WPS7050 Policy Research Working Paper 7050 Local Intermediate Inputs and the Shared Supplier Spillovers of Foreign Direct Investment Hiau Looi Kee Development Research Group Trade and International Integration Team September 2014 Policy Research Working Paper 7050 Abstract Trade liberalizations have been shown to improve domestic direct investment (FDI) firms. An exogenous EU trade policy firms’ performance through the new varieties of imported shock is shown to cause some FDI firms in Bangladesh to intermediate inputs. This paper uses a unique, representative expand, which led to better performance of the domestic sample of Bangladeshi garment firms to highlight that local firms that shared their suppliers. Overall, the shared supplier intermediate inputs may also enhance domestic firms’ per- spillovers of FDI explain 1/4 of the product scope expan- formance, through the shared supplier spillovers of foreign sion and1/3 of the productivity gains within domestic firms. This paper is a product of the Trade and International Integration Team, Development Research Group. It is part of a larger effort by the World Bank to provide open access to its research and make a contribution to development policy discussions around the world. Policy Research Working Papers are also posted on the Web at http://econ.worldbank.org. The author may be contacted at hlkee@worldbank.org. The Policy Research Working Paper Series disseminates the findings of work in progress to encourage the exchange of ideas about development issues. An objective of the series is to get the findings out quickly, even if the presentations are less than fully polished. The papers carry the names of the authors and should be cited accordingly. The findings, interpretations, and conclusions expressed in this paper are entirely those of the authors. They do not necessarily represent the views of the International Bank for Reconstruction and Development/World Bank and its affiliated organizations, or those of the Executive Directors of the World Bank or the governments they represent. Produced by the Research Support Team Local Intermediate Inputs and the Shared Supplier Spillovers of Foreign Direct Investment 1 Hiau Looi Kee 2 JEL Classification: F2 Keywords: Intermediate Inputs, Foreign Direct Investment, Product Scope, Multi-product Firms, Productivity, Local Suppliers, Shared Supplier Spillovers 1 I am grateful to John Sutton for the suggestion of title. I would like to thank Chad Bown, Bruce Blonigen, Dave Donaldson, Davin Chor, Rob Feenstra, Ronald Findlay, Gordon Hanson, Ann Harrison, Keith Head, Beata Javorcik, Amit Khandelwal, Daniel Lederman, Aaditya Mattoo, Marc Melitz, Caglar Ozden, Jennifer Poole, Andres Rodriguez-Clare and seminar/conference participants at the EIIT Rocky Mountain Trade Conference, IGC conference in Penn State, Washington Area International Trade Symposium 2011, Columbia University, University of Virginia, National University of Singapore, Singapore Management University and the World Bank for comments. The data set used in this paper is part of a larger data set, jointly collected by the author and Ana Margarida Fernandes, with the cooperation of the Government of Bangladesh and funding from the World Bank in November 2004. Part of the project is also funded by KCP trustfund (TF#098652). 2 Trade and International Integration, Development Research Group, The World Bank, Washington, DC 20433, USA; Tel. (202)473-4155; Fax: (202)522-1159; e-mail: hlkee@worldbank.org “We make cutter knives in Malaysia. We used to source a special plastic resin from a local supplier to make the handles for our cutter knives. The local supplier also sold the same material to a fax machine plant in Malaysia owned by the Japanese …rm, Panasonic. When Panasonic closed down the fax machine plant, the local supplier also stopped selling the plastic resin, due to insu¢ cient demand. As a result, our cutter knife production su¤ered. Now we are looking to import the material from Taiwan at a higher cost and have to face exchange rate and shipping uncertainties.” – General Manager of SJD Industries (M) Sdn. Bhd., Malaysia, November 2011. “LSI manufactures garment accessories in Bangladesh since 1999. Among other factors, serving FDI garment …rms was an important reason for us to set up our plant in Dhaka, EPZ. At the beginning, the share of FDI garment …rms in our total sales was about 20%. Now it is 35-40%. Many Bangladeshi garment …rms bene…tted from LSI working with FDI garments …rms, and to comply to the standard of FDI garment …rms which requires LSI to upgrade and expand product range, capacity, e¢ ciency, and to reduce our costs and lead time. Moreover, LSI always shares the market intelligence we learned from our FDI garment clients regarding the latest product requirements and fashion trend with our other clients. Thus, the domestic garment …rms that buy from us can further improve themselves based on the information.” – Managing Director of LSI Industries Ltd., Bangladesh, November 2010. 1 Introduction While new intermediate inputs play a critical role in explaining productivity gains and growth in many theoretical models of endogenous growth, empirical evidence in support of this channel has been scant. At the macro level, Feenstra (1994) is the …rst to estimate substantial gains from trade derived from using new import varieties as a measure of new intermediate inputs. Broda and Weinstein (2006) further …nd signi…cant gains in US GDP through increased import variety that pushes down the import price index. Nevertheless, it is not until recently that one begins to see some micro-level evidence linking new imported intermediate inputs to the gains in product scope and productivity of domestic …rms. In the context of input tari¤ reduction due to trade liberalization, Amiti and Konings (2007) show how Indonesian …rms gain in terms of total factor productivity (TFP) due to input tari¤ cuts, which allow the …rms to import more intermediate in- puts. Furthermore, Goldberg, Khandelwal, Pavcnik and Topalova (2010) …nd Indian …rms expand their output variety in response to increased access to new imported intermediate input variety. Most recently, Halpern, Koren and Szeidl (2011) …nd that importing foreign varieties of inputs in- creases Hungarian …rm productivity by 12 percent. However, new varieties of intermediate inputs can also be produced locally, as opposed to being acquired through imports. In fact, there is seldom any distinction made between imported and local intermediate inputs in explaining productivity gains in most models (Ethier, 1982; Romer 1990). For many developing countries with problematic trading infrastructure, promoting viable local intermediate input industries that o¤er high quality and more varieties of intermediate inputs may have signi…cant bene…ts to domestic …nal goods pro- ducers.1 To date, there is no empirical evidence on this important issue which has signi…cant policy implications. Besides lowering tari¤s on imported intermediate inputs, what else can governments do to enhance the product scope and productivity of domestic …rms? This paper shows that local intermediate inputs may also enhance the performance of domestic …rms, through the shared supplier spillovers (SSS) of foreign direct investment (FDI) …rms.2 Plainly stated, SSS of FDI …rms capture the positive, non-pecuniary externalities a FDI …rm may exert on a domestic …rm when they share an intermediate input supplier. In this paper, I focus only on local intermediate input suppliers and, for the ease of discussion, when two …rms share a local intermediate input supplier, they are considered as siblings. FDI …rms may have SSS on their domestic siblings because FDI …rms improve the quality and variety of local intermediate inputs. To identify SSS of FDI …rms, I relate the within-…rm, over time gains in product scope and productivity of each domestic …rm to the exogenous increase of the market shares of their FDI siblings. The exogenous change in market share is the result of a natural experiment due to a one-time trade policy change in a foreign market. At the sample mean, SSS is shown to explain about a quarter of the product scope expansion and a third of the productivity gains within …rms, over the 5 year 1 Unlike imported intermediate inputs, local intermediate inputs are not subjected to tari¤ and exchange rate risks, and could avoid problems due to unreliable customs clearing and shipping delays, which signi…cantly cut down lead time for the downstream …rms. Moreover, …rms may have better control over the quality and speci…cation of these intermediate inputs as they may inspect or supervise the production process of their local suppliers. 2 I am very grateful to John Sutton for suggesting the term “shared supplier spillovers”to me at an IGC conference in spring 2011. 2 sample period, a result that is signi…cant both statistically and economically. I further derive and estimate a simple multiproduct …rm model which shows that, through the use of IV regressions, (i) the increase in FDI presence causes the number of local input suppliers to increase which leads to greater variety of local inputs; and (ii) a larger variety of local inputs promotes product scope and productivity of domestic …rms. Firm level data of the Bangladeshi garment sector is speci…cally collected to study this issue.3 ; 4 The data set consists of a strati…ed random sample of 10 percent of the domestic …rms and 100 percent of the FDI …rms in the apparel sector of Bangladesh.5 Each of these …rms is asked to identify its top three local input suppliers. Given that I have all the FDI …rms in the sample, I have the complete list of the top three local suppliers that work with FDI …rms in Bangladesh. For each domestic …rm it is, therefore, possible to construct the market share of their FDI siblings. The data set also has information on the product scope of each …rm. To measure the unobservable …rm productivity, this paper considers multiple …rm productivity measures, which include sales per worker, output per worker, and estimated TFP (OLS and augmented Olley-Pakes due to Ackerberg, Benkard, Berry and Pakes (2007), which takes into account the FDI status of the …rms). Furthermore, …rm speci…c prices on output and materials are used to de‡ate sales and material costs of …rms for the productivity estimations to avoid bias due to the use of industry price de‡ators. By looking at a wide range of performance indicators, the results of the paper are not speci…c to the way …rm productivity is measured. Nevertheless, these di¤erent performance indicators yield very similar and consistent results. Several robustness checks of the paper further con…rm that these results are not driven by selection bias, omitted variable bias, endogeneity bias, reversed causality, and the alternative spillovers from FDI …rms to domestic …rms when they have 3 In 2004, this data set was jointly collected by Ana Fernandes and me, as part of a World Bank project requested by the government of Bangladesh. I focused on the garment sector while Ana studied the rest of the manufacturing sector. The ultimate purpose of my project is to inform the Government on whether or not it is worthwhile to liberalize the garment sector for more FDI, in anticipation of the end of the Multi-Fiber Agreement in 2005. Many FDI restrictions have been removed since then, given the …ndings of positive spillovers in the previous draft of this paper in the Bank’ s report. 4 Demidova, Kee and Krishna (2012) and Cherkashin, Demidova, Kee and Krishna (2010) also use the same data set to study the sorting behavior of …rms when they face very di¤erent demand shocks and trade policy regimes in di¤erent markets. Both these papers do not look at local intermediate inputs and other factors that would a¤ect product scope and productivity of …rms. 5 There were only about 49 FDI garment …rms in Bangladesh at the time the survey was collected and I made sure that we visited all of them. However, not all …rms provided all the information necessary for the regression analysis. After dropping …rms that have incomplete data, I am left with 41 FDI …rms. 3 common markets or common products. Robustness checks also rule out SSS from other domestic …rms, or from large …rms regardless of FDI status. In a placebo exercise, random matching of domestic …rms with FDI suppliers yields no signi…cant or meaningful results, further lend credits to the …ndings of this paper when it only relies on information of the top three local suppliers at the end of sample period. When the fraction of a …rm’s top three local suppliers that also sell to FDI …rms is used as an alternative measure of the in‡uence of FDI siblings, the results are very similar and consistent with SSS. This paper also con…rms that in the garment sector of Bangladesh, FDI …rms use local intermediate inputs more intensively relative to domestic …rms, which satisfy the necessary condition for FDI spillovers to materialize, according to Rodriguez-Clare (1996). To further supplement the econometric results of this paper, the local suppliers listed in the …rm survey were interviewed. These suppliers indicate that not only does working with FDI clients help them improve their e¢ ciency, product quality and variety, they also sell to their domestic …rm clients these better quality and new variety products. These results are consistent with the …nding of shared-supplier spillovers from FDI …rms. It is important to highlight that my analysis is not about comparing the performance of domestic …rms with more FDI siblings with those that have less or no FDI siblings, as that could be driven by selection bias due to …rm size, and productivity. All the analyses are done with …rm …xed e¤ects and …rm speci…c time trends, so the estimations only rely on variations within a …rm across time that is over and beyond …rm speci…c time trend. In addition, the within estimations also control for …rm-supplier relationship which is …xed at the end of the sample period, and the identi…cation of SSS only depends on the detrend market share movement of the FDI siblings, not the choice and/or the numbers of local suppliers. Finally, while …rms within the same industry often share intermediate input suppliers, some intermediate input suppliers may also serve …rms in di¤erent industries. Thus, while related, SSS is not equivalent to horizontal spillovers. The quote from the general manager of a cutter knife …rm in Malaysia, presented at the beginning of this paper, nicely provides such an example.6 Horizontal spillovers of FDI typically were identi…ed by relating the performance of domestic …rms to the overall industry presence or market shares of all FDI …rms in the same industry, an identi…cation strategy that has been made standard due to an in‡uential 6 Another simple example of intermediate inputs that could be used in di¤erent industries are zippers, which are used by the apparel …rms to make jackets, or shoes …rms to make boots, or luggage …rms to make bags. 4 paper by Aitken and Harrison (1999). Instead, to study SSS of FDI, I focus on the …rm speci…c sibling relationship between domestic and FDI …rms, and I control for the overall presence of all FDI …rms using industry-location-year …xed e¤ects. Industry-location-year …xed e¤ects also control for any macro-level in‡uences on …rm performance, such as trade policies, market competition, demand and productivity shocks.7 Thus, this paper concludes that the governments should attract FDI …rms with signi…cant backward linkages that can help promote the local intermediate input industries, which may then indirectly enhance the performance of domestic …rms that use these local intermediate inputs. Such FDI …rms not only can bene…t those local suppliers by improving their productivity and quality (Javorcik, 2004; Blalock and Gertler, 2008; and Javorcik and Spatareanu, 2009), they will also improve the product scope and productivity of those domestic …rms that use these local intermediate inputs (Rodriguez-Clare, 1996; Carluccio and Fally, 2011).8 The rest of this paper is organized as follows: Section 2 discusses the intuitions behind SSS, shows that there is a conducive environment for SSS to happen in Bangladesh and the related literature on FDI spillovers. Section 3 provides some de…nitions for the main variables used in the empirical analysis. Section 4 describes the data set and the policy environment during the sample period. Reduced form regression results are presented in Section 5, followed by some robustness checks in Section 6. Section 7 concludes. The Appendix of this paper provides some details on the related literature on horizontal spillovers, survey evidence on FDI promoting quality upgrading and variety expansion of local input suppliers, the construction of the …rm speci…c price indexes and the estimation of …rm productivity. 