes yearly balance sheet of banks and there are more than 41,000 balance sheets in the data set. Bank balance sheet data collection is built on Bank of Italy’s earlier work. Due to confidentiality of bank supervision statistics, the data ends in 1973. There are 14 types of assets (liquid assets, bonds, mortgages, etc.) and 9 types of liabilities (capital, reserves, deposits, etc.) included in the data set. Additionally, total costs and total revenues are included from the income statements. The important feature of the data set is that the main balance sheet items are comparable over time since the construction is done with a uniform balance sheet structure. The main categories of banks operating between 1890 and 1973 are banks of national interest (banche di interesse nazionale), cooperative banks (banche popolari ), savings banks (casse di risparmio ordinarie), banking houses (ditte bancarie), central institutes (istituti di credito di categoria), public law banks (istituti di credito di diritto pubblico), first class pledge banks (monti di pieta di prima categoria), and joint-stock or ordinary credit banks (società ordinarie di credito). In addition, other banks are a class of important credit institutions which are not initially but subsequently included in one of the categories above. Table B.5 illustrates the number of credit institutions for each category over the sample period. The data set has information on the location of the headquarters of each bank. Consid- 55 Table B.5 Composition of banks in the Historical Archive of Credit sample Year Other Banks of Cooperative Savings Banking Central Public First Joint-stock Total banks national banks banks houses institutes law class banks and interest banks pledge branches of banks foreign banks 1950 1 3 64 78 13 3 5 8 116 291 1951 1 3 121 78 28 2 5 9 119 366 1952 1 3 136 78 31 3 5 7 123 387 1953 1 3 135 78 31 2 5 8 124 387 1954 1 3 135 78 31 3 5 8 124 388 1955 3 136 78 31 3 6 8 124 389 1956 3 137 77 30 3 6 8 123 387 1957 3 131 77 29 3 6 7 120 376 1958 3 61 78 10 2 6 7 111 278 1959 3 133 78 29 3 6 8 123 383 1960 3 132 78 29 3 6 8 124 383 1961 3 133 78 29 3 6 8 121 381 1962 3 132 78 23 3 6 8 127 380 1963 3 132 78 23 3 6 8 127 380 1964 3 132 78 21 3 6 8 126 377 1965 3 130 78 21 4 6 8 123 373 1966 3 129 78 21 3 6 7 121 368 1967 3 125 78 18 3 6 7 113 353 1968 3 122 78 16 3 6 7 108 343 1969 3 122 78 15 3 6 7 107 341 1970 3 196 80 29 5 6 7 137 463 1971 3 188 80 24 5 6 7 138 451 1972 3 185 80 21 5 6 7 135 442 1973 3 182 80 18 5 6 7 140 441 ering this fact with the restrictive banking legislation of 1936, we can see the provincial and the regional distributions. The representation of banks are low before 1950 but increases from 1951 to 1969. For example, the number of cooperative banks appears in the ASCI sample is low before 1950, especially in the Center and in the South. It is higher than fifty percent in the North. However, visibility of cooperative banks increases above fifty percent from 1951 to 1969. The northern regions have higher coverage rates compared to the southern regions. The main reason is that larger banks are more likely to be included in the ASCI sample due to reporting and recording practices. Typically, the southern banks are smaller on average and less likely to be included in the sample. The main source of the data set from 1937 to 1973 is the Bank of Italy’s supervisory documents. The Banking Act of 1936 requires all banks in a legally defined category to submit interim and annual reports. During this period, the precision of data is higher than 56 previous periods. Official reporting schemes and clear accounting rules since 1948 enable the Bank of Italy to create a more precise and homogenous balance sheet data set. Official guidelines lead to higher quality data with fewer errors and at least 80% of the balance sheets are verified during this period. B.2.1.1 Credit Reallocation Measures In this section, we further provide additional summary statistics for credit reallocation measures at the province level. Following Herrera et al. (2011), we use bank-level loan data to measure credit flows. Table B.6 and Table B.7 present Net Credit Growth and Excess Credit Reallocation at the province level. 57 Table B.6 Net credit growth at the province level Provincia 1950 1951 1952 1953 1954 1955 1956 1957 1958 1959 1960 1961 1962 1963 1964 1965 1966 1967 1968 1969 1970 1971 1972 1973 Agrigento 2.32% -13.77% 7.07% 57.80% 23.14% 23.26% 20.24% -6.71% 3.85% 4.49% 16.51% 8.13% 21.46% 11.78% 0.45% 7.43% 7.55% 11.38% 3.47% 16.94% 9.28% 10.89% 20.50% 11.87% Alessandria 13.31% 18.68% 15.57% 21.99% 27.55% 8.70% 6.87% 7.79% 5.02% 4.72% 37.91% 5.85% -5.46% 13.71% 7.64% 7.51% 4.42% 5.99% 13.92% 16.14% 4.42% -1.64% 8.09% 17.35% Ancona -0.22% -0.02% 22.77% 14.08% 3.78% 18.28% 3.77% 13.03% 9.09% 8.00% 16.94% 12.17% 11.36% 7.51% -1.10% 1.40% 14.45% 25.57% 17.93% 17.87% 3.38% 8.84% 9.51% 23.70% Aosta 57.11% 57.11% 57.11% 8.92% 34.03% 10.01% 21.77% 13.36% 13.83% 14.30% 14.30% 9.92% 9.06% -10.73% -6.54% 9.94% 11.83% 1.08% 14.36% 0.00% -0.64% -20.11% -9.72% -0.34% Arezzo 3.35% 10.27% 24.03% 22.03% 11.32% 13.85% 6.40% 17.52% 10.54% 15.98% 24.81% 15.71% 17.14% 10.36% 5.17% 13.01% 13.82% 6.34% 9.99% 12.14% 4.97% -0.07% 17.10% 6.53% Ascoli Piceno 1.96% -4.61% 10.69% 14.58% 7.78% 12.65% 6.54% 5.52% 4.97% 9.98% 26.99% 10.30% 10.39% 5.19% 15.53% -0.10% 15.40% 13.06% 16.83% 16.34% -0.19% 10.19% 12.45% 6.15% Asti -14.33% 16.98% 19.04% 8.20% 19.49% 9.82% 15.61% 15.00% 4.36% 10.57% 21.47% 21.85% 13.33% 12.04% 0.03% 5.73% 5.49% 4.59% 9.13% 11.00% 12.80% 14.74% 1.34% 3.42% Avellino 18.18% 18.18% 18.18% 13.30% 37.81% 17.59% 6.25% 9.26% 26.40% 43.55% 43.55% 32.76% 25.49% 5.83% -4.80% -46.54% 31.51% 18.68% 43.90% 20.61% 4.40% 2.53% 23.39% -1.52% Bari 24.70% 23.40% 12.21% 10.41% 24.98% 5.19% 14.98% 14.90% 10.83% 20.94% 13.17% 10.88% 17.15% 18.86% -0.01% 16.32% 9.83% 32.06% 14.10% 17.99% 3.41% -1.87% 20.04% 13.74% Belluno 0.00% 7.84% -10.10% 19.02% Benevento 0.54% 11.68% 32.56% 30.24% 7.72% 28.28% 30.70% 9.89% 23.82% 11.42% 8.97% 7.38% 11.07% 20.72% -2.73% 7.87% 0.73% 11.13% 8.98% 21.64% 12.03% 8.43% 7.27% 1.06% Bergamo 0.91% 3.14% 23.15% 16.29% 20.81% 20.28% 10.90% 9.81% 9.93% 16.31% 27.37% 14.38% 17.02% 4.05% -2.05% 1.03% 6.76% 14.55% 10.47% 21.29% 1.01% 4.89% 13.28% 8.25% Bologna 1.45% 3.06% 16.76% 7.68% 12.23% 9.40% 12.80% 8.18% 7.56% 12.71% 15.89% 12.35% 19.91% 16.12% -2.94% 9.52% 11.94% 17.83% 12.55% 11.45% 9.60% -1.37% 8.77% 18.14% Bolzano-Bozen 40.66% -4.97% 0.79% 15.57% 7.28% 2.83% 16.64% 15.35% 9.82% 11.83% 10.15% 10.08% 7.40% 10.77% -16.48% -8.20% 12.17% 19.99% 20.30% 13.35% 4.45% 1.82% 11.66% 16.85% Brescia 7.08% -0.88% 17.61% 16.30% 9.36% 9.58% 18.93% 17.05% 20.23% 16.81% 10.15% 11.42% 15.03% 9.62% -6.07% 1.03% 6.88% 15.11% 13.22% 21.78% 4.91% 15.37% 20.08% 4.51% Brindisi 30.73% 5.68% 22.88% 11.79% 19.74% 3.21% 22.83% -6.24% 21.53% 20.92% -1.36% 3.26% 18.95% 11.97% 5.37% 7.23% 8.87% 20.37% 9.09% -1.78% -2.59% -5.57% -2.63% 14.48% Caltanissetta 5.98% -10.77% 9.52% 2.91% 26.52% -17.12% 32.81% 16.77% 9.95% 26.25% -2.99% -2.99% -2.99% -2.99% 12.11% 12.11% -0.31% 6.38% Campobasso 0.00% 17.01% 22.78% 24.64% Caserta 19.42% 19.42% 19.42% 4.69% 13.86% 13.34% 13.71% 12.61% 16.06% 19.51% 19.51% 18.23% -10.78% 3.90% 42.70% 8.15% 14.53% -5.82% 5.87% 9.60% -7.69% 8.67% 21.20% 27.40% Catania 46.79% 46.79% 46.79% -4.00% 23.80% 5.87% 14.60% 18.13% 20.01% 21.89% 21.89% 14.71% 16.88% 18.67% -2.61% 8.19% 10.80% 12.02% 28.83% 16.81% 11.68% 14.45% 11.23% 12.62% Catanzaro 34.26% 34.26% 34.26% 40.23% 27.30% 33.42% 9.50% 11.34% 12.36% 13.38% 13.38% 26.75% 27.02% 2.33% 13.92% 3.79% 6.90% 16.28% 29.49% 9.45% -0.36% -5.72% 3.78% 3.80% Chieti 5.19% -17.43% 28.41% 3.56% 27.58% 1.08% 13.94% 14.54% 10.15% 27.33% 23.51% 9.19% 21.83% 14.17% 2.61% 18.79% 15.98% 9.12% 10.83% 14.05% 6.60% 6.32% 1.10% 9.72% Como 12.55% 4.06% 15.82% 19.02% 16.48% 17.96% 15.30% 4.45% 10.55% 24.09% 14.16% 14.76% 19.97% -0.37% 0.64% 5.40% 8.43% 8.17% 10.87% 7.10% 0.64% 7.80% 14.28% 3.70% Cosenza 16.01% 30.34% 22.82% 24.19% 18.29% 8.20% 17.32% 11.68% 12.82% 12.34% 9.39% 2.86% 4.43% 6.20% 12.78% 6.79% 13.67% 17.90% 18.43% 13.64% 4.47% 2.56% 5.26% 6.57% Cremona 12.81% 2.51% 19.91% 3.49% 10.57% 20.16% 12.80% 6.22% 7.95% 13.34% 20.85% 15.18% 16.70% 1.17% 2.50% 2.48% 3.27% 12.26% 6.84% 7.47% 8.21% 6.98% 2.73% -8.12% Cuneo 7.90% 11.94% 8.57% 18.30% 12.62% 15.55% 6.28% 12.54% 2.39% 17.52% 16.00% 10.22% 16.16% 13.89% 9.23% 4.30% 4.49% 8.37% 9.64% 18.85% 14.25% 8.77% -3.47% 6.07% Ferrara -14.80% 5.37% 16.77% 23.17% -1.55% 21.58% 9.71% 6.54% 5.61% 6.89% 25.34% 15.86% -0.08% 8.23% -1.17% 4.09% 10.83% 7.10% 10.98% 10.05% -1.69% -3.69% 1.50% 7.22% Firenze 0.17% 2.59% 14.01% 6.41% 10.18% 13.75% 10.46% 11.97% 2.21% 6.03% 17.91% 15.68% 19.51% 17.49% 0.64% 3.18% 7.03% 18.46% 10.41% 14.96% 4.65% 3.60% 10.76% 10.05% Foggia 43.01% 30.56% 15.95% 39.94% 36.85% 23.55% 4.59% 4.32% 21.50% 26.10% 25.36% 33.46% 14.62% 28.48% 7.91% 10.10% 24.61% 15.39% 25.73% 11.25% 6.50% 0.95% 2.75% 22.22% Forli -13.04% -15.66% 17.58% 21.78% 14.14% 19.90% 18.11% 12.58% 8.44% 13.54% 12.97% 20.31% 17.82% 13.25% -8.28% 0.30% 7.84% 16.44% 18.17% 15.64% 1.71% 2.16% 1.60% 7.51% Frosinone 0.00% 6.34% 7.19% 20.31% Genova 4.44% 4.37% 24.37% 19.31% 9.84% 12.28% 4.09% 5.07% -2.51% 11.92% 26.47% 13.03% 18.37% 10.62% 4.59% -8.55% 14.21% 7.44% 11.84% 6.91% 15.39% -3.51% 17.65% 3.46% Gorizia 15.36% 1.31% 4.01% -4.38% 22.32% 16.11% 13.87% 20.00% -2.98% 7.21% 22.33% 6.42% 6.33% 8.22% -13.87% -4.34% 15.99% 13.61% 57.47% 30.74% 23.90% -5.45% 2.99% 6.12% Grosseto 27.42% 12.71% 20.75% 22.36% 14.60% 6.32% 5.84% 23.07% 9.35% 44.65% 70.38% 6.79% 28.97% 7.09% -4.69% 12.82% 40.79% 24.26% 15.99% 11.86% 7.73% 7.73% 1.65% 10.67% Imperia 21.61% 12.92% 2.74% 25.12% 18.24% 31.34% 14.12% 27.71% 19.31% 28.94% 21.54% 19.03% 12.75% 11.68% 4.78% -24.86% -5.89% 29.76% 4.42% 11.17% -10.50% -10.10% 1.61% 5.42% La Spezia 25.17% -2.16% 11.31% 13.57% 12.68% 7.92% 10.63% 3.78% 6.20% 11.78% 14.45% 6.86% 10.35% 11.05% 4.35% 5.07% 7.89% 15.59% 12.66% 14.04% 4.17% 3.31% 6.90% 11.57% L’Aquila 14.11% 19.30% 12.86% 25.43% 10.59% 11.51% 4.38% 10.96% 14.71% 11.13% 14.85% 11.50% 25.61% 18.16% -0.79% 9.55% 9.01% 6.53% 12.98% 16.30% 11.28% 11.64% 11.44% 4.13% Latina 39.49% 13.68% 17.03% 33.74% 41.86% 39.16% 0.75% 10.53% -3.33% 0.00% 12.18% -3.91% 16.79% 6.42% 11.01% 12.43% 11.66% 8.62% 8.66% 23.91% 6.66% 9.52% 3.76% 2.22% Lecce 10.61% -1.83% 13.72% 23.48% 26.69% 0.66% 14.91% 4.78% 0.98% 3.93% 14.01% 21.38% 21.37% 10.94% 8.07% 10.53% 13.61% 10.50% 15.07% 19.54% 10.28% 12.99% 9.24% 12.33% Livorno 8.62% 18.05% 12.51% 12.78% 10.68% 0.82% 6.92% 1.94% -4.30% 3.22% 7.27% 11.74% 12.33% 14.03% 1.47% 2.10% 4.96% 12.30% 22.57% 14.81% 12.95% 13.64% 18.57% -0.98% Lucca 20.33% 1.69% 18.65% 5.30% 22.78% 14.53% 12.50% 14.19% 18.59% 11.17% 11.08% 3.25% 5.01% 19.19% -6.97% 7.99% 6.67% 12.32% 6.02% -0.04% 3.82% -13.45% 2.20% 14.33% Macerata -2.47% -9.17% 4.23% 17.54% 6.44% 8.08% 24.51% -0.35% 19.70% 5.07% 15.50% 16.92% 15.31% 8.02% 9.32% 27.59% 17.81% 22.21% 13.44% 7.38% 5.46% 3.06% 11.92% 7.97% Mantova -11.64% 0.03% 24.50% 9.43% -3.56% 22.56% 15.82% 11.78% 10.92% 11.42% 15.04% -6.21% 12.42% 11.09% -7.66% 8.21% 6.48% 12.63% 12.46% 18.42% 0.89% 1.79% 3.02% 16.04% Massa-Carrara 31.14% 10.19% 21.43% 34.80% 24.35% 26.05% 12.30% 15.71% 7.67% 15.31% 15.54% 6.87% 11.38% 16.96% 6.30% 9.88% 9.34% 18.06% 9.08% 9.58% 19.59% 8.34% 13.11% 9.97% Matera 11.93% -2.25% 20.67% 1.68% 24.07% 13.01% 6.64% 2.13% -13.70% -3.30% 31.26% 10.63% 11.24% 19.67% 14.42% 17.50% 16.48% 17.02% 15.90% 10.51% 5.03% 2.28% 2.93% 24.15% Messina 34.12% 6.95% 25.58% 0.54% 0.60% 9.76% 12.11% 22.13% -14.64% 14.94% 15.89% 37.51% 16.69% 1.74% -8.85% 3.26% 3.59% 9.61% 16.07% 4.01% 16.98% 5.55% -2.78% 19.94% Milano 17.29% 11.05% 23.26% 15.03% 6.09% 12.09% 9.89% 9.15% -1.89% 11.72% 17.42% 16.57% 11.59% 4.27% -2.27% 2.62% 9.97% 6.91% 7.45% 4.36% 12.80% 7.67% 19.63% 2.86% Modena 2.86% 0.81% 17.27% 9.79% 9.30% 9.23% 11.10% 12.22% 12.22% 20.64% 11.80% 10.25% 10.24% 7.71% -2.50% 2.11% 11.04% 14.41% 9.25% 16.03% 6.53% 0.55% 12.17% 22.35% Napoli 5.38% 10.38% 22.46% 13.26% 11.08% 7.80% -0.82% 7.12% 11.52% 13.48% 23.80% 8.15% 15.99% -2.84% -1.57% 13.05% 10.76% 3.10% 8.70% 6.68% 11.86% 1.93% 20.81% -1.85% Novara 8.71% 10.08% 24.95% 10.73% 9.96% 7.19% 16.52% 2.60% 6.41% 14.53% 14.15% 4.08% 12.19% 0.97% -0.77% -0.11% 8.50% 9.08% 1.36% 14.12% 7.16% 3.82% 13.76% -10.71% Padova 18.75% -7.33% 18.76% 22.41% 11.08% 16.88% 16.75% 6.41% 11.11% 15.40% 22.10% 12.95% 13.31% 4.47% -7.06% 4.66% 7.80% 16.29% 21.49% 9.92% 3.14% 4.52% 3.91% 19.06% Palermo 19.44% 5.46% 18.97% 9.32% 5.35% 6.69% 9.73% 13.34% 0.27% 5.56% 10.88% 13.88% 8.45% -3.40% -0.76% 0.14% 8.53% 6.61% 14.63% 6.91% 13.73% 5.24% 3.71% 14.35% Parma -4.73% -0.42% 27.95% 17.89% 11.79% 17.86% 7.74% 4.27% 16.68% 11.49% 15.83% 9.97% 15.72% 10.36% -6.13% 6.58% 7.92% 10.38% 11.15% 10.11% 8.96% 9.59% 7.61% 20.07% Pavia 20.84% 1.61% 21.26% 20.95% 19.60% 10.28% 15.88% 7.28% 12.17% 16.58% 9.49% 16.05% 11.41% 11.93% 2.06% -3.35% 4.78% 6.05% 6.35% 14.06% 10.38% 7.36% 8.54% 21.33% Perugia 3.05% -6.60% 8.14% 14.70% 14.53% 9.95% 11.96% 5.50% 20.26% 4.48% 15.85% 10.68% 12.40% 3.37% 7.63% 7.64% 12.75% 18.21% 15.10% 10.92% 4.11% 8.98% 11.93% 14.75% Pesaro e Urbino 21.38% -20.24% 28.87% 5.03% 10.54% 3.85% 15.91% 11.94% 15.91% 20.58% 17.36% 13.01% 23.17% 4.94% -12.82% -1.61% 12.04% 20.10% 21.60% 20.37% 6.38% 7.05% 6.72% 7.14% Pescara 26.93% -34.88% 5.43% 12.78% 38.86% -16.81% 2.43% 30.23% 17.10% 25.34% 14.41% 12.41% 26.32% 9.09% -10.97% 7.52% 16.67% 16.77% 14.09% 13.63% 15.82% 6.18% 11.97% 1.50% Piacenza 6.84% -12.43% 14.70% 16.04% 4.22% 19.18% 7.56% 5.39% 22.61% 3.00% 3.16% 12.77% 3.65% 25.74% -6.93% -7.58% 1.74% 23.46% 8.39% 15.46% 7.16% 1.95% 2.53% 12.39% Pisa -2.04% -4.03% 9.48% 17.62% 22.47% 17.34% 6.56% 3.62% 8.52% 6.89% 10.79% 12.65% 10.35% 17.15% -4.51% 10.01% 10.75% 16.20% 16.69% 16.78% 11.07% -1.68% 14.54% 13.43% Pistoia 34.32% -2.77% 14.04% 7.60% 7.16% 5.89% 11.84% 24.02% 3.08% 1.88% 14.76% 11.61% 9.73% 27.20% -3.63% -0.66% 11.14% 12.35% 12.90% 10.23% 10.40% 1.25% 12.58% 11.73% Pordenone 1.74% 14.10% 7.97% 7.43% 3.22% 23.69% Potenza 21.31% 22.78% 29.56% 28.80% 33.19% 27.01% 8.60% 23.32% 25.92% 22.23% 38.46% 16.97% 4.00% 2.29% 5.31% 6.36% 0.70% 11.20% 17.38% 3.66% -7.46% -3.48% 7.67% 28.07% Ragusa 10.47% 22.83% 22.48% 19.77% 19.54% 25.66% 25.62% 15.43% 16.68% 11.81% 10.40% 14.55% 14.25% 9.89% 3.66% -0.03% 10.29% 12.41% 16.70% 9.18% 3.95% 0.40% 1.99% 7.43% Ravenna 4.16% -27.75% 5.62% 22.46% 8.40% 24.45% 5.37% 13.80% 3.92% 6.05% 24.11% 15.94% 13.98% 16.83% -2.20% 6.21% 2.48% 15.75% 14.77% 10.99% 