Revisiting finance-trade linkage during the Great Trade Collapse. September 2016 (Work in Progress, not to be cited)

Transcription

1 Revisiting finance-trade linkage during the Great Trade Collapse September 2016 (Work in Progress, not to be cited) Yushi Yoshida Faculty of Economics Shiga University Abstract We re-examine whether Japan s export during the Great Trade Collapse is really affected by financial channel. In this paper we focus on Japan s regional export in which regional exporters rely heavily on regional banks. The damage on local banks in different regions exerted by the global financial crisis were heterogeneous. By examining the same industry across all regions in Japan, we test whether the financial shocks in regional banks affected regional exports differently. Our preliminary empirical evidence questions the validity of notion that financial turmoil in developed countries should diminish their exports. This result suggests that, if there is any finance-trade linkage, export decline is associated with financial shocks in developing countries on the import side. Keywords: Great Trade Collapse, Finance-Trade linkage, Regional banks JEL codes: E44, E32, G21, F40 Shiga University, Faculty of Economics, Banba, Hikone, Shiga, Japan. address: yushi.yoshida@biwako.shiga-u.ac.jp. 1

2 1. Introduction After the global financial crisis, world trade experienced a disproportionate fall from the preceding period. The so-called Great Trade Collapse in international trade has been examined by numerous studies that attempt to find explanations for this unprecedented decline (Ahn, Amiti, and Weinstein, 2011; Amiti and Weinstein, 2011; Bems, Johnson, Yi, 2011; Bussière et al. (2013); and a series of papers in 2010 December issue of IMF Economic Review). As a result, the Japan s fiscal-year trade balance turned deficit in 2008 after continuous 27 years of trade surplus 1. Among several causes of the Great trade Collapse, in this paper we focus on the finance-trade link. Amiti and Weinstein (2011) investigated whether tightening of trade finance during the global financial crisis caused additional decline in international trade. Amiti and Weinstein (2011) find that exports of Japanese manufactures declined if their most associated bank s market-to-book value fell. Following Amiti and Weinstein s (2011) emphasis on the link between trade finance and trade flows, Ahn, Amiti, and Weinstein (2011) turn to transportation modes of international trade, i.e., by air, land, or sea. Trade finance becomes more costly for trade which takes longer time for transportation. For the case of the United States, international trade by sea transportation takes longer time than air or land. Ahn, Amiti, and Weinstein (2011) confirmed that prices of sea-borne traded goods were relatively higher than those of air or land-borne goods during the crisis period. There is a long way from a macroeconomic financial turmoil to a shrinkage of export due to the credit constraint on exporting firms. First, financial turmoil must affect the lending behaviors of financial institutions in significant way. Financial facility supported by governments may mitigate the shocks (Buck and Goldberg, 2014; Correa, Goldberg, Rice, 2014). Second, a change in lending behavior of banks during the financial crisis must put financial constraints on trading firms (exporting firms or importing firms or both). The trade credit mechanism has been re-examined. However, exporting firms in developed countries or those with healthy financial balance sheet in normal times may never face financial constraints. Third and related to the above, those 1 Japan s trade balance had recorded surplus in terms of both the calendar and fiscal (from April to March) year since Except for a few months (mostly in January) in deficit due to seasonality reason, almost all monthly trade balances are in surplus until From October 2008 to March 2009, Japan recorded consecutive deficit months (except for February). After April 2009, the monthly trade balance is generally in surplus until the earthquake and tsunami disaster hit the northeastern part of Japan in March However, even before the earthquake, the annual export value of Japan never come back to the pre-crisis level of 7 trillion Japanese yen. 2

3 exporting or importing firms indeed in credit constraint must constitute substantial part of international trade. In this paper we reexamine the link between international trade and financial constraints by implementing provincial Japanese data. Both bank lending and exports data are collected at the prefecture (equivalent to province or states in other countries) level. We provide that the severity of financial constraint on exporters as well as the degrees of export decline varies widely among prefectures of Japan. Empirical evidence shows that these links are found to be very weak at the best and questions the validity of connecting the great trade collapse with financial constrained exporters. Our preliminary empirical evidence questions the validity of notion that financial turmoil in developed countries should diminish their exports. This result hints that export decline is associated with financial shocks in developing countries on the import side. This findings is more consistent with the emphasis on importers financial constraint in Antràs and Foley (2015) and the limited scope on credit constrained exporters in developed countries as in Bricongne et al. (2012). The structure of this paper is organized as follows. Section 2 reviews previous studies examining the finance-trade linkage. Section 3 examines the relationship between bank lending and exports at the sub-national level. Section 4 estimate, with introduction of financial variables, a variant of the Gravity model in differenced form. Section 5 concludes. 2. Literature Review The recent global financial crisis caused shrinkage in international trade at disproportionate size with respect to the decline in the world income. Some plausible explanations provided for the Great Trade Collapse are the shifts in expenditure components (Bussiére et al., 2013) and the sensitivity of import contents of exported final products (Bems et all, 2011). Given the major disturbance during the financial crisis being credit crunch, one of explanations for the trade collapse focuses on the trade finance. By investigating the firm-level Japanese data, Amiti and Weinstein (2011) show that domestic credit crunch during the global financial crisis exerted financial constraint on exporters, thus leading to the observed trade collapse. The link between international trade and finance are then closely examined at the firm-product level. In general all international trade needs to be financed by one of three types of financing terms: cash-in-advance, open account, and letter of credit. Antràs and Foley (2011) find that, during the financial crisis, exporters are more likely to demand more cash-in-advance terms with new customers and importers with 3