7 Nevertheless, given that, in this paper, domestic and FDI …rms are within the same industries, I can further relate the …nding of SSS to horizontal spillovers. The results show that if an industry consists of mostly domestic …rms that do not have FDI siblings, which could be due to the limited backward linkages among these FDI …rms, the scope of positive horizontal spillovers will be very limited, which could explain the existing …ndings of the literature. Conversely, if an industry consists of mostly domestic …rms that have FDI siblings, the scope of positive horizontal spillovers will likely be big. Please refer to Section 6.4 for the results and the Appendix for the related literature. 8 It should be noted that there are other industries in other countries that where we may expect to …nd similar results. Kee and Tang (2013) looks at the processing exports in China and …nd that the presence of FDI in a downstream industry may promote the output variety in the upstream variety, which leads to an increase in domestic value added in the downstream industry as …rms actively substituting imported materials with domestic materials. Such a result would be consistent with SSS. 5 2 Shared Supplier Spillovers and the Related Literature The intuitions of SSS can be traced back to Marshall (1920) in the classic Principles of Economics. In the book, Marshall detailed possible bene…ts of industry agglomerations, one of which is through backward linkages that the clustering of …rms increases the industry demand for specialized inputs, which may lead to “the growth of subsidiary trades,” proliferation and improvement of local input varieties and create positive externalities on all …rms that use these specialized inputs. Thus, one possible reason for FDI …rms to generate SSS on domestic …rms is by promoting the expansion of the variety of local intermediate inputs. The presence of FDI …rms, through their size and consistent orders, may increase industry demand for specialized intermediate inputs. That will attract entry of …rms into the intermediate input industries, which expands the variety of local intermediate inputs. With a love of variety for inputs technology, those domestic …rms that use these new inputs will experience a productivity boost. This mechanism is modeled in Rodriguez- Clare (1996) where all domestic …rms employ all new varieties of local inputs and, thus, share the local input suppliers with FDI …rms, which leads to spillovers that could be interpreted as SSS. My s theory.9 paper provides the empirical support for Rodriguez-Clare’ To illustrate that FDI …rms promote new local intermediate input varieties, Figure 1 plots the number of FDI …rms in the garment industry alongside the number of local input suppliers in the upstream industries in Bangladesh, from 1984 to 2003.10 The two series are closely correlated, even after controlling for the number of domestic garment …rms and a time trend.11 Thus, at an 9 A recent paper by Carluccio and Fally (2011) also provides a theoretical model showing that those domestic …rms that adopt the technology through the use of specialized local inputs demanded by the FDI …rms may bene…t from the presence of FDI …rms in the industry. 10 Data on the number of local input suppliers is constructed by searching for, on-line, the year of establishment of each of the local input suppliers provided by all the …rms in the current survey. 11 Granger causality tests also suggest that, at this aggregate level, FDI …rms Granger-cause the number of local input suppliers to increase and not the reverse. I did the following two versions of Granger Causality tests (one in level and one is detrend): F DIt = 0 + 1 F DIt 1 + 2 Supplierst 1 + " 1t Supplierst = 0 + 1 F DIt 1 + 2 Supplierst 1 + " 2t (1) F DIt = 0 + 1 F DIt 1 + 2 Supplierst 1 + 3 trendt + " 1t Supplierst = 0 + 1 F DIt 1 + 2 Supplierst 1 + 3 trendt + " 2t (2) In both versions, F-tests reject the null hypothesis that 2 = 0; with a 95% con…dence level, while failing to reject the null hypothesis that 1 = 0: 6 aggregate level, the data seems to suggest that the presence of FDI …rms helps to promote a vibrant local input industry in Bangladesh that is conducive for domestic …rms to expand their product scope and productivity. Another reason for FDI …rms to cause SSS on domestic …rms is by a¤ecting the quality and productivity of their local suppliers. The positive impacts of FDI …rms on their local supplying industries have been identi…ed in Javorcik (2004), Blalock and Gertler (2008) and Javorcik and Spatareanu (2009), based on …rm level data from Lithuania, Indonesia and Czech Republic.12 ; 13 These authors found that the overall presence of FDI in a downstream industry a¤ects the produc- tivity of the domestic …rms in the upstream industry, a …nding that has been widely referred to as the “vertical spillovers” of FDI, or the spillovers of FDI due to backward linkages. I am taking the analysis one step forward by suggesting that these better local suppliers will further improve the productivity and product scope of those downstream domestic …rms that also buy from them. Survey evidence, based on some follow up interviews with the local garment input suppliers in Bangladesh, also supports this point.14 In these interviews, FDI garment …rms are often described as being “pickier” as they demand higher quality inputs. Thus, in order to meet the higher stan- dards of the FDI …rms, these local intermediate input suppliers need to improve their quality and consistency, and to decrease the product reject rates which, inevitably, bene…t their other clients 12 These papers …nd that the productivity of upstream domestic …rms improves with the increased presence of downstream FDI …rms, which suggests that downstream FDI …rms transfer technology to their local suppliers. National industry input-output tables are used to relate upstream domestic …rms to FDI …rms in the downstream industry. Thus, all FDI …rms in the downstream industry are considered as homogenous, regardless of whether or not these FDI …rms indeed use local suppliers. Unlike these papers, the current paper has …rm level information on each FDI …rm, their speci…c local suppliers, and the domestic …rms that are also using the same local suppliers. Thus, while some FDI …rms that do not share local suppliers with domestic …rms may not generate any spillovers to domestic …rms, those FDI …rms that share local suppliers are shown to improve productivity and product scope of domestic …rms. Results of this paper could shed light as to why Javorcik (2004) …nds strong evidence of vertical spillovers with low horizontal spillovers –it could be that these FDI …rms do not share local suppliers with domestic …rms in the same industry. 13 Note that, in addition to validating vertical spillovers, Blalock and Gertler (2008) also …nd signi…cant pecuniary externalities, in terms of pro…tability of downstream domestic …rms buying from an upstream industry that supply to FDI …rms in Indonesia. In their paper, foreign …rms transfer technology to their local suppliers, which, through entry of local suppliers, leads to more competition in the supplier industry and an overall reduction in the price of local intermediate inputs. As a result, domestic …rms downstream of that supply market enjoy lower intermediate input prices which improve their pro…ts, but not necessarily their product scope and productivity. Such pecuniary externalities, while important, are not the focus of this paper and will not a¤ect the results, given that I use …rm level price indices on output and input of each …rm to de‡ ate their sales and material costs. If prices of domestic inputs are lower, due to pecuniary externalities, material costs of …rms will also be lower. De‡ ating the low material costs with the low …rm material price will not bias the quantity of intermediate inputs and, therefore, will not a¤ect …rm productivity. 14 More detailed discussion on the follow up interviews with the suppliers can be found in the Appendix. 7 who are domestic garment …rms.15 3 De…nitions In this section, I will de…ne the main variable of interest in this paper, which is Foreign Sibling Presence or FDI Sibling Presence. Given that this paper is closely related to the FDI spillovers literature, which has a long history of using Industry Foreign Presence to measure the e¤ect of FDI …rms on domestic …rms performance (see e.g. Aitken and Harrison, 1999; Javorcik, 2004), I would like to construct Foreign Sibling Presence in a similar way as Industry Foreign Presence, to allow for easy comparisons. In a nutshell, Industry Foreign Presence captures the collective in‡uence of all the FDI …rms within an industry, while Foreign Sibling Presence only focuses on the in‡uence of those FDI …rms that are related to each domestic …rms through the common local input suppliers. As a result, Industry Foreign Presence is common among all domestic …rms within the same industry while Foreign Sibling Presence varies across domestic …rms. An increase in Industry Foreign Presence could be due to increases in capital investment of some FDI …rms, while such increases will a¤ect Foreign Sibling Presence of a domestic …rm only if these FDI …rms share local input suppliers with this particular domestic …rm. Please refer to Figure 3 for an illustration of how Industry Foreign Presence and Foreign Sibling Presence are calculated in an industry with two domestic …rms, four FDI …rms, and four local suppliers. Note that, in the econometrics exercise below I will focus on relating the within …rm variation of Foreign Sibling Presence (due to the expansion or contraction of the FDI …rms) to the performance of domestic …rms over time, and not looking at cross …rm comparison of Foreign Sibling Presence which could be driven by selection bias, since if a domestic …rm a higher value of Foreign Sibling Presence than another domestic …rm, it could be because the …rst domestic …rm self-select to share local input suppliers with more FDI …rms. I will start by de…ning the Foreign presence of a …rm, which is the share of foreign equity of a …rm. Next, I sum up the foreign presence of all …rms in an industry to obtain Industry Foreign Presence. Both these variables follow the standard de…nitions in the literature. For …rm i in year 15 The SSS of FDI …rms to domestic …rms, while not equivalent, is related to the horizontal spillovers of FDI …rms to domestic …rms, if the FDI and domestic …rms are all in the same industries (Aitken and Harrison, 1999). For readers interested to know more about the literature on horizontal spillovers, please refer to the Appendix. 8 t; its Foreign Presence (F Pit ) is the product of …rm’s foreign ownership share (F Si ) and its capital share in industry j in year t; Kit F Pit = X F Si : (3) Kit i2j F Pit captures how much in‡uence the foreign capital of each …rm has in the industry, with in‡uence being measured by the share of each …rm in industry capital stock. The reason capital share is used to measure the in‡uence of a …rm in an industry is because, by construction, capital is predetermined by the investment in period t 1; unlike employment or output, which is highly endogenous to contemporary changes in …rm productivity.16 When we sum up the F Pit of all …rms in an industry, we have the Industry foreign presence (IF Pjt ) of the industry (j ) in year t, X Kit F Si X i2j IF Pjt F Pit = X : (4) i2j Kit i2j Mathematically, IF Pjt is also equivalent to averaging the foreign ownership share of each …rm in the industry, with weights equal to each …rm’s capital share in the industry. To de…ne Foreign Sibling Presence, I …rst de…ne Sibling …rms, which are …rms that share com- mon local input suppliers. Then Foreign Sibling Presence of each …rm is simply obtained by summing the Foreign Presence of all the sibling …rms of the …rms, similar to how Industry Foreign Presence is de…ned. Note that Foreign Sibling Presence will be proportionate to Industry Foreign Presence if all …rms share the same local input suppliers in an industry. However, if …rms each have a di¤erent set of local input suppliers and therefore di¤erent sibling …rms, their Foreign Sibling Presence will be di¤erent from Industry Foreign Presence. For each …rm i in year t; Let Sit be its set of local suppliers. Then two …rms i and k are considered Siblings in t if they share common local suppliers, i.e. Sit \ Skt 6= ?: Consequently, s ; equals one, if the local for every local supplier s in Sit ; 8s 2 Sit ; I de…ne a dummy variable, Sikt 16 See Appendix on construction of capital and productivity estimation. 9 supplier is also serving …rm k; i.e. s 2 Skt ; or 8 < 1; if s 2 Sit \ S s kt Sikt = : (5) : 0; if s 2 = Sit \ Skt s is a supplier speci…c sibling dummy that indicates whether supplier s is a In other words, Sikt common supplier of i and k in year t: Alternatively, let N Sikt be the total number of common suppliers between i and k in t; X s N Sikt Sikt : (6) s2Sit Then …rms i and k are siblings in t if N Sikt 1: Note that sibling …rms, in the context of this paper, have nothing to do with their ownership structure, and certainly do not imply that they share same parent …rms or are part of the same conglomerate group. Moreover, even though I only have information on the set of local suppliers for each …rm in 2003, there are two local suppliers that are only set up in the latter part of the sample period. For …rms that use these newer suppliers, s are indexed by their sets of local suppliers exhibit year to year variations. This is why Sit and Sikt t: Foreign Sibling P resence or FDI Sibling Presence (F SPit ) of …rm i in year t; is constructed by summing the Foreign Presence of all its siblings from all its local suppliers, X X X X s Kkt F Sk Sikt Kkt F Sk s Sikt X X k2j s2Sit k 2j s2Sit s F SPit F Pkt Sikt = X = X (7) s2Sit k2j Kkt Kkt k 2j k 2j X X X X Kkt N Sikt F Sk Kit F Si Kkt N Sikt F Sk N Sikt Kkt F Sk k 2j k 2j k 2j k 2j = X = X X = IF Pjt X : (8) Kkt Kkt Kit F Si Kit F Si k 2j k 2j k 2j k 2j It is also equivalent to averaging the foreign ownership share of i0 s siblings in j , weighted by each sibling’s capital share in the industry and the number of common input suppliers with the sibling in year t. Comparing (4) with (7) it is easy to see that F SPit is very similar to IF Pjt ; except that it only focuses on the foreign presence of the sibling …rms rather than all …rms in the industry. If a domestic …rm i does not share any local input suppliers with FDI …rms, (i.e. N Sikt F Sk = 0; 8k ); 10 then F SPit = 0: This is the lower bound for F SPit : On the other extreme, if a domestic …rm i shares all its local input suppliers with all FDI …rms, and there are a total of N local suppliers, then F SPit = N IF Pjt : This is the upper bound for F SPit : In general, F SPit increases with the number of local suppliers i shares with FDI …rms, as well as the presence of these FDI …rms in the industry. In short, a domestic …rm i will have a higher F SPit ; if it shares more local suppliers with FDI …rms that have larger presence in the industry. Please refer to Figure 3 for an illustration of how Industry Foreign Presence and Foreign Sibling Presence are calculated in an industry with two domestic …rms, four FDI …rms, and four local suppliers. Those domestic …rms that have at least one FDI sibling is identi…ed by a dummy variable FDI Sibling, F DISit : 8 < 1; if F SPit > 0 F DISit = : (9) : 0; if F SP = 0 it In other words, F DISit indicates wether or not …rm i0 s foreign sibling presence is positive. Note that, unlike IF Pjt , which by construction is common across all …rms in an industry in a given year, F SPit varies by …rm and time. It depends on the foreign presence of each sibling of each …rm as well as the number of common local suppliers with each sibling in each year. Moreover, while F SPit is typically less than IF Pjt ; mathematically it is possible for F SPit to be greater than IF Pjt ; if some siblings have multiple common suppliers with the …rm. Moreover, by construction, given the set of suppliers is …xed at 2003 level for the whole sample period (with the exception of 2 suppliers which only set up after 1999), the year to year within …rm ‡uctuation of F SPit is mainly due to changes in the foreign presence of the sibling …rms, and not the selection of suppliers. Hence the selection bias due to …rms switching and selecting suppliers will be controlled for by …rm …xed e¤ects in within …rm panel estimations. 