4.39% -3.48% -4.40% 3.51% Reggio di Calabria 1.11% 7.14% 25.23% 14.81% 21.81% 19.25% 11.53% 5.06% 10.22% 4.27% 8.63% 10.20% 8.66% 12.69% 16.99% 15.44% 19.10% 12.06% 13.89% 8.68% 4.91% 2.07% 4.08% -1.83% Reggio nell’Emilia -1.15% -2.91% 18.47% 10.91% 9.87% 1.08% 7.79% 13.86% 10.27% 17.64% 21.31% 7.09% 8.67% 5.64% 0.46% 4.58% 14.39% 14.15% 4.25% 8.70% 5.31% 1.89% 3.22% 20.38% Rieti -10.80% -20.18% 25.13% 17.02% 25.68% 23.49% 11.39% 4.98% -1.40% 9.82% 19.02% 23.63% 17.73% 17.88% 0.55% 1.74% 15.85% 10.42% 15.82% 16.68% 9.99% 1.57% 3.44% 11.87% Roma 16.69% 12.81% 17.98% 16.80% 7.84% 13.27% 9.65% -2.10% -7.59% 9.68% 16.91% 15.97% 11.00% 12.36% 0.39% 1.82% 9.52% 11.76% 10.43% 12.28% 13.16% 8.96% 18.19% -0.33% Rovigo -2.25% -4.47% 46.40% -3.07% 28.60% 3.19% 6.57% -14.07% 33.24% 5.68% -13.18% 7.79% 19.47% 19.08% -2.52% -8.17% 10.31% -1.66% 27.88% 0.05% 4.79% -1.43% -2.30% 23.77% Salerno -13.29% 14.11% 30.27% 23.09% 22.49% 17.65% 13.58% 6.40% 11.07% 21.79% 30.79% 22.81% 12.61% 14.03% 12.73% 6.92% 12.07% 22.80% 12.25% 14.75% 7.20% 20.43% 8.37% 21.34% Sassari 5.53% 37.04% 24.24% 30.76% 24.90% 12.87% 11.23% 5.76% 7.17% 25.57% 28.82% 27.78% 28.87% 34.74% 7.23% 8.02% 25.24% 7.96% 16.17% 15.84% 1.22% 5.45% 0.85% -4.81% Savona 15.64% 26.17% 18.40% 30.30% 4.27% 19.26% 9.28% 24.77% 10.90% 11.20% 23.11% 13.16% 11.38% 7.42% 4.18% 0.19% 5.25% 7.71% 5.15% 20.44% 2.24% -2.86% 5.46% 11.54% Siena -2.34% 5.77% 20.03% 19.11% 20.79% 14.30% 7.32% 6.86% 5.66% 18.79% 13.38% 10.67% 3.51% 12.45% 5.18% 8.49% 18.24% 16.20% 19.83% 4.89% 1.32% -4.25% 22.85% -9.34% Siracusa -60.95% 1.39% 27.32% 18.84% 29.42% 23.91% 6.77% 12.41% 18.60% 24.80% 24.80% 18.88% 14.32% 3.80% 7.62% 11.79% 10.06% 16.21% 19.47% 2.11% 18.07% 6.72% 17.68% 10.31% Sondrio 23.66% 2.39% 27.04% 21.25% 14.67% 18.69% 17.53% 12.78% 13.70% 12.03% 16.53% 10.44% 8.89% 4.01% -3.07% 6.04% 7.59% 11.59% 13.88% 8.67% 5.58% 2.75% 5.11% 11.47% Taranto 29.07% 31.71% 42.60% 7.55% 43.00% -5.28% 21.89% 33.97% -9.78% -16.60% 40.11% 12.45% 9.50% 12.17% 19.81% 26.27% 38.45% 10.45% 20.72% 15.19% 6.50% -2.24% 14.83% 12.80% Teramo -3.24% -13.69% 22.20% 2.00% 25.86% 43.50% 22.42% 24.66% 9.28% 12.34% 19.03% 21.94% 36.71% 4.08% -1.66% 13.49% 15.77% 15.07% 7.21% 9.93% 14.00% 22.27% 12.40% 16.11% Terni 7.56% -5.29% 7.77% 17.80% 23.07% 11.98% 3.76% 10.53% 3.82% 5.66% 12.67% 7.59% 15.50% 7.39% 4.90% 1.10% 6.04% 13.17% 22.40% 11.50% 9.44% 4.99% 9.27% 9.33% Torino 13.36% 6.28% 18.49% 11.50% 21.13% 11.46% 16.66% 18.67% 13.97% 14.41% 21.41% 18.70% 15.08% 10.54% -4.32% 1.83% 9.75% 11.16% 15.61% 5.79% 7.72% -2.59% 12.70% 8.13% Trapani 14.35% 9.63% 25.43% 17.72% 21.95% 16.68% 8.82% 8.36% 9.93% 22.45% 19.40% 18.10% 22.14% 8.46% -2.02% 4.88% 18.63% 15.74% 8.40% 15.77% 10.42% 7.84% 6.50% 14.83% Trento 23.84% 13.73% 12.40% 10.94% 10.04% 13.64% 11.64% 13.83% 7.11% 16.05% 15.50% 9.74% 12.88% 6.64% -7.88% 1.85% 12.92% 12.91% 17.22% 5.80% 3.22% 4.55% 12.01% 14.33% Treviso 38.38% 12.52% 13.15% 14.10% 5.33% 12.86% -0.04% 25.96% 6.76% 14.89% 26.71% 13.50% 1.41% 7.45% -2.53% 5.15% 9.94% 14.96% 10.99% 24.26% 0.30% 3.77% 9.06% 19.26% Trieste 35.37% 28.63% 5.62% -6.43% -9.89% 9.40% 20.34% 19.43% 8.44% 21.70% -2.96% 13.67% 15.99% 43.71% -3.07% 7.96% 19.24% 20.19% 34.32% 29.03% 3.34% 10.11% -11.68% 23.29% Udine 11.90% 2.97% 23.96% 15.45% 11.32% 2.38% 12.27% 17.73% 9.93% 12.61% 13.56% 14.22% 9.20% 5.38% 0.66% 5.68% 8.63% 17.82% 14.13% 18.92% 7.70% -0.74% 2.85% 8.22% Varese 8.21% 8.14% 14.42% 3.09% 18.05% 5.96% 11.99% 5.89% 17.45% 16.90% 19.64% 11.30% 14.78% 4.85% -1.04% 0.04% 9.26% 0.16% 5.08% 14.53% 5.69% 12.19% 20.04% 9.63% Venezia 25.86% -4.35% 5.92% 17.43% 10.09% 8.92% 71.23% 7.22% 8.58% 23.46% 16.19% 3.13% 22.25% 4.87% -21.52% 7.56% 20.61% 11.67% 10.17% 15.60% -8.60% -11.98% 2.06% 12.36% Vercelli -7.61% -9.37% 34.55% 20.94% 18.13% -5.05% 5.06% 11.76% 2.37% 9.75% 10.95% 4.18% 9.00% 16.54% 12.67% -0.93% 3.48% 5.38% 5.84% 17.80% 15.88% 8.83% 11.42% 16.82% Verona 3.40% -9.81% 14.23% 13.49% 10.11% 9.81% 15.10% 14.74% 8.21% 14.44% 12.52% 14.65% 16.32% 12.77% -6.18% 4.05% 7.21% 14.65% 23.93% 11.53% 7.82% 5.15% 6.18% 19.51% Vicenza 3.73% 5.29% 25.82% 11.36% 10.00% 8.81% 12.53% 11.22% 9.79% 20.21% 14.30% 7.40% 12.58% 9.32% -2.59% 4.71% 7.53% 17.66% 15.49% 11.63% -3.33% 1.17% 11.32% 21.34% Viterbo -21.30% -0.25% 27.84% 30.10% 5.35% 20.63% -12.68% -1.43% 10.78% 3.98% 12.04% 16.63% 25.73% 0.31% 11.58% 14.66% 17.10% 18.79% 7.53% -1.45% 13.39% 23.47% 14.12% 14.85% 58 Table B.7 Excess credit reallocation at the province level Provincia 1950 1951 1952 1953 1954 1955 1956 1957 1958 1959 1960 1961 1962 1963 1964 1965 1966 1967 1968 1969 1970 1971 1972 1973 Agrigento 0.00% 0.00% 0.00% 0.00% 0.00% 0.00% 0.40% 0.00% 0.00% 0.00% 0.00% 0.00% 0.00% 0.00% 1.97% 0.00% 0.00% 0.00% 0.86% 0.00% 0.00% 0.00% 0.00% 0.00% Alessandria 0.00% 0.66% 0.00% 0.00% 0.00% 0.00% 0.13% 3.97% 11.47% 8.08% 0.00% 4.19% 4.56% 4.75% 14.99% 2.60% 2.79% 0.79% 0.00% 0.00% 1.01% 2.66% 2.32% 0.28% Ancona 13.55% 4.09% 0.00% 1.01% 3.67% 0.00% 1.45% 0.00% 0.00% 3.21% 0.00% 0.00% 0.00% 10.02% 6.47% 2.84% 0.00% 0.00% 0.00% 0.00% 1.40% 0.32% 0.00% 0.00% Aosta 0.00% 0.00% 0.00% 0.00% 0.00% 0.00% 0.00% 0.00% 0.00% 0.00% 0.00% 0.00% 0.00% 0.00% 0.00% 0.00% 0.00% 0.00% 0.00% 0.00% 0.00% 0.00% 0.00% 0.00% Arezzo 0.00% 0.38% 0.00% 0.00% 0.00% 0.00% 0.00% 0.57% 0.00% 0.00% 0.00% 0.00% 0.09% 0.00% 0.00% 0.00% 0.00% 0.00% 5.10% 3.43% 0.00% 0.00% 0.38% 0.40% Ascoli Piceno 1.61% 0.00% 0.00% 0.00% 0.00% 0.00% 0.00% 0.00% 0.81% 0.00% 0.00% 0.00% 0.00% 0.00% 0.00% 2.81% 0.00% 0.00% 0.00% 0.00% 4.33% 0.00% 0.00% 0.00% Asti 2.05% 0.00% 0.00% 0.00% 0.00% 0.00% 0.00% 0.00% 1.27% 0.00% 0.00% 0.00% 0.00% 0.00% 1.47% 0.22% 0.19% 0.00% 0.00% 0.00% 3.90% 0.33% 0.48% 0.00% Avellino 0.00% 0.00% 0.00% 0.00% 0.00% 0.00% 0.00% 0.00% 0.00% 0.00% 0.00% 0.00% 0.00% 0.00% 0.00% 0.00% 0.00% 0.00% 0.00% 0.00% 0.00% 0.00% 0.00% 0.00% Bari 0.49% 4.10% 2.46% 4.08% 0.00% 0.72% 2.33% 4.89% 0.40% 1.07% 0.00% 0.51% 0.00% 0.42% 4.20% 0.45% 0.00% 0.00% 0.31% 0.00% 0.00% 4.28% 0.00% 0.02% Belluno 0.00% 0.00% 0.00% 0.00% Benevento 4.15% 2.25% 0.00% 0.00% 0.00% 0.00% 0.00% 0.00% 0.00% 0.00% 0.00% 0.00% 0.00% 0.00% 0.00% 0.00% 0.00% 0.00% 0.00% 0.00% 0.00% 0.00% 0.00% 7.95% Bergamo 7.00% 0.00% 0.00% 0.00% 0.00% 0.00% 0.00% 0.00% 1.69% 0.00% 0.00% 0.00% 0.00% 0.03% 0.52% 2.56% 0.05% 0.00% 0.02% 0.00% 3.48% 2.06% 0.06% 0.03% Bologna 10.41% 5.04% 0.00% 0.00% 2.96% 1.18% 0.01% 0.00% 0.06% 0.00% 0.00% 0.00% 0.00% 0.72% 1.07% 1.34% 1.53% 0.25% 0.08% 0.36% 0.85% 0.85% 1.84% 0.08% Bolzano-Bozen 0.00% 0.00% 5.20% 0.00% 0.63% 1.83% 0.00% 0.00% 0.00% 0.00% 0.00% 0.00% 0.00% 0.00% 0.00% 2.00% 0.00% 0.00% 0.00% 0.00% 0.36% 0.55% 0.90% 0.00% Brescia 0.00% 16.09% 0.11% 0.35% 0.00% 0.00% 0.00% 0.24% 0.00% 0.00% 0.55% 0.24% 0.00% 0.46% 0.07% 2.33% 0.16% 0.05% 0.00% 0.00% 3.80% 0.11% 0.17% 0.00% Brindisi 0.00% 15.35% 0.00% 0.00% 0.00% 10.96% 0.00% 10.25% 0.00% 0.00% 9.77% 4.78% 0.00% 0.00% 3.71% 6.34% 0.00% 0.00% 0.00% 6.24% 3.88% 13.21% 2.60% 1.38% Caltanissetta 0.00% 0.00% 0.00% 0.00% 0.00% 0.00% 0.00% 0.00% 0.00% 0.00% 0.00% 0.00% 0.00% 0.00% 0.00% 0.00% 5.16% 1.38% Campobasso 0.00% 0.00% 1.08% 0.00% Caserta 0.00% 0.00% 0.00% 0.00% 0.00% 0.00% 0.00% 0.00% 0.00% 0.00% 0.00% 0.00% 0.00% 0.00% 0.00% 0.00% 0.00% 0.00% 0.00% 0.00% 0.00% 1.09% 1.29% 0.00% Catania 4.08% 4.08% 4.08% 0.79% 0.00% 0.27% 0.00% 0.00% 0.00% 0.00% 0.00% 1.46% 0.00% 7.93% 4.04% 2.64% 0.00% 0.00% 0.00% 0.00% 0.00% 0.00% 5.11% 1.12% Catanzaro 0.00% 0.00% 0.00% 0.00% 0.00% 0.00% 0.00% 0.00% 0.00% 0.00% 0.00% 0.00% 0.00% 0.00% 0.00% 0.00% 0.00% 0.00% 0.00% 0.00% 0.00% 1.05% 2.68% 27.36% Chieti 0.00% 0.00% 0.00% 0.00% 0.00% 0.00% 0.00% 0.00% 0.00% 0.00% 0.00% 0.00% 0.00% 0.00% 0.00% 0.00% 0.00% 0.00% 0.00% 0.00% 0.00% 0.00% 0.00% 0.00% Como 0.00% 6.95% 0.00% 0.00% 0.00% 0.00% 0.00% 2.56% 0.00% 0.00% 0.00% 0.00% 0.00% 9.76% 1.72% 1.88% 0.25% 0.00% 0.00% 0.00% 0.68% 1.06% 0.26% 1.00% Cosenza 0.00% 0.00% 0.00% 0.00% 0.00% 0.01% 0.00% 0.00% 0.00% 0.00% 0.00% 0.00% 0.00% 0.00% 0.00% 0.00% 0.00% 0.00% 0.00% 0.00% 0.00% 0.00% 0.12% 0.13% Cremona 0.00% 0.51% 0.00% 0.00% 0.00% 0.00% 0.00% 0.00% 1.62% 0.00% 0.09% 0.00% 0.00% 1.72% 0.86% 0.00% 0.00% 0.00% 0.11% 0.00% 0.00% 0.05% 0.00% 3.80% Cuneo 1.80% 2.53% 11.01% 0.00% 5.26% 2.37% 6.57% 1.12% 4.73% 0.00% 0.00% 0.56% 1.61% 0.00% 1.93% 2.31% 0.62% 0.15% 0.36% 0.00% 0.02% 1.37% 3.99% 1.20% Ferrara 0.29% 3.44% 0.00% 0.00% 3.26% 0.00% 0.16% 4.64% 3.57% 0.93% 0.00% 0.00% 4.38% 0.00% 2.98% 3.27% 0.00% 0.00% 0.00% 0.00% 3.50% 0.70% 0.80% 7.94% Firenze 2.83% 4.46% 0.00% 0.22% 0.03% 0.00% 1.32% 0.66% 8.10% 1.20% 0.17% 0.00% 0.00% 0.05% 3.13% 4.61% 0.01% 0.00% 0.00% 0.00% 0.30% 0.08% 0.00% 0.03% Foggia 0.00% 8.78% 0.00% 0.00% 0.71% 1.98% 14.03% 5.78% 0.00% 0.00% 0.01% 0.00% 0.00% 0.00% 0.00% 5.05% 0.00% 0.00% 0.00% 0.00% 0.27% 5.12% 4.11% 0.00% Forli 5.64% 0.25% 0.00% 0.00% 0.00% 0.00% 0.00% 1.56% 0.00% 0.38% 0.00% 0.00% 0.00% 0.00% 2.54% 4.17% 0.37% 0.00% 0.00% 0.00% 2.58% 3.37% 2.51% 0.00% Frosinone 0.00% 0.00% 5.37% 0.00% Genova 0.01% 0.00% 0.04% 0.01% 0.66% 0.10% 0.13% 0.06% 1.39% 0.07% 0.00% 0.00% 0.01% 0.06% 0.55% 0.29% 0.01% 0.00% 0.44% 0.00% 0.00% 0.09% 0.00% 0.00% Gorizia 0.00% 0.00% 0.00% 0.00% 0.00% 0.00% 0.00% 0.00% 0.00% 0.00% 0.00% 0.00% 0.00% 0.00% 0.00% 0.00% 0.00% 0.00% 0.00% 0.00% 0.00% 0.31% 0.00% 0.00% Grosseto 0.00% 0.00% 0.00% 0.00% 0.00% 0.00% 0.00% 0.00% 0.00% 0.00% 0.00% 0.00% 0.00% 0.00% 0.00% 0.00% 0.00% 0.00% 0.00% 0.00% 0.00% 0.00% 0.00% 0.00% Imperia 0.00% 0.00% 0.00% 0.00% 0.00% 0.00% 0.00% 0.00% 0.00% 0.00% 0.00% 0.00% 0.00% 0.00% 0.00% 0.00% 0.00% 0.00% 0.00% 0.00% 0.00% 0.00% 0.00% 0.00% La Spezia 0.00% 0.00% 0.00% 0.00% 0.00% 0.00% 0.00% 0.00% 0.00% 0.00% 0.00% 0.00% 0.00% 0.00% 0.18% 0.00% 0.00% 0.00% 0.00% 0.00% 0.00% 0.00% 0.00% 0.00% L’Aquila 0.00% 0.00% 0.00% 0.00% 0.00% 0.00% 0.00% 0.00% 0.00% 0.00% 0.00% 0.00% 0.00% 0.00% 3.13% 0.00% 0.00% 0.12% 0.00% 0.85% 0.00% 0.00% 1.09% 0.00% Latina 0.00% 1.26% 0.00% 0.00% 0.00% 0.00% 1.87% 4.30% 0.00% 0.00% 0.00% 0.00% 0.00% 0.00% 0.00% 0.00% 0.00% 0.00% 0.00% 0.00% 0.00% 0.00% 5.32% 14.04% Lecce 0.00% 0.00% 5.35% 1.69% 0.57% 11.93% 1.19% 1.86% 0.00% 0.00% 0.86% 0.00% 6.40% 0.00% 3.06% 2.28% 0.00% 0.97% 0.00% 0.00% 1.30% 2.48% 3.77% 0.35% Livorno 0.00% 0.00% 0.00% 0.00% 0.00% 0.00% 0.00% 0.00% 0.00% 0.00% 0.00% 0.00% 0.00% 0.00% 0.00% 0.00% 0.00% 0.00% 0.00% 0.00% 0.00% 0.00% 0.00% 0.00% Lucca 0.00% 7.87% 0.00% 1.51% 0.51% 8.82% 0.00% 0.00% 0.00% 0.00% 0.00% 4.30% 0.00% 0.00% 0.33% 0.09% 0.00% 0.00% 0.98% 12.94% 0.00% 1.10% 0.00% 0.00% Macerata 0.00% 0.00% 0.00% 0.00% 0.00% 0.00% 0.00% 0.00% 0.00% 0.00% 0.00% 0.00% 0.00% 0.00% 0.00% 0.00% 0.00% 0.00% 0.00% 0.00% 0.00% 0.02% 0.46% 0.34% Mantova 0.00% 0.00% 0.00% 0.00% 0.38% 0.00% 0.00% 0.00% 0.00% 0.00% 0.00% 0.53% 0.00% 0.43% 0.00% 0.00% 0.00% 0.00% 0.00% 0.00% 0.08% 0.40% 0.30% 0.00% Massa-Carrara 0.00% 0.00% 0.00% 0.00% 0.00% 0.00% 0.00% 0.00% 0.00% 0.00% 0.00% 0.00% 0.00% 0.00% 0.00% 0.00% 0.00% 0.00% 0.00% 0.00% 0.00% 0.00% 0.00% 0.00% Matera 0.00% 0.00% 0.00% 2.72% 8.90% 0.00% 9.91% 0.00% 0.00% 0.00% 0.00% 0.00% 0.00% 0.00% 0.00% 0.00% 0.00% 0.00% 0.00% 0.00% 0.00% 0.00% 0.00% 0.00% Messina 0.00% 1.41% 0.00% 0.78% 3.08% 0.00% 0.34% 0.00% 0.53% 0.00% 0.00% 0.00% 0.00% 0.25% 0.61% 0.00% 0.00% 0.12% 0.00% 0.00% 0.00% 0.00% 3.75% 0.00% Milano 0.06% 0.79% 0.09% 1.02% 0.44% 0.59% 0.15% 1.24% 5.12% 0.04% 0.16% 0.40% 1.83% 1.29% 3.59% 0.75% 0.48% 0.74% 0.63% 1.62% 0.53% 4.93% 0.27% 4.88% Modena 2.59% 4.33% 0.46% 1.41% 0.07% 0.17% 0.00% 0.40% 0.06% 0.00% 0.07% 0.29% 0.23% 0.00% 0.90% 3.59% 0.25% 0.00% 0.00% 0.00% 0.89% 1.38% 0.44% 0.00% Napoli 0.03% 0.01% 0.00% 0.16% 0.00% 0.00% 0.82% 0.07% 0.00% 0.00% 0.00% 0.00% 0.02% 1.48% 0.88% 0.84% 0.20% 0.21% 0.00% 0.17% 0.45% 0.48% 0.44% 3.08% Novara 0.00% 0.00% 0.00% 0.00% 0.00% 0.00% 0.00% 0.00% 0.00% 0.00% 0.00% 0.00% 0.00% 0.00% 0.00% 0.65% 0.79% 0.00% 0.00% 0.00% 0.00% 0.00% 0.13% 0.90% Padova 2.22% 2.47% 0.00% 0.00% 0.00% 0.00% 0.00% 0.00% 0.00% 0.00% 0.00% 0.00% 0.00% 1.44% 3.94% 0.00% 0.00% 0.00% 0.00% 0.00% 0.46% 0.00% 0.41% 0.00% Palermo 0.00% 0.00% 0.00% 0.00% 0.00% 0.00% 0.00% 0.00% 0.75% 0.00% 0.00% 0.00% 0.00% 7.25% 2.63% 3.66% 0.00% 0.00% 0.00% 0.00% 0.00% 0.78% 0.00% 0.00% Parma 6.34% 1.19% 0.00% 0.00% 1.04% 0.00% 0.00% 0.00% 0.00% 0.00% 0.00% 0.00% 0.01% 0.00% 2.86% 0.00% 0.00% 0.01% 0.00% 0.00% 0.62% 2.78% 0.19% 0.09% Pavia 0.00% 9.26% 0.00% 0.00% 0.00% 0.00% 1.53% 0.07% 0.00% 0.00% 15.12% 0.00% 3.41% 0.00% 10.64% 0.86% 1.90% 0.00% 0.73% 0.00% 0.61% 3.92% 0.90% 0.00% Perugia 4.93% 3.72% 6.59% 0.00% 0.00% 1.71% 0.00% 0.78% 0.00% 3.50% 0.00% 0.00% 0.00% 4.86% 0.85% 1.36% 0.00% 0.00% 0.00% 0.00% 2.37% 1.27% 0.00% 1.70% Pesaro e Urbino 0.00% 0.79% 0.00% 6.79% 0.00% 0.26% 0.00% 0.00% 0.00% 0.00% 0.00% 0.00% 0.00% 3.89% 0.00% 10.25% 0.00% 0.00% 0.00% 0.00% 1.87% 0.09% 2.88% 0.00% Pescara 0.00% 0.00% 0.00% 0.00% 0.00% 0.00% 0.00% 0.00% 0.00% 0.00% 0.00% 0.00% 0.00% 0.00% 0.00% 0.00% 0.00% 0.00% 0.00% 0.00% 0.00% 0.00% 0.00% 0.00% Piacenza 0.00% 0.00% 0.00% 0.00% 0.00% 0.00% 0.00% 0.00% 0.00% 0.00% 5.49% 0.00% 0.00% 0.00% 0.00% 0.00% 3.88% 0.00% 0.00% 0.00% 0.00% 0.00% 0.00% 0.00% Pisa 6.17% 1.90% 0.00% 0.00% 0.00% 0.00% 2.20% 0.71% 0.00% 0.00% 0.00% 0.98% 0.86% 0.13% 0.00% 0.00% 0.00% 0.32% 0.00% 1.80% 0.00% 0.00% 0.02% 0.00% Pistoia 0.00% 0.00% 0.00% 0.00% 0.00% 0.00% 0.00% 0.00% 0.00% 0.00% 0.00% 0.23% 1.09% 0.00% 1.68% 0.00% 0.00% 0.00% 0.00% 0.00% 0.00% 0.00% 0.00% 0.00% Pordenone 0.00% 0.00% 0.00% 0.51% 8.51% 2.06% Potenza 0.00% 0.00% 0.00% 0.00% 0.00% 0.00% 0.00% 11.79% 0.00% 0.00% 0.00% 0.00% 7.70% 9.15% 0.00% 0.00% 1.87% 0.00% 0.00% 14.20% 10.43% 8.26% 0.42% 0.00% Ragusa 7.20% 0.84% 0.00% 2.80% 0.00% 0.00% 0.98% 0.00% 0.00% 2.88% 0.68% 0.06% 0.00% 0.25% 0.00% 0.00% 0.00% 0.00% 0.00% 0.00% 0.00% 0.71% 0.00% 0.00% Ravenna 4.88% 0.36% 6.19% 0.00% 2.49% 0.00% 7.21% 1.49% 2.85% 0.00% 0.00% 2.45% 0.00% 0.00% 5.43% 3.21% 1.37% 0.00% 0.00% 0.53% 3.70% 0.57% 4.08% 2.17% Reggio di Calabria 0.00% 0.00% 0.00% 0.05% 0.00% 0.00% 4.37% 2.43% 0.00% 0.00% 0.43% 0.00% 0.27% 0.00% 2.34% 0.00% 0.00% 0.00% 0.00% 0.00% 3.47% 8.74% 0.82% 5.30% Reggio nell’Emilia 0.78% 0.00% 0.00% 0.09% 0.16% 4.03% 0.00% 0.00% 0.00% 0.00% 0.35% 0.02% 0.00% 0.51% 4.46% 0.15% 0.00% 0.26% 0.00% 0.16% 0.00% 7.91% 3.89% 0.00% Rieti 0.00% 0.00% 0.00% 0.00% 0.00% 0.00% 0.00% 0.00% 0.00% 0.00% 0.00% 0.00% 5.09% 0.00% 1.03% 8.16% 0.00% 0.00% 0.00% 0.00% 1.71% 5.02% 1.65% 0.00% Roma 0.08% 1.13% 0.33% 0.35% 0.00% 0.02% 0.00% 5.46% 2.93% 0.00% 2.48% 0.23% 2.07% 0.49% 4.40% 3.10% 0.21% 0.00% 1.45% 0.29% 0.24% 0.07% 0.23% 6.36% Rovigo 0.00% 0.00% 0.00% 0.00% 0.00% 2.34% 1.33% 0.00% 0.00% 0.00% 3.37% 0.00% 0.00% 0.00% 0.00% 6.43% 2.78% 1.80% 0.00% 6.97% 0.00% 