77621 Liquidity Constraints and Firms’ Linkages with Multinationals Beata S. Javorcik and Mariana Spatareanu Using a unique data set on the Czech Republic for 1994–2003, this article examines the relationship between a �rm’s liquidity constraints and its supply linkages with multinational corporations (MNCs). The empirical analysis indicates that Czech �rms supplying multinationals are less credit constrained than are nonsuppliers. Closer inspection of the timing of the effect, however, suggests that the result is due to self- selection of less constrained �rms into supplying multinationals rather than to the bene�ts derived from the supplying relationship. As the recent literature �nds that pro- ductivity spillovers from foreign direct investment (FDI) are most likely to take place through contacts between MNCs and their local suppliers, this �nding suggests that well-developed �nancial markets may be needed to take full advantage of the bene�ts associated with FDI inflows. JEL codes: F21, F23, F36 The role of �nancial sector development in fostering economic growth has received considerable attention in recent years. In an influential paper, Rajan and Zingales (1998) demonstrate that industrial sectors that are relatively more in need of external �nance grow disproportionately faster in countries with more-developed �nancial markets, suggesting that �nancial sector development reduces the costs of external �nance to �rms. More recent research has argued that access to �nancing may promote economic growth by allowing �rms to tap into new sources of knowledge from selling in foreign markets or becoming suppliers to multinational corpor- ations (MNCs). In a theoretical contribution, Chaney (2005) shows that if �rms must pay entry costs to sell in a foreign market and if they face liquidity constraints in �nancing these costs, only �rms with suf�cient liquidity will be able to export. While some other �rms could pro�tably export, they are pre- vented from doing so by their lack of suf�cient liquidity. Manova (2006) Beata S. Javorcik (corresponding author) is a reader in economics, University of Oxford, and a research af�liate at the Centre for Economic Policy Research; her email address is beata.javorcik@economics.ox.ac.uk. Mariana Spatareanu is an assistant professor of economics at Rutgers University; her email address is marianas@andromeda.rutgers.edu. THE WORLD BANK ECONOMIC REVIEW, VOL. 23, NO. 2, pp. 323 –346 doi:10.1093/wber/lhp002 Advance Access Publication June 16, 2009 # The Author 2009. Published by Oxford University Press on behalf of the International Bank for Reconstruction and Development / THE WORLD BANK. All rights reserved. For permissions, please e-mail: journals.permissions@oxfordjournals.org 323 324 THE WORLD BANK ECONOMIC REVIEW provides empirical support for this view by showing that countries with better developed �nancial systems tend to export relatively more in industries highly dependent on external capital and in sectors with fewer collateralizable assets. A theoretical model and a calibration exercise by Alfaro and others (2006) suggest that well-developed local �nancial markets are needed in order for host countries to bene�t from spillovers from foreign direct investment (FDI). The reason is that access to �nancing allows local entrepreneurs to start sup- plying multinationals and thus to bene�t from knowledge spillovers from FDI. In a cross-country growth regression, Alfaro and others (2004) �nd that FDI inflows contribute to faster economic growth only where �nancial markets are well developed. The relationship between facing �nancing constraints and supplying MNCs could go either way. If a �rm needs some investment in order to supply multi- nationals (say, to upgrade product quality or increase the scale of production), then the causality goes from the absence of liquidity constraints to becoming a supplier of multinationals. However, it is also possible that receiving a contract from a multinational increases a supplier’s creditworthiness and thus makes it easier to obtain a loan or other outside �nancing. This article uses the approach pioneered by Fazzari, Hubbard, and Petersen (1988) to examine the relationship between facing liquidity constraints and being a supplier to an MNC. The analysis is possible thanks to a unique data set col- lected by the World Bank through two surveys of domestic and foreign compa- nies in the Czech Republic in 2003 and 2004. The surveys make it possible to identify companies selling to multinationals operating in the country and provide detailed information about the duration and characteristics of these relationships. The survey responses are supplemented with panel data on �rms’ balance sheets and pro�t and loss statements from the commercial database AMADEUS (Bureau van Dijk Electronic Publishing 2005). The data set spans 1994–2003 and includes 319 Czech �rms, 88 of which are suppliers of MNCs and are observed both before and after starting the relationship with multinationals. The Czech Republic is a suitable place to study this question for several reasons. Since starting its transition from central planning to a free market economy, it has received large inflows of FDI. At the end of 2003 (the last year of the sample), its stock of FDI reached $45.3 billion or $4,439 per capita. Survey evidence suggests that multinationals are actively engaged in local sour- cing in the Czech Republic, purchasing about half their intermediate inputs (in value terms) from Czech suppliers. The virtual absence of FDI before the beginning of transition also means that supply relationships between multina- tionals and Czech �rms are of a relatively new vintage. Finally, as in all tran- sition economies, many local �rms tend to be liquidity constrained (Konings, Rizov, and Vandenbussche 2003). Survey evidence suggests that before signing a purchase order, multinationals often explicitly require their future Czech suppliers to make some improve- ments or investments. This was the case for more than a quarter of suppliers Javorcik and Spatareanu 325 surveyed in 2004.1 The prospect of a contract from a multinational also induced Czech suppliers to undertake improvements on their own. Some 36 percent of suppliers reported making improvements with the explicit purpose of �nding a multinational customer.2 Also striking, 17 percent of Czech com- panies surveyed reported getting a quality certi�cation (such as ISO 9000) in order to become suppliers to multinationals. These �rms constituted 40 percent of all companies reporting having such a certi�cation. In sum, complying with the expectations or requirements imposed by multinationals may be more dif�- cult or even impossible for potential suppliers without access to credit. And indeed credit constraints faced by Czech companies were mentioned by multi- nationals as one of the top factors preventing them from sourcing more inputs locally (Javorcik and Spatareanu 2005). At the same time, contracts from multinationals (or the prospect of such a contract) may have eased credit constraints for potential or actual suppliers. Almost a quarter (31 of 137) of multinationals surveyed in 2003 reported pro- viding their suppliers with advance payments and �nancing. Similarly, a quarter of suppliers reported that being a supplier of multinationals helped them obtain a bank loan. The results of the empirical analysis, presented in this article, indicate that Czech �rms supplying multinationals tend to be less liquidity constrained than other �rms. However, an examination of the timing suggests that the result is due to the self-selection of less liquidity constrained �rms into supplying relationships rather than to suppliers bene�ting from the links with multina- tional customers. The data suggest that suppliers of multinationals are less liquidity constrained before starting their relationship with a multinational and continue to be less liquidity constrained for the duration of the relationship. This �nding is not driven by multinationals extending credit to their future suppliers, as the results are robust to excluding from the sample suppliers that had received this type of assistance from their multinational customers. The results also hold after excluding from the sample �rms reporting that a supply relationship with a multinational had helped them obtain �nancing from a Czech or a foreign bank. Further, a higher liquidity ratio is found to be a robust predictor of supplying status. Finally, the results from models instru- menting for supplying status show that �rms doing business with multina- tionals do not differ from other �rms in liquidity constraints. Understanding how �rms become suppliers to multinationals has important policy implications in the context of recent empirical �ndings that linkages between multinationals and their local suppliers are the key channel through 1. The most frequent requirements were improvements to the quality assurance process, acquisition of a costly quality certi�cation, improvements to the timeliness of deliveries, use of a new technology, and purchase of new equipment. 