4 cash-in-advance or letter of credit decreased their purchases more than importers with open account. Given the prominent role of financially weak exporters/importers for the cause for the Great Trade Collapse, a closer examination on small and medium exporters that must rely on the small local banks provide a complementary study to Amiti and Weinstein (2011) who examined the listed big companies and fifteen major banks in Japan. Finance-trade link may not mean the same when seen from different scopes. Overall financial conditions captured by financial development at macroeconomic level, Beck (2002) provide evidence that better macroeconomic financial condition increases country s exports. However, at micro level, Berman and Héricourt (2010) find from a large cross-country firm-level study that better financial conditions do not necessarily increase the size of exports. Finance constraint may not only affect exporters whose constraints are binding, but also those exporters without binding constraints because they also have to commit to precautionary saving for future financial shocks (Caggese and Cuñat, 2013). Some empirical studies find that financial constraints were a minor issue at the best during the financial crisis. Bricongne et al. (2012) find using French firm level data that credit constraints mattered only in those industries in high financial dependence and the share of credit constrained firms is small. 3. Financial Depression and Export Decline at Japanese Regions The Great Trade Collapse affected Japan s export as severely as in other part of the world. However, the severity of damage may not be similar across Japanese exporters. Firms in one industry may be more damaged than firms in another industry because of the difference in income elasticity of industry demand. Couple with the fact that some regions are more specialized in specific industries, the degree of decline in export values may look different among regions within Japan. Japan consists of 47 prefectures, which are equivalent political unit of states in the United States. The prefectures are also aggregated as 9 regions. Figure 1 shows one prefecture from each of the three largest economic regions (Kanto, Chubu, and Kinki). Aichi prefecture representing from Chubu region suffered the most in the post-crisis period whereas Hyogo prefecture representing Kinki region was the least prone to the shock. The export growths for other prefectures are shown in Table 1. Even within the same industry, one firm may be more insulated from the crisis than other firms because of its relatively healthier financial condition. If the worsening 4

5 of financial conditions during the financial crisis is also heterogeneous among Japanese 9 regions, the finance-trade linkage may impact exports of Japanese regions differently. Figure 2 indicates the monthly growth of bank loans by year-on-year basis for 9 regions from April 1999 to August Assuming regional bank loans represent financial conditions of those regions, financial conditions widely varies among 9 regions even before the crisis. Figure 3 depicts the year-on-year growth of exports from the three largest economic regions (Kanto, Chubu, and Kinki) for January 2008 to December Kinki suffer from consecutive negative growth throughout the period after the the collapse of the Lehman Brothers in September Chubu also experiences negative growth right after September 2008, but enjoys recovery periods in next year. Unlikely to two regions, Kanto instead experienced larger growth than pre-crisis period and it took more than a year to fall into negative growth. Preliminary comparison of Figure 1 and Figure 3 does not give straightforward evidence for finance-trade link. Aichi prefecture of Chubu region experiences the worst decline in its exports while the financial condition of Chubu region is similar to that of Kinki region in the three months of the post-crisis. This is a clear-cut example of industry concentration. Aichi prefecture encompasses Toyota City, the home of Toyota Automobiles and other small-medium manufacturers in the auto sector. It is pointed out that automobile sector is one of the most severely hit industry. Alessandria, Kaboski, and Midrigan (2011) point out that inventory adjustment played a crucial role in a sharp drop of automobile imports to the United States during the Great Trade Collapse. By reducing the level of inventory of foreign automobiles, the sales in the US market did not drop proportionately to the sharp decline in imports of foreign automobiles. In order to control for industry concentration bias of finance-trade link, the above argument suggests that regional export need to be compared at the same industry. In the following section, we construct industry exports for each 41 prefecures and see whether difference in regional financial conditions can explain the difference in regional exports. Our null hypothesis is therefore that for a give industry exports decline more in a region which is more severely financially constrained. 4. Empirical examination 4.1 Data Japanese regional export data are constructed from the original export data taken from the Japan Custom, Ministry of Finance, henceforth we call this dataset as JC 5

6 data. The JC data provides monthly export at the level of international ports in Japan. These exports data consisting of value and destination country are available only at the most disaggregated HS 9-digt level. The JC data are aggregated over industries and ports to come up with regional exports at HS 2-digit level. Financial variables are monthly bank loans at prefecture level taken from the Bank of Japan. The bank loan is only proxy for financial constraint in general because it includes loans to non-exporting firms as well as consumers. We acknowledge that our financial constraint variable is too broad in comparison with those used in the finance-trade literature However, trade credit should have been also cut down when bank loan precipitated and our financial constraint variable has an advantage of measuring local financial condition within a country. Japanese regional income variables are taken from the Annual Report on Prefectural Accounts, the Cabinet Office, the Government of Japan. The GDP for importing countries are taken from the World Development Indicators of The World Bank. The original GDP series in U.S. dollars are converted to Japanese yen values by period-average foreign exchange rates (JPY/USD). All series are adjusted and expressed in terms of the Japanese yen. The distances between the Japanese regions and importing countries are measured in two steps. First, the distance to each importing country is measured from each prefecture. When there are more than two local ports in one prefecture, the port with the largest value of trade is chosen for the measurement. The distance between each importing country and each Japanese region is then defined as the shortest distance between the country and the prefectures in the region. 4.2 Bivariate analysis of prefectural export and bank loans Before delving into regression models, simple correlations between prefectural exports and prefectural bank loans are examined. Aggregating all export products, Figure 4 plots growth rate of export versus growth rate of bank loans for each prefecture. At this level of aggregation in Figure 4, we do not find systematic positive correlation between exports and bank loans. For many cases in fact, plots are clustered horizontally, suggesting very low correlation. Export-finance correlations for each prefecture are shown in Table 2. Out of 41 prefectures, the null of no correlation cannot be rejected for 30 prefectures. For the rest of prefectures, correlations are positive (negative) and statistically significant at the one percent level for three (eight) prefectures. In search of possible link between trade and finance, we expect positive correlation between these two variables: Exports decline when credit is tight in financial 6