2.27% 9.39% 0.00% Salerno 9.45% 1.36% 0.00% 0.00% 0.00% 1.42% 0.00% 0.13% 0.00% 0.00% 0.00% 0.00% 1.28% 0.00% 0.77% 1.46% 0.00% 0.00% 0.00% 0.00% 0.00% 3.33% 1.75% 0.38% Sassari 0.00% 0.00% 0.00% 0.00% 0.00% 0.00% 0.00% 3.08% 0.00% 0.00% 0.00% 0.00% 0.81% 0.00% 0.00% 0.00% 0.00% 0.00% 0.00% 0.00% 2.85% 0.00% 1.82% 1.54% Savona 0.00% 0.00% 0.00% 0.00% 1.56% 0.00% 0.00% 0.00% 0.00% 0.00% 0.00% 0.00% 0.00% 2.92% 9.10% 2.12% 0.00% 0.00% 0.00% 0.00% 0.00% 0.00% 0.00% 5.15% Siena 0.00% 0.00% 0.00% 0.00% 0.00% 0.00% 0.00% 0.00% 0.00% 0.00% 0.00% 0.00% 0.00% 0.00% 0.00% 0.00% 0.00% 0.00% 0.00% 0.00% 0.40% 0.00% 0.00% 0.00% Siracusa 0.00% 0.00% 0.00% 0.00% 0.00% 0.00% 0.00% 0.00% 0.00% 0.00% 0.00% 0.00% 0.00% 0.00% 0.00% 0.00% 0.00% 0.00% 0.00% 0.00% 0.26% 0.64% 0.00% 7.15% Sondrio 0.00% 0.76% 0.00% 0.00% 0.00% 0.00% 0.00% 0.00% 0.00% 0.00% 0.00% 0.00% 0.00% 0.00% 0.00% 0.00% 0.00% 0.00% 0.00% 0.00% 0.00% 0.00% 0.00% 0.00% Taranto 0.00% 0.00% 0.00% 0.00% 0.00% 0.00% 0.00% 0.00% 0.00% 0.00% 0.00% 0.00% 0.00% 0.00% 0.00% 0.00% 0.00% 0.00% 0.00% 0.00% 0.00% 2.87% 0.00% 0.00% Teramo 0.00% 0.00% 0.00% 7.15% 0.00% 0.00% 0.00% 0.00% 0.00% 0.00% 0.00% 0.00% 0.00% 4.27% 10.22% 0.00% 0.00% 0.00% 0.00% 0.00% 0.00% 0.00% 0.00% 0.00% Terni 3.82% 2.15% 0.00% 0.00% 0.00% 0.00% 0.00% 0.00% 10.59% 0.20% 0.00% 0.00% 0.00% 2.25% 0.00% 4.37% 1.04% 0.00% 0.00% 0.00% 0.00% 0.00% 0.00% 0.51% Torino 0.40% 0.40% 0.11% 0.94% 0.16% 0.41% 0.91% 0.00% 0.69% 0.00% 0.00% 0.17% 0.81% 1.75% 0.51% 1.63% 0.00% 1.67% 0.00% 1.13% 0.73% 0.80% 0.67% 0.47% Trapani 0.00% 0.00% 0.00% 1.63% 0.00% 0.00% 0.74% 0.54% 0.92% 0.00% 0.00% 0.00% 0.00% 0.00% 4.44% 0.11% 0.00% 0.00% 2.48% 0.00% 0.00% 2.25% 0.91% 0.00% Trento 0.00% 0.00% 0.00% 0.00% 0.00% 0.00% 0.00% 0.00% 0.05% 0.00% 0.00% 0.00% 0.00% 0.50% 0.00% 1.41% 0.00% 0.00% 0.00% 0.00% 0.00% 1.01% 0.85% 0.60% Treviso 0.00% 0.14% 0.00% 0.00% 0.09% 0.05% 4.89% 0.00% 0.00% 0.00% 0.00% 0.00% 10.20% 0.27% 0.84% 2.86% 1.20% 0.00% 0.00% 0.00% 3.91% 0.61% 2.33% 0.00% Trieste 0.53% 0.00% 0.00% 0.00% 1.57% 0.00% 0.57% 0.00% 0.00% 0.00% 4.03% 0.00% 0.00% 0.84% 0.65% 0.00% 0.00% 0.00% 0.00% 0.54% 0.00% 0.00% 0.27% 0.00% Udine 0.00% 0.00% 0.00% 0.31% 0.03% 0.68% 0.14% 0.22% 0.00% 0.00% 0.00% 0.06% 0.00% 0.10% 4.57% 1.06% 0.99% 0.00% 0.02% 0.00% 2.29% 3.60% 2.91% 5.40% Varese 0.00% 1.08% 7.43% 2.69% 0.00% 3.94% 0.00% 1.70% 0.40% 0.00% 0.00% 2.23% 0.83% 0.00% 2.16% 2.84% 0.00% 1.93% 0.00% 0.65% 0.00% 0.00% 0.00% 0.00% Venezia 0.00% 7.42% 10.94% 0.02% 4.46% 0.00% 0.25% 0.00% 0.00% 2.37% 0.00% 8.33% 0.00% 0.18% 0.00% 4.57% 0.00% 0.00% 0.00% 0.91% 1.22% 0.18% 16.12% 7.59% Vercelli 35.80% 3.66% 0.00% 0.00% 1.02% 6.06% 0.00% 0.00% 3.88% 0.00% 0.00% 2.80% 0.00% 2.06% 2.54% 3.42% 0.75% 0.00% 4.52% 0.00% 0.00% 0.98% 0.00% 0.00% Verona 0.00% 0.00% 0.00% 0.00% 0.00% 0.43% 0.00% 0.03% 0.00% 0.00% 0.00% 0.00% 0.00% 0.00% 0.15% 0.00% 0.00% 0.00% 0.00% 0.00% 0.00% 0.02% 0.00% 0.00% Vicenza 1.89% 0.03% 0.00% 0.00% 0.76% 0.00% 0.65% 0.00% 7.04% 0.00% 0.22% 0.02% 0.01% 1.64% 0.83% 0.17% 0.58% 0.14% 0.00% 0.00% 0.59% 0.54% 0.51% 0.00% Viterbo 17.22% 8.96% 0.00% 0.00% 18.27% 0.00% 4.10% 0.00% 0.00% 9.05% 0.00% 0.00% 0.00% 8.46% 0.00% 0.00% 0.00% 0.00% 0.00% 0.00% 0.00% 0.63% 0.00% 0.00% 59 B.2.2 Province Characteristics We collect data from historical censuses held in 1951, 1961, and 1971. The main problem is that the data is not digitally available. Only scanned census documents are accessible at the Italian National Institute of Statistics’ (ISTAT) website.1 We manually extracted data for province characteristics using scanned census documents. Particularly, we use general population censuses (“Censimento Generale Della Popolazione") and industry and commerce censuses (“Censimento Generale Dell’Industria E Del Commercio") to obtain province char- acteristics. Using general summary data (“Dati Generali Riassuntivi") from censuses, we can extract a good amount of useful data at province level. We obtain economic province characteristics from industry and commerce censuses. We use share of individual firms as an indicator for economic development2 . We also get number of firms, workers, and bank branches from industry and commerce censuses. We add number of workers per firm and number of bank branches per firm to control for economic and financial characteristics of provinces. Figure B.9 presents an example from the 1951 census of industry and commerce. We acquire population and education related characteristics from general population censuses. We use share of active population as an indicator of labor force participation and share of higher education degrees as an indicator for level of education at a province. Figure B.10 presents an example from the 1951 census of population related with education. 1 ISTAT catalog can be accessed at ebiblio.istat.it. 2 Guiso et al. (2004a) show that individuals are more likely to start a business in more developed regions in Italy. 60 150 Segue T AV. 2 - Unità looali amministrative - Unità ' looali 'opem ti ve oon e senza forza motrice c- Provi ncie, per r am o di attività economica : TOTALE RAMI DI AT TIVITÀ ECONOAUCA U NrrA LOCAL TOTAl.E AM~IL"ISTR . UN IT À LO CALI OPER ATIV E UNITÀ LOCALI al ~ con e senza forza motrice co n forza motrice :a... c PROV INCIE t otal e a rtigi ane totale a rtigiane ;" ad - Z N d etti a dd et t i addetti I potenza I p o ten za N addetti N I N I operai e manovali N I a d de t t i N I N I utl· operai e lizzablle m anovali HP N I a d de tti liz~~~l1e HP 1 Alessandrta . 83 359 1 9 .597 6~ 273 40 .561 7 .275 11.64J 3 .405 50 . 197 37.272 170. 629 1.937 4 .263 8 .821 9 .680 62 .632 2 ARti 16 35 4.662 18 .770 10.223 3 .792 5 .731 1.406 13 .428 9 .269 31. 706 881 1.889 5 .124 4 .678 18.805 3 Cuneo. 94 299 10.834 44 .745 24610 8673 13 316 3 .688 31.509 21 .441 136895 2.470 5 .028 13.797 10.928 45 .044 4 Nov ara. 64 581 8873 93 .801 69 716 6 .193 9 .915 3 .572 81. 631 64.801 332 .022 1.892 4217 9 .764 8 .937 94 .382 5 Torino 592 12 .904 29.305 341.584 250 .178 21. 579 37 .200 12 .493 301. 535 235 .094 1.012 .696 6 .990 16 .215 33 .529 29 897 354. 488 6 Y'e rcelli • 64 224 8 .153 95 .299 73 .559 5956 9 .173 3 .286 85.043 69 .932 228 .127 1.737 3643 8 .012 8217 95 .523 7 ' Ao sta (Vall o d i) . 20 88 1.452 19.191 14 .954 922 1.456 636 17 .067 14.086 331. 178 336 596 1.852 1.472 19.279 8" B ergamo 107 893 9 .392 113.360 81 .308 6 804 11. 151 3 .600 100 .309 82 .566 383 .287 1.934 4 .509 10.053 9.499 114.253 9 B resci a . 156 1.027 12455 103 .088 72 .965 9 .521 15 .184 4603 , 85 .900 67 .514 325 886 2 .768 5938 13 .384 12 611 104.115 ' 10 Como . 142 1. 151 12535 129570 93674 8646 . _ 14887 5838 114.152 88 .028 " 290 .117 . 3 .159 7 .096 16.791 12 .677 130 .721 I l Cremon~ 38 238 . 7 438 34 840 20 785 6 .159 8894 2 :097 25003 18 .314 74552 1.390 2 .766 6 .730 7 .476 35.018 12 Mantova 51 164 9 559 31.682 13. 747 8 093 12 387 . 2 326 17576 10 .734 46 .171 1.491 3473 7562 9 .610 31.846 13 MlIano 2 .054 34 .383 52 .102 648504 461.707 36 262 67068 24 383 574. 714 432 006 1.916 .471 12575 31.061 49 .250 54 . 156 682 .887 14 Pavia. 55 320 11.330 13765 48803 8714 13.392 3810 58 371 43 .889 151.562 2 .192 4853 10.626 11. 385 74 .085 15 So ndrio . 37 113 2 .251 14 .953 10 .414 l 567 2440 1.0 04 lO 897 8 .261 133413 625 1.137 3 .249 2288 . 15 .066 16 Vare se 86 442 9 487 145 .149 113..113 6 037 10.652 4 .981 133 .761 108.396 380 .830 2 347 5 .526 12 .123 9 .573 145 .591 17 Bo lzano. 116 719 7 .258 35 . 134 21 612 5 550 8847 2 .972 25 .644 18 427 195 .556 1.951 3 .100 12.103 7 .374 35 .853 18 Trèn to 143 504 8022 42565 27 .597 5840 8 .783 3 587 31.743 23242 174 .383 2 .349 4188 12.605 8 .165 43 .069 19 .Belluuo . 74 309 3 .518 17559 10.793 2 .567 4 .009 1.391 12437 8 .964 92706 823 1.616 4 .254 3 .592 17.868 20 P adova. 141 1 333 10.036 52530 30018 8 .070 14.014 2 .501 35 680 25 617 104 .212 1.3 85 3 .569 9 .022 10.177 53 863 21 Rovigo. 34 . 125 5 .681 19.248 9 .376 4 846 7 .536 1.064 .10 .702 7 .572 99 .294 612 1.611 4 .849 5 .715 19.373 22 Trev is o . 53 19é 8 740 54 .725 33 229 6876 11861 2 426 42269 30 790 94 319 1.262 3.391 7 .220 8 .793 54 .9 23 23 166 U~ Ven ezi a . 7867 67 .089 45 .486 5 .969 10.545 2. 316 50 943 38 960 244 .679 1. 224 3 .156 8664 8 .033 70 .322 24 10 .451 g ~ 32 .2.907 V erona . 118 57 .621 35 .072 8 . 180 41.562 30503 109 .103 1. 656 3 .777 7 .253 10.569 59 . 125 25 Vi cenza. 85 691 9 .542 79 888 55 .230 1 .456 19 66 .608 51. 805 137 .776 1.892 4522 ' 10 .290 9 .627 80.579 2b 22 27 Gor iz ia. Udinc . 119 l~i 11.630 59 1.733 ' ~~:6~ 15675 42 469 1.295 9 .139 1~:g~ 3 .461 459 18 484 50 .740 14789 38 254 68081 163 609 212 1.972 521 4 578 1.235 12 .481 1.755 11.749 21.341 67.673 28 Trieste (T err it ori o d i ) 345 2 .724 4 .123 47.208 31. 824 2 .766 5.31 6 1.225 32 .254 22 .951 112 .765 564 I.m 2 .933 4 .468 49 .932 29 Genova . 786 14 .045 13.782 155.412 109 574 9383 15448 5 .224 120 .970 95230 518429 2 674 5495 19.496 14.568 169 .457 30 I m per ia . 70 3304 12026 6 088 2 544 ~ 3 .850 1. 058 7 .085 4 .533 20 471 645 1.190 2 .143 3 .374 12.32O 31 L a Sp ez ia : 64 3 .143 23 '561 16.053 2 277 3 .456 1.127 17.645 13 .475 77 .620 594 1.141 4 .796 3 .207 24. 017 32 Sa vona. 108 52< 3 .922 36. 827 26 .244 2 .758 4 .263 1. 500 30 .405 23 .582 164.871 799 1.572 3 .018 4 .030 37.351 33 B ol og na. '. 436 4 64C 14.910 88 274 56. 898 I l. m 18.52' 4 .257 64.524 47 .664 142.936 2 .345 5 .455 13.634 15.346 92 .914 34 F erra ra. 64 - 62C 6 .766 31. 155 17.435 5 509 8481 1.328 18.597 .13 .426 100 .021 685 1.563 3 .571 6 .830 31. 775 35 F orli . ' . 79 231 8686 32 .784 17.074 7 .184 10961 2242 19.290 13 .068 60 582 1.505 3 .218 9. 067 8765 33 .015 ~~ 36 Mod ena. 69 9526 42 527 23 .990 7 .654 11.904 2 .726 29 .193 20 .7 15 79 .486 1.594 3 .556 9 .846 9 .595 43 .187 37 Parma . 109 7808 31.814 17.124 6 277 9 607 2 550 21.095 14.037 68 .562 1.580 3 .230 9 .001 7 .917 32 .794 38 Piacenza 61 35' 5 .720 27 .741 16.347 4502 6 .962 1.942 19.205 13.224 81. 843 1. 231 2 499 6 .790 5 .781 28 .100 39 Ravenna . 41 15 5 .913 21.850 11.442 4926 6895 1.6 56 13.439 8 .927 44 .500 1.066 2 .030 6 .606 5 .954 22 .007 40 R eggio nelJ'E~iJia 135 394 7 .835 31.004 15.891 6 .520 9 .872 2 .218 19 .567 12 .509 72 .452 1.445 3 .134 8 .212 7 .970 31.398 41 A rezzo 1~ 33 5 .683 23 .929 14.272 4 .592 6 .602 1. 515 16 .376 12.276 50 .090 949 1.842 6 .342 5.71 6 24 .05 2 42 F ir enze : 366 6 .3 15.809 116.360 79 .326 11.480 20 .573 6399 · 91.512 69 .314 198 .768 3 .520 8388 17.749 16. 175 122.716 43 Gro sseto. 16 55 3 .569 19.411 13.073 2 .732 3 .894 1. 039 13 867 11.007 46 429 610 1.168 3. 715 3 .585 19 .466 44 Liv or no . 82 390 3 .767 37 .724 27 .439 2 665 4 .006 1. 091 29 .645 23 .644 186 .717 538 1. 170 3 .090 3 .849 38 .114 45 L u cca . 73 270 6 . 134 . 38 .757 26898 4620 7.289 2 . 141 29 .998 23 .976 83896 1.222 2 .635 6 .643 6 .207 39 .027 46 Massa'C~r~a~a : 59 378 2820 20.243 14.190 1. 896 2 .954 999 15470 12 .199 130 .388 503 1. 018 3 .955 2 .879 20 .621 47 Pisa 46 482 6 .772 36 . 193 22 .341 5 .351 8 .432 2235 27.064 19 .588 108 .185 1.355 3 .069 9 .280 6818 36.675 48 P is t ola . 35 94 4 . 191 21. 119 12.906 3 .300 5 .304 1.382 15 .270 l Ll 04 46 .302 810 1.769 4 .751 4 .226 21.213 49 Si ena . 68 342 5 . 134 20 .555 12.017 3. 903 5 .752 1.5 38 13 .774 10.221 32.698 892 1.779 4 .660 5 .202 20 .897 50 Peme ìa . 62 489 8 .700 37 .487 22 .554 6 .933 9 .953 2 .300 25 .041 19 .227 64 .631 1.405 2 .701 8 .058 8 .762 37 .976 51 T e rni . 86 794 3 .411 • 24 .9 11 17.794 2 .637 3 .868 980 19 .945 16 .334 274 .045 591 1.124 3 .726 3. 497 25 .705 52 An cona, 138 1.619 7 .857 38 .738 23 .692 6 . 196 8 .776 1.946 27.185 20.302 57 .547 1. 142 2 387 6560 7 .995 40.357 il 53 Asco li Pic~n~ 68 165 6684 18.214 7 .322 5 .445 8 .440 1.659 8764 5 .051 26. 184 1.092 2 311 5.312 6 .752 18 .379 54 Ma cera ta , 36 . 5 .721 19.077 9 .761 4 .549 6632 1.670 11.671 7 .820 38306 1.072 2 .117 5 .830 5 .757 19. 265 55 P es a ro e Urbi~o : 77 4 .540 16.032 8 .273 3 .582 5 .366 1.477 9 .607 6 .293 24 .486 1.029 1.919 5 .274 4 .617 16.24 2 56 Frosinone , 35 103 5 .703 20 .593 l Ll46 4561 6 .735 1.032 l O 483 7.949 49 .940 548 1. 102 6 383 5 .738 20.696 57 L a tina . 26 65 3 .287 12 .156 6.664 2572 3 .808 701 5.550 3 .842 21 .818 371 804 3 .083 3 .313 12.221 58 Rieti • lO 27 2819 8 610 4 . 147 2 316 3 .209 551 4 .643 3 .306 49 900 357 676 2.418 2829 8 .637 59 R oma . 798 25 . 0~9 23215 195.523 137 .013 17 .075 32 . 185 6 .517 116 .576 86 .941 359 .715 3 .608 9 .378 24 .016 24. 013 220 .602 60 V iterbo " 36 5 .074 14 .451 6.286 4 .017 5 .896 1. 107 7.035 4 .536 18.401 647 1.261 3 .858 5 .110 . 14.547 bi Cam po basso . 41 154 7 .138 15 .782 5 .226 6 .048 8 .608 1. 016 5.674 3 .559 22645 632 1.338 6 .122 7 .179 15.936 62 Ch iet i. 73 23 6 .566 19.270 8 .372 5 .271 8 .021 1.292 9 .078 5 .858 32 .481 810 1.886 6 .219 6 .639 19 . 506 L'Aqull~ 63 64 P es cara . 53 41 lli 4 .967 3 658 15.906 15.699 8 .374 9 .036 4 .091 2 .854 5 .371 4 .358 646 796 6 .760 10 .243 5 .127 7.675 39 .896 30. 165 404 354 717 903 3 .634 2 .549 5 .020 3 .699 16 .282 15 .953 65 66 T era m o . Avellino. 20 21 10~ 4. 738 114 7. 467 13 .870 17.501 6 .218 6 .458 3 .99. 6 358 6 .072 8. 664 999 653 7 .104 5 .275 4. 734 3 .685 106 . 116 16.570 657 364 1.419 744 4.667 4 .962 4 .758 7 .488 13 .976 17.62 1 67 B enevento: 68 Caserta . 17 62 11 I 4 153 7 .584.831 12 .313 20 898 4 .954 9 . 142 4 . 105 6 .395 5 .932 8 883 708 787 5 .685 8 .242 3 .941 5 .846 14.788 25 496 407 425 967 932 5 .782 4 .951 4 .848 7 .646 12.432 21.05 1 69 N apoli 813 8523 21.359 142 .755 " 92 .435 16 .984 30 .806 4 .260 94 .230 73 .286 380. 277 2 . 187 6. 226 13.13 1 22 . 172 151.27 8 70 Salerno : 152 587 13.287 . 53 .811 31.28i 10 .781 15 .748 2. 183 33 .092 26.256 76 043 1. 056 2 .520 9.388 13.439 54.458 71 B ari 133 3 .12t 15 . 143 57 .949 29 .259 11.638 18 .630 3 .262 31.735 22 .564 102 .947 1.243 3 . 127 8 .469 15.276 61.0 75 72 Brindisi: 29 120 4 .6 04 12.358 4.940 3 .506 5 .110 947 4 .832 2 .841 14982 297 694 2 .524 4 .633 12 .478 73 Foggia , 44 624 8 .666 30 .689 14 .685 7 .031 I l .405 1. 152 12:268 8. 994 31 .514 508 1.301 6 .076 8 .710 31. 31.3 74 L ecce . 8 255 9 .688 31.422 16 .277 7 .280 11.097 1.599 9 .345 6 .284 24 -.484 609 1.544 4.619 9 .770 31. 677 75 Taranto: 34 207 5 .321 17 .709 8 .561 4 . 138 6 .468 840 8 .375 5 .808 25. 540 406 1.079 3 . 109 5 .355 17.9 16 76 Matera. 34 II I 2. 958 7 .892 3 .420 2 .509 3 .728 383 2 .646 1. 831 7 .395 201 472 2. 427 2 .992 8 . 003 77 Potenza. 43 394 7.575 17 .463 6 .323 6 .617 9 .519 860 4 .114 2 .387 11.849 571 1. 143 6 .076 7 .618 17 .857. 78 Catanzaro . 61 378 11. 624 28.461 11. 