2. These improvements included investing in new machinery and equipment, improving product quality, increasing staff training, raising production volume, reducing the share of defective units produced, and reorganizing manufacturing lines. 326 THE WORLD BANK ECONOMIC REVIEW which indigenous �rms bene�t from inflows of FDI (Moran 2001; Javorcik 2004; Moran, Graham, and Blomstro ¨ m 2005; Blalock and Gertler 2008; Javorcik and Spatareanu 2008). The �ndings of this study suggest that in the absence of well-functioning credit markets, local �rms may �nd it dif�cult to start business relationships with multinationals and thus may not be able to reap the bene�ts of productivity spillovers that such relationships can bring. Caution is warranted, however, when interpreting these �ndings. While the results suggest that well-functioning credit markets are important in facilitating business relationships between local �rms and multinationals, they do not suggest that a well-developed �nancial market is a suf�cient condition for such relationships. Other factors, such as a certain level of sophistication of the local manufacturing sector, may be needed for these relationships to materialize. This study is structured as follows. Section I presents the data and the summary statistics. Section II reviews the related literature. Section III discusses the estimation strategy and the results. Section IV presents some policy implications. I . D ATA AND S U M M A RY S TAT I ST I C S Examining the questions asked by this study poses big data challenges. Information on the type of customers supplied by �rms (and hence their sup- plying status for multinationals) is typically not collected by statistical agencies, tax authorities, or commercial databases. While time-varying information on relationships with multinational customers can be obtained through �rm-level surveys, such surveys cannot be used to collect long spans of historical data on �rm balance sheets and pro�t and loss statements. Thus to conduct this study, enterprise survey information was combined with historical �rm-level data from a commercial database, creating a unique data set that can be used to examine the relationship between �nancial constraints and supplying multinationals. The enterprise surveys were conducted in the Czech Republic in 2003 and 2004 by a professional polling company in face-to-face interviews with senior managers at respondents’ workplaces. All respondents were guaranteed full anonymity. The data were collected for 857 Czech �rms and 256 foreign- owned �rms operating in the country. The �rst survey focused on manufactur- ing �rms, and the second covered both manufacturing and services industries. About one-�fth of respondents were located in the capital city of Prague, while the rest were distributed across all regions of the country. As the primary inter- est of this article is the linkages between local �rms and multinationals, the analysis relies on the data for Czech �rms only. The survey data identify �rms that supply multinationals operating in the Czech Republic and contain information on the duration of these relationships and other company characteristics. The 2003 survey asked respondents to Javorcik and Spatareanu 327 indicate the year in which they became suppliers to multinationals. The 2004 survey distinguished between the date of signing the contract and the date of making the �rst delivery. The date of signing the contract is used in this study as the date of becoming a supplier to multinationals. Of 857 Czech �rms in the sample, 390 are suppliers to multinationals (331 suppliers operate in the manu- facturing sector, while 59 are services �rms).3 The analysis focuses on the man- ufacturing sector because new investment in physical assets is more likely to be important for manufacturing �rms wanting to become suppliers to multina- tionals than for services companies. The surveys are supplemented with �nancial information on interviewed �rms, taken from the commercial database AMADEUS, compiled by Bureau van Dijk Electronic Publishing (2005). The additional �nancial information, including �gures on sales, tangible �xed assets, depreciation, and pro�t (loss), is available for approximately two-thirds of surveyed �rms. This rich database contains detailed �rm-level information for 1994–2003. Deleting incomplete or inconsistent data and extreme outliers4 leaves 2,136 �rm-year observations on 386 Czech manufacturing �rms, 155 of them were suppliers to multina- tionals. Because of concerns about the self-selection of �rms into supplying relationships, suppliers that cannot be observed before they start their relation- ship with multinationals are excluded, leaving 1,735 �rm-year observations on 319 Czech �rms, 88 of them supplying to multinationals. All suppliers to mul- tinationals are observed both before and after starting their relationship. Suppliers to multinationals are distributed across many industries, including food products and beverages, machinery and equipment, fabricated metal pro- ducts, and rubber and plastic products. Summary statistics on suppliers and nonsuppliers show that suppliers are some- what larger in terms of employment, tend to invest more (relative to their capital stock), and have a higher debt to capital stock ratio and a higher liquidity ratio (table 1). They also tend to be older and have higher labor productivity. They are more likely to export, have an ISO certi�cation, and employ managers speaking a foreign language. However, they tend to experience slower sales growth. The Czech Republic is an ideal setting for this analysis for three reasons. First, as mentioned, the country has received large inflows of FDI. In the early years of transition, food, beverage, and tobacco sectors and some other consumer goods industries received large FDI inflows as many multinationals entered the country hoping to secure a �rst mover advantage in a newly open market. Because of the Czech Republic’s central location, reputation for high- quality engineers, and the fast progress of reforms, the country also attracted many multinationals wishing to establish export platforms supplying the 3. The high percentage of suppliers to multinationals in the data set reflects deliberate oversampling, which was done through a phone prescreening of potential survey respondents. 4. Negative values of tangible �xed assets, sales, and depreciation were dropped, as were the 1 percent tails of the following variables: sales growth, tangible �xed assets growth, and cash flow deflated by tangible �xed assets. 