7 markets or exports expand in the lax financial conditions. Finding only a few regions with positive correlation is consistent with our research motive, but finding negative correlations are surprising and we need to find justifiable explanations for this result. The first possible explanation is that aggregating all export products may bias the underlying link between finance and trade. Export may not be prone to financial stress for some industries as shown in Bricongne et al. (2012) that French exporting firms are affected from credit constraint only in selected industries. We circumvent this aggregation bias by disaggregating exports into 97 HS2-digit industries. The correlations for 3,977 industry-prefecture pairs are shown in Table 3. For 1,307 industry-prefecture pairs, correlations cannot be calculated because of too few or no export data. The data restriction is severer because correlations are calculated with year-on-year growth rate. For the rest of 2,670 pairs, they are not evenly distributed between positive and negative correlations. At the one percent statistical significance level, positive (negative) correlations are for 141 (265) industry-prefecture pairs. Table 4 summarizes by prefectures the correlations at the HS 2-digit level. Surprisingly, for Aichi, Kanagawa, and Hyogo, positive correlations dominates among the statistical significant cases. These three prefectures are among the largest prefectures in terms of prefecture income and the values of international trade. On the other hand, Tokyo experiences negative correlations (with statistical significance at the one percent level) between bank loans and exports for almost one third of all industries. At this preliminary stage, we report that there is heterogeneity in finance-trade linkage among prefectures. The second aspect is the lead and lag relationship between finance and trade. Concurrent correlation may not be appropriate when finance stress only affects the export contracts in a few or several months later. In the literature of the Great Trade Collapse, the presumption of causality direction is one way from financial turmoil to trade collapse. Therefore, the correlation between the current trade and the lagged finance stress is the most important linkage. The cross-correlations for each prefecture are shown in Figure 5. Similarly, the cross-correlations are also examined at the HS 2-digit level; HS95 (toys) in Figure 6, HS40 (glasses and glass ware) in Figure 7, and HS85 (electrical machinery). The third point is that the link between finance and trade only matters for more financial constrained importers. On the other hand, the US and European countries are the ones most severely affected by the US subprime loan crisis and the consequent global financial crisis. The above considerations give fair ground for an argument that finance-trade linkage may be distinct among importing countries. Table 6 provides for 7

8 concurrent correlation between bank loans and export by importing countries. Importing countries are grouped by the income level provided by the World Bank. Figure 8 provides the cross-correlation results for Vietnam, categorized in the 2016 August WDI as in the lower-middle income group. In Table 6-a, correlations between bank loans and exports of prefectures at monthly frequency are calculated for each importing country in high income group. Negative correlations, which is inconsistent with the notion that drying-up credit depresses exports, dominate especially for Asian countries. Out of 41 prefectures, correlations are negative and statistically significant for 12 cases in Korea, 7 in Hong Kong, and 6 in Singapore. For these three countries, the numbers of positive correlations with statistical significance are much smaller. On the other hand, The US, Germany, Hungary, Italy, and Netherlands have negative correlations (with statistical significance) for 5 or 6 prefectures, but the number of positive correlations are only one or two prefectures shy. In table 6-b, correlation for importin countries in upper middle income group are summarized. The outstanding case is 18 negative correlations (with statistical significance) for China. In Table 6-c, correlations for importing countries in lower middle income group are shown. They are less drastic in comparison to two income groups, however, both Pakistan and Indonesia have negative correlations for 5 prefectures and only one or less positive correlations. For importing countries in low income group, most of correlations are not statistically significant. The fourth possible explanation to pervasive negative correlations is that the sample period is too long. This is not to answer why there are many cases with negative correlations, but it raises a possible notion that finance only matters for a shorter period after negative shocks. The sample period covering the normal periods may hide a possible positive correlation between bank loans and exports. We pursue this issue in the following regression model. 4.3 Regression model As widely conducted in empirical international trade literature, the basic regression model for regional export follows a variant of Gravity-model. ln Eijkt = α 1 lnyit + α 2 lny jt + α 3 lntij + ε ijkt (1) All variables are in natural logarithmic forms. Dependent variable ( E ijkt ) is export from prefecture i, to country j, of industry k at time t. Explanatory variables are income of 8

9 exporting prefecture ( Y it ) and importing country ( Y ) and distance between two jt economies ( T ij ). ε ijkt is disturbance term. In order to focus on the effect of financial shock, i.e., large negative shocks, on regional exports, equation (1) is modified in log difference form as in equation (2). In addition, fixed effects for prefecture-importer-industry triplets and (time) are included. Note that distance variable is dropped because of time-invariant nature of the variable. ln Eijkt = λi + η j + µ k + ν t + α1 lnyit + α 2 lny jt + ε ijkt (2) The variable of interest in this study is financial variable ( F it ) for region i at time t. The straightforward way to introduce financial variable in export equation is to simply add it as another explanatory variable as in equation (3). ln E ijkt = λ + η + µ + ν i + α lny 1 j it k + α lny 2 t jt + β ln F it + ε ijkt (3) The effect of financial constraint may interact with other explanatory variables. For example, tighter financial constraint reduces in a region may reduce exports more proportionately if a drop in income of importing country is substantial. Equation (4) includes interaction terms between financial variable and other explanatory variables. ln E ijkt = λ + η + µ + ν i + α j 1 lnyit + α 2 lny jt + β ln Fit (4) + γ ln F lny 1 it k it t + γ ln F lny 2 it jt + ε ijkt 4.4 Estimation results Some caveats need to be noted before estimation results are presented. First, regression analysis is conducted at annual frequency because income variables are only available at lower frequency. Therefore, monthly data of exports and bank loans are also aggregated to annual frequency. Second, the actual number of observations is substantially reduced because there are many zero trades disaggregated at the level of prefecture, industry, and importing country. These zero trades are dropped from the 9