495 9 .588 13 .214 1.907 11.5 35 7 .489 52 .904 1. 049 2.017 8 .37 1 11.685 . 28. 839 79 Cose n za . 65 224 9 .839 30 .798 15.751 7 .974 I l .447 1.482 13.070 9 .768 44 896 795 1. 654 7 .3 16 9 .904 . 31.022 80 R eggio d i 'C;"l~b~ia : 101 1.497 7 .965 22 .773 9 .84 1 6 .175 9 .308 1. 340 9 .115 6 .093 19 .730 626 1. 359 3 .820 8 .066 24 .27 O 81 Agrigento . 74 235 7 .292 20 .463 9.369 6 .019 8 .119 988 7 .873 5 .623 26 .303 534 1. 164 4 .351 7 .366 20 .69 8 82 CaitaniBBetta: 28 140 3 .696 12.575 6 .201 . 3 .090 4 .541 390 6 347 4 .843 15.404 202 500 2 .188 3 .724 12.715 83 Catan ia . 84 699 12 .728 37.772 16.753 10. 637 15 .704 2 . 186 17.618 11.507 45 .580 1.282 3.149 9 .160 12 .812 38.471 '8 4 E n na . 38 93 2 .701 10.046 5 .734 2 .172 3 . 114 346 5 .211 4 345 19 .M9 183 423 2 .437 2 .739 10 .13 9 85 Measìna.. 176 584 9 .603 31.598 15 .437 7 .567 11.161 1. 824 15.027 10 .096 39 .880 967 2 .259 6 .716 9 .779 32 .182 86 P a ler mo. lI5 3 . 100 12.760 47 .033 24 .167 10.599 16.691 2 .483 23 .935 16 .738 68 .500 1.509 3 .793 11.520 12 .875 50 .133 87 Ragusa . ~ ,~, 10 .427 3 .973 3 .421 5068 556 3 .983 2.423 13 .045 366 927 3 .759 3 .953 10 .486 88 Si rac nsa. 25 204 4 .5 17 11.338 4 .034 3 .781 5363 722 4.976 2 .891 15 .592 376 901 2 .931 4 .542 lI .542 89 Tra panI. H 66 2 6 .729 17 598 6 .880 5 .318 7 .613 1.193 . 7 . 167 4 .247 22 834 533 1.344 4 .565 6 .795 17.804 90 CagUar i . 89 1.996 8 . 176 52 .645 37 .133 6 .464 9 .547 l 613 36 .928 31.038 178 .002 875 1. 850 6 297 8 .265 54 .64 1 91 N u oro. 92 Sassar i. : 12 40 34 3 590 183 5 .173 9.335 16.438 4 .047 7 .053 2 .834 3.884 4 .091 I 6 .062 1.. 191 604 2 .978 7.415 1.730 4 .338 9 .877 27 .120 380 696 765 1.474 2 .846 4 .950 3 .602 5 .213 9 .369 16.621 IT A LI A 11. 924151.077 756 .116 4 .670 .126 3.018 .987 585 .66 9 926 .112 1216 .037 3.432 .526 2 .59 2 .69 2 12 .4 10 .648 120 .5 54 271 .903 710 .397 768 .040 4 .8 21 . 203 (a ) (a) Cfr. n ot a a pago 13 9 Figure B.9 An example from the 1951 census of industry and commerce 61 306 VOL. VII - DATI GENERALI RIASSUNTIVI ISTRUZIONE 307 TA v. 37 - Popolazione residente di oltre 6 anni, per sesso, grado di istruzione e provincia FORNITI DI TITOLO DI STUDIO PRIVI DI TITOLO DI STUDIO Segue TAV. 31 - Popolazione residente di oltre 6 anni, per sesso, grado di istruzione e provincia PROVINCIE TOTALE E Laureati Diplomati Licenza di scuola Licenza di scuola Alfabeti Analfabeti FORNITI DI TITOLO DI STUDIO PRIVI DI TITOLO DI STUDIO REGIONI media inferiore elementare PROVINCIE TOTALE MF M MF M MF I I MF I M I M MF I M MF I M MF I M E Laureati Diplomati Licenza di scuola media inferiore Licenza di scuola elementare Alfabeti Analfabeti REGIONI MF M MF M MF M MF M MF M MF M MF M Alessandria 3.620 2.844 16.576 9.142 27.751 16.097 325.480 156.801 57.860 26.421 16.217 7.212 447.504 218.517 I I I I I I I Asti 1.425 1.165 5.628 2.793 10.093 5.526 160.046 78.928 25.716 12.181 5.571 2.595 208.479 103.188 Cuneo 3.197 2.516 13.598 6.912 19.546 11.212 412.577 202.280 66.415 33.085 15.794 8.042 531.127 264.047 - Novara. 2.983 2.472 12.853 7.238 27.408 15.094 300.532 140.660 37.585 16.768 9.678 3.848 391.039 186.080 Perugia 3.924 3.107 14.472 7.807 20. 318 12.740 313.132 162.028 94.917 46. 379 75.768 28.496 522.531 260 . 557 Torino 19.879 16.116 58.919 33.841 151.179 77.278 941. 969 436.206 142.700 63.476 24.097 9.611 1.338.743 636.528 Terni l. 333 1.075 6.044 3.555 iO.212 6.299 122.830 63.513 33.695 16.702 27.030 9.491 201.144 100.635 Vercelli 2.355 1.963 10.905 6.356 22.574 12.910 262.500 121.333 43.583 19.393 12.150 4.833 354.067 166.788 Umbria 5.257 4.182 20.516 11.362 30.530 19.039 435.962 225 .541 128.612 63.081 102.798 37.987 723.675 361.192 Piemonte. 33.459 27.076 118.479 66.282 258.551 138.117 2.403.104 1.136.208 373.859 171.324 83.507 36.141 3.270.959 1.575.148 Valle d'Aosta 511 431 2.300 1.225 4.438 2.416 64.431 32.187 11.611 5.654 2.235 983 85.526 42.896 Ancona 3.060 2.391 12.469 7.214 19 .389 11.807 237.966 117.209 50.965 23.980 36.897 11.932 360.746 174 .533 Ascoli Piceno 2.051 1.644 8.206 4.710 10.836 6.670 163.417 84.697 57.280 28.698 53.416 16.198 295.206 142.617 Macerata 2.055 1.672 7.602 4.074 10 .019 6.232 168 .396 85.788 44.548 21.987 38.698 11.575 271.318 131.328 Bergamo 3.493 2.791 15.432 8.133 30.921 17.459 475.106 226.280 73.212 36.473 12.839 6.404 611.003 297.540 Pesaro e Urbino 1.985 1. 553 8 .633 4.634 10.976 6.698 184.747 94.092 50.982 25.041 41. 572 16 . 734 298.89;; 148.752 Brescia. 4.434 3.556 18.560 9.348 34.726 19.152 582.935 281.092 95.882 48.215 20.775 10.807 757.312 372.170 Como. 3:638 2.945 15.142 8.053 37.767 19.799 395.747 186.913 47.314 22.012 9.142 3.979 508.750 243.701 Marché 9.151 7.260 36.910 20.632 51.220 31.407 754.526 381.786 203.775 99.706 170.583 56.439 1.226.165 597.230 Cremona 2.363 1.837 10.458 5.462 18.693 9.925 254.865 124.268 H.734 21.813 14.287 7.132 345.400 170.437 Mantova 2.178 1.750 9.175 4.557 16.675 9.547 279.238 138.937 55.441 27.030 20.677 8.688 383.384 190.509 Frosinone .. 1.931 1.661 6.716 4 .059 11.189 7.752 210.228 116.668 84.979 42 .827 90.723 24.971 · 405.766 197 .938 Milano 37.726 30.505 119.514 67.877 297.613 146.609 1.561. 707 724.429 237.747 107.655 52.717 23.448 2,307.024 1.100.523 Latina 1.123 943 5.086 3.046 8.604 5.782 130.008 68.646 56.475 28.011 40.227 14.116 241.523 120.544 Pavia. 3.894 2.930 16.773 8.774 29.879 16.536 345.667 167.550 57.595 27.175 18.187 8.164 471.995 231.129 Rieti. 847 735 3.404 1. 832 4. 864 3.111 89. 846 48.522 34 .833 17.594 25.650 8 .678 159.444 80.472 Sondrio 813 650 3.846 1.864 6.771 3.709 107.385 51.661 13.828 6.837 2.816 1.318 135.459 66.039 Roma 59.163 48.051 139.037 75.824 234.132 127.960 1.085.958 514.852 299 . 380 126.578 114.304 37.087 l. 931. 974 930 .352 Varese 3.275 2.618 15.374 8.723 38.530 21.154 329.356 151.306 42.278 19.141 7.548 3.173 436.361 206.115 Viterbo l. 174 l. 006 5.299 3.262 6. 831 4.495 133.679 69.00.7 50.921 25 .759 33.268 12.580 23ì.172 116.1Og Lombardia 61.814 49.582 224.274 122.791 511.575 253.890 4.332.006 2.052.436 668.031 316.351 158.988 73.113 5.956.688 2.878.163 Lazio. 64.238 52.396 159.542 88.023 265.620 149.100 1.649.719 817.695 526.588 240.769 304.172 97 . 432 2.969.879 1.445 415 Bolzano 2.711 2.369 9.704 5:469 22.859 11.985 227.625 109.884 29.363 14.770 3.434 1.642 295.696 146.119 Campobasso 2.042 J .742 7.033 3.986 9.361 6.029 178.180 92.083 87.098 44.895 74.176 24.171 357.890 172.906 Trento '. 2.662 2.256 11.852 6.535 21.104 10.945 287.026 137.120 27.087 13.503 2.425 1.188 352.156 171.547 .Chieti 2.031 l. 668 7.301 4 .145 10.625 6.865 178 .213 94.866 73.438 36.772 80.414 23.215 352.022 167.531 T:rentino-Alto Adige 5.373 4.625 21.556 12.004 43.963 22.930 514.651 247.004 56.450 28.273 5.859 2.830 L'Aquila 2.115 l. 704 8 .569 4.708 11.951 7.444 189.658 96.389 72 . 666 35.418 38.631 12.197 323.590 157.860 647.852 317.666 Peseara l. 617 1.297 6.630 3.943 9.719 5.978 107.526 55.905 46.927 23.341 40.187 12.949 212.606 103.413 Teramo 1.272 l. 065 5 .241 2 .973 6.798 4.429 120.561 64.294 53 . 110 27.064 54.513 18.121 241.495 117.946 Belluno 999 831 4.846 2.670 7.963 5.020 158.782 77.555 34.665 16.334 6.724 2.621 213.979 105.031 Padova 4.765 3.805 16.680 9.287 28.344 16.733 425.688 209.180 115.607 54.107 43.663 18.187 634.747 311.299 Abruzzi e Molise 9.077 7.476 34.774 19.755 48.454 30.745 774.138 403.537 333.239 167.490 287.921 90.653 1.487.603 719.656 Rovigo. 1.309 1.073 5.549 2,943 9.071 6.079 192.170 100.208 69.234 31. 776 37.023 13.068 314.356 155.147 Treviso. 2.690 2.282 12.526 6.952 21.141 13.178 395.520 193.337 82.890 38.988 29.120 11.730 543.887 266.467 Avellino 2.738 2.333 9.353 5.948 12 .155 8 .534 190.312 105.498 105.683 52. 892 108,799 33.858 429.040 209.063 Venezia 5.602 4.610 19.400 10.721 41.031 24.032 404.706 199.978 127.540 59.400 57.723 23.646 656.002 322.387 Benevento 1.674 1.452 5.968 3.716 7 .950 5.420 131.737 72 . 570 65 .181 33. 338 . 77 .913 23.795 290.423 140.291 Verona. 4.025 3.356 16.177 9.000 29.284 17.519 421.347 205.423 78.676 36.662 27.170 10.846 576.679 282.806 Caserta. . 3.501 2.963 12.264 7.195 18 .822 12.798 222.094 117.385 124.201 59.909 136.074 49.297 516.956 249 .547 Vicenza 2.942 2.340 12.731 6.928 22.780 13.447 398.690 192.720 81.156 38.102 20.564 7.950 538.863 261.487 Napoli 26 . 281 21.194 72.065 40 .675 126.407 77. 200 789.805 382.916 409 . 584 186.110 355.064 144.019 l. 779.206 852.114 Veneto 22.332 18297 48.501 159'."814 Salerno. 5.408 4.422 17 . 268 10.458 26.762 17.602 311.823 161.728 179.949 90.431 180. 527 66.271 721. 737 350.n2 87.909 96.008 2.396.903 1.178.401 589.768 275.369 221.987 88.048 3.478.513 1.704.624 Campania. 39.602 32 .364 116.918 67.992 192.096 121.554 1.645.771 840.097 884 .598 422.680 858.377 317 : 240 3.737.362 1.801.927 Gorizia. 1.025 856 6.183 3.274 15.261 8.639 86.500 39.542 9.717 4.552 3.511 1.331 122.197 58.194 Udine 3.637 3.073 18.078 10.200 31.137 19.893 528.207 261 :629 100.800 46.755 35.595 10.626 717.454 352.176 Bari 10.473 8.593 26 .598 16.067 48.679 30. 855 474.561 226.902 231.106 111.402 234. 817 106.769 l.026.234 500.588 Brindisi 1.602 l. 327 4.724 2.999 8.592 5.844 108.738 57.398 68 .924 35.331 76.351 28; 702 268. 931 13l.601 Friuli-Venezia G. 4.662 3.929 24.261 13.474 46.398 28.532 614.707 301.171 110.517 51.307 39.106 11.957 839.651 410.370 Foggia 3.434 2.905 11.937 7 .404 18 ,341 11 949 257.945 131.216 128.772 64.093 138.652 57.581 559.081 275.148 Trieste (Territ. di) 5.255 4.366 21.865 12.254 60.444 29.523 166.960 73.480 17.523 7.516 6.712 1.980 278.759 129.119 Lecce 4.075 3.439 11.776 7.137 . 18 .313 12.069 235 .909 118.851 137.060 69.603 131.026 52.082 538.159 263.181 Taranto 2 .613 2,221 8 .497 5.327 17.162 11.496 166.054 84.319 86 -070 42 .833 81.076 361.472 178.302 I 32. 106 1 Genova 14.828 11.788 53.477 30.759 97.220 53.824 519.375 237.799 150.170 64;981 31.247 13.355 866.317 ,412.506 Puglia. 22.197 18.485 63.532 38 .934 111.087 .72.213 1.243.207 618 686 651.932 323.262 661.922 277.240 2.753.877 1.348.820 Imperia 1.821 1.467 6.944 3.675 14.100 8.161 101.161 48.126 24.187 10.802 6.793 2.518 155.006 74.749 La Spezia 1.868 1.501 7.945 4.626 17.248 11.399 143.197 68.663 29.692 13.499 14.786 4.695 214.736 104.383 Matera 869 746 2.810 l. 718 3.741 2.639 70.313 36.102 33.207 16.898 45 .046 19.719 155.986 77.822 Sayona 2.325 1.851 8.508 4.656 18.107 10.982 151.437 71.291 32.183 15.029 8.574 3.823 221.134 107.632 Potenza l. 949 l. 672 6.443 3.994 8.909 6.230 161. 700 83.995 91. 904 48. 305 111.977 43 .631 382.882 187.827 Liguria 20.842 16.6(17 76.874 43.716 146.675 84.366 915.170 425.879 236.232 104.311 61.400 24.391 1.457.193 699.270 Basilicata 2.818 2.418 9.253 5.712 12.650 8.869 232.013 120 .. 097 125 .111 65.203 157.023 63.350 I 538.868 265.6411 Bologna 9.843 7.539 28.461 15.563 52.133 28.733 481.025 231:474 92.424 41.735 40.348 16.322 704.234 341.366 Catanzaro 3.845 3.287 11.348 7.079 15 . 220 10. 317 206.331 111.296 169.426 88.431 198.394 70.663 604.564 291.07 8 Ferrara 2312 1.805 8.263 4.540 14.488 8.684 236.870 120.104 69.724 33.466 45.208 17.606 376.865 186.205 Cosenza 3.405 2.968 11.936 7.387 15.910 10. 854 211.587 112 .036 159 .939 81. 341 18l. 713 66.537 584.490 281.12'3 Forli • 3.001 2.275 13.732 7.475 21.625 12.676 268.242 133.531 72.746 35.045 56.056 23.947 435.402 2U.949 Modena 3.336 2.568 11.521 6.270 20.963 12.287 321.647 159.584 63.569 30.867 30.362 12.331 451.398 223.907 Reggio di Calabria 4.503 3.496 13.429 8.192 18 . 259 12.194 199.462 106.358 138.722 72.537 172.650 59 . 985 li 547.025 262.762 Parma 3.245 2.530 11.140 5.577 18.469 10.559 251. 983 123.684 52.811 25.423 22.348 9.727 359.996 177 .500 Calabria. PiaceIlza " 11. 753 9.751 36 .713 22 .658 49.389 33.365 617.380 329.690 468.087 242.309 552.757 197.185 1.736.079 834 . 958 1.923 1.561 9.065 4.722 14.324 7.806 194.195 94.908 39.766 19.454 16.122 8.342 275.395 136.793 I Ravenna 2.083 1.585 9.228 4.751 13.361 7.926 170.645 83.903 43.822 20.832 30.274 13.675 269.413 132.672 Agrigento 2.736 2.181 4.329 7.607 10.509 6.779 169.778 81.542 102.749 51.276 116.171 57.617 . 409.550 203.72 4 Reggio nell'Emilia. 2.295 1.835 9.499 5.141 15.340 8.949 258.750 130.009 46.635 22.274 22.144 8.439 354.663 176.647 Cal tanissetta 1.596 1.277 4.749 2.842 6.704 4.282 110.990 54.071 61.493 30.787 71.076 34.862 I 256.608 128.121 Emilia-Romagna. 110.389 3.227.366 1.590.039 Catania 9.234 7.417 23 .820 14.362 31. 708 19.593 289.610 134.755 160 . 253 77.430 181.172 83 .991 I 695.797 337.54 8 28.038 21.698 100.909 54.039 170.703 97.620 2.183.357 1.077.197 481.497 229.096 262.862 Enna l. 178 962 3.854 2.289 4.869 3.124 89.124 42.324 43.654 21.877 65.944 33.585 208.623 104 . 161 Messina 6.994 5.137 19.156 11.314 26 . 207 16.728 289. 180 140.445 119.020 58.336 125.789 47.496 586. 346 279.45 6 Arezzo 1.643 1.316 7.215 4.004 11.589 7.498 173.015 91.698 56.941 27.809 46.204 16.772. 2g6.6'07 149.097 Palermo 12.731 9.984 29.008 15.690 44.207 25.410 388 .523 180 .982 227 .170 108.045 183.526 90.080 885.165 430.19 l Firenze 11.506 8.909 32.771 17.868 60.999 35.985 519.308 250.150 139.548 61.826 79.491 30.102 843.623 404.840 Ragusa ' l. 846 1.518 5 .156 3.178 7.479 4.834 93.774 44.809 43.296 21.260 60.005 27.851 21l.556 103.45o Grosseto 1.063 873 3.889. 2.150 6.500 4.381 112.350 57.256 41.583 21.528 26.434 11.510 191.819 97.698 Siracusa 2.478 2.027 7.451 4 . 601 11.820 7. 380 123.961 59.934 65.471 33.117 69.255 32.512 280.436 139.57 Livorno 2.416 1.922 9.132 5.108 17.831 11.614 160.673 79.894 46.294 ' 20.571 20.812 7.401 257.158 126.510 Trapani 3.050 2.358 9.185 5 . 355 12.313 7.459 170.927 80.211 89.257 43.692 86.367 41.608 371.099 180.68 Lucca 2.489 1.967 10.478 5.904 15.827 9.796 227.065 107.584 53.041 24.463 25.294 9.134 334.194 158.848 Massa-Carrara . 1.249 978 5.378 2.969 8.457 5.460 122.522 61.432 27.525 12.786 18.595 6.320 183.726 89.945 Sicilia. 41.843 32.861 109 .986 63.960 155.816 95.589 1. 725.867 819.073 912.363 445.820 959.305 449.602 3.905 .180 . 1.906.905 Pisa 2.966 2.330 9.320 5.455 16.495 11.265 198.313 99.211 56.519 26.776 36.588 12.392 320.201 157.429 Pistoia 1.200 956 5.275 2.914 9.463 6.216 132.920 66.164 33.098 15.374 ' 19.489 7.110 201.445 98.734 Cagliari 4.044 3.201 13.105 7.181 22.237 13.163 242.976 124.248 152.842 79.302 133.639 57.931 568. 843 285 .02o Siena 2.122 1.738 6.808 3.715 11.019 7.090 142.782 73.966 47.235 23.647 44.358 16.591 254.324 126.747 Nuoro 936 814 3 .347 1. 796 4.495 2.776 99.505 50 .065 65.564 34.439 48.591 21.216 222.438 11l.106 Sassari 2 . 468 l.975 6.522 3.379 10 .339 6.050 145.098 75 .019 81.639 41.215 58.996 25.134 305.062 152.772 Toscana 26.654 20.989 99.266 50.987 158.180 99.305 1.788.948 887.355 501.784 234.780 317.265 117.832 2.883.097 1.409.848 Sardegna 7 . 