328 THE WORLD BANK ECONOMIC REVIEW T A B L E 1 . Summary Statistics on Czech Firms Supplying and Not Supplying Multinationals Variable Number of observations Mean Standard deviation Supplying �rms I/K (investment to capital stock ratio) 405 0.192 0.392 DSales 405 0.077 0.339 CF/K (cash flow to capital stock ratio) 405 0.281 0.537 Number of employees 405 339 550 Debt/K (debt to capital stock ratio) 405 0.123 0.166 Liquidity ratio 261 0.190 0.237 ln(Gross pro�t) 247 5.328 1.678 ln(Age) 394 1.939 0.702 ln(Value added per worker) 289 3.219 3.066 ln(Total factor productivity) 273 1.240 0.438 ln(Total factor productivity Olley – Pakes) 273 1.378 0.258 Exporter 405 0.874 0.332 State-owned enterprise 277 0.047 0.212 Manager’s foreign language 88 0.773 0.421 Manager’s foreign experience 88 0.227 0.421 ISO certi�cate 88 0.739 0.442 Nonsupplying �rms I/K (investment to capital stock ratio) 1,330 0.158 0.413 DSales 1,330 0.082 0.386 CF/K (cash flow to capital stock ratio) 1,330 0.257 0.573 Number of employees 1,328 314 508 Debt/K (debt to capital stock ratio) 1,330 0.115 0.163 Liquidity ratio 628 0.144 0.239 ln(Gross pro�t) 597 4.922 1.354 ln(Age) 1,314 1.749 0.729 ln(Value added per worker) 1,070 2.969 3.282 ln(Total factor productivity) 873 1.231 0.404 ln(Total factor productivity Olley – Pakes) 873 1.352 0.254 Exporter 1,330 0.72 0.449 State-owned enterprise 1,082 0.059 0.236 Manager’s foreign language 231 0.714 0.453 Manager’s foreign experience 231 0.234 0.424 ISO certi�cate 231 0.602 0.491 Multinationals in the same sector 107 0.199 0.227 Potential multinational customers 107 0.026 0.023 Source: Authors’ analysis based on data from two World Bank surveys of Czech �rms in 2003 and 2004 and Bureau van Dijk Electronic Publishing (2005); see text for details. neighboring European Union. By the end of the sample period (2003), 21 percent of manufacturing FDI stock was in the automotive industry; 14 percent in petroleum, chemical, rubber, and plastic products; and 12 percent in other nonmetallic products. The opening of services industries to FDI stimulated massive inflows into �nancial intermediation, real estate, and wholesale and retail trade. From the mid-1990s on FDI flows into services exceeded those into manufacturing. Javorcik and Spatareanu 329 Second, MNCs operating in the Czech Republic appear to be relying heavily on Czech suppliers. Most (90 percent) multinationals interviewed in the 2003 survey reported purchasing inputs from at least one Czech company.5 The median MNC had a sourcing relationship with 10 Czech suppliers, while a multinational in the top quartile had such a relationship with at least 30. Asked about the share of inputs purchased from each type of supplier (in value terms), multinationals indicated sourcing on average 48.3 percent of inputs from Czech enterprises, 33.3 percent from �rms in the European Union/Eastern Europe, and 12.6 percent from multinationals in the Czech Republic. The share of inputs sourced from other regions appeared to be negligible. Fifty-�ve of 114 MNCs that answered this question reported buying at least half of their inputs from Czech suppliers. More than one-tenth of respondents acquired all of their intermediates from Czech enterprises. Around 40 percent of multina- tionals expected to purchase more inputs from Czech suppliers in the future.6 Third, while the Czech Republic possessed reasonably developed �nancial markets during the period under study, their sophistication and the level of competition (at least in the �rst half of the sample period) were still below those in industrialized countries. For instance, during the �rst year covered by the sample (1994), the ratio of bank deposits to GDP, a common measure of �nancial intermediation, was 0.58 in the Czech Republic, much higher than the average of 0.36 for upper middle-income countries that same year but lower than the average of 0.67 for high-income economies (Beck, Demirgurc-Kunt, and Levine 1999). The Czech private bond market was much less well developed. The ratio of private bond market capitalization to GDP (the ratio of total outstanding domestic debt securities issued by private dom- estic entities to GDP) was 0.02, compared with 0.07 in upper middle-income countries and 0.34 in high-income countries. In the ratio of bank overhead costs to total assets, the Czech Republic ranked with high-income countries (0.03) and appeared much more ef�cient than an average upper middle-income economy (0.05). The banking sector in the Czech Republic appeared to be highly concentrated, however, with the ratio of the three largest banks’ assets to total banking sector assets at 0.78, compared with 0.64 in high-income economies and 0.67 in upper middle-income economies. During the period under study, the Czech banking sector experienced signi�cant restructuring, privatization, and entry of foreign investors. 5. The question speci�cally asked respondents not to include suppliers of services, such as catering and cleaning. 6. These �gures are similar to those collected in other surveys. For instance, the Opinion Window survey commissioned by CzechInvest in 2002 found that multinationals in the Czech Republic sourced on average 32.2 percent of their inputs locally in 2000 and 34.7 percent in 2001. This share was expected to increase to 35.8 percent in 2002. Similarly, CzechInvest reported that 57 percent of multinationals indicated an ability to increase local content (CzechInvest 2002). 330 THE WORLD BANK ECONOMIC REVIEW II. THE ROLE OF CA S H FLOW Ever since the influential paper by Fazzari, Hubbard, and Petersen (1988), numerous studies have examined the effects of liquidity constraints on invest- ment. These studies challenged the neoclassical theory of investment that posits that the decision to invest is driven solely by relative prices and that a �rm’s �nancial structure is irrelevant since external funds provide a perfect substitute for internal capital. Or, as Modigliani and Miller (1958) put it, with perfect capital markets, a �rm’s investment decision is independent of its �nancial con- dition. The alternative research agenda, proposed by Fazzari, Hubbard, and Petersen (1988), was based on the burgeoning literature on information asym- metries: in an environment with information asymmetries, external funds may be more costly and thus provide an imperfect substitute for internal capital. The difference arises to compensate lenders for the adverse selection and moral hazard problems associated with borrowers. If this is the case, investment should respond positively to increases in internal funds available for investment. The primary way of testing this hypothesis is to estimate the investment equation including a measure of the expected pro�tability of the �rm along with a measure of its net worth. Researchers have concluded that to the extent that the measure of net worth (usually cash flow) predicts investment behavior, �nancing constraints exist. The link between investment and cash flow is a subject of ongoing debate. One thread of the literature—starting with Fazzari, Hubbard, and Petersen (1988) and followed by Hoshi, Kashyap, and Scharfstein (1991), Lizal and Svejnar (2002), and others—argues that investment sensitivities to cash flow can be interpreted as evidence of �nancial constraints. However, Kaplan and Zingales (1997, 2000) question that approach and provide evidence that because of nonmonotonicities investment sensitivity to cash flow is not a measure of liquidity constraints. Fazzari, Hubbard, and Petersen (2000) chal- lenge that conclusion and derive the conditions under which the relationship between investment and cash flow is monotonic. They argue that if the a priori classi�cation of �rms is based on criteria that result in large differences in the marginal cost of external funds across groups, constrained �rms with a large cost of external �nancing will have larger investment sensitivity to cash flow than the relatively unconstrained �rms with very small cost of external funds. Although the debate is still unresolved, this study follows the Fazzari, Hubbard, and Petersen (1988) argument.7 7. This article is also related to the literature on the relationship between country-level FDI inflows and �rm-level �nancing constraints. In a cross-country study, Harrison, Love, and McMillan (2004) show that FDI inflows are associated with a reduction in �nancing constraints. In contrast, in a �rm-level analysis of Cote d’Ivoire, Harrison and McMillan (2003) �nd that borrowing by foreign �rms exacerbates the credit constraints of domestic �rms. This article can be viewed as an examination of one of the many channels through which FDI inflows can affect �nancing constraints of domestic �rms in host countries. Javorcik and Spatareanu 331 II I. E M P I R I CA L A N A LY S I S The empirical strategy for this study is to estimate the traditional accelerator speci�cation (see also Gelos and Werner 2002; Konings, Rizov, and Vandenbussche 2003). The growth rate of sales is the accelerator variable, which is expected to be a reasonable proxy for short-term changes in expected pro�tability. Cash flow is included to capture liquidity constraints, and an interaction of cash