4 Data Firm level survey was conducted from the period of November 2004 to April 2005, which covers a strati…ed random sample of 350 …rms, which is about 10 percent of the total population of the domestic …rms and 100 percent of FDI …rms currently operating in the Bangladeshi garment sector. The sample is strati…ed to re‡ect the population distribution of …rms by size, by industry (woven 11 garments versus non-woven garments), and by location (Chittagong, Chittagong-EPZ, Dhaka, and Dhaka-EPZ). After cleaning up the data to exclude outliers and …rms with incomplete information, there are a total of 297 …rms in the …ve year unbalanced panel data set of 1213 observations, from 1999 to 2003.17 In this unbalanced panel data set, the composition is 68 percent in woven industry and 32 percent in non-woven industry, roughly re‡ecting the population of …rms in the garment sector. Among the sampled …rms, 14 percent have positive foreign equity, while the remaining 86 percent are purely domestic owned. Table 1 presents the sample means of the key variables by woven and non-woven industries and by equity ownership. It is clear that in both industries FDI …rms are, in general, larger in sales and exports, purchase more material inputs, including imported materials, hire more employees, and have more capital. All these presumably suggest that foreign …rms are more productive. Note that, to promote the improvement and expansion of locally available inputs, it is necessary for FDI …rms to increase the industry demand for these inputs through greater local input intensity, a point emphasized greatly in Rodriguez-Clare (1996), but not necessarily more productive. Given that FDI …rms are much larger on average in the current data set, this is likely to be the case. In fact, given its size, a typical FDI …rm in the current sample source 83 percent more local inputs than domestic …rms, even though only 20 percent of their inputs are locally supplied, while the comparable …gure for a domestic …rm is 32 percent. More evidence on FDI …rms increase the industry demand of local intermediate inputs in Bangladesh is presented in Section 6.4, where we examine the local intermediate input intensity of FDI …rms. This is essential to support SSS, whereby, the presence of foreign …rms signi…cantly increases the industry demand for local inputs, which may then lead to quality upgrading and variety expansion in the intermediate input industry. Table 2 presents the sample means of industry foreign presence and FDI sibling and foreign sibling presence of domestic …rms in the sample, by industries. On average, there is more foreign presence in the woven industry than in the non-woven industry, judging by their industry foreign presence, although the di¤erence is only about 8 percentage points. The contrast is larger between two industries when focussing on the siblings. On average, 52 percent of domestic …rms in the woven industry have FDI siblings, while only 16 percent of domestic …rms in the non-woven industry 17 In the survey, the accounting managers of the …rms were asked to provide …nancial information from 1999 to 2003, which is what I used to conduct the analysis. Some of these …rms stopped production and exited the industry during the sample period but have resumed production in 2004. These …rms are the exiters in the sample. 12 have FDI siblings. Furthermore, the average foreign sibling presence in the woven industry is 6.1 percent, nearly 10 times higher than that of the non-woven industry. This is true even if I restrict the comparison to only those domestic …rms with FDI siblings. The foreign sibling presence for domestic …rms with FDI siblings in the woven industry is 11.3 percent, while the same variable for the non-woven industry is only 3.3 percent. Di¤erences between the two industries may be driven by other industry level variables, such as trade policies and demand shocks. I will, instead, rely only on the within …rm variations in foreign sibling presence in the regressions. 5 Reduced Form Regression Results The two industries in the garment sector of Bangladesh, namely, woven and non-woven have very di¤erent production structures and techniques. The Appendix discusses how two separate industry speci…c production functions are estimated using Ackerberg et al (2007) in a three step procedure that takes into account endogeneity of labor and material inputs, and how input and investment decisions may depend on the FDI status of …rms. This technique is similar to that of De Loecker (2007), to allow production function to depend on exporter status. Here, I focus on relating foreign sibling presence to product scope and sales per worker, output per worker and the estimated …rm productivity. The Appendix also contains a discussion on the construction of the very crucial …rm speci…c price indexes. Output and material inputs of …rms used in the production function estimation are constructed by de‡ating total revenue and cost of materials using these …rm speci…c output and material price indexes. This signi…cantly improves what Haskel, Pereira and Slaughter (2007) describe as a “pervasive problem in the literature on micro panels ”that uses industry prices in place of the often missing …rm level prices. 5.1 Shared supplier spillovers: Do FDI …rms improve the performance of their domestic siblings? To identify SSS, I relate the performance of domestic …rms to their foreign sibling presence, as de…ned in equation (7) ; in a domestic …rm only panel data set: ln yit = i + jkt + F SP F SPit + Xit + i T rendit + it ; (10) 13 where the dependent variable (yit ) includes the product scope, sales per worker, output per worker, TFP estimated via OLS and TFP estimated via augmented Olley-Pakes procedures of domestic …rms in our sample. A positive estimate of F SP suggests the performance of domestic garment …rms is enhanced, due to the increased presence of their FDI siblings. I control for …rm …xed e¤ects, i; in the panel regressions, (10) ; and only rely on the within …rm variations of performance and foreign sibling presence to identify the coe¢ cient. In other words, between-…rm productivity changes, such as the exiting of ine¢ cient …rms as the market toughens due to the increased presence of FDI …rms, whilst important, should not a¤ect or explain the within coe¢ cient on foreign sibling presence. In addition, it is important to highlight that, since I only rely on the within …rm variations of performance and foreign sibling presence to identify the coe¢ cient, factors and characteristics of …rms, such as …rm size (large vs small …rms) and …rm productivity (more productive vs less productive …rms) will be controlled for and will not a¤ect the estimation of F SP . As previously mentioned, …rm …xed e¤ects also controlled for the set of suppliers which is kept constant at 2003, and the within …rm movement of F SPit only re‡ects the changes in foreign presence of the sibling …rms of each domestic …rms. Selection of suppliers, as well as the number of local suppliers, while important, will not a¤ect the within estimations, since by construction, within …rm movement of F SPit is not a¤ected by the selection of suppliers. In addition, …rm speci…c time trend, T rendit ; is also included to soak up any …rm trend on ln yit and F SPit that is speci…c to each …rm. This would include some …rm speci…c exogenous shocks, such as a shock to the common suppliers, that move ln yit and F SPit up a time trend, simultaneously. In other words, i will pick up the e¤ect of F SPit on ln yit if for whatever reason …rms that perform better over time (rising ln yit ) happen to have F SPit that is either rising or falling over time. For example, a domestic …rm may receive a positive productivity shock by hiring a better manager, and the manager sources from a better local supplier that also serves FDI …rms with expanding market shares, T rendit will be able to pick this up and will not contaminated the estimation of F SP : In short, with T rendit and …rm …xed e¤ects, only the non-trending within …rm co-movement of F SPit and ln yit is used to identify F SP . 14 5.1.1 Omitted variable bias Equation (10) controls for industry-location-year speci…c e¤ects, jkt ; to wipe out any macro omit- ted variables which are common among all …rms within the same industry, location, and year and which may a¤ect the performance of domestic …rms and foreign sibling presence. Such variables may include industry speci…c demand and productivity shocks, government policies that favor do- mestic …rms, investment climate change in the export processing zones, or trade policy changes of the main markets such as the EU and the US. Equation (10) also controls for industry foreign presence and the resulting market competition speci…c to an industry in a given year. In addition, …rm level control variables, Xit , are also included, which are age, the share of imported materials in total material cost, and the share of materials in total sales. This is because overseas buyers may request Bangladeshi …rms to use imported fabrics to ensure quality of the …nal products. Such business practices are typical among …rms that export to the US and these …rms could be more productive as the US market is more competitive. Using imported fabrics decreases the demand for domestic materials, which may decrease the number of FDI siblings and cause the within …rm year to year change in foreign sibling presence to be smaller and, in turn, inducing a downward bias on the coe¢ cient for foreign sibling presence. To control for this, the share of imported materials in the total materials of …rms is included. Another possible omitted variable pertains to production techniques. Ine¢ cient …rms tend to waste material, which leads to a high material-to-sales ratio. The more materials a …rm uses, the more likely it is that this …rm has more FDI, as they may demand more domestic materials. This leads to larger within-…rm year to year change in foreign sibling presence among unproductive …rms that have high materials-to-sales ratio, which in turn leads to a downward bias on the coe¢ cient on foreign sibling presence. Equation (10) also controls for materials-to-sales ratio in the regression. Finally, the age of a …rm may also bias the estimate. Speci…cally older …rms tend to be more productive, and older …rms tend to work with the more established local suppliers, which could be also preferred by FDI …rms. This causes an upward bias on the coe¢ cient of foreign sibling presence. 5.1.2 Selection bias, endogeneity and reverse causality While the beauty of using foreign sibling presence is that F SPit is …rm speci…c and time varying, which allows us to control for industry-location-year …xed e¤ects to wipe out the in‡uence of the 15 macro variables, the short coming of using F SPit is, also that, it is a …rm level variable that is subjected to selection bias, endogeneity, reverse causality and measurement errors. One may worry that, as a domestic …rm performs better over time, it may choose to buy from local suppliers that also work with FDI …rms. Such self-selection will cause an upward bias in the least squares estimate of F SP : However, given that I only use local supplier information in 2003 to construct F SPit ; within …rm variation of F SPit is driven by changes in the market presence of FDI siblings, rather than through adding or dropping of local suppliers. Thus, while potentially large or more productive domestic …rms may self-select to buy from local suppliers that also serve FDI …rms, given the nature of within …rm estimation and the …xed set of local suppliers for each …rm, selection bias may not be too relevant.18 In other words, within estimations limit the comparison of …rm performance to be within …rm over time, so the issue of cross …rm comparison of performance driven by …rm size while could be important is not relevant in this context. There is also a concern that if a local supplier becomes exogenously better (which could be because it is a FDI …rm), it improves the performance of all its clients, and some FDI client …rms may expand their market presence, as a result. Such simultaneity problems will also cause an upward bias in the least squares estimate of F SP : On the other hand, as a domestic …rm becomes better over time, it may expand its own market share, causing the market share of FDI …rms to decrease and lead to a smaller F SPit . In other words, within …rm performance changes may cause F SPit to change. This reverse causality will result in a downward bias in the least squares estimate of F SP : Another source of downward bias in the least squares estimate of F SP is measurement error in F SPit . Given that I only use local supplier information for 2003 to construct F SPit , local suppliers are assumed to stay the same through our the sample period. While this limits the size of select bias once …rm …xed e¤ects are used, may nonetheless introduce measurement errors in F SPit that potentially may bias the least squares estimate of F SP towards zero. The overall bias in least squares estimate of F SP is not clear, it depends on whether or not 18 Note that, it is in fact not that case that large or more productive domestic …rms self select to buy from local suppliers that also serve FDI …rms. The reader may be reassuring to know that when we regress the productivity of domestic …rms on their foreign sibling status, it shows that …rms that have foreign siblings are signi…cantly less productive than …rms that do not have foreign siblings, controlling for industry-location-year …xed e¤ects and other …rm characteristics such as the share of the US and the EU market in total export, and the share of imported material in total material. These domestic …rms that have FDI siblings are also not signi…cantly larger than other domestic …rms that do not have FDI siblings. These results are available upon request due to space constraints. 