448 5.990 22.974 12.356 37.071 21.989 487.579 249.332 300.045 154.956 241.226 104.281 1.096.343 548.9 04 ITALIA 42%.324 340.783 1.319.8Il 175.151 2.514"14 1.446.517 24.946.399 12.216.852 1 .581.622 3.649.251 5.456.005 2.158.51'3 .42.300.635 20.581 .799 ,. I I Figure B.10 An example of educational degrees from the 1951 census of population 62 B.2.3 Patent Data The purpose of this section is to clarify the data collecting process for patents. We will compare two patent data sets collected for this study and explain the reason why we end up using the patent data set provided by Bianchi and Giorcelli (2020) in the final version. First, they are able to match the names on patents with individuals and location. They start with matching the names of high school graduates to the inventors of patents. Then, to refine and improve the matching they use work histories provided by Italy’s Social Security Administration. In addition, they manually check and confirm the matched names on patents to increase precision. As a result, the data set has more accurate information and more complete picture at the province level. They collect patent data using the Italian Patent Office (IPO) between 1950 and 2010, and the international patents included in the European Patent Office’s (EPO) PATSTAT database. The data set provides number of patents at each province in Italy during the given time period. Table B.9 presents the distributions of patents per province. Second, we collect raw patent data using EPO’s portal PATSTAT for Italy for the period between 1950 and 1982 in the earlier versions of this study. We exclude utility models and designs as it is a common practice in the literature. Using a matching algorithm, we were able to create a data set based on the name on the patent applications. The raw data has some information available including an application and person identifier, name of the applicant, year of application and location information. The first problem we face is that not all applications have location information available. To deal with this problem, we use a simple matching algorithm. Matching non-standardized names on patent applications with firms available in a standardized database is a widely studied topic. Thoma et al. (2010) set a list of rules and discuss different methods on how to combine patent data sets with each other and other sources of data. Lotti and Marin (2013) follow their methodology and study how to match Italian patents from PATSTAT database with a commercial database on Italian firms. 63 Our approach to match patent applications with a location is much simpler compared to those. We only need location information for a patent application since we aggregate the number of patent applications at province level. The main problem is that almost all patent applications before 1977 do not have available location information. This is because EPO was established in 1977. The data before this year is gathered from national patent offices and there are blanks in the applications. However, most of these firms filed for a patent before 1977 also applied for a patent after 1977. Thus, we can match these firms with a location since firm or individual specific identifiers are available. We create a list of firms and individuals with their locations. Then, using the person identifier (person can be a firm or individual) we are able to match and create a database with location information. Furthermore, we benefit from de Rassenfosse et al. (2019) to expand our patent database. They provide geographic coordinates for inventor and applicant locations. The database starts at 1980 and spans over more than 30 years. The application number is available in their publicly available data. We use patent applications from 1980 since it is the only year overlapping with our time period. Doing so, we are able to add almost 5,000 more patent applications to our database (Table B.8). We should note that most of the unmatched patent applications belongs to individuals rather than firms. After this simple matching, we end up with 66,520 patent applications corresponding to 2,927 unique person identifiers (almost all of them are firms). The right panel of Table B.8 shows that around 90% of matched applications made by firms that have a patent application both before and after 1977. Hence, our sample consists of firms keep innovating over the time period selected. However, the missing information (i.e. location) on patent applications seriously affected the data collecting process. Table B.10 presents the distributions of patent applications for the earlier data set. Milan has the most patent applications and around 15 provinces comprises almost 85% of all patent applications between 1950 and 1980. Furthermore, these top 15 provinces have almost complete data during the period. The rest of provinces has a 64 Table B.8 The number of patent applications matched Using only raw patent data Using raw patent data and de Rassenfosse et al. (2019) Type Unmatched Matched Total Unmatched Matched Total before77 72,081 993 73,074 72,081 993 73,074 both 32,659 56,788 89,447 29,995 59,452 89,447 after77 35,122 3,948 39,070 32,995 6,075 39,070 Total 139,862 61,729 201,591 135,071 66,520 201,591 Note: before77 represents firms or individuals only filed a patent application before 1977. after77 represents firms or individuals only filed a patent application after 1977. both represents firms or individuals filed a patent application before and after 1977. lot of gaps in the data. Overall, the reason why we do not use the earlier patent data set becomes more clear comparing Table B.9 and Table B.10. The completeness of the data set from Bianchi and Giorcelli (2020) provides a balanced panel to conduct our analysis. 65 Table B.9 The number of patents per province from Bianchi and Giorcelli (2020) Provincia 1950 1951 1952 1953 1954 1955 1956 1957 1958 1959 1960 1961 1962 1963 1968 1969 1970 1971 1972 1973 1974 1975 1976 1977 1978 1979 1980 Agrigento 4 1 2 0 3 3 8 5 2 7 6 6 3 0 2 1 0 1 1 0 2 0 1 0 3 4 1 Alessandria 38 50 44 64 60 43 52 63 46 47 48 42 37 31 23 22 19 52 60 29 32 42 50 47 38 33 38 Ancona 32 41 49 59 64 70 75 46 57 72 57 44 41 16 15 16 11 41 35 25 31 25 41 53 49 38 35 Aosta 12 12 13 12 2 6 10 7 2 5 2 6 3 0 3 2 4 5 6 6 3 2 5 7 3 1 5 Arezzo 7 10 15 16 10 17 37 17 16 21 21 9 17 5 12 8 6 10 32 16 20 19 31 46 65 44 29 Ascoli Piceno 5 16 17 25 16 18 22 22 19 21 16 10 15 1 24 14 13 36 42 28 27 30 33 49 54 23 35 Asti 9 18 11 21 8 15 16 14 6 16 14 11 9 8 5 10 7 22 28 26 13 10 19 20 41 17 33 Avellino 1 2 2 5 7 11 3 5 6 7 6 4 4 1 3 2 0 2 5 6 4 2 2 3 7 4 4 Bari 33 25 27 34 38 38 49 37 37 41 29 35 23 12 6 12 9 28 22 20 10 21 15 22 22 33 18 Belluno 14 12 9 11 23 20 25 14 17 18 12 8 9 3 4 4 4 17 19 17 9 13 4 13 25 14 10 Benevento 6 6 1 4 5 5 8 2 8 1 2 3 2 0 3 3 2 6 7 7 8 8 6 4 9 3 12 Bergamo 71 89 76 91 86 84 98 75 88 100 79 76 77 40 49 52 28 108 110 64 79 77 99 84 122 112 112 Bologna 296 225 251 275 298 331 304 299 261 294 294 247 262 126 119 89 76 261 300 196 218 253 269 302 249 222 201 Bolzano-Bozen 30 43 37 55 59 56 70 39 49 42 43 33 28 19 17 16 17 54 71 43 39 30 30 49 62 38 63 Brescia 134 119 111 108 148 129 128 105 114 164 166 133 107 59 49 41 40 150 167 114 106 137 119 185 196 127 159 Brindisi 5 1 1 4 7 9 6 7 5 3 3 4 5 0 4 1 1 3 2 6 4 0 1 8 3 1 3 Cagliari 5 16 10 12 19 14 13 12 17 28 17 8 9 3 4 5 4 16 10 4 8 8 12 10 21 15 11 Caltanissetta 2 3 3 1 0 3 8 3 2 0 2 1 4 0 1 1 1 1 2 1 2 1 2 3 2 2 1 Campobasso 1 9 4 3 4 5 5 5 7 6 7 0 1 2 4 2 1 1 6 1 1 1 4 7 3 3 2 Caserta 10 5 9 10 13 14 11 12 7 10 10 6 7 2 9 3 1 14 13 10 12 7 8 10 15 12 5 Catania 27 14 20 30 31 39 32 35 31 28 21 21 19 11 10 6 4 29 19 28 15 21 15 21 34 26 12 Catanzaro 9 3 1 9 6 14 15 8 12 15 12 7 7 1 2 2 1 12 7 7 5 2 2 11 7 7 5 Chieti 13 9 8 12 11 8 8 12 9 3 11 9 7 2 6 1 0 4 5 7 12 14 7 17 16 12 9 Como 118 119 93 120 132 116 94 99 92 115 96 83 85 54 44 27 29 107 133 74 93 108 77 79 88 86 96 Cosenza 4 1 1 3 4 7 8 6 3 4 3 6 7 2 5 3 4 8 11 6 4 6 6 13 10 5 5 Cremona 0 0 0 0 0 0 0 0 0 0 0 0 0 0 13 10 9 17 35 16 16 42 20 19 30 21 30 Cuneo 30 37 26 31 39 32 32 41 35 38 30 34 25 10 14 10 3 41 35 28 26 30 34 47 46 38 30 Enna 0 3 2 3 1 4 3 2 1 1 2 0 3 1 1 0 0 0 1 1 0 1 0 0 1 0 2 Ferrara 54 28 26 40 65 44 40 35 51 39 44 34 44 17 0 0 0 0 0 0 0 0 0 0 0 0 0 Firenze 244 238 239 278 285 286 264 292 208 237 256 195 204 89 85 67 41 201 225 155 129 140 145 182 179 158 173 Foggia 11 6 7 9 11 19 20 6 9 11 10 6 9 2 1 0 2 10 10 3 3 4 4 9 6 2 3 Forli 31 25 37 40 32 49 47 30 42 53 51 26 28 11 8 14 4 30 33 21 29 18 19 31 43 19 32 Frosinone 6 5 6 2 6 15 11 11 5 7 4 7 4 4 9 9 4 19 13 10 7 6 5 11 18 14 7 Genova 334 284 332 303 310 295 275 220 233 259 192 175 165 80 94 56 39 136 150 99 106 126 99 105 123 101 87 Gorizia 21 8 15 18 12 10 12 17 7 15 9 2 7 2 3 4 4 12 13 12 9 7 11 8 8 11 6 Grosseto 4 12 12 6 5 10 17 15 3 6 6 9 8 0 2 4 2 17 3 10 6 9 7 7 15 9 4 Imperia 60 45 55 51 64 43 60 62 82 92 69 53 63 28 7 7 2 20 24 20 11 12 7 10 6 8 5 La Spezia 46 42 35 52 51 48 40 43 36 33 37 22 22 7 14 10 7 28 16 12 10 16 13 30 12 13 13 L’Aquila 4 4 4 13 13 10 16 7 7 5 2 4 9 1 5 3 1 3 7 11 5 6 3 6 14 8 13 Latina 5 8 6 8 8 14 9 13 7 5 6 10 9 4 5 4 9 23 17 22 7 14 10 13 13 16 8 Lecce 4 7 6 6 16 9 13 9 4 9 1 4 4 1 6 6 3 6 5 10 6 1 11 10 13 11 2 Livorno 28 27 24 29 28 33 35 26 19 35 24 19 12 14 9 5 8 19 19 17 13 11 11 30 24 21 16 Lucca 32 32 38 40 28 37 40 54 40 53 31 34 33 12 12 12 11 26 29 13 18 23 17 14 25 21 15 Macerata 11 9 17 11 18 32 21 29 18 20 21 12 10 8 13 7 5 16 16 16 11 14 9 21 20 30 29 Mantova 51 43 52 75 62 58 48 39 49 28 47 35 44 23 24 7 7 23 26 16 21 17 33 21 41 44 33 Massa-Carrara 13 11 12 14 14 28 13 12 9 10 17 4 11 7 6 7 6 16 16 10 11 9 4 14 17 13 13 Matera 0 1 0 0 1 1 1 1 0 2 0 0 0 0 0 1 0 0 2 1 2 0 3 2 3 0 2 Messina 19 25 17 16 16 15 9 24 19 24 14 12 9 4 6 6 2 13 14 8 5 6 5 10 20 13 7 Milano 1622 1581 1667 1833 1869 1941 1878 1813 1751 1938 1818 1579 1551 870 753 594 426 1590 1619 1082 1270 1203 1099 1267 1265 1051 996 Modena 54 63 56 64 85 92 98 105 117 122 121 104 134 66 83 63 29 180 173 111 101 148 134 113 139 117 72 Napoli 117 117 108 140 124 149 117 121 108 116 106 95 73 35 48 21 30 84 67 47 60 56 43 71 82 37 42 Novara 89 55 67 69 71 60 79 65 54 68 60 61 39 22 25 27 17 58 86 22 31 45 45 40 37 46 41 Nuoro 5 3 1 0 3 4 6 0 3 3 6 2 1 1 1 1 0 0 0 0 1 0 0 2 3 0 0 Padova 65 92 89 90 102 91 113 78 115 78 53 85 80 31 35 31 28 149 122 98 87 76 94 100 104 53 60 Palermo 41 33 25 43 41 40 39 37 38 40 32 25 37 11 13 14 7 36 32 26 15 22 28 14 26 32 30 Parma 68 77 72 62 87 91 64 56 73 54 50 51 50 23 34 27 15 70 84 46 49 59 58 52 69 67 51 Pavia 72 64 82 88 90 137 97 75 73 90 73 80 77 36 23 26 18 68 62 41 31 61 39 52 53 44 68 Perugia 31 25 30 28 32 36 43 32 34 36 36 24 13 16 11 16 14 40 39 24 32 26 14 39 34 22 21 Pesaro e Urbino 17 17 21 22 23 29 23 22 33 26 23 17 20 11 12 8 7 39 52 26 36 23 23 25 18 27 15 Pescara 8 5 11 18 22 14 16 21 17 13 17 11 17 4 3 7 6 23 16 9 16 7 8 20 12 6 8 Piacenza 34 20 38 29 37 50 39 31 37 35 36 28 21 20 16 7 6 26 37 27 18 23 24 39 40 17 24 Pisa 22 22 17 26 31 28 27 32 22 32 29 26 35 15 10 7 7 36 47 23 24 28 24 21 32 41 20 Pistoia 17 15 15 25 26 35 32 48 29 51 28 23 27 20 16 10 6 34 29 23 28 35 32 25 32 21 23 Potenza 12 2 9 8 2 7 10 4 5 2 1 3 4 1 5 1 2 3 4 0 1 1 4 4 12 3 5 Ragusa 6 2 3 0 13 2 10 3 3 5 3 4 0 2 1 2 4 9 6 4 2 5 3 16 5 3 3 Ravenna 19 20 21 25 34 25 25 23 30 32 45 18 26 17 13 16 16 50 35 19 23 21 28 31 55 30 33 Reggio di Calabria 11 10 9 16 6 7 14 13 11 13 16 7 8 2 8 7 1 10 8 9 7 6 4 9 9 9 2 Reggio nell’Emilia 52 56 55 97 63 110 83 86 109 89 87 109 98 41 57 45 20 101 125 63 47 83 83 82 68 71 77 Rieti 2 7 5 2 1 4 6 3 6 0 1 2 2 2 0 0 0 4 3 2 2 3 3 6 8 13 4 Roma 603 529 552 599 702 656 651 598 483 638 440 410 355 176 183 148 123 474 444 295 311 343 274 396 439 346 290 Rovigo 12 5 8 21 16 10 14 13 15 20 21 12 17 4 5 5 4 11 13 7 10 8 7 11 8 6 12 Salerno 11 20 21 20 15 27 24 24 18 16 19 10 21 9 12 12 10 36 24 17 14 18 11 27 25 8 11 Sassari 7 2 7 3 7 8 5 6 1 6 16 8 1 1 3 2 3 3 10 4 2 5 24 32 22 22 13 Savona 39 33 23 33 21 31 30 27 29 35 24 18 15 11 6 6 3 19 20 18 19 13 12 21 21 14 17 Siena 23 18 21 19 16 21 18 21 13 13 18 21 13 10 8 5 4 23 31 19 19 18 19 17 21 15 12 Siracusa 8 6 3 6 17 10 3 4 11 8 4 5 11 1 4 2 3 5 10 5 1 2 3 5 9 3 1 Sondrio 5 2 3 5 8 8 6 9 7 15 8 11 5 6 1 6 2 0 3 2 3 5 8 5 5 4 3 Taranto 11 6 7 6 12 9 9 3 5 7 8 2 9 3 5 8 1 13 5 4 8 3 3 8 13 12 5 Teramo 5 8 2 6 8 2 3 10 4 8 2 6 3 2 0 1 0 7 9 4 2 1 2 6 1 3 2 Terni 10 17 16 18 16 25 23 15 13 15 5 6 8 7 6 2 7 9 12 6 7 6 5 8 9 4 9 Torino 766 692 708 691 692 722 699 630 650 570 513 439 496 213 200 186 143 613 588 471 481 478 530 530 511 414 425 Trapani 5 1 2 4 5 2 5 5 4 8 11 6 8 1 4 4 3 6 4 4 1 6 3 0 10 6 3 Trento 28 28 31 36 36 36 30 31 40 24 32 25 20 9 16 21 14 40 43 40 34 25 28 44 50 42 46 Treviso 50 57 51 62 83 68 55 50 48 55 62 45 63 41 33 29 29 97 100 76 79 88 70 103 98 71 85 Trieste 62 46 77 88 84 55 56 43 34 43 52 25 30 16 21 4 8 35 23 29 23 17 8 16 31 26 15 Udine 61 54 42 49 79 86 87 87 72 77 59 56 36 27 65 27 22 158 160 144 125 126 114 167 119 121 110 Varese 107 108 135 122 142 112 127 94 97 101 107 107 89 48 62 39 42 176 166 128 101 129 111 106 120 94 95 Venezia 59 45 45 57 58 70 53 68 58 76 70 53 54 29 25 27 14 69 62 38 49 37 36 47 46 46 61 Vercelli 66 56 59 53 51 38 50 46 25 44 35 25 33 16 20 14 18 53 50 26 22 30 25 34 34 30 27 Verona 67 98 93 123 97 122 115 120 101 93 96 78 72 37 48 40 18 99 92 87 63 59 59 94 67 63 57 Vicenza 78 74 78 83 90 110 97 87 122 109 107 95 76 52 60 62 28 139 133 111 109 138 106 155 159 95 122 Viterbo 5 2 12 9 14 10 14 11 3 9 6 8 8 5 6 1 0 12 6 1 3 3 4 13 11 9 3 66 Table B.10 The number of patents per province from raw patent documents Province 1950 1951 1952 1953 1954 1955 1956 1957 1958 1959 1960 1961 1962 1963 1964 1965 1966 1967 1968 1969 1970 1971 1972 1973 1974 1975 1976 1977 1978 1979 1980 Agrigento 1 Alessandria 1 8 4 14 2 2 14 13 7 21 5 29 15 10 12 34 14 8 188 397 518 578 661 756 713 684 Ancona 1 1 2 1 4 3 6 8 27 Aosta 4 2 20 Arezzo 13 2 4 9 Ascoli Piceno 1 1 3 3 8 7 14 14 15 25 34 23 46 52 34 57 79 48 38 37 49 83 Asti 1 2 2 11 25 36 38 62 Avellino 3 Bari 3 11 20 Barletta-Andria-Trani 1 6 1 7 Belluno 1 2 2 12 2 Benevento 2 Bergamo 1 2 2 1 6 1 1 1 3 8 15 4 21 11 37 30 68 74 99 Biella 5 5 5 3 12 6 Bologna 1 4 1 1 3 2 2 1 6 2 3 2 4 5 8 38 17 41 21 87 78 130 389 286 307 456 288 430 617 Brescia 2 5 9 11 1 2 16 7 11 45 75 86 92 139 113 222 266 275 287 379 Brindisi 2 Cagliari 2 Caltanissetta 3 38 18 54 71 264 535 279 313 187 284 