flow with a multinational supplier dummy variable is included to examine whether multinational suppliers are subject to liquidity constraints that are different from those of other �rms. The baseline speci�ca- tion is as follows: Iit =KitÀ1 ¼ a0 þ a1 DSit =SitÀ1 þ a2 CFit =KitÀ1 þ a3 CFit =KitÀ1 à Supplierit þ a4 Supplierit þ a5 CFit =KitÀ1 à lnðVA=LÞ þ a6 lnðVA=LÞ ð1Þ þ a7 CFit =KitÀ1 à Exporterit þ a8 Exporterit þ a9 CFit =KitÀ1 à SOEit þ a10 SOEit þ a11 lnðSizeit Þ þ a12 lnðAgeit Þ þ a13 Debt=KitÀ1 þ ai þ at þ 1it where Iit is gross investment by �rm i at time t and is de�ned as a change in tangible �xed assets plus depreciation, Kit is real capital stock and is proxied by deflated tangible �xed assets, Sit is real sales, and CFit is the real cash flow as reported in the AMADEUS database, which de�nes it as the sum of pro�t (or loss) after taxation, extraordinary pro�t (or loss), and depreciation (Bureau van Dijk Electronic Publishing 2005). Investment and cash flow variables are normalized by the capital stock to control for the size effect. Sales and cash flow are deflated by wholesale price deflators speci�c to three-digit NACE sectors, obtained from the Czech Statistical Of�ce. A deflator for tangible �xed assets obtained from the Czech Statistical Of�ce is used for tangible �xed assets and depreciation. Supplierit is a time-varying dummy variable taking a value of 1 if �rm i is a multinational supplier at time t, and 0 otherwise. It is de�ned based on the information obtained from enterprise surveys. The coef�- cient a2 captures the sensitivity of �rm-level investment to internal funds. If a �rm is liquidity constrained (if the desired investment level is constrained by the availability of internal �nance), the coef�cient is expected to be positive and statistically signi�cant. With perfect capital markets, the �rm and lender would be indifferent between internal and external �nancing and hence the coef�cient would be expected to equal 0. The goal of the analysis is to examine the link between access to credit and multinational supplier status. A priori, having a contract from a well-known MNC would be expected to increase the creditworthiness of Czech suppliers and thus ease their �nancing constraints. Therefore, multinational suppliers 332 THE WORLD BANK ECONOMIC REVIEW would be less dependent on their internal cash flow than nonsuppliers. To examine this effect, cash flow is interacted with the indicator variable for mul- tinational suppliers. If �rms supplying MNCs are not liquidity constrained, the sum of the coef�cients a2 and a3 would be expected to equal 0. It is also possible that �rms’ ability to obtain external �nancing and to become multinational suppliers is driven by other factors. For instance, more productive �rms may be better positioned to become suppliers and may be identi�ed by lenders as lower risk borrowers. To attenuate this concern, the analysis controls for labor productivity (de�ned as the log of the value added per worker) and its interaction with cash flow. Similarly, exporters may possess qualities that make it easier for them to obtain multinational contracts while their relationships with buyers abroad may make them lower risk borrowers. Therefore, survey data are used to control for the �rm’s exporting status and its interaction with cash flow. Finally, a dummy variable is used to capture different investment behavior of state-owned enterprises and an interaction of the dummy variable with cash flow is used to capture the possibility that state enterprises may enjoy soft budget constraints (for evidence, see Lizal and Svejnar 2002). State enterprises in the sample are identi�ed by responses to survey questions on whether a company was established as a state enterprise and whether (and when) it was privatized. The model also includes several �rm-speci�c time-varying factors that might influence the level of investment. The analysis controls for a �rm’s size, measured by employment and expressed in log form, the log of a �rm’s age, and the level of long-term debt normalized by capital stock. To control for unobserved heterogeneity across �rms, a model is estimated using �rm �xed effects (ai). Year �xed effects (at) are also included. They capture aggregate conditions affecting the cost of capital in a particular year, so controls for interest rates or tax rates are unnecessary. A common concern with the cash flow sensitivity approach is that the cash flow variable may pick up more than pure liquidity effects. However, this article focuses on comparing cash flow sensitivity across �rms, and so as long as such a bias does not vary systematically by multinational supplier status, it is not a major concern. The estimation results from the baseline speci�cation are presented in table 2. Regression 1 tests for the direct effect of cash flow on investment. The results suggest that �rms operating in the Czech Republic are liquidity con- strained. The coef�cient on the cash flow variable is positive and statistically signi�cant at the 1 percent level, reflecting that internal funds are an important determinant of the investment decision. As expected, the sales growth coef�- cient is also positive and statistically signi�cant. Regression 2 repeats the exercise using lagged cash flow. The conclusion is the same, although the magnitude of the coef�cient is somewhat smaller. While it might be preferable to employ lagged rather than contemporaneous values of cash flow, doing so would have signi�cantly reduced the sample size. T A B L E 2 . Baseline Speci�cation Variable (1) (2) (3) (4) (5) (6) DSales 0.087*** (0.027) 0.123*** (0.038) 0.081** (0.032) 0.074** (0.033) 0.073** (0.033) 0.074** (0.033) CF/K 0.325*** (0.024) 0.446*** (0.047) 0.447*** (0.084) 0.395*** (0.050) 0.447*** (0.084) CF/K lagged 0.205*** (0.025) CF/K * Supplier 2 0.360*** (0.064) 2 0.318*** (0.065) 2 0.325*** (0.065) 2 0.319*** (0.065) Supplier 0.079 (0.061) 0.063 (0.061) 0.066 (0.061) 0.064 (0.061) CF/K * ln(VA/L) 0.008 (0.007) 0.010 (0.008) 0.011 (0.008) 0.010 (0.008) ln(VA/L) 2 0.008 (0.005) 2 0.012* (0.007) 2 0.012* (0.007) 2 0.012* (0.007) CF/K * Exporter 2 0.059 (0.075) 2 0.059 (0.075) Exporter 0.067 (0.083) 0.065 (0.083) CF/K * SOE 2 0.085 (0.310) 2 0.075 (0.310) SOE 2 0.068 (0.217) 2 0.057 (0.218) Debt/K 0.011 (0.007) 0.011 (0.007) 0.011 (0.007) ln(Employment) 2 0.064 (0.059) 2 0.064 (0.059) 2 0.063 (0.059) ln(Age) 2 0.087 (0.070) 2 0.085 (0.070) 2 0.089 (0.070) Intercept 0.066** (0.030) 0.122*** (0.028) 0.074 (0.089) 0.447 (0.331) 0.493 (0.328) 0.448 (0.332) Number of observations 1735 1398 1382 1359 1359 1359 Number of �rms 319 301 314 307 307 307 R-squared 0.14 0.07 0.18 0.15 0.15 0.15 F-test CF/K þ CF/K * Supplier ¼ 0 1.80 1.75 1.20 1.73 p-value 0.18 0.19 0.27 0.19 CF/K þ CF/K * Exporter ¼ 0 58.88 58.45 p-value 0.00 0.00 CF/K þ CF/K * SOE ¼ 0 1.00 1.37 p-value 0.32 0.24 Note: All speci�cations include �rm and year �xed effects. Numbers in parentheses are standard errors. *Signi�cant at the 10 percent level. Javorcik and Spatareanu **Signi�cant at the 5 percent level. ***Signi�cant at the 1 percent level. Source: Authors’ analysis based on data from two World Bank surveys of Czech �rms in 2003 and 2004 and Bureau van Dijk Electronic Publishing 333 (2005); see text for details. 