10 sample because of log specifications. The actual number of observations is less than ten percent of possible 10,881,072 observations (41 prefectures, 228 importers, 97 industries, and 12 years). Table 7 presents the estimation results for baseline regression model. Specification (i) corresponding with equation (3) is shown along specification (ii) corresponding with equation (4). The foreign income ( ln(fgdp) ) is positive in both specifications but statistically significant only in (ii). The magnitude of coefficient is much smaller than those usually found for the Gravity-type model regressions, but the log-difference form in our model is different from the standard log form. More troubling results are negative and statistically significant coefficients of prefecture income ( ln(pgdp) ). However, our data includes only one-way trade and the demand related variable is only foreign income. The bank loan ( ln(loan) ) is negative and statistically significant. Even after controlling for income variables, the regression model results reject the null hypothesis that credit crunch reduces exports. The interaction terms between incomes and bank loan are positive and statistically significant. As discussed in subsection 4.2, the contemporaneous specification of bank loan may be problematic. In Table 8, the lagged bank loan is substituted for the contemporaneous variable. The change for income variables are in favorable direction. The estimated coefficients for foreign income becomes much larger whereas the absolute value of prefecture income effect is smaller. However, the effect of bank loan is still negative and statistically significant. We turn to the last point raised in subsection 4.2 that the normal period may obscure the positive correlation between bank loans and exports in financial turmoil. From specification (v) to (xi) in Table 9 provides 5-year window regression results. Foreign income remains positive and statistically significant in most of subsample periods except for first three specifications not covering post-crisis years. Prefecture income become positive and statistically significant for specifications (vii) through (ix). Specifications (viii) and (ix) cover immediately consecutive years (2009 and 2010) after the disrupt of financial crisis marked on September 2008 but exclude less turbulent time in the post-crisis periods. Bank loan remains negative in specifications (v) through (viii), but turns positive and statistically significant in specifications (ix) and (x), respectively, and For specification (xii) covering , it should be noted that the earthquake in the northeastern region on March 2011 and consequent breakdown of domestic production network affected the link between bank loans and Japanese prefecture export. 10

11 Further, we investigate whether bank loan only matters for importing countries in low or lower middle income groups. The shrinkage of domestic bank loan may only affect exports through withholding the letter of credit to importing firms in developing countries and emerging economies. Bank loan variables are interacted with dummy variables for lower middle income group, low income group, and both groups. The estimated results are shown in Table 10. Interestingly, bank loan has offsetting effect (i.e., positive coefficients of interaction terms) on loan-export correlation for developing countries although bank loan coefficients remain negative. 5. Conclusions In this paper we reexamine the link between international trade and financial constraints by implementing provincial Japanese data. Both bank lending and exports data are collected at the province level. We provide that the severity of financial constraint on exporters as well as the degrees of export decline varies widely among provinces of Japan. Empirical evidence shows that these links are found to be very weak at the best and questions the validity of connecting the great trade collapse with financial constrained exporters. Our preliminary empirical evidence questions the validity of notion that financial turmoil in developed countries should diminish their exports. This result hints that export decline is associated with financial shocks in developing countries on the import side. This finding is more consistent with the emphasis on importers financial constraint in Antràs and Foley (2015). 11

12 Appendix: A1. Descriptions of Chapters (Two-digit HS classification codes) 12

13 A2. Classification of Regions (1) Hokkaido (Hokkaido); (2) Tohoku (Aomori, Iwate, Miyagi, Akita, Yamagata, Fukushima); (3) Kanto (Ibaragi, Tochigi, Gunma, Saitama, Chiba, Tokyo, Kanagawa); (4) Chubu (Niigata, Toyama, Ishikawa, Fukui, Yamanashi, Nagano, Gifu, Shizuoka, Aichi); (5) Kinki (Mie, Shiga, Kyoto, Osaka, Hyogo, Nara, Wakayama); (6) Chugoku (Tottori, Shimane, Okayama, Hiroshima, Yamaguchi); (7) Shikoku (Tokushima, Kagawa, Ehime, Kochi); (8) Kyushu (Fukuoka, Saga, Nagasaki, Kumamoto, Oita, Miyazaki, Kagoshima); (9) Okinawa (Okinawa). Note: Region names are in italics and member prefectures are in parentheses. Acknowlegements: Financial support from Grant-in-Aid for Scientific Research (C ), JSPS, is gratefully acknowledged. 13

14 References: Alessandria, G., Kaboski, J.P., and Midrigan, V., US trade and inventory dynamics. American Economic Review, Papers & Proceedings, 101(3), Amiti, M. and Weinstein, D.E., Exports and Financial Shocks. Quarterly Journal of Economics, 126, Antràs, Pol, and C. Fritz Foley Poultry in Motion: A Study of International Trade Finance Practices. Journal of Political Economy 123 (4): Beck, T., Financial development and international trade, Is there a link? Journal of International Economics, 57, Bems, R., Johnson, R.C., and Yi, K.M., Vertical Linkages and the Collapse of Global Trade. American Economic Review: Papers & Proceedings, 101(3), Berman, N. and Héricourt, J., Financial factors and the margins of trade: Evidence from cross-country firm-level data. Journal of Development Economics, 93, Bernanke, B.S., Nonmonetary effects of the financial crisis in the propagation of the Great Depression. American Economic Review, 73(3), Bricogne, J.C., Fontagné, L., Gaulier, G., Taglioni, D., and Vicard, V., Firms and the global crisis: French exports in the turmoil. Journal of International Economics, 87, Buch, C.M. and Goldberg, L.S., International banking and liquidity risk transmission: Lessons from across countries, Federal Reserve Bank of New York Staff Reports, No Bussière, M., Callegari, G., Ghironi, F., Sestieri, G., and Yamano, N., Estimating Trade Elasticities: Demand Composition and the Trade Collapse of American Economic Journal: Macroeconomics, 5(3), Caggese, A. and Cuñat, V., Financing constraints, firm dynamics, export decisions, and aggregate productivity. Review of Economic Dynamics, 16, Chaney, T., 2016, Liquidity Constrained Exporters, Journal of Economic Dynamics and Control, forthcoming Correa, R., Goldberg, L., and Rice, T., Liquidity risk and U.S. bank lending at home and abroad, Federal Reserve Bank of New York Staff Reports, No Ellison, G. and Gleaser, E.L., Geographic concentration in U.S. manufacturing industries: A dartboard approach, Journal of Political Economy, 105(5), Paravisini, D., Local bank financial constraints and firm access to external finance, Journal of Finance, 53(5),