Campobasso 7 Caserta 2 1 2 16 15 90 71 58 50 50 25 Catania 2 3 2 5 4 3 4 2 4 4 Catanzaro 1 4 Chieti 2 20 36 46 9 1 12 71 Como 3 4 2 7 10 1 6 1 14 26 18 10 15 49 37 40 47 74 41 105 121 190 Cosenza 1 1 1 4 Cremona 1 2 2 2 8 27 32 46 25 5 36 47 Crotone 1 Cuneo 13 4 4 8 11 32 10 4 13 12 16 46 50 59 121 57 38 31 16 73 65 57 66 101 97 50 120 Fermo 1 3 8 Ferrara 1 3 1 1 1 2 10 6 20 18 36 31 Florence 3 5 1 1 3 1 2 2 3 3 2 4 2 53 18 12 75 8 13 59 95 240 120 306 113 162 191 297 Foggia 3 Forli-Cesena 1 3 4 10 12 13 Frosinone 1 3 21 17 Genoa 2 1 1 1 2 1 2 1 1 4 2 10 3 2 10 13 42 39 37 33 69 77 72 60 118 Gorizia 1 2 2 1 1 6 6 24 18 30 24 10 1 10 34 48 Imperia 2 La Spezia 1 5 2 1 2 3 2 7 L’Aquila 3 2 4 3 2 1 4 6 10 1 5 Latina 1 2 9 Lecce 1 1 5 10 5 7 12 Lecco 1 1 2 2 1 2 1 6 16 16 52 32 66 129 200 Livorno 1 1 2 20 35 54 51 63 Lodi 9 12 76 204 168 123 81 49 160 231 Lucca 1 2 15 1 2 4 Macerata 1 1 3 2 1 1 2 6 Mantua 1 3 6 25 22 Massa and Carrara 2 4 6 21 45 30 57 Matera 1 Messina 5 7 Milan 12 35 34 44 34 58 81 68 77 128 242 278 220 182 142 183 309 676 676 596 798 988 1342 1410 1886 2768 2741 2358 2367 2638 3029 Modena 2 1 1 4 2 5 16 18 29 6 32 65 83 Monza and Brianza 2 7 2 8 10 29 51 37 60 49 66 71 116 117 Naples 1 1 3 4 5 35 8 14 33 54 Novara 2 2 2 1 2 7 6 2 4 11 13 21 20 36 54 Padua 2 1 1 2 15 14 30 18 22 21 48 42 83 Palermo 1 3 Parma 3 3 2 3 2 9 5 4 15 65 Pavia 3 2 2 4 73 63 32 32 18 62 56 110 171 226 305 83 439 412 370 391 269 171 238 306 469 Perugia 1 2 1 2 9 11 9 26 Pesaro and Urbino 2 1 19 11 17 5 4 11 8 Pescara 1 1 5 Piacenza 2 3 1 2 5 6 16 11 10 14 46 Pisa 4 5 2 14 1 1 2 8 13 1 2 10 4 1 9 16 11 19 30 Pistoia 1 15 45 Pordenone 1 3 1 6 14 64 39 25 9 11 20 19 23 43 47 68 43 96 Potenza 3 7 Prato 2 1 Ragusa 1 3 2 3 7 Ravenna 4 6 4 5 1 3 5 21 13 30 Reggio Calabria 4 4 Reggio Emilia 8 4 16 9 14 41 Rieti 1 Rimini 3 1 3 8 3 7 10 32 Rome 2 2 1 6 9 21 10 11 11 14 30 34 34 28 40 74 140 177 199 260 426 561 818 915 Rovigo 2 4 1 1 1 2 2 3 2 2 1 9 Salerno 1 1 2 2 1 1 35 Sassari 1 3 Savona 1 1 2 7 9 4 7 13 26 Siena 1 8 4 3 13 7 7 3 13 Sondrio 2 1 1 2 1 9 1 South Sardinia 1 6 South Tyrol 1 2 4 7 24 20 21 39 34 Syracuse 11 Taranto 4 1 8 Teramo 1 2 1 2 2 Terni 1 2 9 6 Trapani 1 Trento 2 5 31 4 40 24 7 29 20 Treviso 1 1 1 2 3 2 10 10 19 15 14 48 36 57 Trieste 1 4 8 14 3 1 9 17 19 6 Turin 33 19 10 21 21 15 12 20 13 21 21 16 10 14 16 26 47 50 63 99 88 139 201 186 308 267 464 529 718 492 836 Udine 3 2 1 1 3 6 5 44 32 26 50 47 54 Varese 1 1 2 2 14 1 1 1 3 2 2 7 17 21 28 34 44 45 60 88 176 Venice 2 1 2 1 4 11 10 33 55 Verbano-Cusio-Ossola 1 2 2 2 3 10 3 3 3 1 1 2 1 3 Vercelli 1 1 3 1 13 Verona 3 1 1 2 3 2 4 3 2 5 18 11 7 57 30 Vicenza 2 2 1 5 7 2 1 24 21 2 6 3 4 35 18 73 8 29 41 60 73 28 74 143 115 Viterbo 1 2 2 67 B.2.4 Additional Patent Data from 1968 to 1973 The bank loan data covers the period between 1890 and 1973 and the patent data is from 1950 to 2010 with a gap between 1963 and 1968. Thus, the final data set comprises the period between 1950 and 1963. However, to shed more light we include patent data from 1968 to 1973 to the sample. We try to provide better summary statistics. Figure B.11 and Table B.11 displays how credit reallocation measures changes over time compared to real GDP growth since there is no gap in bank loan data. We take the average of credit reallocation measures for each province in a given year. In the early 1950s, gross credit reallocation and real GDP growth move in the opposite directions. On the other hand, credit destruction, consequently excess credit reallocation, moves hand in hand with the real GDP growth in the early 1950s. Gross credit reallocation and net credit growth declined in the early 1950s and then increased towards the mid-1950s. However, they gradually decreased until late 1950s. Starting in the 1960s, gross credit reallocation and net credit growth started to follow a more similar pattern with real GDP growth. Lastly, credit destruction and excess credit reallocation stay relatively low during sample period. Overall, credit creation, gross credit reallocation, and net credit growth closely follow each other over time, while credit destruction and excess credit reallocation display a similar movement. These results are not unexpected considering that the time coincides with the greatest development of the Italian economy. Also, we work with bank loans instead of firm debts and we expect banks to increase the amount of loans during an economic expansion period. Figure B.12 presents the relationship between innovation and credit reallocation over time. Again we take the average of credit reallocation measures and number of patents for each province for a given year. Patents increase towards the end of 1950s after a slight decline in the early 1950s. This period coincides with the Italian economic boom. However, after this prosperous period, there is a large decline in the number of patents in the early 1960s. Nuvolari and Vasta (2015) argue that scientific activities prevail patenting during this period. 68 Next, we try to explore more how innovation and credit reallocation are related at the province level. We examine how provinces are distributed using number of patents and credit reallocation measures. We take the average of number of patents and credit reallocation measures for the whole sample period to draw the scatter plots. The inclusion of data between 1968 and 1973 does not substantially affect the distribution of provinces (See Figure B.13, Figure B.14, Figure B.15, Figure B.16, and Figure B.17). Hence, the plots suggest a negative relationship between innovation and credit reallocation as earlier data. Furthermore, we present Table B.12 with the inclusion of data between 1968 and 1973 to examine province characteristics considered in our analysis. We take the average of all considered variables for all provinces at a given year. Data collected from censuses are presented only at the year the census held. First, the number of patents follows a path similar to an inverted-U shape between 1950 and 1963. Table B.12 reveals that the number of patents significantly decreases until 1971 and after a substantial increase in 1971. This result is not surprising because Nuvolari and Vasta (2015) claims that scientific activities prevail patenting between 1960 and 1970. Thus, the sudden increase in the number of patents in 1971 can be the fruit of scientific activities performed during this period. We measure productivity as the total value added per firm in a province. Productivity gradually decreases until 1961 and starts to increase after. However, Table B.12 shows that it declines again until 1971 and a very large increase in 1971 happened in productivity. The evidence suggests that innovation and productivity follow a similar path over time. The number of banks is stable over time moving around 4 banks on average in each province, while number of bank branches on average increases substantially over time. There are 96 branches on average in each province in 1951, while the number of bank branches reaches 146 on average in 1971. Additionally, credit market in Italy is highly concentrated between 1950 and 1963, but the concentration starts to decrease after 1970. Average number of workers for each firm increases from 3.64 in 1951 to 4.02 in 1971, while the share of active population decreases from 46.2% in 1951 to 36.9% in 1971. Italy’s 69 great economic development period pays out as share of higher education degrees increases from 3.8% in 1951 to 8% in 1971. B.2.4.1 Tables and Figures Figure B.11 Credit reallocation measures and the real GDP growth rate (from 1950 to 1973) 70 Figure B.12 Credit reallocation measures and the average number of patents per firm (from 1950 to 1973) Figure B.13 Distribution of provinces - Gross credit reallocation and the number of patents per 1,000 firms (from 1950 to 1973) 71 Figure B.14 Distribution of provinces - Net credit growth and the number of patents per 1,000 firms (from 1950 to 1973) Figure B.15 Distribution of provinces - Excess credit reallocation and the number of patents per 1,000 firms (from 1950 to 1973) 72 Figure B.16 Distribution of provinces - Credit creation and the number of patents per 1,000 firms (from 1950 to 1973) Figure B.17 Distribution of provinces - Credit destruction and the number of patents per 1,000 firms (from 1950 to 1973) 73 (a) Number of Patents (b) Net Credit Growth (c) Excess Credit Reallocation Figure B.18 Regional overview of variables of interest (from 1950 to 1973) Note: This figure plots the regional overview of three main variables of interest. Panel (a) displays the average number of patents for each region. The Northern regions have higher number of patents. Panel (b) presents the regional distribution of net credit growth. The Southern regions have higher net credit growth as expected. Panel (c) shows the regional differences in the excess credit reallocation measure. Overall, the Northern regions have higher levels of excess credit reallocation, but two of the Southern regions have the highest levels. 74 Table B.11 Summary statistics for credit reallocation measures (from 1950 to 1973) Year Gross Net Excess Credit Credit Real Credit Credit Credit Creation Destruction GDP Reallocation Growth Reallocation Growth 1950 17.61% 11.44% 1.84% 14.53% 3.09% 8.41% 1951 13.73% 5.79% 1.65% 9.76% 3.97% 9.68% 1952 20.38% 19.68% 0.70% 20.03% 0.35% 4.75% 1953 16.73% 15.87% 0.45% 16.30% 0.43% 7.35% 1954 17.62% 16.55% 0.73% 17.08% 0.54% 3.80% 1955 14.86% 13.11% 0.72% 13.99% 0.88% 6.97% 1956 13.58% 12.44% 0.82% 13.01% 0.57% 4.97% 1957 12.68% 11.11% 0.84% 11.89% 0.78% 5.72% 1958 11.79% 9.52% 0.82% 10.65% 1.13% 5.94% 1959 14.72% 13.88% 0.38% 14.30% 0.42% 7.12% 1960 18.75% 17.77% 0.51% 18.26% 0.49% 7.71% 1961 13.62% 12.91% 0.41% 13.27% 0.36% 8.47% 1962 14.98% 13.91% 0.62% 14.45% 0.54% 6.98% 1963 11.96% 10.41% 1.08% 11.19% 0.77% 6.22% 1964 7.94% 1.14% 1.67% 4.54% 3.40% 3.96% 1965 8.89% 4.78% 1.40% 6.84% 2.05% 4.60% 1966 11.75% 11.33% 0.29% 11.54% 0.21% 6.68% 1967 13.49% 13.20% 0.11% 13.35% 0.14% 7.71% 1968 14.74% 14.53% 0.21% 14.64% 0.11% 7.32% 1969 13.37% 12.67% 0.62% 13.02% 0.35% 6.59% 1970 8.38% 6.55% 0.83% 7.47% 0.92% 6.04% 1971 7.78% 4.02% 1.27% 5.90% 1.88% 1.61% 1972 10.37% 8.30% 1.21% 9.33% 1.04% 3.43% 1973 12.92% 10.57% 1.42% 11.75% 1.17% 6.72% 75 Table B.12 Summary statistics for province characteristics (from 1950 to 1973) Year Number of Productivity Number of Credit Number of Number of Share of Share of Share of Patents (000 lire) Banks Market Workers Bank Individual Higher Active Concentration per Firm Branches Firms Education Population Degrees 1950 73.43 269.43 3.45 0.68 1951 68.53 248.85 4.26 0.62 3.64 96.33 91.37% 3.79% 46.24% 1952 71.49 240.87 4.50 0.61 1953 79.02 232.74 4.50 0.61 1954 82.72 225.95 4.51 0.61 1955 84.83 218.88 4.52 0.61 1956 82.10 210.35 4.50 0.61 1957 76.31 206.52 4.37 0.62 1958 72.88 202.40 3.34 0.68 1959 79.30 203.61 4.45 0.61 1960 71.60 200.43 4.45 0.61 1961 61.73 349.22 4.44 0.61 3.99 118.60 91.45% 4.94% 40.44% 1962 60.93 331.51 4.43 0.61 1963 31.15 305.50 4.43 0.61 1964 - 287.96 4.44 0.61 1965 - 278.67 4.39 0.61 1966 - 272.11 4.33 0.61 1967 - 263.97 4.16 0.61 1968 31.36 260.61 4.02 0.62 1969 25.02 251.20 4.00 0.62 1970 18.56 235.84 5.28 0.56 1971 73.47 498.07 5.14 0.56 4.02 146.42 90.22% 7.98% 36.85% 1972 74.73 469.82 5.03 0.57 1973 51.83 415.81 5.02 0.57 76 B.3 Robustness This section includes the robustness tests to further test the validity of our estimation results. B.3.1 Weak Instruments First-stage regression results suggest that we may suffer from weak instruments. If our instruments are weakly correlated with the endogenous regressors, IV estimators can be biased due to the poor properties of two-stage least squares when instruments are weak. Hence, further investigation is required. We need to perform a weak instrument test to detect weak correlation with the endogenous regressor. Then, we need to make weak-instrument robust inference in case our instruments fail to pass the test.3 A widely accepted common rule for testing the strength of an instrument is an F-statistic greater than 10 from the first-stage regression (Staiger and Stock (1997)). On a cursory look, we can say that our instruments perform well, especially for excess credit reallocation. But there are some model specifications in which the instrument fails to pass a weak instrument test, particularly for net credit growth. Therefore, we need to investigate further. Andrews et al. (2019) survey the literature on detecting weak instruments and making weak-instrument robust inference. They conclude that the efficient F-statistic from Olea and Pflueger (2013) should be used for detecting weak instruments. In a just-identified setting, the efficient F-statistics coincide with the usual F-statistic from the first-stage regression. Furthermore, they indicate that the efficient F-statistic should be compared to Stock and Yogo (2005) critical values in just-identified settings, and to Olea and Pflueger (2013) crit- ical values in over-identified settings. In addition, Keane and Neal (2021) draw the same conclusion about detecting weak instruments. On a cursory look at the lower panels of Table B.13 and Table B.14, we can say that our instruments perform well, especially for excess credit reallocation. The F-statistic from the first-stage regression is above the rule thumb of 10. But there are some model specifications 3 Please see Andrews et al. (2019) and Keane and Neal (2021) for a detailed discussion of detecting and treating weak instruments. 77 in which the instrument fails to pass the threshold of 10, particularly for net credit growth. Thus, we perform a weak instrument test to ensure that our instruments do well. We compare the F-statistic from the first-stage regression to the critical values of Stock and Yogo (2005). Furthermore, we need to make weak-instrument robust inference in case the instrument fails to pass weak instrument test. Keane and Neal (2021) claim that in a just-identified setting with one endogenous regressor, the F-statistic for weak instrument test from Anderson and Rubin (1949) is uniformly the most powerful test. We present the relevant Anderson- Rubin (AR) F-statistic in Table B.13 and Table B.14. However, the instrument for excess credit reallocation does well and can be regarded as a “strong" instrument. Thus, we check the AR F-statistics for our instrument for net credit growth. AR test confirms that we can make weak-instrument robust inference. 