16 reverse causality and measurement errors dominate selection and endogeneity biases, while omitted variable bias will have ambiguous e¤ects on the least squares estimate of F SP . One way to address all these issues is to introduce more control variables, and by using lag F SPit to limit the contemporaneous spurious relationship between within …rm performance and F SPit : Here I exploit an unanticipated change in the EU trade policy and use it to build an instrumental variable (IV) that has within …rm non-trending exogenous changes for F SPit : Note that the results based on more control variables and lag F SPit are very similar to the IV estimation and are not reported here due to space constraint but is available upon request. 5.1.3 Instrumental variable estimations In 2000, the EU announced that it will implement the “Everything-But-Arms”(EBA) initiative in 2001, which provides duty-free, quota-free access to imports from all 48 Least Developed Countries, Bangladesh being one of them. However, to enjoy such trade preference, rules of origin (ROOs) requirements of the products must be met. There are two sub-industries within the garment sector of Bangladesh, one consisting of …rms producing woven apparels and the other consisting of …rms producing non-woven apparels, such as knitwear and sweaters. These two industries have very distinct production techniques and, while any of the nonwoven apparel producing …rms can easily satisfy ROOs, only the larger woven …rms, many of which are FDI …rms, …nd it pro…table to meet ROOs by using local inputs that are, typically, more expensive.19 Thus, the announcement of EBA in 2000 prompted di¤erential impacts on the investment and capital share of the …rms, depending on the sub-industry they are in and whether or not they export to the EU. In other words, the announcement of EBA in 2000 prompted the woven FDI …rms that export to the EU to investment and expand their market share, and at the same time increase their demand for local inputs to meet ROOs. Figure 2 presents the share of FDI …rms in the industry capital. While FDI …rms that export to the EU, generally, have a larger presence in the industries relative to FDI …rms that do not export to the EU, the presence increases only in the woven sub-industry. The news of EBA caused the market share of FDI …rms that export to the EU to increase from 38 percent in 1999 to 43 percent in 2000 and stabilized to 42 percent in 2003. Conversely, the share of those 19 Please refer to Demidova, Kee and Krishna (2012) for a discussion of how ROOs of the EU add an additional layer of …xed and marginal costs for …rms exporting to the EU. 17 FDI …rms in the woven sub-industry that do not export to the EU dropped from 6 percent in 1999 to 0.7 percent in 2000, and barely increased to 1.6 percent in 2003. On the other hand, market presence of FDI …rms in the non-woven sub-industry did not follow this pattern. Such distinct movements of market shares among di¤erent FDI …rms in di¤erent sub-industries were a result of an unanticipated exogenous policy change in the EU that may have a¤ected foreign sibling presence of some domestic …rms. I will use the impact of the EBA announcement on the market presence of those FDI …rms in the woven industry as an instrument for the exogenous increase in the foreign sibling presence. The exclusion restriction here is that the announcement and implementation of EBA has no direct impact on the productivity of domestic …rms. This exclusion restriction is motivated by the …ndings in the literature that, while the more productive …rms may self select into exporting, further exporting may not have feedback e¤ects on the productivity of exporters (Clerides, Lach and Tybout, 1998; Bernard and Jensen, 1999). However, some recent papers …nd that exporting may further promote productivity gains (Van Biesebroeck, 2006; De Loecker, 2007; Fernandes, 2007). As a robustness check, I run both the …rst stage and second stage IV regressions on a subset of domestic …rms that do not export to the EU, given that in this case, trade policy of the EU should not directly a¤ect the productivity and performance of these …rms. In addition, given that EBA a¤ects all …rms that export to the EU, thus the expansion of market shares may not limit to the FDI …rms. In the next section on robustness check, I also examine how domestic …rms may be bene…tted from their domestic siblings that expand their market shares as a result of EBA. Please refer to the next section, and particularly footnote 18 for this discussion. Table 3 presents the …rst stage estimations, where I regress F SPit on a dummy variable which equals one if domestic …rm i has a FDI sibling that exports to the EU in year t; F DIS _EUit ; and the triple interaction term between F DIS _EUit ; an EBA dummy that equals to one for 2000 onwards and a woven industry dummy variable: F SPit = 1 F DIS _EUit + 2 F DIS _EUit woveni EBAt + Zit + it ; (11) where Zit has all the right-hand side variables of (10) ; which includes …rm …xed e¤ects, …rm speci…c time trend, …rm level control variables, and industry-location-year …xed e¤ects. I expect 1 and 2 to be positive, which would suggest that, conditional on domestic …rm i having a FDI sibling that 18 is exporting to EU in year t; foreign sibling presence of …rm i is higher if …rm i is in the woven industry, in the years following the announcement of EBA. Column (1) of Table 3 presents the results based on a subset of domestic …rms that do not export to the EU, and Column (2) shows the …rst stage regression based on the full sample of domestic …rms who may or may not export to the EU. Given that the instrumental variables only vary by industry and year, I cluster the standard errors by industry-year in all the columns. The estimated 1 and 2 are positive and statistically signi…cant, with F-statistics that are greater than 10, suggesting that these instrumental variables have explanatory power on F SPit . Tables 4 and 5 present the second stage regressions according to (10), for the restricted sample of domestic …rms that do not export to the EU and for the full sample of domestic …rms that may or may not export to the EU. These tables also present the least squares estimations and compare them to the IV estimates. In both the tables, the IV estimates for F SP are larger than the LS estimates, suggesting that the downward biases, due to reverse causality between the performance of domestic …rms and their foreign sibling presence, as well as measurement errors in F SPit , dominate the upward biases due to selection and endogeneity. For the restricted sample of domestic …rms that do not export to the EU, an exogenous increase in foreign sibling presence due to EBA causes these domestic …rms to have better performance in terms of a higher product scope, sales per worker, output per worker, and TFP (estimated with OLS and the augmented Olley Pakes procedure). This is the sample of …rms, upon whose performance EBA should not have had a direct impact, other than through their FDI siblings that export to the EU, thus satisfying the exclusion restriction. These results are very similar in the full sample of domestic …rms. In summary, by exploiting exogenous changes in foreign sibling presence due to EBA, I show that domestic …rms bene…tted from the increased presence of their FDI siblings, a result that is driven by the improved access to better and new local input variety as the FDI garment …rms push up their demand for local inputs. Based on the estimates in Columns (2) and (10) of Table 5, a one percent increase in foreign sibling presence is associated with a 1 percent gain in product scope and 3 percent gain in productivity for domestic …rms on average. From 1999 to 2003, the average within …rm gain in product scope and productivity among domestic …rms is about 4 and 8 percent, respectively, while the average change in foreign sibling presence is 1 percent. A back-of-an-envelope calculation would then suggest that the increase in foreign sibling presence throughout the sample 19 period could explain about a quarter of the within …rm product scope expansion and a third of the within …rm productivity gains. These results are important, statistically and economically. 6 Robustness Checks 6.1 Alternative Interpretations Could the above results be driven by the linkage between FDI and domestic …rms when they produce the same products or export to the same market, and not necessarily due to SSS? To study these other channels, I construct two variables to capture the market presence of those FDI …rms that share common products or common market with each domestic …rm. Speci…cally, product foreign presence (P F Pit ) of each domestic …rm i in industry j and year t is de…ned as the following: X X Kkt p X Kkt X p P F Pit F Sk Rik = F Sk Rik ; (12) Kjt Kjt p2Pi k2j k 2j p2Pi p where Pi is the set of products (HS 6 digit goods) for i in t; and Rik is a dummy variable which p equals one if i and k both produce product p. Note that there is no time index for Rik since I only have information of the product mix of …rms in 2003. So P F Pit for each …rm i is the weighted average of the market presence of all the FDI …rms that have common products with i in industry j; with weights re‡ecting their shares of capital in j and the number of common products with i: Similarly, market foreign presence (M F Pit ) of each domestic …rm i in industry j and year t is constructed as the following: X X Em kt m M F Pit = m F Sk Rikt ; (13) Ejt m2Mit k2j m is the value of export of …rm k to market m in where Mit is the set of export markets for i in t; Ekt m is the total value of export of industry j of Bangladesh to market m in year t; and Rm year t; Ejt ikt is a dummy variable which equals one if i and k both export to market m in year t: Table 2 presents p m by industry. There are about 90 percent of the sample average of P F Pit ; M F Pit ; Rik and Rikt domestic …rms that have at least one common product with a FDI …rm and more than 97 percent of domestic …rms that have common output markets with FDI …rms. This is not too surprising 20 since most …rms export to the EU, the US or both, and produce similar products. Relative to foreign sibling presence, product and market foreign presence are also signi…cantly higher, which potentially may explain more of the within …rm productivity gains over the sample period. Alternatively, could domestic …rms bene…t from their domestic siblings? To understand this, I construct the following domestic sibling presence variable (DSPit ) for each domestic …rm i in year t: X X s DSPit (1 F Pkt )Sikt : (14) s2Sit k2j Finally, could SSS merely capture the large …rm impact on suppliers and not necessarily because these …rms are FDI …rms? Given that FDI …rms are on average much larger than domestic …rms within an industry, it is plausible that our empirical results are driven by …rm size and not the FDI linkages. Please note that some domestic …rms are just as large as the FDI …rms within an industry. If …rm size is the driving reason behind the …nding of SSS, we would expect domestic sibling presence variable to have some explanatory power on domestic …rms’performance. However, to properly address this, I construct the following …rm size sibling presence variable (F SSPit ) for each domestic …rm i in year t; which capture the market shares of all …rms (both domestic and foreign sibling …rms) that each domestic …rm share a local input suppliers: X Kkt Saleskt s F SSPit = X X Sikt : (15) s2Sit Kkt Saleskt k 2j k 2j F SSPit is similar to F SPit ; but rather than focusing on the foreign ownership share, F SSPit uses the market shares of each sibling …rms as weights. Thus, if …rm i has a large sibling …rm, FDI or Saleskt domestic, its market share X will be large and have a large in‡uence on F SSPit : Saleskt k 2j Columns (1) to (4) and (6) to (9) of Table 6 present the regression results when I relate product foreign presence, market foreign presence, domestic sibling presence and …rm size sibling presence to product scope and TFP of domestic …rms. In all cases, these other possible channels are not statistically signi…cant, suggesting that the performance of domestic …rms does not improve simply because they share a common product or market with FDI …rms, or when they share common 21 local suppliers with other domestic …rms, or any large …rms.20 In particular, the fact that both domestic sibling presence and …rm size sibling presence are not statistically signi…cant suggest that our …nding of SSS is not driven by large …rms.21 It should also be noted that when F SPit is included in these columns, the coe¢ cients on F SPit remain positive and signi…cant, similarly to the estimates listed in Columns (1) and (9) of Table 5. So including these other possible channels do not a¤ect the …ndings of SSS. Due to space constraints, this set of results are not reported but is available upon request.22 6.2 Placebo experiment –random siblings Another concern could be that the sibling relationship is, somehow, random and the previous result is just coincidental. There is also a concern that using the top supplier list of 2003 to estimate foreign sibling presence for the previous years, I may introduce large and systematic measurement errors that may a¤ect the regression results. Finally, for each …rm, I only have information on their top three local suppliers, which may leave out other local suppliers if the …rms have more. To address these concerns, I use an arti…cial foreign sibling presence variable, constructed when domestic …rms are randomly assigned FDI siblings. I repeated this placebo experiment 50 times and report the mean and standard deviation of the estimated coe¢ cients from these 50 regressions in Columns (5) and (10) of Table 6. The randomized foreign sibling presence does not have any explanatory 20 In addition, we also use EBA as an IV for domestic sibling presence, given that when EBA was announced in 2000, both FDI and domestic …rms may a¤ect their investment decision which a¤ects the foreign sibling presence and domestic sibling presence of the same domestic …rms. In other words, we have two “treated groups,” and it is important to show that sibling foreign presence did not pick up the in‡uence of domestic sibling presence. In both the restricted sample of non-EU exporters and the full sample, domestic sibling presence is not statistically signi…cant in the second stage across all dependent variables. In the …rst stage, while the interaction term is positive and signi…cant, domestic sibling that export to the EU is not statistically signi…cant. When I also include FDI siblings that export to the EU in the …rst stage, together with its interaction term with EBA and woven dummies, only the FDI siblings variables are positive and signi…cant and the domestic sibling variables have wrong sign and insignicant. Overall, this result shows that domestic siblings do not have a positive impact on the performance of other domestic …rms and only FDI siblings are important. 21 It should also be noted that in the survey data, each …rm also reports whether their top manager has worked in a FDI …rm prior to joining the …rm. I tried to explore whether there is any spillover from FDI …rms to domestic …rms through such movement of workers. I failed to …nd any statistically meaningful results. Across …rms, those that have top manager with FDI experience do not out perform …rms that do not have top manager with FDI experience. Within …rm, there is no robust evidence indicating those FDI trained managers make their …rms grow faster. Due to space constraint, these results are not reported but are available upon request. 