334 THE WORLD BANK ECONOMIC REVIEW Regression 3 examines whether the link between cash flow and investment differs between multinational suppliers and other �rms. The model includes a dummy variable that takes the value of 1 in each year in which the �rm supplies an MNC operating in the Czech Republic and 0 otherwise. The dummy variable is also interacted with cash flow. If �rms with linkages to mul- tinationals �nd it easier to obtain credit, the sum of the coef�cients on cash flow and the interaction term should not be statistically signi�cant. While cash flow continues to bear a positive and statistically signi�cant coef�cient, the interaction term is negative and statistically signi�cant at the 1 percent level. The F-test indicates that the hypothesis that the sum of the two coef�cients is equal to 0 cannot be rejected, suggesting that, unlike nonsuppliers, multina- tional suppliers do not face liquidity constraints. Neither labor productivity nor its interaction with cash flow reaches conventional signi�cance levels. The supplier dummy variable is not statistically signi�cant, suggesting that multina- tional suppliers do not differ in their investment behavior from other �rms.8 Next, the analysis tests for whether the �nding that multinational suppliers are less credit constrained is due to �rms being exporters rather than to their being multinational suppliers. Exporting �rms may be less credit constrained because of a steady stream of income from more creditworthy foreign custo- mers, and their experience dealing with foreign buyers may better position them to become multinational suppliers. Potential �rm-level determinants of investment behavior (size, age, and debt level) are also controlled for. The �ndings are robust to these additional controls. The coef�cient on the interaction between the multinational supplier dummy variable and cash flow remains negative and statistically signi�cant at the 1 percent level. As before, the F-test suggests that multinational suppliers do not face liquidity constraints. In contrast, exporters appear to be as liquidity constrained as other Czech �rms. The interaction term is not statistically signi�cant, and the F-test rejects the absence of a link between investment and cash flow.9 The likely expla- nation is that many Czech �rms that continued to sell to their Slovak customers after Czechoslovakia split in 1993 are considered to be exporters, yet their Slovak buyers are unlikely to be more creditworthy than Czech buyers. This also explains why such a high percentage of observations in the sample pertain to exporters.10 The additional controls for size, age, and debt level do not appear to be statistically signi�cant. Lizal and Svejnar (2002) �nd that state enterprises in the Czech Republic were facing soft budget constraints in the 1990s. As there are only 19 state enterprises in the sample, many of which were privatized during the period considered, there is little concern that their presence affects the main �ndings. Nevertheless, 8. Some differences may be captured by �rm �xed effects included in the model. 9. Excluding the supplier dummy variable and its interaction with cash flow from the model would not change this conclusion. 10. The Slovak Republic is the second largest export market for Czech �rms. Javorcik and Spatareanu 335 regression 5 adds a state enterprise dummy variable and its interaction with cash flow. Neither variable appears to be statistically signi�cant, but as expected the F-test cannot reject the hypothesis that state enterprises are not credit con- strained. The �nding on multinational suppliers remains unchanged. The last column in table 2 includes all the controls listed in equation (1) and con�rms the earlier conclusions. The cash flow variable has a positive and stat- istically signi�cant coef�cient, and its interaction with the multinational sup- plier dummy variable is negative and signi�cant at the 1 percent level. Based on these coef�cients and the F-test, Czech �rms in general appear to be liquid- ity constrained, but multinational suppliers do not. As before, the results suggest that state enterprises may be subject to soft budget constraints. Are Future Multinational Suppliers Less Credit Constrained? As mentioned, it is possible that less liquidity constrained �rms self-select as suppliers to MNCs. Because multinational customers tend to have higher requirements for quality, technological sophistication, and on-time delivery than domestic buyers in developing and transition economies, becoming a mul- tinational supplier is likely to be associated with some �xed cost for local �rms. Thus, it may well be the case that only �rms not facing liquidity con- straints are able to become multinational suppliers. This possibility is examined by checking whether multinational suppliers appear to be less liquidity con- strained than other �rms before they start their contracts with multinationals, as estimated by the following model: Iit =KitÀ1 ¼ b0 þ b1 DSit =SitÀ1 þ b2 CFit =KitÀ1 þ b3 CFit =KitÀ1 à Supplierit þ b4 Supplierit þ b5 CFit =KitÀ1 à 1 yr beforeit þ b6 1 yr beforeit ð2Þ þ b7 CFit =KitÀ1 à 2 yrs beforeit þ b8 2 yrs beforeit þ b9 CFit =KitÀ1 à lnðVA=LÞ þ b10 lnðVA=LÞ þ b11 lnðSizeit Þ þ b12 lnðAgeit Þ þ b13 Debt ratioit þ ni þ nt þ uit where 1 yr beforeit equals 1 at time t if �rm i will become a multinational sup- plier at t þ 1, and 0 otherwise, and 2 yrs beforeit equals 1 at time t if �rm i will become a multinational supplier at t þ 2, and 0 otherwise. A sum of b2 and b7 equal to 0 would indicate that multinational suppliers were not credit constrained two years before starting their relationship with an MNC. A sum of b2 and b5 equal to 0 would suggest that multinational suppliers were not facing credit constraints one year before starting their relationship with an MNC. Either or both �ndings would suggest self-selection of unconstrained �rms into becoming multinational suppliers. The estimation results of equation (2) are presented in table 3. Regression 1 looks at whether multinational suppliers were liquidity constrained one year T A B L E 3 . Current Suppliers, Future Suppliers and Nonsuppliers 336 Variable (1) (2) (3) (4) (5) DSales 0.085*** (0.032) 0.084*** (0.032) 0.078** (0.033) 0.078** (0.033) 0.063* (0.034) CF/K 0.482*** (0.048) 0.482*** (0.048) 0.432*** (0.051) 0.432*** (0.051) –0.326 (0.636) CF/K * 2 yrs before 2 0.26 (0.506) 2 0.234 (0.506) 2 0.232 (0.507) 2 0.258 (0.498) CF/K * 1 yr before 2 0.510*** (0.133) 2 0.515*** (0.133) 2 0.465*** (0.134) 2 0.465*** (0.134) 2 0.622*** (0.164) CF/K * Supplier 2 0.439*** (0.067) 2 0.440*** (0.067) 2 0.399*** (0.068) 2 0.400*** (0.069) 2 0.395*** (0.111) 2 yrs before – 0.048 (0.102) – 0.065 (0.102) –0.065 (0.102) –0.061 (0.100) 1 yr before 0.02 (0.079) – 0.014 (0.087) – 0.037 (0.087) –0.037 (0.087) –0.023 (0.086) Supplier 0.03 (0.073) – 0.01 (0.084) – 0.032 (0.084) –0.032 (0.088) –0.034 (0.083) CF/K * ln(VA/L) 0.008 (0.007) 0.008 (0.007) 0.011 (0.008) 0.011 (0.008) –0.015* (0.009) ln(VA/L) –0.009* (0.005) – 0.009 (0.005) – 0.013** (0.007) – 0.013** (0.007) –0.002 (0.007) Debt/K 0.011 (0.007) 0.011 (0.007) 0.015** (0.007) ln(Employment) – 0.069 (0.059) –0.069 (0.059) –0.028 (0.058) ln(Age) – 0.072 (0.070) –0.073 (0.070) –0.064 (0.069) Supplier * Year 1999 –0.009 (0.066) Supplier * Year 2000 0.010 (0.072) THE WORLD BANK ECONOMIC REVIEW Includes interactions of CF/K with two-digit industry �xed effects Number of observations 1382 1382 1359 1359 1359 Number of �rms 314 314 307 307 307 R2 0.19 0.19 0.16 0.16 0.22 F-test CF/K þ CF/K * Supplier ¼ 0 0.46 0.44 0.26 0.25 1.25 p-value 0.50 0.51 0.61 0.62 0.26 CF/K þ CF/K * 1 yr before ¼ 0 0.04 0.06 0.06 0.06 2.09 p-value 0.83 0.80 0.80 0.80 0.15 CF/K þ CF/K * 2 yrs before ¼ 0 0.19 0.15 0.16 0.53 p-value 0.66 0.70 0.69 0.47 Note: All speci�cations include �rm and year �xed effects and a constant. Numbers in parentheses are standard errors. *Signi�cant at the 10 percent level. **Signi�cant at the 5 percent level. ***Signi�cant at the 1 percent level. Source: Authors’ analysis based on data from two World Bank surveys of Czech �rms in 2003 and 2004 and Bureau van Dijk Electronic Publishing (2005); see text for details. Javorcik and Spatareanu 337 before they started their relationship with an MNC. As before, the coef�cient on cash flow is positive, though slightly larger, and statistically signi�cant at the 1 percent level. The interaction terms between the multinational supplier dummy variable and cash flow and between future supplier and cash flow are both negative and statistically signi�cant at the 1 percent level. F-tests suggest that, unlike Czech �rms in general, neither current nor future multinational suppliers face liquidity constraints. Regression 2 considers the two-year period before starting a relationship with an MNC. The interactions of cash flow with 1 yr before and supplier remain negative and statistically signi�cant. The coef�cient on the interaction with 2 yrs before is negative, though not statistically signi�cant. F-tests cannot reject the hypothesis that multinational suppliers are not liquidity constrained and that this lack of constraints is already present in the two-year period before becoming a supplier. Regression 3 shows that the �ndings are robust to controlling for �rm size, age, and