15 Paravisini, D., Rappoport, V., Schnabl, P., and Wolfenzon, D., Dissecting the effect of credit supply on trade: Evidence from matched credit-export data. Review of Economic Studies, 82(1), Schmidt-Eisenlohr, T., Towards a theory of trade finance. Journal of International Economics, 91,

16 Figure 1. Changes in monthly export values, year on year by selected prefecture Jan-08 Feb-08 Mar-08 Apr-08 May-08 Jun-08 Jul-08 Aug-08 Sep-08 Oct-08 Nov-08 Dec Hyogo The collapse of the Lehman Brothers Kanagawa Aichi Note: Export values from each international port are taken from the Japan Custom and are aggregated by the author at the prefecture level. 16

17 Figure 2. Growth in monthly bank loan by region, year on year Apr-99 Aug-99 Dec-99 Apr-00 Aug-00 Dec-00 Apr-01 Aug-01 Dec-01 Apr-02 Aug-02 Dec-02 Apr-03 Aug-03 Dec-03 Apr-04 Aug-04 Dec-04 Apr-05 Aug-05 Dec-05 Apr-06 Aug-06 Dec-06 Apr-07 Aug-07 Dec-07 Apr-08 Aug-08 Dec-08 Apr-09 Aug-09 Dec-09 Apr-10 Apr-11 Aug-11 Dec-11 Apr-12 Aug-12 Aug-10 Dec-10 Dec-12 Apr-13 Aug-13 Note: Banking loan by Japanese regions are taken from the Bank of Japan. 17 Hokkaido Tohoku Hokuriku Kanto Chubu Kinki Chugoku Shikoku Kyushu-Okinawa

18 Figure 3. Growth in monthly bank loan, year on year by selected Japanese regions Jan-08 Feb-08 Mar-08 Apr-08 May-08 Jun-08 Jul-08 Aug-08 Sep-08 Oct-08 Nov-08 Dec-08 Jan-09 Feb-09 Mar-09 Apr-09 May-09 Jun-09 Jul-09 Aug-09 Sep-09 Oct-09 Nov-09 Dec-09 Jan-10 Feb-10 Mar-10 Apr-10 May-10 Jun-10 Jul-10 Aug-10 Sep-10 Oct-10 Nov-10 Dec-10 TheCollpase of the Lehman Brothers Kanto Chubu Kinki Note: Banking loan by Japanese regions are taken from the Bank of Japan. 18

19 Figure 4-a. Export versus bank loan, year on year change from January 2000 to August 2013 Export versus Bank Loan year on year change January August 2013 Aichi Ehime Ibaraki Okayama Okinawa Iwate Miyazaki Miyagi Kyoto Kumamoto yoy_exp Hiroshima Kagawa Kochi Saga Mie Yamagata Yamaguchi Shiga Kagoshima Akita yoy_loan 19

20 Figure 4-b. Export versus bank loan, year on year change from January 2000 to August 2013 Export versus Bank Loan year on year change January August 2013 Niigata Kanagawa Aomori Shizuoka Ishikawa Chiba Osaka Oita Nagasaki Tottori yoy_exp Shimane Tokyo Tokushima Tochigi Toyama Fukui Fukuoka Fukushima Hyogo Hokkaido Wakayama yoy_loan 20

21 Figure 5-a. Cross-correlations between Export and bank loan, year on year change from January 2000 to August 2013 Aichi Ehime Ibaraki Okayama Okinawa Iwate Miyazaki Miyagi Kyoto Kumamoto Hiroshima Kagawa Kochi Saga Mie Yamagata Yamaguchi Shiga Kagoshima Akita 21

22 Figure 5-b. Cross-correlations between Export and bank loan, year on year change from January 2000 to August 2013 Niigata Kanagawa Aomori Shizuoka Ishikawa Chiba Osaka Oita Nagasaki Tottori Shimane Tokyo Tokushima Tochigi Toyama Fukui Fukuoka Fukushima Hyogo Hokkaido Wakayama 22

23 Figure 6. Cross-correlations between Export and bank loan, HS95 (Toys) Aichi Ehime 23 Ibaraki Okayama Okinawa Miyagi Cross-correlations of HS95(Toys) Kyoto Kumamoto Hiroshima Kagawa Saga Yamaguchi Shiga

24 Figure 6. Continued, HS95 (Toys) 24 Niigata Kanagawa Shizuoka Ishikawa Chiba Osaka Tokyo Toyama Fukui Fukuoka Cross-correlations of HS95(Toys) Hyogo Hokkaido

25 Figure 7. Cross-correlations between Export and bank loan, HS40 (Rubber and rubber products) Cross-correlatoins of HS40(Rubbe and rubber productsr) Aichi Ehime Ibaraki Okayama Miyazaki Miyagi Kyoto Kumamoto Hiroshima Kagawa Saga Mie Yamagata Yamaguchi Shiga Kagoshima Akita 25