22 Note that another alternative hypothesis could be that foreign sibling presence captures the e¤ect of “exporter sibling presence,”–may be by sharing local suppliers with an exporter, domestic …rms may perform better. However, given that all …rms in our sample are exporters, there will be no meaningful variation in the exporter sibling presence variable to explain the performance of domestic …rms. 22 s product scope and productivity. This is in sharp contrast to the previous power in a¤ecting …rm’ …nding where foreign sibling presence is consistently important in explaining …rm performance. This suggests that the previous …nding may not have been a ‡uke, and the measurement errors based on supplier information of 2003 may not have been empirically important, and the list of top three suppliers may just be enough to relate the …rms. 6.3 Alternative Measure of FDI In‡uence In this section, we explore using another variable to capture the in‡uence of FDI siblings on domestic …rm performance, which is the fraction of a …rm’s top 3 local suppliers that also sell to FDI …rms (FDIsuppliersh it ).23 If a domestic …rm has more local suppliers that also sell to FDI …rms, it is likely that this domestic …rm will bene…t more from the presence of these FDI siblings. Similar to F SPit , this variable is endogenous, which we will instrument with the EBA variables the same way as we have instrumented for F SPit , (see Section 5.1.3). Table 7 presents the empirical results for the full sample of domestic …rms, including the highly signi…cant …rst-stage F-statistics at the bottom of the table. Results based on the restricted sample of domestic …rms who do not export to the EU are very similar and are available upon request. All regressions control for …rm …xed e¤ects, …rm-speci…c time trend, industry-location-year …xed e¤ects, age, the share of intermediate inputs in total sales and the share of imported intermediate input in total materials, similar to Table 5, with standard errors clustered by industry-year. The results are qualitatively very similar to the previous results using F SPit , which is not surprising, since FDIsuppliersh it and F SPit are highly correlated in the sample, both within …rms as well as across …rms.24 Overall, a within s performance, in terms …rm increase in FDIsuppliersh it is positively associated with better …rm’ of product scope and productivity. In particular, according to the IV estimate in Column (8), a one standard deviation increase FDIsuppliersh it causes …rm’s productivity to improve by nearly 30 percent, which is signi…cant statistically as well as economically. These results are consistent with the previous …ndings based on F SPit ; which support the hypothesis of shared supplier spillovers of FDI …rms. 23 I thank the editor, Nathan Nunn, for this suggestion. 24 Results available upon request. 23 6.4 Horizontal Spillovers What can the …nding of shared supplier spillovers of FDI …rms within the same industries, while not equivalent, inform us about the overall horizontal spillovers of FDI …rms? Here I follow the standard approach in identifying horizontal spillovers, by relating the productivity of domestic …rms to the industry foreign presence of all FDI …rms in the same industry:25 ln T F Pit = i + t + IF P IF Pjt + Xit ; where the log of TFP of domestic …rms, ln T F Pit ; is regressed on the industry foreign presence, IF Pjt ; controlling for …rm and year …xed e¤ects, i and t; as well as industry controls for demand shocks (tari¤s set by the EU and US on the garment imports from Bangladesh), and productivity shocks (the average industry level TFP). Given that IF Pjt is time varying industry speci…c, I no longer can control for industry-location-year …xed e¤ects, and it is necessary to cluster the standard errors by industry-year to avoid the classic macro-variable-in-micro-unit problems (Moulton, 1990). Table 8 presents the results based on industry foreign presence. These results are more com- patible with the existing literature in terms of methodology. Column (1) shows that, controlling for productivity and demand shocks, an increase in industry foreign presence tends to depress the productivity of domestic …rms. This strong negative result echoes the existing …ndings of Aitken and Harrison (1999). Columns (2) and (3) split the sample of domestic …rms into those that share local suppliers with FDI …rms and those that do not. Here, one can see some contrasting picture. While the e¤ect of industry foreign presence on the productivity of domestic …rms that have foreign siblings is ambiguous –it has a positive sign but is statistically insigni…cant, those domestic …rms that do not share local suppliers with the FDI …rms clearly su¤er from the increased presence of FDI …rms in the industries. In other words, productivity spillovers depend on whether or not the domestic …rms are connected to FDI …rms through local suppliers. In Columns (4) to (5) I further replace industry foreign presence with its two components, DS N DS IF Pjt = IF Pjt + IF Pjt ; 25 Please refer to the Appendix for a discussion on the related literature on this area. 24 DS is the industry foreign presence of those FDI …rms that have domestic siblings. where IF Pjt N DS is the These are the FDI …rms that share local suppliers with domestic …rms. Likewise, IF Pjt industry foreign presence of those FDI …rms that do not have domestic siblings. These are the FDI …rms that do not share local suppliers with domestic …rms. Column (4) shows that, for those domestic …rms that have foreign siblings, the increased industry presence of FDI that share local suppliers with domestic …rms clearly improve their productivity. In other words, positive horizontal spillovers happen between domestic …rms and FDI …rms that share local input suppliers. This result is consistent with our previous …ndings of shared supplier spillovers. On the other hand, Column (5) shows that the presence of both types of FDI …rms unambiguously hurt those domestic …rms that do not share local suppliers with FDI …rms. Presumably, this is because, without a channel for productivity spillovers, the presence of FDI …rms only hurts the market share of these domestic …rms and forces them to operate ine¢ ciently. This validates the market-stealing …nding of Aitken and Harrison (1999). Overall, the results suggest that di¤erent domestic …rms may be a¤ected very di¤erently from the presence of FDI …rms. While the domestic …rms that are not connected to FDI …rms through local suppliers mostly hurt by the presence of FDI …rms in the industry, those domestic …rms that share local suppliers with FDI …rms unambiguously bene…t from the presence of these FDI …rms. The overall e¤ect on the industry depends on the proportion of domestic …rms that have FDI siblings. In the current data set it is 36 percent. The strongly negative e¤ect of industry foreign presence gradually fades away when the proportion of domestic …rms that have FDI siblings increases. In other words, shared supplier spillovers may not show up at an industry level as horizontal spillovers if the proportion of domestic …rms with FDI siblings is small. Thus, this result may shed light as to why most papers in the past do not …nd strong supports for horizontal spillovers. First, those FDI …rms may not have backward linkages which foreclose the possibility of any productivity spillovers to domestic …rms in the same industry. Second, the proportion of domestic …rms that share local suppliers with FDI …rms could be low, which limits the in‡uence of FDI …rms on the domestic industry. In other words, the …nding of shared supplier spillovers can still be consistent with overall negative horizontal spillovers in the industry, which may consist of a large proportion of domestic …rms that do not have FDI siblings. 25 6.5 Greater Local Input Intensity of FDI Our analysis thus far shows that the presence of FDI in the garment industry leads to higher productivity and product scope of domestic …rms in the same industry when these …rms share the common local input suppliers. For such results to make sense, it is necessary to show that FDI …rms have a greater local input intensity than an otherwise identical domestic …rms. Speci…cally, according to Proposition 1 of Rodriguez-Clare (1996), “if the linkage coe¢ cient of multinationals is higher than the linkage coe¢ cient of domestic …rms, then an increase in M (NEs) generates an increase in n and w (more bene…cial to host country).” The linkage coe¢ cient is de…ned as “the level of employment generated in the intermediate-goods sector per unit of labor hired directly by domestic …rms and multinationals, respectively.” In other words, for each unit of labor hired by the FDI …rms, if it generates more employment in the upstream sector due to the use of domestic materials, then the linkage coe¢ cient of FDI …rms is considered higher than that of the domestic …rms, and may lead to greater variety of domestic materials and bene…t other domestic …rms that use these new varieties of domestic materials.26 To exam the hypothesis that in the garment sector of Bangladesh, the linkage coe¢ cient of FDI …rms is higher than that of domestic …rms, I constructed two variables. The …rst is domestic material per total employment of each …rm and the second is the share of domestic material in total material of each …rm. Table 9 presents the empirical evidence. Controlling for industry- location-year …xed e¤ects, …rm characteristics such as age, size (log of output, log of material and log of employment), capital intensity (capital-labor ratio) and export markets (export share of EU and US as EU required more domestic materials to satisfy ROOs), FDI …rms on average use more domestic materials per unit of employment, and has a higher domestic material share in total materials than otherwise identical domestic …rms. To the extent that higher domestic material per employee and higher domestic material share in total materials for FDI imply higher employment generated in the intermediate goods sector per unit of labor hired directly by the FDI …rms, then the regression results suggest that FDI …rms are bene…cial to the garment sector of Bangladesh as their presence lead to more variety of domestic materials being produced. The results are robust to whether I only focus on cross sectional comparisons (Columns (1), (2), (5) and (6)), or allowing 26 I thank Gordon Hanson for suggesting this test. 26 for time series panel correlation (Columns (3), (4), (7) and (8)). These results suggest that in the garment sector of Bangladesh the domestic material intensity of FDI …rms is indeed higher than that of domestic …rms, which satis…es the su¢ cient condition of Proposition 1 of Rodriguez-Clare (1996) and is therefore consistent with the …nding of shared supplier spillovers from FDI …rms to domestic …rms.27 ; 28 Due to space constraint, in the Appendix, I present and estimate a simple model directly links the local intermediate input variety to product scope expansion and productivity gain of domestic …rms when local intermediate input variety depends on the presence of FDI …rms in the industry. Even though the model is highly stylized, it is consistent with the previous results of the reduced form regressions and support the …nding of SSS. 6.6 Structural Regressions To formally study the role of FDI in promoting the variety of local input which causes productivity of domestic …rms to increase, I rely on the following structural model, motivated by Ethier (1982) and Rodriguez-Clare (1996). There are two sectors in the economy, a di¤erentiated intermediate input sector, producing N variety of input, mn ; n = 1; :::; N; and a di¤erentiated …nal goods industry, producing output Y; based on a production function which depends on labor, L; capital, K; and all the intermediate inputs, mn ; with a constant elasticity of substitution, > 1 among the di¤erent varieties of intermediate input. The …nal goods industry has i = 1; :::; I …rms, and some of these …rms are FDI …rms. The number of FDI …rms are exogenously given in the model (regulated by the government). Speci…cally, a typical …rm i in the …nal goods sector has the following production function (year subscript omitted), " N # M X 1 1 Yi = i mni Li L Ki K : (16) n=1 27 Please note that in the previous version of the paper, I have compared the level of domestic materials of FDI …rms to that of domestic …rms(Table 10). I wasn’ t trying to compare the linkage coe¢ cient of FDI to that of domestic …rms. I was just trying to show that FDI …rms use more domestic materials in “level” , not necessarily intensively. Even if the result on Table 10 could be a¤ected by …rm size, it does not support or invalidate the results I have just discussed which are based on the linkage coe¢ cient of …rms. In addition, I have also included Firm Size Sibling Presence as a robustness check in Table 7, Columns (3) and (7), to show that the result of shared supplier spillovers of FDI is not driven by …rm size. 28 The reasons why FDI …rms in the garment industry of Bangladesh use local intermediate input more intensively is beyond the scope of this paper. 27 In a symmetric equilibrium where mni = mi ; (16) can be rewritten as M M Yi = iN 1 Mi Li L Ki K ; (17) where Mi = N mi ; is the total amount of intermediate inputs used in the production of Yi : Holding Mi …xed, (17) shows that an increase in N raises Yi : Taking logs on both sides of (17) ; and de…ning the total factor productivity (TFP) of …rm i as the following: ln T F Pi ln Yi M ln Mi L ln Li K ln Ki ; then it is clear that an increase in N will raise i0 s TFP, given that >1: M ln T F Pi = ln i + ln N: (18) 1 In an open economy, the total variety of intermediate inputs available for the …nal goods sector is the sum of the locally produced variety, N D ; and the imported variety, N I ; N ND + NI; which implies that an increase in the local variety of input will increase the productivity of the …nal good sector, M ln T F Pi = ln i + ln N D + N I : (19) 1 In equilibrium, N D depends on the aggregate demand of the …nal good industry, which could increase due to the entry of FDI …rms, N D = f (F DI ) : (20) Equation (19) presents the structural relationship between …rm productivity and the number of input variety. This equation can be easily estimated, based on data for the number of local and imported inputs. I proxy N D using the number of local input suppliers and N I based on the number 28 of imported intermediate inputs variety.29 The sum of the number of local input suppliers and the number of imported input variety gives me N: Given that the number of local input suppliers and the number of imported input variety probably measure N D and N I with errors, the least square estimate could have a downward bias. Columns (1) and (2) of Table 10 present the least squares results. Column (1) ignores the number of imported input variety, N I ; and only focuses on the relationship between productivity and local input variety, while Column (2) includes both local and imported input variety in the regression. Firm …xed e¤ects are used to control for ln i; and given that N D and N are common across all …rms within a year, the standard errors are clustered by year. Both columns show that there is a positive and signi…cant relationship between the productivity of domestic …rm and the number of input varieties. However, these results can be downward biased given that N D and N are measured with errors. To show, empirically, that an increase in the number of FDI …rms in the …nal good industry may lead to an increase in the number of local input variety and, thus the TFP of domestic …rm in the …nal sector, I instrument N D ; using the number of FDI …rms in a …rst stage regression. In addition, I use the international price of cotton fabrics as an instrument for N I :30 Here, the exclusion restriction is that the total number of FDI …rms in the Bangladeshi garment sector is exogenous and has no direct impact on the productivity of individual domestic …rms other than through local intermediate inputs. This is justi…able given that during the sample period, foreign investment in the garment sector was highly restrictive under the Bangladesh Investment Policy (1999). While existing FDI …rms may invest and expand their capacity, new FDI …rm entry was tightly regulated by the government, which makes the total number of FDI …rms de facto exogenous during the sample period. From 1999 to 2003, only 6 new FDI …rms were established, all were green …eld investments, and half were new additions to the existing FDI conglomerates in Bangladesh from South Korea. Comparatively, 54 percent of the domestic …rms in the data were set up during the sample period, suggesting that the number of FDI …rms were, in fact, tightly controlled by the 29 In the survey, …rms report the HS 6 digit codes for the inputs they used for production. For each of these HS 6 digit inputs, I consider imports from di¤erent countries as di¤erent varieties. I match these HS codes with Bangladesh bilateral import data from Comtrade to construct the number of unique imported input variety for each year, from 1999 to 2003, respectively. 