debt level. In sum, the �ndings are suggestive of unconstrained �rms self-selecting into becoming multinational suppliers. To take into account a currency crunch that took place in the Czech Republic in 1999–2000 following a banking crisis (see Pruteanu 2004), an interaction of the supplier dummy variable with a dummy variable for year 1999 (and 2000) is added to the speci�cation. Doing so will shed light on whether multinational suppliers were affected differently by the credit crunch: multinationals with their global distribution networks are less affected by changes in the Czech market and thus less likely to adjust their relationships with their suppliers. As evident from regression 4, however, there is no indi- cation of any different investment behavior among multinational suppliers than among other �rms during the credit crunch period. Neither interaction term is statistically signi�cant. Other conclusions remain unchanged. To account for the possibility that �rms in growing sectors might be more likely to be both multinational suppliers and not liquidity constrained, inter- actions between dummy variables for two-digit NACE codes (18 in total) and the cash flow variable are added. Only two of these interaction terms are stat- istically signi�cant (furniture; computer, electronic, and optical products). The results con�rm the previous �ndings that suppliers to MNCs are not liquidity constrained and that the effect is already present two years before signing a contract with an MNC. This speci�cation also �nds a signi�cant positive coef- �cient on the debt variable and a signi�cant negative coef�cient on the inter- action between cash flow and labor productivity. One may wonder about the results of F-tests based on the interaction of cash flow and labor productivity as well as the interaction of cash flow and current (or future) supplying status. F-tests taking into account the average labor productivity among current (or future, as appropriate) suppliers support the earlier conclusions: both current and future multinational suppliers do not appear to be credit constrained. 338 THE WORLD BANK ECONOMIC REVIEW Finally, additional robustness checks (not reported to save space) show that the conclusions are not affected by dropping observations with negative values for cash flow or by including industry-year �xed effects. Another way to shed light on the link between credit constraints and multinational supplying status is to estimate a probit model that aims to explain the supplying status with the lagged liquidity ratio, gross pro�t (logged), and debt (normalized by capital). Supplierit is the dependent variable. Liquidity ratio is de�ned as the difference between current assets and current liabilities divided by total assets. This speci�cation also controls for �rm size (number of employees), age, and labor productivity (all in logs) as well as three-digit industry and year �xed effects. The results show a positive and statistically signi�cant link between lagged liquidity ratio, lagged gross pro�t, and the probability of being a multinational supplier (table A-1). Coef�cients on debt, employment, and labor productivity are not statistically signi�cant. As this last �nding is somewhat puzzling, �rm performance was also measured using total factor productivity estimated by the sector-speci�c production function (ordinary least squares or the Olley– Pakes 1996 method). Once liquidity ratio, gross pro�t, and debt are controlled for, �rm productivity is not a statistically signi�cant predictor of supplying status. Finally, the data also indicate that younger �rms are more likely to supply MNCs.11 In sum, the �ndings suggest that �rms not facing liquidity constraints self-select into becoming multinational suppliers. This is consistent with the observation that to obtain contracts from MNCs �rms need to meet the strin- gent requirements of multinational customers and that only �rms with access to �nancing may be able to do so. The survey data are in line with these con- clusions. Most suppliers make improvements within the 12-month period before signing a contract with an MNC. The most frequent changes include improvements to product quality, staff training, and productivity enhance- ments. Many of these changes are probably made to obtain ISO certi�cations. More than 40 percent of suppliers reported being required by prospective mul- tinational customers to obtain ISO certi�cation. As the certi�cation process is quite costly, usually involving the services of a specialized consulting �rm, it would not be surprising if only �rms that were not liquidity constrained were able to complete it. Robustness Checks To eliminate the possibility that the �ndings could be driven by MNCs extending credit to future suppliers, the 15 Czech �rms that reported receiving some �nancial help from their multinational customers were removed from the 11. In a probit model predicting the decision of Czech �rms to become multinational suppliers rather than the decision to supply MNCs in a given year, liquidity ratio and �rm size were the main predictors of the decision to become a multinational supplier. Javorcik and Spatareanu 339 sample. The results con�rm the earlier pattern. Multinational suppliers were not liquidity constrained two years before supplying an MNC, and they remained unconstrained while supplying the multinational (table 4, regressions 1 and 2). To examine whether the �ndings are due to the possibility that future multi- national suppliers have a lower credit risk because of a contract with an MNC, Czech suppliers that reported that having a relationship with an MNC helped them obtain �nancing are dropped from the sample. Eliminating these 24 �rms does not affect the results (regressions 3 and 4). The �nding that multinational suppliers are less credit constrained is thus con�rmed, and the evidence suggests that less constrained �rms self-select into becoming multinational suppliers. Instrumental Variable Approach With the evidence suggesting self-selection by less credit constrained �rms into supply relationships with MNCs and the possibility that some explanatory variables are endogenous, the �nal step is to apply an instrumental variable approach. The analysis uses the generalized method of moments (GMM) system estimation (proposed by Blundell and Bond 1998) and instruments for sales growth, labor productivity, supplier status, cash flow, and cash flow inter- actions with supplier status and with labor productivity. The GMM estimator combines a differenced and a level equation. Lagged levels of endogenous vari- ables are used as instruments for contemporary differences, and lagged differ- ences are used as instruments for the level equation. Several additional instruments are also used. Firms whose managers speak a foreign language or who have worked for foreign companies before are likely to be better positioned to obtain contracts from multinationals. Thus, dummy variables reflecting these two characteristics are used as instruments for supply status. Level of language pro�ciency was determined by whether the manager can conduct business negotiations in a foreign language or can understand a business agreement in a foreign language, as reported in surveys. As exporters may �nd it easier to become multinational suppliers because of their experience of dealing with foreign customers, the second lag of exporting status is also used as an instrument. As it is also likely that proximity to MNCs facilitates business relationships, the instrument set includes proxies for the presence of multinationals in the same industry and in downstream industries. The share of sector output pro- duced by foreign �rms is the proxy for the presence of MNCs in the same sector. It is calculated by weighting the output of each �rm f in sector j (Yft) by the share of the �rm f ’s equity owned by foreigners (Foreign shareft) and dividing it by the total output of sector j: P f for all f [j Foreign shareft à Yft ð3Þ MNCs in the same sector jt ¼ P : f for all f [j Yft T A B L E 4 . Excluding Suppliers Bene�ting from Multinational Assistance 340 Excluding �rms reporting easier access to credit Excluding �rms receiving �nancial assistance from because of their relationship with multinational multinational corporations