26 Figure 7. continued, HS40(Rubber and rubber products) Niigata Kanagawa 26 Aomori Shizuoka Ishikawa Chiba Osaka Oita Nagasaki Tottori Tokyo Tokushima Tochigi Toyama Fukui Fukuoka Fukushima Cross-correlations of HS40(Rubber and rubber products) Hyogo Hokkaido

27 Figure 8. Cross-correlations between Export and bank loan, HS85 (Electrical Machinery) Aichi Ehime 27 Ibaraki Okayama Okinawa Iwate Miyazaki Miyagi Kyoto Kumamoto Hiroshima Kagawa Saga Mie Yamagata Yamaguchi Shiga Cross-correlations of HS85(Electrical machinery) Kagoshima Akita

28 Figure 8. continued Niigata Kanagawa 28 Aomori Shizuoka Ishikawa Chiba Osaka Oita Nagasaki Tottori Tokyo Tokushima Tochigi Toyama Fukui Fukuoka Fukushima Cross-correlations of HS85(Electrical machinery) Hyogo Hokkaido Wakayama

29 Figure 9. Cross-correlations between Export and bank loan, Vietnam Aichi Ehime 29 Ibaraki Okayama Okinawa Iwate Miyazaki Miyagi Kyoto Kumamoto Hiroshima Kagawa Kochi Saga Mie Yamagata Yamaguchi Shiga Finance-Trade cross-correlations for exports to Vietnam Kagoshima Akita

30 Figure 9. continued, Vietnam Niigata Kanagawa 30 Aomori Shizuoka Ishikawa Chiba Osaka Oita Nagasaki Tottori Tokyo Tokushima Tochigi Toyama Fukui Fukuoka Fukushima Finance-Trade cross-correlations for exports to Vietnam Hyogo Hokkaido Wakayama

31 Table 1. Changes in monthly export values, year on year by prefecture Prefecture Jan-08 Feb-08 Mar-08 Apr-08 May-08 Jun-08 Jul-08 Aug-08 Sep-08 Oct-08 Nov-08 Dec-08 Aichi 7.2% 6.3% -2.2% 2.3% -4.5% -8.2% 3.8% -13.9% -5.9% -17.6% -40.1% -50.8% Akita -33.9% 4.5% -12.5% 7.8% 2.5% -19.5% 0.0% -49.8% -32.0% -54.9% -66.5% % Aomori -13.1% -6.3% 24.2% -20.8% 2.1% -31.3% 97.4% 39.8% 40.1% 26.5% -16.6% 73.9% Chiba 4.1% 3.0% 0.4% -3.6% -6.5% -5.9% -0.2% -4.2% -9.0% -11.8% -39.4% -58.8% Ehime 15.7% 17.5% -4.8% 18.5% -3.2% 10.1% -32.1% -6.9% -6.8% -7.8% 18.4% -24.0% Fukui 39.6% 44.9% 34.8% 32.1% 20.1% 10.6% 29.2% 15.7% 20.5% 6.2% -30.8% -21.6% Fukuoka 13.0% 14.5% 2.5% 16.6% 21.4% 11.8% 26.7% 25.3% 18.6% 6.5% -21.1% -48.5% Fukushima -18.7% 24.3% 5.3% -18.7% 25.5% -8.3% 26.3% 14.5% -76.1% -10.5% -33.9% -51.9% Hiroshima 10.7% 17.2% 23.4% 7.0% 17.3% 6.7% 12.4% 15.2% 11.2% 12.6% -35.4% -17.2% Hokkaido 12.0% 26.6% 19.9% 8.1% 16.2% 16.1% 36.4% 35.6% 18.6% 16.3% -34.8% -41.4% Hyogo 6.5% 8.6% -0.6% 4.8% 3.3% -1.0% 5.6% -0.6% 7.1% -1.4% -21.9% -23.6% Ibaragi 31.9% 3.4% 15.0% 1.5% 13.3% 22.9% 21.9% 24.7% 55.2% 18.1% -12.1% -35.5% Ishikawa 29.8% 48.6% 2.3% 61.9% 51.0% -14.2% 23.2% -6.0% -5.0% -27.3% 7.8% -41.0% Iwate -18.9% 16.4% 10.3% -37.9% -8.3% -2.7% 2.3% -54.6% 7.0% -52.5% -36.7% -51.2% Kagawa 12.4% 140.0% 18.2% 89.7% 2.6% -13.1% 22.9% 6.9% 119.0% -30.7% -19.0% 18.1% Kagoshima 6.4% -55.2% -39.9% -23.6% -25.2% 6.2% 11.0% 24.5% -13.4% 4.9% -52.1% -42.1% Kanagawa 11.3% 10.4% 8.8% 9.8% 12.8% 0.5% 14.0% 0.7% 8.2% -4.5% -29.5% -37.7% Kochi 21.7% 18.1% 59.0% 31.7% -20.3% 64.1% -22.2% 17.4% 15.8% 72.4% -6.6% -47.4% Kumamoto 15.2% 26.1% 17.4% -14.6% 23.9% -7.0% 0.9% 4.0% 70.5% -52.6% -37.8% -76.1% Kyoto -64.6% -26.5% -52.7% -6.5% -5.4% -29.3% -13.4% -2.3% -41.5% -17.0% -85.4% -79.0% Mie -2.6% 8.8% -14.5% 5.4% -0.2% -1.5% 18.6% 1.5% 0.5% 7.2% -18.9% -27.0% Miyagi 1.5% 5.8% 10.0% -16.7% -0.2% 0.2% -6.3% 4.9% -15.3% -36.5% -36.3% -59.4% Miyazaki 35.0% -9.1% 37.6% 13.0% 2.3% -12.3% -0.1% -6.1% -3.5% -11.8% -30.9% -64.3% Nagasaki 17.5% 25.2% 94.4% 32.0% 4.6% -49.2% -12.8% 51.4% -67.3% -10.3% -11.4% 24.8% Niigata 17.9% 19.2% 25.1% 13.3% 7.1% -3.5% 27.9% 12.9% 41.4% -9.0% -28.3% -49.3% Ohita -5.7% 6.8% -0.3% -19.8% -10.3% -24.3% -12.5% -35.5% -39.1% -64.4% % -71.5% Okayama 21.9% 21.2% 16.3% 13.3% 8.7% 17.7% 19.8% 29.9% 38.2% 8.8% -14.5% -48.8% Okinawa 40.6% 102.5% 175.0% 116.7% 78.9% -52.1% -75.6% 81.2% 15.6% 67.4% -74.3% -86.4% Osaka 10.9% 8.7% 6.4% 5.4% 10.1% 3.2% 9.5% 5.7% 4.3% -4.8% -24.1% -43.7% Saga 32.5% 4.5% -52.5% 9.3% % 220.3% % -24.1% 115.7% 28.0% -31.4% 53.9% Shiga 1.6% -1.5% -5.4% -10.1% -9.3% -5.7% 19.0% -37.4% -26.7% -19.4% -30.2% -61.4% Shimane 51.1% 20.6% 1.6% 50.5% 4.5% 7.1% 7.1% -6.0% -8.0% 12.2% -42.6% 44.2% Shizuoka 5.2% -2.0% -6.7% -11.4% -2.6% -9.4% -3.2% -6.1% -13.9% -26.2% -26.6% -54.7% Tochigi -3.4% 2.4% -9.6% -19.8% 0.6% -12.7% -14.3% -14.4% -6.9% -8.3% -38.6% -71.8% Tokushima 25.5% -1.5% -32.2% 79.1% -3.9% -83.5% 68.3% 36.1% -50.2% -22.6% % -72.8% Tokyo 3.3% 9.0% -5.7% 2.4% 6.1% -2.3% 7.5% 2.6% -1.4% -5.0% -26.4% -36.8% Tottori 13.9% 24.0% -0.3% 10.7% 15.4% 9.2% 13.0% 10.8% -5.3% 7.1% -38.4% -75.6% Toyama -13.8% -12.6% 0.6% 7.1% 6.1% -1.6% -2.9% -25.9% -9.3% -25.4% -56.2% -35.9% Wakayama 13.0% 16.0% -7.1% 8.9% 10.7% 21.2% 28.8% 22.8% 26.7% -1.9% 18.9% -22.7% Yamagata -3.0% -2.9% -3.6% -4.1% 13.1% -24.6% -0.6% -15.9% -17.9% -10.6% -58.0% % Yamaguchi -1.5% 14.2% -3.2% 8.1% 12.1% 5.4% 20.2% 12.4% 4.1% -14.3% -30.6% -52.4% Simple Ave. 8.5% 14.7% 8.6% 11.2% 3.9% 0.3% 3.7% 4.6% 4.5% -6.5% -34.9% -41.1% Note: Export values from each international port are taken from the Japan Custom and are aggregated by the author at the prefecture level. 31