30 The international price of cotton fabrics is constructed using the unit value of Indian’ s export of cotton fabrics to the world, according to data from Comtrade. 29 government. In other words, the exclusion restriction assumes that the productivity and product scope of any one domestic owned …rms do not a¤ect the number of FDI …rms in the industry, which is determined only by the government. On the other hand, the world price of cotton fabrics clearly should not a¤ect the productivity of domestic garment …rms other than through its negative impact on imported fabrics variety. Columns (3) and (4) of Table 10 present the second stage results. The IV estimates are both positive and statistically signi…cant. These estimates are also larger than the least squares estimates, suggesting that the IV estimates are better in handling measurement errors in N D and N I : Due to the space constraint, the …rst stage results are not reported but is available upon request. Instead, I report the …rst stage F statistics in Table 10. In both cases, the …rst stage regressions have F statistics that are greater than 10 and the coe¢ cients have the expected signs. These results con…rm that an increase in the number of FDI …rms raises the number of local input variety and the total input variety, which leads to higher productivity for domestic …rms. To study the e¤ect of FDI on product variety of the …nal goods sector, consider that for each …rm i; Yi represents a composite output of di¤erent …nal good varieties, " Vi # X 1 1 Yi = yvi ; < 0: (21) vi =1 One can think of Yi as the production possibility frontier (PPF) of …rm i (e.g. GAP, Old Navy), and each …rm i produces many varieties of the …nal good (e.g. T-shirts, sweaters). The concavity of Yi is ensured by < 0; which is the constant elasticity of substitution in production between the di¤erent varieties of yvi ; vi = 1; :::; Vi : Combining (17) with (21) shows that an expansion of the variety of intermediate inputs works much like a positive productivity shock which causes a outward shift in …rm i0 s PPF and, at given prices of each …nal good variety, may lead to an expansion in the output variety as some previously not pro…table varieties may now become pro…table. Figure 4 demonstrates this for a two variety case. Under the …xed price level, in the original equilibrium, …rm i only produces variety 1; but as the PPF shifts out due to an increase in input variety, …rm i also produces variety 2 in the new equilibrium. To formally show that, I consider a symmetric equilibrium, where within each …rm i; the price for each variety of Y is the same ensuring that the quantity produced for each variety is also the 30 same, pvi = pi ; and yvi = yi : (22) This implies that the aggregate bundle of goods produced by i equals the quantity of each variety times the total output variety of i raised to a positive power: 1 Yi = Vi yi : One can, therefore, rewrite the production function, in terms of output per variety, as the following, M 1 M yi = Vi iN 1 Mi Li L Ki K ; (23) which shows that, given the same amount of inputs, if …rm i produces more varieties of output, the quantity for each variety is smaller. To produce each unit of yi ; …rm i minimizes the cost of production, which results in the following unit cost function (assuming M + L + K = 1); h M i 1 1 ci = Vi iN 1 PMM PL L PKK ; (24) where Pj ; 8j = fM; L; K g ; is the price of intermediate input, labor and capital, and is a constant which depends on the 0 s. Equation (24) implies that an increase in the variety of intermediate inputs pushes down the unit cost of producing yi : In contrast, given input prices and variety, an increase in output variety raises the cost for each variety. Given ci ; to maximize pro…t, …rm i will set the price for each variety to be a …xed markup over ci; h M i 1 1 pi = ci = Vi iN 1 PMM PL L PKK (25) where > 1 depends on the constant elasticity of substitution between di¤erent …rm i0 s: Equation (25) implies that, given prices of inputs and output, an increase in input variety leads to an increase 31 in output variety: h M i pi 1 1 Vi = iN 1 M L K ) PM PL PK 2 3 14 M X ln Vi = + ln i + ln N + ln pi j ln Pj 5 : (26) 1 j =fM;L;K g Equation (26) presents a structural relationship between product scope and the input variety of a multi-product pro…t maximizing …rm. It shows that an increase in input variety leads to the expansion of product scope of a …rm, controlling for productivity, output price, and the industry prices of materials, labor, and capital. It also neatly shows that a rise in input variety has the similar expansionary e¤ect on product scope as a positive productivity shock that increases the productivity of a …rm. Recalling Figure 4, an increase in input variety or productivity will both shift the PPF out such that, under constant prices, a …rm will …nd it pro…table to produce more output varieties. Given that I have shown that more FDI …rms lead to more local input varieties, more FDI …rms, therefore, cause the product scope of domestic …rms to be larger. Equation (26) also shows that any reduction in input prices will also lead to an expansion in product scope for domestic …rms. The …nding here, that increases in input variety and reduction in input prices lead to the proliferation of output variety, is very similar to that of Goldberg, Khandelwal, Pavcnik and Topalova (2010). In their paper, the authors show that trade liberalization in India in the 1990s caused an explosion in the variety of imported intermediate inputs and a reduction in the prices of these inputs, which led to an expansion in product scope within …rms. Here, I show that a more liberalized FDI regime will also lead to an increase in local input variety, which causes domestic …rms in the same industry to be more productive and has a higher product scope. Given the linear structure, (26) can easily be estimated using the following log linear speci…ca- tion (the time subscript is reintroduced for clarity): ln Vit = i + N ln Nt + TFP ln it + p ln pit + M ln PM t + L ln PLt + uit ; (27) where one would expect N, TFP and p to be positive, and the coe¢ cients for input prices to be negative. The regression error in (27) includes the price of capital which is unobserved. To estimate (27) ; I use the …rm speci…c output price index to proxy pit ; the augmented_OP estimates of TFP 32 for it ; the average …rm speci…c input price index for PM t ; and wages for PLt : However, it is clear that in addition to Nt being endogenous, which I will instrument using the number of FDI …rms, some other right-hand side variables are also endogenous, and may depend on the number of FDI …rms in the garment sector too. At least one independent instrument for each of the right-hand side variables is needed for (27) to be identi…ed. Here I use the following instrumental variables: average productivity of the industry for it ; and the international prices of cotton and fabrics for pit and PM t : Wages are assumed to be exogenous due to the tremendous hidden unemployment or under-employment in Bangladesh which provide a large pool of workers relative to the size of the industry. Table 10 presents the results. Columns (5) and (6) …rst present the least squares estimates when I only include …rm …xed e¤ects and the number of local suppliers or the number of total input variety on the right-hand side. While the coe¢ cients are positive and signi…cant, they are likely to be contaminated with measurement errors. The IV estimates are presented in Columns (7) and (8), which are positive and signi…cant. Here again the reported …rst stage F statistics are larger than 10 in both regressions, and the coe¢ cients have the expected signs. Columns (9) to (12) estimate (27) : Columns (9) and (10) present the least squares estimates. While the least squares estimates of Nt are positive and signi…cant, the majority of the remaining coe¢ cients either have wrong signs or are insigni…cant. Columns (11) and (12) show the second stage of the IV estimates. Now, all the coe¢ cients have the correct signs and are mostly signi…cant. Most importantly, the results con…rm that an increase in the number of FDI …rms leads to increases in local input variety and total input variety, which raise the product scope of domestic …rms. The IV estimates for N are smaller than the least squares estimates due to reverse causality – larger product scope may cause an increased demand for locally produced intermediate inputs which causes an upward bias in the least squares estimates. On the contrary, the IV estimates are based on exogenous increases in local input variety as the number of FDI …rms rises to pin down the e¤ect on domestic product scope. The …rst stage results are also good, where all the coe¢ cients have the expected signs. Here instead of reporting all the 4 …rst stage F statistics for each regression, I only report the lowest …rst stage F statistics in Table 10. The point here is to show that if the lowest F statistics are still larger than 10, then all the …rst stage F statistics are larger than 10, suggesting that all the …rst stage regressions have signi…cant explanatory power. 33 Overall, the results con…rm that FDI …rms in the garment sector cause the number of local input variety to increase, which leads to signi…cant gains for domestic …rms, in terms of product scope and productivity. 7 Conclusions This paper studies and …nds support for shared supplier spillovers of FDI …rms, whereby, the product scope and productivity of domestic …rms improve, due to the increased presence of their FDI siblings. Siblings are …rms that share local intermediate input suppliers. This e¤ect is, primarily, driven by increased …rm access to new or better local input varieties as the expanding FDI …rms push up industry demand for local intermediate inputs, and not through selection or pecuniary externalities. I present some empirical evidence, based on reduced form regressions, showing that when FDI and domestic …rms share common local input suppliers, an exogenous increase in the presence of FDI …rms in the industry will cause domestic …rms to perform better in terms of product scope, sales per worker, output per worker and productivity. In the Appendix, I present a simple theoretical model of a multi-product …rm with a love of variety for intermediate inputs. The model predicts that productivity and product scope of the …rm rise with the expansion of intermediate inputs in the industry. Given that FDI …rms increase industry demand for intermediate inputs, which leads to the proliferation of local input variety, more FDI …rms will, therefore, lead to higher productivity and product scope for domestic …rms in the same industry. Structural regressions based on the model con…rm the results. Thus, the results of this paper provide support to endogenous growth models which emphasize the importance of new intermediate inputs in explaining productivity growth. Most importantly, this paper suggests a new channel via which the presence of FDI …rms may bring non-pecuniary externalities to domestic …rms. Finally, besides lowering tari¤s on imported intermediate inputs, what else can the governments do to enhance the product scope and productivity of domestic …rms? This paper would suggest that a well designed FDI policy that attracts FDI with signi…cant backward linkages, may come a long way in promoting the intermediate input industries, and may also bene…t the domestic …nal good …rms. 34 References [1] Ackerberg, Daniel, Benkard, C. Lanier, Berry, Steven, and Ariel Pakes (2007). “Econometric Tools for Analyzing Market Outcomes”. In J.J. Heckman (Ed.) The Handbook of Econometrics, Volume 6. [2] Aitken, Brian J. and Ann E. Harrison (1999). “Do Domestic Firms Bene…t from Direct Foreign Investment? Evidence from Venezuela,” American Economic Review 89(3), 605-618. [3] Amiti, Mary and Jozef Konings (2007). “Trade Liberalization, Intermediate Inputs, and Pro- ductivity: Evidence from Indonesia,” American Economic Review 97(5), 1611-1638. [4] Blalock, Garrick and Paul Gertler (2008). “Welfare gains from Foreign Direction Investment through Technology Transfer to Local Suppliers,” Journal of International Economics 74, 402-421. [5] Bernard, Andrew and J. Jensen (1999). “ Exceptional Exporter Performance: Cause, E¤ect, or Both?” Journal of International Economics 47(1), 1-25. [6] Broda, Christian and David E. Weinstein (2006). “Globalization and the Gains from Variety,” Quarterly Journal of Economics 121(2), 541-585. [7] Carluccio, Juan and Thibault Fally (2011). “Foreign Entry and Spillovers with Technological Incompatibilities in the Supply Chain,” mimeo. [8] Cherkashin, Ivan, Svetlana Demidova, Hiau Looi Kee and Kala Krishna (2010). “Firm Het- erogeneity and Costly Trade: A New Estimation Strategy and Policy Experiments,” NBER working paper #w16557. [9] De Loecker, Jan (2007). “Do Exports Generate Higher Productivity? Evidence from Slovenia,” Journal of International Economics 73, 69-98. [10] Demidova, Svetlana, Hiau Looi Kee and Kala Krishna (2012). “Do Trade Policy Di¤erences Induce Sorting? Theory and Evidence From Bangladeshi Apparel Exporters,” Journal of International Economics 87(2), 247-261. 35 [11] Ethier, Wi‡red J. (1982). “National and International Returns to Scale in the Modern Theory of International Trade.” American Economic Review 72(3), 389-405. [12] Feenstra, Robert C. (1994). “New Product Varieties and the Measurement of International Prices,” American Economic Review 84(1), 157-177. [13] Fernandes, Ana M. (2007). “Trade Policy, Trade Volumes and Plant-Level Productivity in Colombian Manufacturing Industries,” Journal of International Economics 71(1), 51-72 . [14] Goldberg, Pinelopi K., Amit Kumar Khandelwal, Nina Pavcnik, and Petia Topalova (2010). “Imported Intermediate Inputs and Domestic Product Growth: Evidence from India,” Quar- terly Journal of Economics. [15] Halpern, Laszlo, Miklos Koren and Adam Szeidl (2011). “Imported Inputs and Productivity,” mimeo. [16] Javorcik, Beata Smarzynska (2004). “Does Foreign Direct Investment Increase the Productivity of Domestic Firms? In Search of Spillover Through Backward Linkages,” American Economic Review 94(3), 605-627. [17] Javorcik, Beata and Mariana Spatareanu (2009). “Tough Love: Do Czech Suppliers Learn from Their Relationships with Multinationals?” Scandinavian Journal of Economics 111(4). [18] Marshall, Alfred (1920). Principles of Economics (8e), Macmillan: London. [19] Moulton, Brent R. (1990). “An Illustration of a Pitfall in Estimating the E¤ects of Aggregate Variables on Micro Unit,” Review of Economics and Statistics 72(2), 334-338. [20] Olley, G. Steven and Ariel Pakes (1996). “The Dynamics of Productivity in the Telecommu- nications Equipment Industry,” Econometrica 64(6), 1263-1297. [21] Rodriguez-Clare, Andres (1996). “Multinationals, Linkages, and Economic Development,” American Economic Review 86(4), 852-873. [22] Romer, Paul M. (1990). “Endogenous Technological Change.” Journal of Political Economy 98(5), S71-S102. 