78 Table B.13 The effect of credit reallocation on innovation with first stage results (1) (2) (3) (4) (5) (6) VARIABLES Patents Patents Patents Patents Patents Patents per firms per firms per firms per firms per firms per firms Excess Credit Reallocation -0.098*** -0.027** -0.102*** -0.111*** -0.026** -0.027** (0.028) (0.012) (0.030) (0.033) (0.012) (0.013) Share of Active Population 0.000*** 0.000*** 0.000*** (0.000) (0.000) (0.000) No. of Bank Branches per Firms -0.136*** -0.136*** -0.135*** (0.040) (0.040) (0.040) Share of Individual Firms -0.035*** -0.035*** -0.035*** (0.004) (0.004) (0.004) Share of Higher Education Degrees 0.049*** 0.049*** 0.049*** (0.006) (0.006) (0.006) Productivity (Total Value Added per Firm) 0.000*** 0.000*** 0.000*** (0.000) (0.000) (0.000) Excess Credit Reallocation (First lag) 0.004 0.004 -0.000 -0.000 (0.004) (0.004) (0.002) (0.002) Excess Credit Reallocation (Second lag) 0.011 0.001 (0.007) (0.002) Observations 1,204 1,204 1,204 1,204 1,204 1,204 Region FE YES YES YES YES YES YES Year FE YES YES YES YES YES YES Instruments Weak Instrument Test by Stock and Yogo (2005) F-Stat 17.43 19.02 15.74 14.27 17.36 15.95 Stock-Yogo Critical Values 10% of Maximal IV Size 16.38 16.38 16.38 16.38 16.38 16.38 15% of Maximal IV Size 8.96 8.96 8.96 8.96 8.96 8.96 20% of Maximal IV Size 6.66 6.66 6.66 6.66 6.66 6.66 25% of Maximal IV Size 5.53 5.53 5.53 5.53 5.53 5.53 Weak Instrument Test by Olea and Pflueger (2013) Efficient F-Stat 17.43 19.02 15.74 14.27 17.36 15.95 Critical Values 5% of Worst Case Bias 37.42 37.42 37.42 37.42 37.42 37.42 10% of Worst Case Bias 23.11 23.11 23.11 23.11 23.11 23.11 20% of Worst Case Bias 15.06 15.06 15.06 15.06 15.06 15.06 30% of Worst Case Bias 12.04 12.04 12.04 12.04 12.04 12.04 Weak Instrument Robust Inference - AR Test (Anderson and Rubin (1949)) AR F-Stat 32.04 6.196 33.17 34.34 5.705 5.272 First Stage Regression VARIABLES EXC EXC EXC EXC EXC EXC No. of Savings Banks in 1936 (per 100,000 inhabitants) 0.015*** 0.015*** 0.015*** 0.014*** 0.015*** 0.014*** (0.004) (0.003) (0.004) (0.004) (0.004) (0.004) Robust standard errors in parentheses *** p<0.01, ** p<0.05, * p<0.1 79 Table B.14 The effect of credit growth on innovation with first stage results (1) (2) (3) (4) (5) (6) VARIABLES Patents Patents Patents Patents Patents Patents per firms per firms per firms per firms per firms per firms Net Credit Growth 0.010** 0.014** 0.012* 0.012* 0.017** 0.017** (0.005) (0.006) (0.006) (0.006) (0.007) (0.007) Share of Active Population -0.001 -0.001 -0.001 (0.000) (0.000) (0.000) No. of Bank Branches per Firms -0.067 -0.070 -0.069 (0.068) (0.074) (0.074) Share of Individual Firms -0.037*** -0.036*** -0.036*** (0.005) (0.005) (0.005) Share of Higher Education Degrees 0.046*** 0.045*** 0.045*** (0.008) (0.009) (0.009) Productivity (Total Value Added per Firm) 0.000*** 0.000*** 0.000*** (0.000) (0.000) (0.000) Net Credit Growth (First lag) -0.003** -0.003** -0.003** -0.003** (0.001) (0.001) (0.001) (0.001) Net Credit Growth (Second lag) -0.000 0.000 (0.001) (0.001) Observations 1,162 1,162 1,162 1,162 1,162 1,162 Region FE YES YES YES YES YES YES Year FE YES YES YES YES YES YES Instruments Weak Instrument Test by Stock and Yogo (2005) F-Stat 7.631 9.369 5.853 5.898 7.182 7.345 Stock-Yogo Critical Values 10% of Maximal IV Size 16.38 16.38 16.38 16.38 16.38 16.38 15% of Maximal IV Size 8.96 8.96 8.96 8.96 8.96 8.96 20% of Maximal IV Size 6.66 6.66 6.66 6.66 6.66 6.66 25% of Maximal IV Size 5.53 5.53 5.53 5.53 5.53 5.53 Weak Instrument Test by Olea and Pflueger (2013) Efficient F-Stat 7.631 9.369 5.853 5.898 7.182 7.345 Critical Values 5% of Worst Case Bias 37.42 37.42 37.42 37.42 37.42 37.42 10% of Worst Case Bias 23.11 23.11 23.11 23.11 23.11 23.11 20% of Worst Case Bias 15.06 15.06 15.06 15.06 15.06 15.06 30% of Worst Case Bias 12.04 12.04 12.04 12.04 12.04 12.04 Weak Instrument Robust Inference - AR Test (Anderson and Rubin (1949)) AR F-Stat 9.519 19.79 10.39 10.60 21.14 21.33 First Stage Regression VARIABLES NET NET NET NET NET NET Inverse Credit Market Concentration in 1936 0.008*** 0.009*** 0.007** 0.007** 0.008*** 0.008*** (0.003) (0.003) (0.003) (0.003) (0.003) (0.003) Robust standard errors in parentheses *** p<0.01, ** p<0.05, * p<0.1 80 B.3.2 Robustness to the North-South Divide In this section, we show that our results are not driven by the North-South divide. Histor- ically, the Southern regions and provinces tend to be financially underdeveloped in Italy. In addition, the structure of the banking industry in 1936 was a result of the Banking legislation of 1936. The structure in the Northern regions was more likely to be the outcome of histori- cal events and forced consolidation regardless of the level of economic development in 1930s. Therefore, excluding the Southern regions provides more exogeneity for our instruments. We drop the provinces in the Southern regions from the sample. The results hold even more strongly for excess credit reallocation. However, the significance of net credit growth seems to disappear. Table B.15 and Table B.16 present the results. 81 Table B.15 The effect of credit reallocation on innovation in the Northern provinces (1) (2) (3) (4) (5) (6) VARIABLES Patents Patents Patents Patents Patents Patents per firms per firms per firms per firms per firms per firms Excess Credit Reallocation -0.188*** -0.060** -0.198*** -0.218*** -0.059** -0.061** (0.057) (0.025) (0.064) (0.066) (0.026) (0.027) Share of Active Population -0.000* -0.000* -0.000* (0.000) (0.000) (0.000) No. of Bank Branches per Firms -0.223*** -0.223*** -0.223*** (0.071) (0.071) (0.072) Share of Individual Firms -0.028*** -0.028*** -0.028*** (0.005) (0.005) (0.005) Share of Higher Education Degrees 0.036*** 0.036*** 0.036*** (0.011) (0.011) (0.011) Productivity (Total Value Added per Firm) 0.000*** 0.000*** 0.000*** (0.000) (0.000) (0.000) Excess Credit Reallocation (First lag) 0.010 0.010 -0.001 -0.001 (0.009) (0.010) (0.003) (0.003) Excess Credit Reallocation (Second lag) 0.023 0.002 (0.020) (0.006) Observations 546 546 546 546 546 546 Region FE YES YES YES YES YES YES Year FE YES YES YES YES YES YES Instruments Weak Instrument Test by Stock and Yogo (2005) F-Stat 11.53 12.21 9.978 10.95 10.60 12.76 Stock-Yogo Critical Values 10% of Maximal IV Size 16.38 16.38 16.38 16.38 16.38 16.38 15% of Maximal IV Size 8.96 8.96 8.96 8.96 8.96 8.96 20% of Maximal IV Size 6.66 6.66 6.66 6.66 6.66 6.66 25% of Maximal IV Size 5.53 5.53 5.53 5.53 5.53 5.53 Weak Instrument Test by Olea and Pflueger (2013) Efficient F-Stat 11.53 12.21 9.978 10.95 10.60 12.76 Critical Values 5% of Worst Case Bias 37.42 37.42 37.42 37.42 37.42 37.42 10% of Worst Case Bias 23.11 23.11 23.11 23.11 23.11 23.11 20% of Worst Case Bias 15.06 15.06 15.06 15.06 15.06 15.06 30% of Worst Case Bias 12.04 12.04 12.04 12.04 12.04 12.04 Weak Instrument Robust Inference - AR Test (Anderson and Rubin (1949)) AR F-Stat 44 8.863 44.91 45.16 7.884 6.857 First Stage Regression VARIABLES EXC EXC EXC EXC EXC EXC No. of Savings Banks in 1936 (per 100,000 inhabitants) 0.017*** 0.020*** 0.017*** 0.015*** 0.019*** 0.018*** (0.005) (0.006) (0.005) (0.005) (0.006) (0.005) Robust standard errors in parentheses *** p<0.01, ** p<0.05, * p<0.1 82 Table B.16 The effect of credit growth on innovation in the Northern provinces (1) (2) (3) (4) (5) (6) VARIABLES Patents Patents Patents Patents Patents Patents per firms per firms per firms per firms per firms per firms Net Credit Growth 0.039* 0.029 0.050 0.048 0.036 0.034 (0.022) (0.018) (0.035) (0.032) (0.026) (0.023) Share of Active Population -0.002 -0.002 -0.002 (0.001) (0.001) (0.001) No. of Bank Branches per Firms -0.247** -0.255** -0.257** (0.107) (0.126) (0.120) Share of Individual Firms -0.039*** -0.041*** -0.041*** (0.009) (0.011) (0.011) Share of Higher Education Degrees 0.020 0.017 0.019 (0.019) (0.024) (0.022) Productivity (Total Value Added per Firm) 0.000*** 0.000*** 0.000*** (0.000) (0.000) (0.000) Net Credit Growth (First lag) -0.008 -0.008 -0.005 -0.005 (0.006) (0.006) (0.004) (0.004) Net Credit Growth (Second lag) 0.003 0.003 (0.003) (0.003) Observations 518 518 518 518 518 518 Region FE YES YES YES YES YES YES Year FE YES YES YES YES YES YES Instruments Weak Instrument Test by Stock and Yogo (2005) F-Stat 3.645 3.403 2.188 2.442 2.239 2.550 Stock-Yogo Critical Values 10% of Maximal IV Size 16.38 16.38 16.38 16.38 16.38 16.38 15% of Maximal IV Size 8.96 8.96 8.96 8.96 8.96 8.96 20% of Maximal IV Size 6.66 6.66 6.66 6.66 6.66 6.66 25% of Maximal IV Size 5.53 5.53 5.53 5.53 5.53 5.53 Weak Instrument Test by Olea and Pflueger (2013) Efficient F-Stat 3.645 3.403 2.188 2.442 2.239 2.550 Critical Values 5% of Worst Case Bias 37.42 37.42 37.42 37.42 37.42 37.42 10% of Worst Case Bias 23.11 23.11 23.11 23.11 23.11 23.11 20% of Worst Case Bias 15.06 15.06 15.06 15.06 15.06 15.06 30% of Worst Case Bias 12.04 12.04 12.04 12.04 12.04 12.04 Weak Instrument Robust Inference - AR Test (Anderson and Rubin (1949)) AR F-Stat 13.32 8.407 13.81 13.76 8.789 8.832 First Stage Regression VARIABLES NET NET NET NET NET NET Inverse Credit Market Concentration in 1936 0.009* 0.008* 0.007 0.007 0.007 0.007 (0.005) (0.005) (0.005) (0.005) (0.005) (0.005) Robust standard errors in parentheses *** p<0.01, ** p<0.05, * p<0.1 83 B.3.3 Robustness to an Alternative Specification We test the robustness of baseline results by controlling for province characteristics and adding lags of the variable of interest. We further perform more tests to assess the robustness of the baseline results. Table B.17 and Table B.18 present the results. We change the definition of the dependent variable to “Patents per 100,000 people," as widely used in the literature, and repeat the same exercises with the new definition. The results are robust to a change in the definition of the dependent variable. The signs of excess credit reallocation and net credit growth remain the same as all province characteristics. However, the magnitudes of the coefficient estimates change drastically with new definition, which is expected due to change in the denominator of the dependent variable. 84 Table B.17 The effect of credit reallocation on innovation - Alternative specification (1) (2) (3) (4) (5) (6) VARIABLES Patents Patents Patents Patents Patents Patents per 100,000 per 100,000 per 100,000 per 100,000 per 100,000 per 100,000 people people people people people people Excess Credit Reallocation -393.191*** -141.073*** -409.612*** -447.789*** -140.352** -144.709** (112.978) (53.241) (121.219) (136.262) (54.996) (58.856) Share of Active Population 1.870*** 1.870*** 1.871*** (0.580) (0.580) (0.580) No. of Bank Branches per 100,000 People -0.037 -0.037 -0.038 (0.051) (0.051) (0.051) Share of Individual Firms -119.726*** -119.725*** -119.567*** (18.139) (18.130) (18.204) Share of Higher Education Degrees 132.263*** 132.253*** 132.511*** (24.512) (24.497) (24.622) Productivity (Total Value Added per capita) 1.522*** 1.522*** 1.520*** (0.156) (0.156) (0.156) Excess Credit Reallocation (First lag) 18.617 18.434 -0.824 -0.826 (16.636) (17.752) (7.680) (7.776) Excess Credit Reallocation (Second lag) 48.030 5.268 (29.938) (10.993) Observations 1,204 1,204 1,204 1,204 1,204 1,204 Region FE YES YES YES YES YES YES Year FE YES YES YES YES YES YES Instruments Weak Instrument Test by Stock and Yogo (2005) F-Stat 17.43 20.65 15.74 14.27 18.86 17.14 Stock-Yogo Critical Values 10% of Maximal IV Size 16.38 16.38 16.38 16.38 16.38 16.38 15% of Maximal IV Size 8.96 8.96 8.96 8.96 8.96 8.96 20% of Maximal IV Size 6.66 6.66 6.66 6.66 6.66 6.66 25% of Maximal IV Size 5.53 5.53 5.53 5.53 5.53 5.53 Weak Instrument Test by Olea and Pflueger (2013) Efficient F-Stat 17.43 20.65 15.74 14.27 18.86 17.14 Critical Values 5% of Worst Case Bias 37.42 37.42 37.42 37.42 37.42 37.42 10% of Worst Case Bias 23.11 23.11 23.11 23.11 23.11 23.11 20% of Worst Case Bias 15.06 15.06 15.06 15.06 15.06 15.06 30% of Worst Case Bias 12.04 12.04 12.04 12.04 12.04 12.04 Weak Instrument Robust Inference - AR Test (Anderson and Rubin (1949)) AR F-Stat 31.50 9.654 32.81 34.21 9.012 8.300 First Stage Regression VARIABLES EXC EXC EXC EXC EXC EXC No. of Savings Banks in 1936 (per 100,000 inhabitants) 0.015*** 0.015*** 0.015*** 0.014*** 0.015*** 0.014*** (0.004) (0.003) (0.004) (0.004) (0.003) (0.003) Robust standard errors in parentheses *** p<0.01, ** p<0.05, * p<0.1 85 Table B.18 The effect of credit growth on innovation - Alternative specification (1) (2) (3) (4) (5) (6) VARIABLES Patents Patents Patents Patents Patents Patents per 100,000 per 100,000 per 100,000 per 100,000 per 100,000 per 100,000 people people people people people people Net Credit Growth 39.293** 51.718** 47.595* 47.975* 60.679** 60.184** (18.963) (21.676) (24.427) (24.506) (27.106) (26.762) Share of Active Population -1.014 -0.983 -1.004 (1.463) (1.605) (1.601) No. of Bank Branches per 100,000 People -0.102* -0.110* -0.109* (0.059) (0.062) (0.062) Share of Individual Firms -147.844*** -148.395*** -148.325*** (21.718) (22.722) (22.649) Share of Higher Education Degrees 144.856*** 141.821*** 142.138*** (29.499) (30.941) (30.892) Productivity (Total Value Added per capita) 1.259*** 1.261*** 1.262*** (0.187) (0.195) (0.195) Net Credit Growth (First lag) -10.720** -10.652** -11.495** -11.575** (4.938) (4.981) (5.109) (5.093) Net Credit Growth (Second lag) -0.797 1.026 (2.258) (2.355) Observations 1,162 1,162 1,162 1,162 1,162 1,162 Region FE YES YES YES YES YES YES Year FE YES YES YES YES YES YES Instruments Weak Instrument Test by Stock and Yogo (2005) F-Stat 7.631 9.185 5.853 5.898 7.182 7.318 Stock-Yogo Critical Values 10% of Maximal IV Size 16.38 16.38 16.38 16.38 16.38 16.38 15% of Maximal IV Size 8.96 8.96 8.96 8.96 8.96 8.96 20% of Maximal IV Size 6.66 6.66 6.66 6.66 6.66 6.66 25% of Maximal IV Size 5.53 5.53 5.53 5.53 5.53 5.53 Weak Instrument Test by Olea and Pflueger (2013) Efficient F-Stat 7.631 9.185 5.853 5.898 7.182 7.318 Critical Values 5% of Worst Case Bias 37.42 37.42 37.42 37.42 37.42 37.42 10% of Worst Case Bias 23.11 23.11 23.11 23.11 23.11 23.11 20% of Worst Case Bias 15.06 15.06 15.06 15.06 15.06 15.06 30% of Worst Case Bias 12.04 12.04 12.04 12.04 12.04 12.04 Weak Instrument Robust Inference - AR Test (Anderson and Rubin (1949)) AR F-Stat 9.705 16.32 10.67 10.92 17.56 17.73 First Stage Regression VARIABLES NET NET NET NET NET NET Inverse Credit Market Concentration in 1936 0.008*** 0.009*** 0.007** 0.007** 0.008*** 0.008*** (0.003) (0.003) (0.003) (0.003) (0.003) (0.003) Robust standard errors in parentheses *** p<0.01, ** p<0.05, * p<0.1 86 B.3.4 Robustness to Inclusion of Additional Data Second, we use the patent data from 1950 to 1963 in the main estimations because there is a gap between 1963 and 1968 in the patent data. To check whether the data from 1968 to 1973 change the results, we include this data in our sample and reestimate all the models. The results are robust to the inclusion of further data. The coefficient estimates for excess credit reallocation slightly increase, while the coefficient estimates for net credit growth remain virtually unaltered. 