corporations Variable (1) (2) (3) (4) DSales 0.076** (0.033) 0.068** (0.035) 0.104*** (0.034) 0.098*** (0.035) CF/K 0.419*** (0.049) 0.353*** (0.052) 0.492*** (0.049) 0.442*** (0.052) CF/K * 2 yrs before 2 0.23 (0.505) 2 0.193 (0.505) 2 0.45 (0.631) 2 0.46 (0.631) CF/K * 1 yr before 2 0.471*** (0.134) 2 0.411*** (0.134) 2 0.530*** (0.135) 2 0.480*** (0.136) CF/K * Supplier 2 0.410*** (0.071) 2 0.358*** (0.072) 2 0.454*** (0.069) 2 0.416*** (0.070) 2 yrs before 2 0.091 (0.106) 2 0.11 (0.106) 2 0.002 (0.121) 2 0.018 (0.121) 1 yr before 2 0.059 (0.094) 2 0.083 (0.094) 0.028 (0.099) 0.002 (0.099) Supplier 2 0.05 (0.088) 2 0.074 (0.088) 0.038 (0.097) 0.012 (0.098) CF/K * ln(VA/L) 0.01 (0.007) 0.014* (0.007) 0.007 (0.008) 0.01 (0.008) ln(VA/L) 2 0.008 (0.005) 2 0.012* (0.006) 2 0.008 (0.005) 2 0.015** (0.007) Debt/K 0.007 (0.007) 0.012 (0.007) ln(Employment) 2 0.065 (0.058) 2 0.110* (0.062) THE WORLD BANK ECONOMIC REVIEW ln(Age) 2 0.056 (0.071) 2 0.088 (0.075) Intercept 0.103 (0.092) 0.511 (0.327) 0.061 (0.099) 0.742** (0.346) Number of observations 1311 1288 1267 1244 Number of �rms 299 292 290 283 R2 0.16 0.13 0.21 0.17 F-test CF/K þ CF/K * Supplier ¼ 0 0.02 0.01 0.32 0.15 p-value 0.90 0.94 0.57 0.70 CF/K þ CF/K * 1 yr before ¼ 0 0.15 0.19 0.08 0.08 p-value 0.70 0.66 0.78 0.78 CF/K þ CF/K * 2 yrs before ¼ 0 0.14 0.10 0.00 0.00 p-value 0.71 0.75 0.95 0.98 Note: All speci�cations include �rm and year �xed effects. Numbers in parentheses are standard errors. *Signi�cant at the 10 percent level. **Signi�cant at the 5 percent level. ***Signi�cant at the 1 percent level. Source: Authors’ analysis based on data from two World Bank surveys of Czech �rms in 2003 and 2004 and Bureau van Dijk Electronic Publishing (2005); see text for details. Javorcik and Spatareanu 341 The proxy for the presence of multinationals in downstream sectors (sectors supplied by �rm i operating in sector j ) is de�ned following Javorcik (2004) as: P X f for all f [ k Foreign shareft à Yft Potential MNC customersjt ¼ a jk à P : k if k=j f for all f [ k Yft ð4Þ The proportion of sector j’s output supplied to a downstream sector k based on the 1999 input–output matrix of the Czech Republic (ajk) is used to weight multinational presence in each downstream sector k. As the formula indicates, inputs supplied within the sector are not included. Thus, the greater the foreign presence in sectors supplied by industry j and the larger the share of output supplied to industries with a multinational presence, the higher is the value of the variable.12 The calculations are based on all �rms included in the AMADEUS database, not just the �rms in the sample. Cash flow interactions with the instruments mentioned above are used to instrument for the inter- action of cash flow with the multinational supplier dummy variable. Table 5 lists the instruments included in a given speci�cation. The number of observations in GMM regressions is smaller than in the previous speci�cations. Because the model is expressed in �rst differences an additional year of data is lost. Further years of data are lost because the instru- ments are based on second and further lags. While the results should be treated with caution because of the small number of observations, they are nevertheless informative. The Hansen test for overiden- ti�cation restrictions shows that the null hypothesis cannot be rejected at con- ventional signi�cance levels (see table 5). The Arellano–Bond test shows that the null hypothesis of no second-order serial correlation also cannot be rejected. These speci�cation tests suggest that the regressions yield consistent estimates. The GMM results suggest that supplier status has no signi�cant impact on a �rm’s liquidity constraints, once self-selection is taken into account. The inter- action term between cash flow and supplier status is not statistically signi�cant in any of the regressions (or in many other regressions estimated but not reported here to save space). In all speci�cations, the F-test rejects the absence of a relationship between cash flow and investment for multinational suppliers. As expected, the cash flow variable remains statistically signi�cant in all regressions, suggesting that domestic �rms are liquidity constrained. In summary, the evi- dence suggests that suppliers differ from nonsuppliers in liquidity constraints, but the effect appears to be due to self-selection rather than to a relationship with an MNC leading to an easing of the supplier’s �nancial constraints. 12. To illustrate the meaning of the variable, suppose that the sugar industry sells half of its output to jam producers and half to chocolate producers. If no multinationals are producing jam but half of all chocolate production comes from foreign af�liates, Potential MNC customersjt will be calculated as follows: 1 1 1 1 2 * 0 þ 2 * 2 ¼ 4. T A B L E 5 . Generalized Method of Moments Regressions 342 Variable (1) (2) (3) (4) (5) I/K lagged 0.119** (0.051) 0.119** (0.051) 0.111** (0.050) 0.121** (0.051) 0.124** (0.052) DSales 0.021 (0.054) 0.017 (0.055) 0.010 (0.055) 0.015 (0.053) 0.017 (0.054) CF/K 0.322*** (0.067) 0.323*** (0.066) 0.336*** (0.068) 0.324*** (0.066) 0.323*** (0.066) CF/K * Supplier 0.121 (0.100) 0.121 (0.100) 0.11 (0.104) 0.123 (0.100) 0.124 (0.099) Supplier 0.034 (0.032) 0.03 (0.032) 0.033 (0.031) 0.028 (0.032) 0.028 (0.031) CF/K * ln(VA/L) 0.017* (0.010) 0.017* (0.010) 0.016 (0.010) 0.017* (0.010) 0.017* (0.010) ln(VA/L) 2 0.015*** (0.005) 2 0.015*** (0.005) 2 0.016*** (0.005) 2 0.015*** (0.005) 2 0.015*** (0.005) Debt/K 0.013 (0.012) 0.013 (0.012) 0.012 (0.012) 0.013 (0.012) 0.012 (0.011) ln(Employment) 2 0.016 (0.021) 2 0.018 (0.022) 2 0.019 (0.020) 2 0.012 (0.021) 2 0.01 (0.021) ln(Age) 2 0.002 (0.002) 2 0.002 (0.002) 2 0.002 (0.002) 2 0.002 (0.002) 2 0.002 (0.002) Intercept 0.238 (0.166) 0.246 (0.166) 0.241 (0.153) 0.201 (0.149) 0.195 (0.145) Number of observations 728 728 728 728 728 Number of �rms 243 243 243 243 243 Additional instrumental CF/Kt22*Manager’s CF/Kt22*Manager’s CF/Kt22*Manager’s CF/Kt22*Potential MNC CF/Kt22*Manager’s THE WORLD BANK ECONOMIC REVIEW variables foreign language foreign language foreign language customerst22 foreign experience CF/Kt22*Potential MNC CF/Kt22*Potential MNC Potential MNC Potential MNC CF/Kt22*Potential MNC customerst22 customerst22 customerst22 customerst22 customerst22 Potential MNC Potential MNC MNCs in the same CF/Kt22*MNCs in the Potential MNC customerst22 customerst22 sectort22 same sectort22 customerst22 CF/Kt22*MNCs in the Exportert22 MNCs in the same CF/Kt22*MNCs in the same sectort22 sectort22 same sectort22 MNCs in the same Exportert22 MNCs in the same sectort22 sectort22 Exportert22 F-test CF/K þ CF/K * Supplier ¼ 0 17.58 17.75 17.64 18.3 18.27 p-value 0.00 0.00 0.00 0.00 0.00 AR(1) test p-value 0.03 0.03 0.03 0.03 0.03 AR(2) test p-value 0.94 0.95 0.98 0.93 0.93 Hansen test p-value 0.91 0.91 0.94 0.92 0.92 Note: Numbers in parentheses are standard errors. *Signi�cant at the 10 percent level. **Signi�cant at the 5 percent level. ***Signi�cant at the 1 percent level. Source: Authors’ analysis based on data from two World Bank surveys of Czech �rms in 2003 and 2004 and Bureau van Dijk Electronic Publishing (2005); see text for details. Javorcik and Spatareanu 343 I V. P O L I C Y I M P L I C A T I O N S Many countries around the world strive to attract FDI, believing that foreign investors not only bring capital but also serve as a channel of knowledge trans- fer across international borders. Policymakers, expecting some of this knowl- edge to result in externalities that bene�t domestic producers, are willing to offer often generous incentive packages to foreign investors. For instance, 59 of 108 countries surveyed by the World Bank reported offering some type of incentives for FDI in 2004 (Harding and Javorcik 2007). A recent survey of the empirical literature on spillovers from FDI concludes that such spillovers are most likely between MNCs and their local suppliers (Go¨ rg and Greenaway 2004). Thus, understanding what factors allow local �rms to become suppliers to MNCs could have strong implications for under- standing knowledge spillovers and public policy choices. Two main �ndings emerge from the study. First, in contrast to Czech �rms in general, which face �nancial