36 Figure 1: Numbers of Garment FDI …rms and Local Suppliers In Bangladesh (1984=1) Figure 2: The Share of FDI Firms in Bangladesh’s Apparel Sector, 1999-2003 37 Figure 3: An Example of the Calculations of Industry Foreign Presence and Foreign Sibling Presence Figure 4: Output variety increases as PPF shifts out due to an increase in input variety y2 y1 38 Table 1: Sample Averages Non-woven Woven Domestic FDI Domestic FDI Sales 2648.90 3894.15 2656.05 14200.00 Export 2538.41 3662.36 2620.61 14200.00 Material 1722.67 2527.50 1874.64 9665.94 Imported material 1013.16 2150.88 1494.03 8393.14 Employee (number) 639.55 946.57 571.81 1877.64 Investment 138.69 137.59 49.04 266.04 Capital 580.10 1582.38 734.65 4103.32 Age (year) 5.23 6.10 7.98 7.29 Number of …rms 89 15 167 26 Note: All values are in US$000’ s, except where otherwise speci…ed. Table 2: Sample Averages for Domestic …rms Non-woven Woven Industry foreign presence (IF Pjt ) 28.68 36.43 FDI siblings (F DISit = 1) 15.57 51.91 Foreign sibling presence (F SPit ) 0.48 6.08 Share of domestic …rms that have a common product with a FDI …rm 89.52 92.93 Product foreign presence (P F Pit ) 13.48 36.61 Share of domestic …rms that have a common market with a FDI …rm 97.60 97.31 Market foreign presence (M F Pit ) 2.12 10.58 Note: All values are in percent. Table 3: First Stage Regressions: Dependent variable –Sibling Foreign Presence (1) (2) FDI siblings that export to EU 0.09*** 0.04*** (0.01) (0.01) FDI siblings that export to EU* 0.04*** 0.01*** Woven*EBA (0.01) (0.00) Observations 104 1000 F-stat 202.79*** 13.46*** Notes: All columns include …rm …xed e¤ects, industry-region-year …xed e¤ects, …rm age, share of imported materials, share of material in sales, and …rm speci…c time trends. *, **, *** indicate statistical signi…cance at 90%, 95%, and 99% con…dence levels, respectively. Column (1) consists of Bangladeshi …rms that do not export to the EU; Column (2) consists of Bangladeshi …rms that may or may not export to the EU. 39 Table 4: Restricted Sample of Domestic Firms with Firm speci…c Time Trends (1) (2) (3) (4) (5) (6) (7) (8) (9) (10) Dependent Variables Product Scope Sales per Worker Output per Worker TFP_OLS TFP_AOP Estimation Methods LS IV LS IV LS IV LS IV LS IV Foreign sibling presence 2.21* 7.30*** 4.85* 18.43*** 4.92* 18.40*** 2.47** 9.05*** 2.53** 8.33*** (1.07) (0.52) (2.23) (1.07) (2.27) (1.15) (1.03) (0.55) (1.00) (0.55) Age 1.52 0.59*** 1.43 -1.06 1.51 -0.96 0.57 -0.63** 0.45 -0.62* (1.08) (0.23) (3.28) (0.67) (3.23) (0.71) (1.45) (0.28) (1.43) (0.34) Imported Materials/Materials 0.17 -0.06 -0.21 -0.81*** -0.22 -0.81*** -0.21 -0.50*** -0.19 -0.45*** 40 (0.36) (0.16) (0.85) (0.15) (0.84) (0.16) (0.43) (0.10) (0.43) (0.11) Material/Sales -0.17 -0.01 0.92 1.34*** 0.73 1.14*** -0.41 -0.21 -0.44 -0.29* (0.64) (0.31) (0.59) (0.22) (0.56) (0.18) (0.36) (0.18) (0.28) (0.16) Observations 116 113 116 113 116 113 108 104 104 100 Notes: All dependent variables are in logs. TFP_OLS is from (39) and (40) ; TFP_AOP is from (37) and (38). Firm …xed e¤ects and industry-region-year …xed e¤ects are included in all columns. Robust standard errors in parentheses are clustered by industry-year, with degree of freedom adjustment for small sample. *, **, *** indicate statistical signi…cance at 90%, 95%, and 99% con…dence levels, respectively. Sample only consists of Bangladeshi …rms that do not export to the EU. All columns include …rm speci…c time trends. Table 5: Full Sample of Domestic Firms with Firm Speci…c Time Trends (1) (2) (3) (4) (5) (6) (7) (8) (9) (10) Dependent Variables Product Scope Sales per Worker Output per Worker TFP_OLS TFP_AOP Estimation Methods LS IV LS IV LS IV LS IV LS IV Foreign sibling presence 0.94** 1.19*** 2.15*** 5.30*** 1.97*** 5.32*** 1.08*** 2.76*** 1.09*** 2.64*** (0.32) (0.20) (0.20) (0.39) (0.17) (0.38) (0.09) (0.12) (0.09) (0.08) Age 0.06 0.06* 0.52*** 0.49*** 0.61*** 0.58*** 0.44*** 0.42*** 0.27** 0.26** (0.05) (0.03) (0.15) (0.10) (0.13) (0.09) (0.11) (0.07) (0.11) (0.07) Import Materials/Materials -0.07** -0.07*** 0.39 0.39*** 0.33 0.33** 0.16 0.16** 0.06 0.06 41 (0.03) (0.02) (0.21) (0.14) (0.21) (0.14) (0.11) (0.08) (0.08) (0.06) Material/Sales -0.00 -0.00 0.43 0.43** 0.23 0.23 -0.78** -0.79*** -1.00** -1.01*** (0.08) (0.05) (0.26) (0.18) (0.24) (0.16) (0.30) (0.20) (0.34) (0.23) Observations 1034 1034 1034 1034 1034 1034 1013 1013 1000 1000 Notes: All dependent variables are in logs. TFP_OLS is from (39) and (40) ; TFP_AOP is from (37) and (38). Firm …xed e¤ects and industry-region-year …xed e¤ects are included in all columns. Robust standard errors in parentheses are clustered by industry-year, with degree of freedom adjustment for small sample. *, **, *** indicate statistical signi…cance at 90%, 95% and 99% con…dence levels, respectively. Sample consists of Bangladeshi …rms that may or may not export to the EU. All columns include …rm speci…c time trends. Table 6: Robustness Checks –Other Sources of Spillovers (1) (2) (3) (4) (5) (6) (7) (8) (9) (10) Dependent Variables Product Scope TFP_AOP Product Foreign Presence 0.61 0.26 (1.03) (0.46) Market Foreign Presence -0.06 -0.58 (0.76) (0.42) Domestic Sibling Presence -0.89 -0.93 (0.68) (1.35) Firm Size Sibling Presence 27.17 -1.47 (27.87) (47.35) Randomized Foreign Sibling Presence -0.01 0.00 (0.23) (0.29) Age -0.21* -0.20* -0.18 0.06 -0.00 0.28** 0.28** 0.31** 0.28** 0.18 42 (0.10) (0.10) (0.11) (0.05) (0.08) (0.11) (0.11) (0.11) (0.11) (0.17) Imported Materials/Materials -0.04 -0.04 -0.05 -0.07* -0.06 -0.06 0.07 0.06 0 .06 -0.01 (0.12) (0.12) (0.11) (0.03) (0.05) (0.09) (0.09) (0.07) (0.08) (0.15) Material/Sales -0.21 -0.20 -0.21 0.01 -0.05 -1.00** -0.99** -1.01** -1.00** -1.20*** (0.18) (0.18) (0.18) (0.09) (0.15) (0.35) (0.35) (0.35) (0.34) (0.35) Observations 1034 1034 1034 1034 1034 1000 1000 1000 1000 1000 Notes: All dependent variables are in logs. TFP_AOP is from (37) and (38). Firm …xed e¤ects and industry-region-year …xed e¤ects are included in all columns. Unless otherwise stated, robust standard errors in parentheses are clustered by industry-year, with degree of freedom adjustment for small sample. For (5) and (10) only, random matching of siblings are repeated 50 times. Coe¢ cients and standard error reported are the average and standard deviation of the 50 regressions. *, **, *** indicate statistical signi…cance at 90%, 95%, and 99% con…dence levels, respectively. Sample consists of Bangladeshi …rms that may or may not export to the EU. All columns include …rm speci…c time trends. Table 7: Full Sample of Domestic Firms with Firm speci…c Time Trends (1) (2) (3) (4) (5) (6) (7) (8) (9) (10) Dependent Variables Product Scope Sales per Worker Output per Worker TFP_OLS TFP_AOP Estimation Methods LS IV LS IV LS IV LS IV LS IV FDIsuppliersh it 0.28*** 0.34*** 0.19 1.82*** 0.17 1.83*** 0.17** 0.96*** 0.22*** 0.91*** (0.05) (0.04) (0.15) (0.34) (0.14) (0.34) (0.06) (0.15) (0.07) (0.16) Age 0.04 0.04 0.47** 0.47** 0.56*** 0.56*** 0.32*** 0.32** 0.17 0.17 (0.03) (0.03) (0.16) (0.16) (0.15) (0.15) (0.10) (0.10) (0.11) (0.11) Imported Materials/Materials -0.07* -0.07** 0.38 0.36 0.32 0.30 0.15 0.14 0.05 0.04 43 (0.03) (0.03) (0.22) (0.23) (0.22) (0.23) (0.12) (0.12) (0.09) (0.09) Material/Sales -0.09 -0.09 0.37 0.23 0.19 0.05 -0.90*** -0.97*** -1.11*** -1.18*** (0.07) (0.07) (0.29) (0.35) (0.26) (0.31) (0.23) (0.21) (0.29) (0.27) Observations 1012 1012 1012 1012 1012 1012 992 992 980 980 Notes: All dependent variables are in logs. TFP_OLS is from (39) and (40) ; TFP_AOP is from (37) and (38). Firm …xed e¤ects and industry-region-year …xed e¤ects are included in all columns. Robust standard errors in parentheses are clustered by industry-year, with degree of freedom adjustment for small sample. *, **, *** indicate statistical signi…cance at 90%, 95%, and 99% con…dence levels, respectively. Sample only consists of Bangladeshi …rms that may or may not export to the EU. . All columns include …rm speci…c time trends. Instruments: F DIS _EUit ; F DIS _EUit woveni EBAt ; 1st-stage F-Stat: 9.82*** with both coe¢ cients positive and signi…cant. Table 8: Robustness Checks –Horizontal Spillovers (1) (2) (3) (4) (5) Dependent Variables TFP_AOP Industry Foreign Presence -2.13*** 0.31 -2.49*** (0.24) (0.61) (0.25) Industry Foreign Presence 10.89** -4.82*** for FDI …rms with domestic siblings (4.96) (1.29) Industry Foreign Presence 1.29 -2.60*** for FDI …rms without domestic siblings (0.74) (0.31) Tari¤s of EU -0.10 -0.14 -0.15** 0.67* -0.35** (0.06) (0.18) (0.06) (0.33) (0.13) Tari¤s of US 0.20** -0.30 0.30*** -1.91** 0.65** 44 (0.06) (0.16) (0.08) (0.77) (0.21) Industry Average Productivity 1.13** -0.20 1.33** -0.36 1.34** (0.49) (0.47) (0.56) (0.51) (0.56) Observations 1041 419 622 419 622 Notes: All dependent variables are in logs. TFP_AOP is from (37) and (38); Firm …xed e¤ects and year …xed e¤ects are included in all columns; Robust standard errors in parentheses are clustered by industry-year; *, **, *** indicate statistical signi…cance at 90%, 95%, and 99% con…dence levels, respectively; Column (1) consists of all Bangladeshi …rms; Columns (2) and (4) consists of all Bangladeshi …rms with FDI siblings; Columns (3) and (5) consists of all Bangladeshi …rms with no FDI siblings; Table 9: Robustness Checks –Greater Local Input Intensity of FDI (1) (2) (3) (4) (5) (6) (7) (8) Dependent Variable Domestic Material per Employee Share of Domestic Material in Total Material FDI 781.83** 604.24* 503.93** 503.93* 0.12* 0.11* 0.12*** 0.12* (391.36) (349.94) (218.46) (293.33) (0.07) (0.07) (0.04) (0.07) Export Share of EU 167.41 113.07 -115.47 -115.47 0.085 0.00 0.03 0.03 (631.60) (551.19) (250.87) (435.66) (0.114) (0.11) (0.07) (0.12) Export Share of US -385.35 -391.57 -580.81** -580.81 -0.031 -0.11 -0.10 -0.10 (662.60) (588.52) (248.34) (445.67) (0.119) (0.11) (0.08) (0.13) Log of Employment -2058.54*** -1498.94*** -1346.67*** -1346.67*** -0.055* -0.02 -0.01 -0.01 (669.29) (148.56) (246.84) (272.16) (0.031) (0.03) (0.01) (0.03) Capital-Labor Ratio -0.01 -0.01 -0.01*** -0.01*** 0.000 0.00 0.00*** 0.00*** (0.07) (0.01) (0.00) (0.00) (0.000) (0.00) (0.00) (0.00) Log of Materials 588.63** 610.92*** 550.63*** 550.63*** -0.020 -0.01 -0.02 -0.02 (264.40) (96.51) (111.98) (139.11) (0.021) (0.02) (0.01) (0.02) 45 Log of TFP -523.04 -142.83 -216.77 -216.77 -0.001 0.03 0.01 0.01 (462.65) (143.17) (149.70) (175.10) (0.027) (0.03) (0.01) (0.02) Age -191.70* -29.43 -190.80*** -190.80** -0.047** -0.02 -0.05*** -0.05*** (103.92) (117.62) (64.78) (77.03) (0.019) (0.02) (0.01) (0.02) Log of Output 981.98* 448.19*** 480.05** 480.05** 0.017 -0.01 0.00 0.00 (571.81) (125.20) (196.67) (199.31) (0.025) (0.02) (0.01) (0.02) Industry-Location Fixed E¤ects Yes No No No Yes No No No Industry-Location-Year Fixed E¤ects No Yes Yes Yes No Yes Yes Yes Number of Observations 263 1155 1155 1155 263 1155 1155 1155 Notes: (1) and (5) are based on cross sectional data in 2003. The rest of columns are based on panel data from 1999 to 2003. Standard errors in parenthesis. (1) and (5) report the OLS estimates with bootstrap standard errors. (2) and (6) report the between estimates. (3) and (7) report panel estimates with industry-location-year clustered standard errors. (3) and (7) report panel estimates with …rm clustered standard errors. Table 10: Structural Estimations (1) (2) (3) (4) (5) (6) (7) (8) (9) (10) (11) (12) Dependent Variables TFP_AOP Product Scope Estimation Methods LS LS IV IV LS LS IV IV LS LS IV IV Number of local input suppliers 0.27* 0.36*** 0.20*** 0.20** 0.26*** 0.09** (0.12) (0.14) (0.01) (0.01) (0.01) (0.04) Number of total intermediate inputs 0.10* 0.12** 0.11*** 0.12*** 0.15*** 0.04* (0.05) (0.05) (0.01) (0.02) (0.01) (0.02) TFP 0.06 0.060.47*** 0.66** (0.03) (0.03) (0.16) (0.33) Price of output 0.16** 0.16** 0.04 0.40 46 (0.05) (0.05) (0.70) (1.12) Price of intermediate inputs 0.02 -0.18*** -0.20*** 0.14 (0 .04) (0.07) (0.04) (0.06) Wages 0.04 -0.14*** -0.20*** 0.04 (0.05) (0.05) (0.05) (0.07) F-Statistics 29.74 11.92 29.11 11.50 158.66 158.66 Observations 1041 1041 1041 1041 1165 1165 1165 1165 1041 1041 1041 1041 Notes: Full sets of …rm …xed e¤ects are included in all columns. Standard errors in parentheses are clustered by year. TFP_AOP is from (36) and (38). All variables are in logs. Excluded instruments for IV estimations are: number of FDI …rms, international cotton fabrics price (in (4), (8) and (12) only), industry average TFP (in (11) and (12) only), and international cotton price (in (11) and (12) only). For (11) and (12), I only report the lowest F-statistics from the 4 …rst stage regressions. First stage results are available upon request. *, **, *** indicate statistical signi…cance at 90%, 95%, and 99% con…dence levels, respectively. A Appendix for On-line Publication A.1 Horizontal Spillovers of FDI There is a vast literature on the topic of horizontal spillovers. Theoretical papers in this area tend to conclude positive spillovers. Besides Rodriguez-Clare (1996) and Carluccio and Fally (2011), Findlay (1978) also provides a dynamic model to show the role of FDI …rms in transferring tech- nology from the advanced to the backward countries. Markusen and Venables (1999) presents an analytical model where FDI …rms may act as a catalyst for industrial development if they generate enough demand to support the upstream industries through backward linkages, which further foster the downstream industries through forward linkages. However, empirical evidence for horizontal spillovers are mixed. Earlier papers, based on case studies (e.g. Caves, 1974), or cross industry evidence (e.g. Blomstrom and Persson, 1983; Blom- strom and Wol¤, 1994), tend to conclude that there exists a positive correlation between the presence of FDI in an industry and the average productivity of domestic …rms, while papers based on …rm or plant level statistics of developing countries have found the opposite (Aitken and Har- rison, 1999; Haddad and Harrison, 1993; Djankov and Hoekman, 2000; Konings, 2001). The way horizontal spillovers are identi…ed in this literature is to associate the productivity of domestic …rms with the presence of all FDI …rms in the industries. It is not until recently that the some fresh results started to emerge. Haskel, Pereira, and Slaughter (2007) found small but statistically sig- ni…cant evidence of positive spillovers in a study of UK manufacturing plants. Based on a matched establishment-worker database from Brazil, Poole (2013) …nds some convincing evidence showing that worker turnover from FDI …rms to domestic …rms is an important channel for knowledge transfer. Given that I have information on which are the FDI …rms in the industries that have backward linkages and which are not, I can study the di¤erential impacts of the presence of these di¤erent FDI …rms in the industries on the productivity of domestic …rms, who may or may not share common local suppliers with the FDI …rms. In other words, this paper is able to take into account the heterogeneity of FDI …rms in terms of their backward linkages, as well as the heterogeneity of domestic …rms in terms of their connection with the FDI …rms, through the use of common local input suppliers. The results are presented in Table 8 of this paper which I discuss in detail in 47 Section 6. In a nutshell, the results show that the negative horizontal spillover e¤ects that have been predominantly highlighted in the literature only a¤ect those domestic …rms that do not have a FDI sibling. For these …rms, the presence of FDI …rms unambiguously hurt their productivity, regardless of whether or not the FDI …rms have backward linkages. On the other hand, for those domestic …rms that have FDI siblings, the presence of those FDI …rms that have backward linkages clearly improves their productivity. A.2 Survey Evidence on FDI Promoting Quality Upgrading and Variety Ex- pansion of Local Input Suppliers The whole premises of SSS hinge on the assumptions/assertions that FDI …rms help promote a more vibrant local intermediate input industry with better quality (Javorcik, 2004) and more variety (Rodriguez-Clare, 1996) of local intermediate inputs. Is there any evidence suggesting that this is the case for the garment sector of Bangladesh? Between 2010 and 2011, a survey were conducted on the local suppliers that were identi…ed by the garment …rms in our sample, including those suppliers that were named top three by the FDI …rms.31 Products of these local suppliers include all kinds of yarn, thread, fabric, padding, buttons, labels, elastics, fastening products (belts, buckles, laces, etc.), printing, dyeing, sewing, hangers and hanger accessories, tape, corrugated paper and cartons, back boards, enzymes, softeners, staining agents, silicone, synthetic …ber, polythene bags, etc. Among the survey …rms, 84 percent indicating that they served FDI …rms during the …rst three years of their plants. The average share of FDI …rms in their sales during the …rst three years is 35 percent. 