87 Table B.19 The effect of credit reallocation on innovation (from 1950 to 1973) (1) (2) (3) (4) (5) (6) VARIABLES Patents Patents Patents Patents Patents Patents per firms per firms per firms per firms per firms per firms Excess Credit Reallocation -0.132*** -0.039** -0.146*** -0.161*** -0.043** -0.048** (0.043) (0.018) (0.052) (0.061) (0.021) (0.023) Share of Active Population 0.000* 0.000* 0.000* (0.000) (0.000) (0.000) No. of Bank Branches per Firms -0.053 -0.056 -0.055 (0.035) (0.036) (0.037) Share of Individual Firms -0.030*** -0.030*** -0.030*** (0.003) (0.003) (0.003) Share of Higher Education Degrees 0.037*** 0.038*** 0.037*** (0.005) (0.005) (0.005) Productivity (Total Value Added per Firm) 0.000*** 0.000*** 0.000*** (0.000) (0.000) (0.000) Excess Credit Reallocation (First lag) 0.015* 0.015* 0.005 0.005 (0.008) (0.008) (0.003) (0.003) Excess Credit Reallocation (Second lag) 0.016 0.005 (0.010) (0.003) Observations 1,718 1,718 1,717 1,716 1,717 1,716 Region FE YES YES YES YES YES YES Year FE YES YES YES YES YES YES Instruments Weak Instrument Test by Stock and Yogo (2005) F-Stat 11.18 11.58 8.940 7.856 9.324 8.207 Stock-Yogo Critical Values 10% of Maximal IV Size 16.38 16.38 16.38 16.38 16.38 16.38 15% of Maximal IV Size 8.96 8.96 8.96 8.96 8.96 8.96 20% of Maximal IV Size 6.66 6.66 6.66 6.66 6.66 6.66 25% of Maximal IV Size 5.53 5.53 5.53 5.53 5.53 5.53 Weak Instrument Test by Olea and Pflueger (2013) Efficient F-Stat 11.18 11.58 8.940 7.856 9.324 8.207 Critical Values 5% of Worst Case Bias 37.42 37.42 37.42 37.42 37.42 37.42 10% of Worst Case Bias 23.11 23.11 23.11 23.11 23.11 23.11 20% of Worst Case Bias 15.06 15.06 15.06 15.06 15.06 15.06 30% of Worst Case Bias 12.04 12.04 12.04 12.04 12.04 12.04 Weak Instrument Robust Inference - AR Test (Anderson and Rubin (1949)) AR F-Stat 41.08 7.333 42.27 43.30 7.663 7.785 First Stage Regression VARIABLES EXC EXC EXC EXC EXC EXC No. of Savings Banks in 1936 (per 100,000 inhabitants) 0.009*** 0.009*** 0.008*** 0.008*** 0.009*** 0.008*** (0.003) (0.003) (0.003) (0.003) (0.003) (0.003) Robust standard errors in parentheses *** p<0.01, ** p<0.05, * p<0.1 88 Table B.20 The effect of credit growth on innovation (from 1950 to 1973) (1) (2) (3) (4) (5) (6) VARIABLES Patents Patents Patents Patents Patents Patents per firms per firms per firms per firms per firms per firms Net Credit Growth 0.011** 0.016*** 0.013** 0.014** 0.020** 0.020** (0.005) (0.006) (0.007) (0.007) (0.008) (0.008) Share of Active Population -0.001** -0.001* -0.001* (0.000) (0.000) (0.000) No. of Bank Branches per Firms 0.020 0.019 0.020 (0.063) (0.071) (0.072) Share of Individual Firms -0.030*** -0.030*** -0.030*** (0.004) (0.004) (0.004) Share of Higher Education Degrees 0.036*** 0.036*** 0.036*** (0.006) (0.007) (0.007) Productivity (Total Value Added per Firm) 0.000*** 0.000*** 0.000*** (0.000) (0.000) (0.000) Net Credit Growth (First lag) -0.003** -0.003** -0.004** -0.004** (0.001) (0.001) (0.002) (0.002) Net Credit Growth (Second lag) -0.000 -0.000 (0.001) (0.001) Observations 1,658 1,658 1,657 1,656 1,657 1,656 Region FE YES YES YES YES YES YES Year FE YES YES YES YES YES YES Instruments Weak Instrument Test by Stock and Yogo (2005) F-Stat 8.668 10.50 6.505 6.365 7.990 7.930 Stock-Yogo Critical Values 10% of Maximal IV Size 16.38 16.38 16.38 16.38 16.38 16.38 15% of Maximal IV Size 8.96 8.96 8.96 8.96 8.96 8.96 20% of Maximal IV Size 6.66 6.66 6.66 6.66 6.66 6.66 25% of Maximal IV Size 5.53 5.53 5.53 5.53 5.53 5.53 Weak Instrument Test by Olea and Pflueger (2013) Efficient F-Stat 8.668 10.50 6.505 6.365 7.990 7.930 Critical Values 5% of Worst Case Bias 37.42 37.42 37.42 37.42 37.42 37.42 10% of Worst Case Bias 23.11 23.11 23.11 23.11 23.11 23.11 20% of Worst Case Bias 15.06 15.06 15.06 15.06 15.06 15.06 30% of Worst Case Bias 12.04 12.04 12.04 12.04 12.04 12.04 Weak Instrument Robust Inference - AR Test (Anderson and Rubin (1949)) AR F-Stat 11.59 23.59 12.52 12.67 25.47 25.61 First Stage Regression VARIABLES NET NET NET NET NET NET Inverse Credit Market Concentration in 1936 0.006*** 0.007*** 0.005** 0.005** 0.006*** 0.006*** (0.002) (0.002) (0.002) (0.002) (0.002) (0.002) Robust standard errors in parentheses *** p<0.01, ** p<0.05, * p<0.1 89 B.3.4.1 The North-South Divide The results are not driven by the North-South divide, since they hold (even more strongly) when we drop Southern regions from the sample with the inclusion of additional data. Fi- nancially underdeveloped regions tend to be in the South. In sum, the 1936 law froze the Italian banking system at a very peculiar time. If we exclude the South, the structure of the banking industry in 1936 was the result of historical accidents and forced consolidation, with no connection to the level of economic development at that time. Table B.21 and Table B.22 present the results. 90 Table B.21 The effect of credit reallocation on innovation in the Northern provinces (from 1950 to 1973) (1) (2) (3) (4) (5) (6) VARIABLES Patents Patents Patents Patents Patents Patents per firms per firms per firms per firms per firms per firms Excess Credit Reallocation -0.202*** -0.049** -0.217*** -0.237*** -0.051** -0.054* (0.060) (0.024) (0.070) (0.074) (0.026) (0.028) Share of Active Population -0.000** -0.000** -0.000** (0.000) (0.000) (0.000) No. of Bank Branches per Firms -0.186*** -0.187*** -0.186*** (0.057) (0.058) (0.059) Share of Individual Firms -0.032*** -0.032*** -0.031*** (0.004) (0.004) (0.004) Share of Higher Education Degrees 0.023*** 0.023*** 0.023*** (0.008) (0.008) (0.008) Productivity (Total Value Added per Firm) 0.000*** 0.000*** 0.000*** (0.000) (0.000) (0.000) Excess Credit Reallocation (First lag) 0.017 0.017 0.002 0.003 (0.011) (0.012) (0.003) (0.003) Excess Credit Reallocation (Second lag) 0.026 0.003 (0.020) (0.005) Observations 780 780 780 780 780 780 Region FE YES YES YES YES YES YES Year FE YES YES YES YES YES YES Instruments Weak Instrument Test by Stock and Yogo (2005) F-Stat 11.76 13.18 9.893 10.45 11.07 12.60 Stock-Yogo Critical Values 10% of Maximal IV Size 16.38 16.38 16.38 16.38 16.38 16.38 15% of Maximal IV Size 8.96 8.96 8.96 8.96 8.96 8.96 20% of Maximal IV Size 6.66 6.66 6.66 6.66 6.66 6.66 25% of Maximal IV Size 5.53 5.53 5.53 5.53 5.53 5.53 Weak Instrument Test by Olea and Pflueger (2013) Efficient F-Stat 11.76 13.18 9.893 10.45 11.07 12.60 Critical Values 5% of Worst Case Bias 37.42 37.42 37.42 37.42 37.42 37.42 10% of Worst Case Bias 23.11 23.11 23.11 23.11 23.11 23.11 20% of Worst Case Bias 15.06 15.06 15.06 15.06 15.06 15.06 30% of Worst Case Bias 12.04 12.04 12.04 12.04 12.04 12.04 Weak Instrument Robust Inference - AR Test (Anderson and Rubin (1949)) AR F-Stat 51.75 5.440 52.70 52.87 5.161 4.687 First Stage Regression VARIABLES EXC EXC EXC EXC EXC EXC No. of Savings Banks in 1936 (per 100,000 inhabitants) 0.013*** 0.016*** 0.012*** 0.012*** 0.015*** 0.014*** (0.004) (0.004) (0.004) (0.004) (0.004) (0.004) Robust standard errors in parentheses *** p<0.01, ** p<0.05, * p<0.1 91 Table B.22 The effect of credit growth on innovation in the Northern provinces (from 1950 to 1973) (1) (2) (3) (4) (5) (6) VARIABLES Patents Patents Patents Patents Patents Patents per firms per firms per firms per firms per firms per firms Net Credit Growth 0.058 0.054 0.078 0.077 0.066 0.065 (0.041) (0.041) (0.073) (0.073) (0.059) (0.057) Share of Active Population -0.004 -0.004 -0.004 (0.003) (0.003) (0.003) No. of Bank Branches per Firms -0.166 -0.164 -0.166 (0.156) (0.186) (0.182) Share of Individual Firms -0.039*** -0.040*** -0.041*** (0.011) (0.013) (0.013) Share of Higher Education Degrees 0.008 0.004 0.005 (0.024) (0.030) (0.029) Productivity (Total Value Added per Firm) 0.000*** 0.000** 0.000** (0.000) (0.000) (0.000) Net Credit Growth (First lag) -0.015 -0.015 -0.009 -0.009 (0.015) (0.015) (0.009) (0.009) Net Credit Growth (Second lag) 0.001 0.004 (0.004) (0.004) Observations 740 740 740 740 740 740 Region FE YES YES YES YES YES YES Year FE YES YES YES YES YES YES Instruments Weak Instrument Test by Stock and Yogo (2005) F-Stat 2.018 1.855 1.098 1.108 1.248 1.304 Stock-Yogo Critical Values 10% of Maximal IV Size 16.38 16.38 16.38 16.38 16.38 16.38 15% of Maximal IV Size 8.96 8.96 8.96 8.96 8.96 8.96 20% of Maximal IV Size 6.66 6.66 6.66 6.66 6.66 6.66 25% of Maximal IV Size 5.53 5.53 5.53 5.53 5.53 5.53 Weak Instrument Test by Olea and Pflueger (2013) Efficient F-Stat 2.018 1.855 1.098 1.108 1.248 1.304 Critical Values 5% of Worst Case Bias 37.42 37.42 37.42 37.42 37.42 37.42 10% of Worst Case Bias 23.11 23.11 23.11 23.11 23.11 23.11 20% of Worst Case Bias 15.06 15.06 15.06 15.06 15.06 15.06 30% of Worst Case Bias 12.04 12.04 12.04 12.04 12.04 12.04 Weak Instrument Robust Inference - AR Test (Anderson and Rubin (1949)) AR F-Stat 18.94 18.17 19.06 19.02 18.17 18.14 First Stage Regression VARIABLES NET NET NET NET NET NET Inverse Credit Market Concentration in 1936 0.005 0.005 0.004 0.004 0.004 0.004 (0.004) (0.004) (0.004) (0.004) (0.004) (0.004) Robust standard errors in parentheses *** p<0.01, ** p<0.05, * p<0.1 92 B.3.4.2 An Alternative Specification We perform the estimation with the dependent variable of “Patents per 100,000 people", including additional data. Table B.23 and Table B.24 present the results. The results are robust to a change in the definition of the dependent variable. The signs of excess credit reallocation and net credit growth remain the same as all province characteristics. However, the magnitudes of the coefficient estimates change drastically with new definition, which is expected due to change in the denominator of the dependent variable. 93 Table B.23 The effect of credit reallocation on innovation - Alternative specification (from 1950 to 1973) (1) (2) (3) (4) (5) (6) VARIABLES Patents Patents Patents Patents Patents Patents per 100,000 per 100,000 per 100,000 per 100,000 per 100,000 per 100,000 people people people people people people Excess Credit Reallocation -533.406*** -202.366** -594.176*** -655.353*** -226.240** -247.446** (176.771) (84.992) (215.094) (249.404) (98.357) (110.530) Share of Active Population 1.372*** 1.378*** 1.395*** (0.434) (0.442) (0.452) No. of Bank Branches per 100,000 People -0.012 -0.015 -0.017 (0.042) (0.044) (0.045) Share of Individual Firms -127.659*** -127.584*** -127.129*** (15.661) (15.982) (16.305) Share of Higher Education Degrees 128.329*** 129.252*** 129.534*** (20.664) (21.468) (22.233) Productivity (Total Value Added per capita) 0.840*** 0.837*** 0.832*** (0.143) (0.148) (0.153) Excess Credit Reallocation (First lag) 62.751** 63.925* 26.028* 26.521* (32.007) (34.884) (14.196) (14.994) Excess Credit Reallocation (Second lag) 66.671* 23.029 (40.470) (16.944) Observations 1,718 1,718 1,717 1,716 1,717 1,716 Region FE YES YES YES YES YES YES Year FE YES YES YES YES YES YES Instruments Weak Instrument Test by Stock and Yogo (2005) F-Stat 11.18 13.10 8.940 7.856 10.66 9.382 Stock-Yogo Critical Values 10% of Maximal IV Size 16.38 16.38 16.38 16.38 16.38 16.38 15% of Maximal IV Size 8.96 8.96 8.96 8.96 8.96 8.96 20% of Maximal IV Size 6.66 6.66 6.66 6.66 6.66 6.66 25% of Maximal IV Size 5.53 5.53 5.53 5.53 5.53 5.53 Weak Instrument Test by Olea and Pflueger (2013) Efficient F-Stat 11.18 13.10 8.940 7.856 10.66 9.382 Critical Values 5% of Worst Case Bias 37.42 37.42 37.42 37.42 37.42 37.42 10% of Worst Case Bias 23.11 23.11 23.11 23.11 23.11 23.11 20% of Worst Case Bias 15.06 15.06 15.06 15.06 15.06 15.06 30% of Worst Case Bias 12.04 12.04 12.04 12.04 12.04 12.04 Weak Instrument Robust Inference - AR Test (Anderson and Rubin (1949)) AR F-Stat 37.68 9.749 38.95 40.10 10.24 10.32 First Stage Regression VARIABLES EXC EXC EXC EXC EXC EXC No. of Savings Banks in 1936 (per 100,000 inhabitants) 0.009*** 0.010*** 0.008*** 0.008*** 0.009*** 0.008*** (0.003) (0.003) (0.003) (0.003) (0.003) (0.003) Robust standard errors in parentheses *** p<0.01, ** p<0.05, * p<0.1 94 Table B.24 The effect of credit growth on innovation - Alternative specification (from 1950 to 1973) (1) (2) (3) (4) (5) (6) VARIABLES Patents Patents Patents Patents Patents Patents per 100,000 per 100,000 per 100,000 per 100,000 per 100,000 per 100,000 people people people people people people Net Credit Growth 50.058** 60.528** 60.328** 60.981** 71.974** 72.238** (21.672) (23.861) (28.357) (28.885) (30.290) (30.453) Share of Active Population -2.525 -2.325 -2.321 (1.605) (1.725) (1.720) No. of Bank Branches per 100,000 People -0.042 -0.047 -0.047 (0.049) (0.052) (0.053) Share of Individual Firms -139.148*** -138.657*** -138.699*** (16.628) (17.358) (17.359) Share of Higher Education Degrees 134.859*** 134.192*** 134.152*** (23.095) (24.408) (24.480) Productivity (Total Value Added per capita) 0.720*** 0.725*** 0.725*** (0.147) (0.156) (0.156) Net Credit Growth (First lag) -13.492** -13.262** -14.586** -14.584** (6.019) (5.999) (5.848) (5.825) Net Credit Growth (Second lag) -1.695 -0.315 (2.457) (2.504) Observations 1,658 1,658 1,657 1,656 1,657 1,656 Region FE YES YES YES YES YES YES Year FE YES YES YES YES YES YES Instruments Weak Instrument Test by Stock and Yogo (2005) F-Stat 8.668 10.64 6.505 6.365 8.241 8.223 Stock-Yogo Critical Values 10% of Maximal IV Size 16.38 16.38 16.38 16.38 16.38 16.38 15% of Maximal IV Size 8.96 8.96 8.96 8.96 8.96 8.96 20% of Maximal IV Size 6.66 6.66 6.66 6.66 6.66 6.66 25% of Maximal IV Size 5.53 5.53 5.53 5.53 5.53 5.53 Weak Instrument Test by Olea and Pflueger (2013) Efficient F-Stat 8.668 10.64 6.505 6.365 8.241 8.223 Critical Values 5% of Worst Case Bias 37.42 37.42 37.42 37.42 37.42 37.42 10% of Worst Case Bias 23.11 23.11 23.11 23.11 23.11 23.11 20% of Worst Case Bias 15.06 15.06 15.06 15.06 15.06 15.06 30% of Worst Case Bias 12.04 12.04 12.04 12.04 12.04 12.04 Weak Instrument Robust Inference - AR Test (Anderson and Rubin (1949)) AR F-Stat 14.44 19.26 15.39 15.50 20.87 21.09 First Stage Regression VARIABLES NET NET NET NET NET NET Inverse Credit Market Concentration in 1936 0.006*** 0.007*** 0.005** 0.005** 0.006*** 0.006*** (0.002) (0.002) (0.002) (0.002) (0.002) (0.002) Robust standard errors in parentheses *** p<0.01, ** p<0.05, * p<0.1 95