constraints, multinational suppliers do not appear to be liquidity constrained. Second, the data suggest that the lack of liquidity constraints is present before �rms enter into a supplier relationship with MNCs, which is consistent with unconstrained �rms self-selecting into supplying multinationals. Caution is required, however, in interpreting these �ndings. While the �ndings are robust to a number of controls that may be driving both access to credit and the ability of �rms to supply multinationals, the possibility remains that the list of controls is incomplete. Further, even though the results suggest that well- functioning credit markets are important in facilitating business relationships between local �rms and MNCs, they do not suggest that a well-developed �nan- cial market is a suf�cient condition for such relationships. Many other factors, such as a certain level of sophistication of the local manufacturing sector, a match between the skill endowment of the host economy and the sourcing needs of MNCs, and a good business environment, may be needed in order for these relationships to materialize. Thus, the �ndings could plausibly be generalized to other upper middle-income economies, but probably not to low-income economies. AC K N OW L E D G M E N T S The authors thank Thorsten Beck, Steve Fazzari, Jose Luis Groizard, Leonardo Iacovone, Yue Li, Inessa Love, Jan Svejnar, three anonymous reviewers, partici- pants in the workshops Regional and Micro-level Effects of Globalization in Tu¨ bingen, FDI and the Consequences in Ghent, Eastern Economic Association Annual Meetings in New York City, Midwest Conference on Economic Theory and International Trade in Columbus, OH, and the LICOS seminar at Catholic University Leuven for helpful comments and suggestions. 344 THE WORLD BANK ECONOMIC REVIEW FUNDING The authors are grateful to the World Bank’s Research Support Budget for �nancial assistance for the project “Vertical Relationships between Multinationals and Local Firms in the Czech Republic.� APPENDIX A T A B L E A - 1 . Probit Model Predicting a Firm’s Supplying Status Variable Liquidity ratio 0.743*** (0.191) 0.830*** (0.223) 0.823*** (0.248) 0.809*** (0.253) lagged ln(Gross pro�t) 0.079** (0.037) 0.083* (0.049) 0.151** (0.061) 0.129** (0.057) lagged Debt/K lagged 0.131 (0.236) 0.074 (0.248) 0.446 (0.317) 0.428 (0.325) ln(Employment) 0.095* (0.050) 0.075 (0.069) 0.094 (0.066) 0.106 (0.067) lagged ln(Age) lagged 2 0.054 (0.086) 2 0.035 (0.090) 2 0.470*** (0.132) 2 0.480*** (0.135) ln(VA/L) 0.014 (0.019) lagged ln(Total factor 0.105 (0.320) productivity) lagged ln(Total factor 0.27 (0.195) productivity Olley 2 Pakes) lagged Intercept 2 0.561 (0.604) 2 1.444** (0.677) 2 1.223 (0.823) 2 1.983*** (0.725) Number of 1350 1051 949 887 observations Note: All speci�cations include industry and year �xed effects. Numbers in parentheses are robust standard errors. *Signi�cant at the 10 percent level. **Signi�cant at the 5 percent level. ***Signi�cant at the 1 percent level. Source: Authors’ analysis based on data from two World Bank surveys of Czech �rms in 2003 and 2004 and Bureau van Dijk Electronic Publishing (2005); see text for details. REFERENCES Alfaro, Laura, Areendam Chanda, Sebnem Kalemli-Ozcan, and Selin Sayek. 2004. “FDI and Economic Growth: The Role of Local Financial Markets.� Journal of International Economics 65(2):89– 112. ———. 2006. How Does Foreign Direct Investment Promote Economic Growth? NBER Working Paper 12522. Cambridge, MA: National Bureau of Economic Research. Beck, Thorsten, Asli Demirgurc-Kunt, and Ross Levine. 1999. “A New Database on Financial Development and Structure.� World Bank Policy Research Working Paper 2146. World Bank, Washington, D.C. Blalock, Garrick, and Paul J. Gertler. 2008. “Welfare Gains from Foreign Direct Investment through Technology Transfer to Local Suppliers.� Journal of International Economics 74(2):402– 21. Javorcik and Spatareanu 345 Blundell, Richard, and Steve Bond. 1998. “Initial Conditions and Moment Restrictions in Dynamic Panel Data Models.� Journal of Econometrics 87(1):115–43. Bureau van Dijk Electronic Publishing. 2005. AMADEUS: A Pan-European Database of Comparable Financial Information for 9 Million Public and Private Companies. Brussels. Chaney, Thomas. 2005. Liquidity Constrained Exporters. Working Paper, Department of Economics, University of Chicago. CzechInvest Factsheet No. 3, January 2002. Prague: CzechInvest. Fazzari, Steven M., R. Glenn Hubbard, and Bruce C. Petersen. 1988. “Financing Constraints and Corporate Investment.� Brookings Papers on Economic Activity 1:141–95. ———. 2000. “Financing Constraints and Corporate Investment: Response to Kaplan and Zingales.� Quarterly Journal of Economics 115(2):695 –705. Gelos, R. Gaston, and Alejandro Werner. 2002. “Financial Liberalization, Credit Constraints, and Collateral: Investment in the Mexican Manufacturing Sector.� Journal of Development Economics 67(1):1–27. ¨ rg, Holger, and David Greenaway. 2004. “Much Ado about Nothing? Do Domestic Firms Really Go Bene�t from Foreign Direct Investment?� World Bank Research Observer 19(2):171–97. Harding, Tor�nn, and Beata S. Javorcik. 2007. “Developing Economies and International Investors: Do Investment Promotion Agencies Bring Them Together?� Policy Research Working Paper 4339. World Bank, Washington, D.C. Harrison, Ann, and Margaret McMillan. 2003. “Does Direct Foreign Investment Affect Domestic Firms’ Credit Constraints?� Journal of International Economics 61(1):73– 100. Harrison, Ann E., Inessa Love, and Margaret S. McMillan. 2004. “Global Capital Flows and Financing Constraints.� Journal of Development Economics 75(1):269–301. Hoshi, Takeo, Anil Kashyap, and David Scharfstein. 1991. “Corporate Structure, Liquidity, and Investment: Evidence from Japanese Industrial Groups.� Quarterly Journal of Economics 106(4):33– 60. Javorcik, Beata S. 2004. “Does Foreign Direct Investment Increase the Productivity of Domestic Firms? In Search of Spillovers through Backward Linkages.� American Economic Review 94(3): 605–27. Javorcik, Beata S., and Mariana Spatareanu. 2005. “Disentangling FDI Spillover Effects: What Do Firm Perceptions Tell Us?� In Theodore Moran, Edward Graham, and Magnus Blomstro ¨ m, eds., Does Foreign Direct Investment Promote Development? Washington, D.C.: Institute for International Economics. ———. 2008. “To Share or Not To Share: Does Local Participation Matter for Spillovers from FDI?� Journal of Development Economics 85(1– 2):194–217. Kaplan, Steven, and Luigi Zingales. 1997. “Do Investment-Cash Flow Sensitivities Provide Useful Measures of Financing Constraints?� Quarterly Journal of Economics 112(1):169 –215. ———. 2000. “Investment-Cash Flow Sensitivities Are Not Valid Measures of Financing Constraints.� Quarterly Journal of Economics 115(2):707 –12. Konings, Josef, Marian Rizov, and Hylke Vandenbussche. 2003. “Investment and Credit Constraints in Transition Economies: Micro Evidence from Poland, the Czech Republic, Bulgaria and Romania.� Economic Letters 78(2):253– 58. Lizal, Lubomir, and Jan Svejnar. 2002. “Investment, Credit Rationing, and the Soft Budget Constraint: Evidence from Czech Panel Data.� Review of Economics and Statistics 84(2):353– 70. Manova, Kalina. 2006. Credit Constraints in Trade: Financial Development and Export Composition. Cambridge, MA: Harvard University, Department of Economics. Modigliani, Franco, and Merton Miller. 1958. “The Cost of Capital, Corporation Finance, and the Theory of Investment.� American Economic Review 48(3):261– 97. Moran, Theodore. 2001. Parental Supervision: The New Paradigm for Foreign Direct Investment and Development. Washington, D.C.: Institute for International Economics. 346 THE WORLD BANK ECONOMIC REVIEW Moran, Theodore, Edward Graham, and Magnus Blomstro ¨ m. 2005. Does Foreign Direct Investment Promote Development? Washington, D.C.: Institute for International Economics. Olley, Steven G., and Ariel Pakes. 1996. “The Dynamics of Productivity in the Telecommunications Equipment Industry.� Econometrica 64(6):1263 – 97. Pruteanu, Anca. 2004. “Was There Evidence of Credit Rationing in the Czech Republic?� Eastern European Economics 42(5):58–72. Rajan, Raghuram, and Luigi Zingales. 1998. “Financial Development and Growth.� American Economic Review 88(3):559 – 86.