32 Table 2. Correlation between exports and bank loans at the prefecture level Prefecture Correlation Prefecture Correlation Aichi Niigata Ehime * Kanagawa Ibaraki Aomori Okayama Shizuoka * Okinawa Ishikawa Iwate * Chiba Miyazaki * Osaka Miyagi 0.14 Oita Kyoto * Nagasaki Kumamoto Tottori Hiroshima Shimane Kagawa Tokyo * Kochi Tokushima Saga Tochigi Mie * Toyama Yamagata * Fukui Yamaguchi * Fukuoka * Shiga * Fukushima Kagoshima Hyogo Akita Hokkaido Wakayama Note: Exports are monthly total export values. Both exports and bank loans are year on year growth rate. * indicates statistical significance at the one percent level. 32

33 Table 3. Finance-trade correlations for industry-prefecture pairs number of pairs statistically significant positive correlation 1, negative correlation 1, no data 1,307 Note: The total number of industry-prefecture pairs is 3,977 (41 prefectures times 97 industries). For 1,307 industry-prefecture pairs, correlations cannot be calculated because of too few or no export data. The data restriction is severer because correlations are calculated with year-on-year growth rate. The statistical significance level is one percent. 33

34 Table 4. finance-trade correlations at industry-prefecture pairs by prefecture Prefecture sig+ sig- share Prefecture sig+ sig- share Aichi % Niigata % Ehime % Kanagawa % Ibaraki % Aomori % Okayama % Shizuoka % Okinawa 2 3 7% Ishikawa 1 1 4% Iwate % Chiba % Miyazaki % Osaka % Miyagi % Oita % Kyoto % Nagasaki 1 0 2% Kumamoto % Tottori 2 2 8% Hiroshima % Shimane 3 0 9% Kagawa 3 3 9% Tokyo % Kochi % Tokushima % Saga % Tochigi % Mie % Toyama % Yamagata % Fukui % Yamaguchi % Fukuoka % Shiga % Fukushima % Kagoshima 1 1 4% Hyogo % Akita 2 2 8% Hokkaido % Wakayama 1 3 8% Note: The numbers of correlations with statistical significance at the one percent level are shown. Sig+ (sig-) denotes for statistical significant and positive (negative). The total number of correlations differs among prefectures because exports are too infrequent for some industry-prefecture pairs to calculate the year-on-year growth rate. The share is the number of significant correlations divided by the number of available correlations. 34

35 Table 5. finance-trade correlations at industry-prefecture pairs by HS2-digit industry HS2 sig+ sig- share HS2 sig+ sig- share HS2 sig+ sig- share % % % % % % % % % % % % % % % % % % % % % % % % % % % % % % % % % % % % % % % % % % % % % % % % % % % % % % % % % % % % % % % % % % % % % % % % % % % % % % % % % % % % % % % % % % % % % % % % % Note: see the notes for Table 5. 35

36 Table 6-a. finance-trade correlations by importing country (high income group) country sig+ sig- available share country sig+ sig- available share Korea % Greece % Hong Kong SAR, China % Cyprus % Singapore % Estonia % Brunei % Latvia % Macao SAR, China % Lithuania % Bahrain % Croatia % Saudi Arabia % Slovenia % Kuwait % Czech Republic % Qatar % Slovak Republic % Oman % Greenland % Israel % Canada % United Arab Emirates % United States % Iceland % Bermuda % Norway % The Bahamas % Sweden % Turks and Caicos Islands % Denmark % Barbados % United Kingdom % Trinidad and Tobago % Ireland % Puerto Rico % Netherlands % Virgin Islands % Belgium % Cayman Islands % Luxembourg % Antigua and Barbuda % France % British Virgin Islands % Monaco % St. Kitts and Nevis % Andorra % Chile % Germany % Uruguay % Switzerland % Channel Islands % Portugal % Seychelles % Spain % Australia % Gibraltar % New Zealand % Italy % Nauru % Malta % New Caledonia % Finland % French Polynesia % Poland % Guam % Austria % Northern Mariana Islands % Hungary % Note: The numbers of correlations with statistical significance at the one percent level are shown. Sig+ (sig-) denotes for statistical significant and positive (negative). The total number of correlations denoted as (available) differs among importing counties because exports are too infrequent for some country-prefecture pairs to calculate the year-on-year growth rate. The share is the number of significant correlations divided by the number of available correlations. 36

37 Table 6-b. finance-trade correlations by importing country (upper middle income group) country sig+ sig- available share country sig+ sig- available share China % Cuba % Thailand % Dominican Republic % Malaysia % Grenada % Maldives % St. Lucia % Iran % Dominica % Iraq % St. Vincent and the Grenadines % Jordan % Colombia % Lebanon % Venezuela % Azerbaijan % Guyana % Kazakhstan % Suriname % Turkmenistan % Ecuador % Georgia % Peru % Russia % Brazil % Serbia % Paraguay % Albania % Algeria % Romania % Libya % Bulgaria % Equatorial Guinea % Turkey % Gabon % Belarus % Angola % Bosnia and Herzegovina % Mauritius % Macedonia % Namibia % Montenegro % South Africa % Mexico % Botswana % Belize % Fiji % Costa Rica % American Samoa % Panama % Tuvalu % Jamaica % Marshall Islands % Palau % Note: The numbers of correlations with statistical significance at the one percent level are shown. Sig+ (sig-) denotes for statistical significant and positive (negative). The total number of correlations denoted as (available) differs among importing counties because exports are too infrequent for some country-prefecture pairs to calculate the year-on-year growth rate. The share is the number of significant correlations divided by the number of available correlations. 37

38 Table 6-c. finance-trade correlations by importing country (lower middle income group) country sig+ sig- available share country sig+ sig- available share Mongolia % El Salvador % Vietnam % Nicaragua % Philippines % Bolivia % Indonesia % Morocco % Cambodia % Tunisia % Lao PDR % Sudan % Myanmar % Mauritania % India % Côte d'ivoire % Pakistan % Ghana % Sri Lanka % Cabo Verde % Bangladesh % Nigeria % Timor-Leste % Cameroon % Bhutan % Congo % Syrian Arab Republic % São Tomé and Principe % Yemen % Djibouti % Armenia % Kenya % Uzbekistan % Lesotho % Kyrgyz Republic % Zambia % Tajikistan % Swaziland % West Bank and Gaza % Papua New Guinea % Ukraine % Samoa % Moldova % Vanuatu % Guatemala % Solomon Islands % Honduras % Tonga % Kiribati % Micronesia % Note: The numbers of correlations with statistical significance at the one percent level are shown. Sig+ (sig-) denotes for statistical significant and positive (negative). The total number of correlations denoted as (available) differs among importing counties because exports are too infrequent for some country-prefecture pairs to calculate the year-on-year growth rate. The share is the number of significant correlations divided by the number of available correlations. 38

39 Table 6-d. finance-trade correlations by importing country (low income group) country sig+ sig- available share Dem. People's Rep. Korea % Afghanistan % Nepal % Haiti % Senegal % The Gambia % Guinea-Bissau % Guinea % Sierra Leone % Liberia % Togo % Benin % Mali % Burkina Faso % Niger % Rwanda % Chad % Central African Republic % Dem. Rep. Congo % Burundi % Ethiopia % Somalia % Uganda % Tanzania % Mozambique % Madagascar % Zimbabwe % Malawi % Comoros % Eritrea % Note: The numbers of correlations with statistical significance at the one percent level are shown. Sig+ (sig-) denotes for statistical significant and positive (negative). The total number of correlations denoted as (available) differs among importing counties because exports are too infrequent for some country-prefecture pairs to calculate the year-on-year growth rate. The share is the number of significant correlations divided by the number of available correlations. 39

40 Table 7. Panel estimation of prefecture exports Note: The top figure is the estimated coefficient and the bottom figure is the standard deviation. (i) (ii) Δln(fgdp) Δln(pgdp) Δln(loan) Δln(loan)Δln(fgdp) Δln(loan)Δln(pgdp) beginning year ending year number of observations 806, ,436 adjusted R

41 Table 8. ged financial variable Note: The top figure is the estimated coefficient and the bottom figure is the standard deviation. (iii) (iv) Δln(fgdp) Δln(pgdp) laggedδln(loan) laggedδln(loan)δln(fgdp) laggedδln(loan)δln(pgdp) beginning year ending year number of observations 764, ,802 adjusted R