26 percent of these …rms suggesting that serving FDI …rms was important in their decision to set up plants in Bangladesh. All these evidence indicating that FDI garment …rms play an important role supporting the local intermediate input industry in Bangladesh. With regard to quality upgrading and variety expansion, 68 percent of the surveyed local sup- pliers agree that FDI clients are known to be “pickier,”that demand highest quality products with lowest rejection rates, so to meet the expectation of FDI clients, suppliers need to improve their quality and e¢ ciency. This provides evidence supporting the quality upgrading hypothesis of Ja- vorcik (2004). On the other hand, 62 percent of the surveyed local suppliers agree that their FDI clients suggest new products for the suppliers to manufacture, which expand the product scope of 31 The other local suppliers either have closed down, refused to participate or cannot be located in Bangladesh. 48 the suppliers, which validates the variety hypothesis of Rodriguez-Clare (1996). Finally, 68 percent of the respondents agree that FDI clients may have better procurement practices, better inventory management (bar codes and computerization), so the suppliers learn these technologies and apply them to their own businesses. This is consistent with the …nding of Javorcik (2004) that FDI …rms transfer technology to their local suppliers which makes them more productive. With regards to the direct evidence of SSS, 82 percent of the respondents agree that their domestic garment …rm clients bene…t from them working with the FDI garment …rms. Speci…cally, 35 percent of the sampled local suppliers indicating that they have sold some new products to the domestic garment …rms which were originally suggested to them by their FDI clients; 68 percent of these suppliers apply the same procurement practice, inventory management they may have learned from their FDI clients to their domestic garment …rm clients; and 35 percent of the respondents agree that their FDI clients make them upgrade their product quality and they subsequently also sell the better quality products to a domestic garment …rm. Finally, 75 percent of the suppliers concur that FDI garment …rms help promote and support a more vibrant supplier industry, which now have better quality garment accessories and fabrics, as well as more variety of accessories and fabrics. So domestic garment …rms now have better access to more quality accessories and fabrics, cut down lead time and business costs for them. Overall the survey evidence is consistent with the econometric …ndings of this paper. These evidence indicates that both the quality upgrading and variety expansion channels are relevant in the garment sector of Bangladesh, which help explain the shared-supplier spillovers of FDI. A.3 Firm Level Price Indexes To estimate the …rm’s productivity, one needs to measure …rm output and material input. Output and material input variables are constructed by de‡ating total value of sales and materials with output and material input price indexes, respectively. Due to the lack of data, industry level price indexes have long been used in the literature in place of …rm price indexes. There are obvious problems in using industry price indexes to de‡ate …rm sales and material costs. For example, many heterogenous …rm models would suggest that more productive …rms will charge a lower price. As such, using an industry price index, which re‡ects the average price level of all …rms in the industry, to de‡ate sales of the more productive …rms will underestimate the output level, which 49 leads to an under-estimation of …rm productivity. A unique strength of my data is the fact that there is information on prices at the …rm level, which allows me to construct …rm speci…c price indexes that are consistent across years and …rms. Eslava, Haltiwanger, Kugler, and Kugler (2004) construct a Tornqvist price index for each …rm which is consistent within …rms over time. The …rm price index is a weighted average of unit value changes for each of the product the …rm produces in each year, with weights that re‡ect the average share of the product in total sales of the …rms in two consecutive years. However, by setting each …rm price index equal to 1 in the base year, cross …rm variation is ignored. This can hide …rm heterogeneity in terms of productivity. In this …rm survey, I have information on the value and quantity of the …ve main products for each …rm in 2003. I can, therefore, construct a weighted average unit value of products for each …rm in 2003 with weights re‡ecting the share of each product in the total sales of the …rm. This will be the …rm product price level in 2003. The industry price level in 2003 is constructed by taking the weighted average of the …rm price level with weights re‡ecting the size of the …rm in the industry. By dividing the …rm price level by the industry price level, I obtain a cross sectional …rm price index for 2003. Firms that have a …rm price level higher than the industry price level will have a …rm price index in 2003 exceeding unity. Conversely, …rms that have a price level less than that of the industry in 2003, will have a …rm price index below unity. In this manner, the cross sectional price index will capture …rm heterogeneity in 2003. Finally, to extend the …rm price index to the previous years, I rely on the information provided by the …rms in the survey regarding the annual change in price of their main product. In this way, the constructed multi-year …rm price index will be consistent within …rms across years, as well as across …rms within a year. A similar procedure is used to construct …rm speci…c material price index. I use these …rm level product and material price indexes to de‡ate total sales and material costs of the …rms to obtain output and material inputs of the …rms for the production function estimation.32 32 There may be a concern that …rm speci…c prices may convey information on the quality of the …rm. Firms that have higher quality products (or more services per good), and thus, higher prices will have a higher …rm price index. By de‡ ating total sales using this …rm price index, I obtain an output measurement that is quality free, i.e., is in terms of “e¤ective units” of the good. Thus, the productivity estimates will not be contaminated with the quality of the …rm’ s products, which is a known problem in the existing literature, which uses an industry price index to de‡ate …rm sales. 50 A.4 The Production Function I assume that the following Cobb-Douglas production function holds separately for woven and non-woven industries (industry subscripts are omitted): L Yit = it Lit Mit M KitK ; (28) where i and t are the indexes for …rm and year, respectively, and Yit ; Lit ; Mit and Kit are the output, labor, materials, and capital of …rm i in year t: Output and material input are obtained by de‡ating total sales and material cost using …rm speci…c price indices which are constructed using detailed price information from the …rm survey. The total factor productivity (TFP) of …rm i in year t is it : Assume that in log, it can be decomposed linearly into the following, ln it ! it + t + A ait + F F DIit + it ; (29) where ! it is observable to the …rms at the beginning of each period before variable input choices are made, but not to the researchers. The year speci…c productivity, t; may capture the e¤ects of time and others factors that are common to all …rms during a year (within an industry) and A ait ; is the e¤ect of (log of) age on productivity.33 I further allow FDI …rms to have a di¤erent productivity than domestic …rms by including a FDI dummy variable in (29). Whether or not age and FDI status have a direct impact on the productivity of a …rm remains an empirical question. While older …rms may be more established and therefore can withstand a low productivity shock, they may also be more organized and therefore more productive. Likewise, FDI …rms may be able to weather low productivity draw, but they may also be more productive due to the transfer of technology from the parent …rms. These scenarios cause A and F to have ambiguous signs a priori. I will be able to test the e¤ect of age and FDI status on productivity in the empirical section. The last term, it ; is the truly unobserved classical error term. Taking log of (28) and using (29) ; I have yit = t + A ait + F F DIi + L lit + M mit + K kit + ! it + it ; (30) 33 Given that all …rms are exporters in this data set, aA ait ; may also capture the e¤ect of export experience on productivity, due to possibly learning-by-exporting. 51 where all lower case letters are in logs. In logs, output is linearly related to the two variable inputs, labor and materials, as well as the …xed input, capital stock. Given that ! it is observable to the …rms (but not to the researchers) before the variable input choices are made, it could be positively correlated with lit and mit , which would cause the least squares estimates of L and M to be biased upward. However, for the woven industry, ! it and mit could be negatively correlated since more productive …rms could manage to use less material while satisfying ROOs; and this would cause the least squares estimate of M to be downward biased. In addition, if larger, older …rms tend to stay in business despite low productivity, while younger, smaller …rms tend to quit more easily, such endogenous exit decisions on the part of …rms will bias the least squares estimates of the A and K downwards. A.5 Estimating Productivity To address such input endogeneity and selectivity bias, Olley and Pakes (1995) (OP) derive a 3-step procedure to obtain consistent estimates of the 0 s: In their model, …rms choose to exit or not once they know their productivity. If they do not exit, they decide on how much to invest and also make other output and input decisions. The productivity, ! it ; is assumed to be the only unobserved state variable in each year t that follows a common exogenous Markov process, which, jointly with …xed input, kit , and its age, determines the exit decision and investment demand, iit ; of the …rms in each period. They consider the Markov perfect Nash equilibrium, so …rm’s expectations match the realization of future productivity. Then a polynomial function of iit ; kit ; and (the log of ) age, ait ; can be used to proxy for the unobserved productivity, ! it . This is possible because, given kit and ait , iit is an increasing function of ! it ; which makes the investment function invertible. The assumption that investment is monotonically increasing with the unobserved productivity is crucial, since without it, invertibility is likely not possible. Furthermore, to control for the exit decision, they estimate a Probit regression to obtain the surviving probability and use that to control for the part of unobserved productivity that is negatively correlated with kit : s In the current data set, it is likely that (in addition to the unobserved productivity) …rm’ investment decisions also depend on the FDI status of the …rms, since FDI …rms may choose to stay in business and continue to investment despite low productivity draws. This is quite evident from Table 1, where FDI …rms are shown to be larger and invest more than the domestic …rms. This 52 may also suggest that FDI …rms face di¤erent market structure and factor prices as the domestic …rms. To accommodate such facts, I modify OP along the lines suggested by Ackerberg, Benkard, Berry and Pakes (2007) and De Loecker (2007).34 Speci…cally, when studying the e¤ect of exporting on …rm productivity, De Loecker (2007) allows exporters to have a di¤erent investment function.35 In the current context, given that all …rms are exporters, but only some …rms are FDI …rms, I allow the investment function to be indexed by their FDI status,36 iit = iF DI;t (kit ; ait ; ! it ) : This allows FDI …rms to react di¤erently from domestic …rms when it comes to investment decision, as capital, age, or productivity of the …rms change. Controlling for capital, age and FDI status, the investment function is assumed to be invertible, as in the original OP set up, such that I can use a separate polynomial function of investment, capital and age as controls for the unobserved productivity, for the FDI …rms and domestic …rms.37 1 ! it = iF DI;t (kit ; ait ; iit ) = hF DI;t (kit ; ait ; iit ) : (31) In other words, I can proxy the unobserved …rm productivity parsimoniously with a polynomial function hF DI;t (kit ; ait ; iit ). In addition to the FDI status, I also allow the polynomial function to be di¤erent in di¤erent time periods, which explains why I index the function with t. This is because the EU, the main market for garment exporters from Bangladesh, introduced the “Everything- but-Arms” (EBA) initiative in 2001, which o¢ cially removed all quota restrictions and tari¤s 34 I thank Ariel Pakes for a detailed discussion on the productivity estimation technique and for the suggestions of using Ackerberg et al (forthcoming). 35 Alternatively, one could have modeled FDI status as a state variable, similar to capital, age, and productivity, as the past exporter status in Van Biesebroeck (2006). However, this requires that FDI status changes within …rms over time for some …rms in the sample. This is not the case for the current data set. All …rms are observed to either have no foreign ownership for the whole sample period, or to have the same FDI status throughout the sample period. Without the evolution of FDI status, it is not possible to model it as a state variable. 36 FDI dummy equals one when the …rms have any foreign equity. In the current sample, the minimum foreign ownership is 25 percent. 37 Using the same data set, Demidova, Kee, and Krishna (2012) estimate …rm productivity, allowing for …rm-market speci…c demand shocks. In their context, it is crucial to control for market demand shocks as they are trying to explain the breakdown of the hierarchy of …rm in terms of productivity in sorting themselves into di¤erent markets. In the current application, I am most concerned about how FDI …rms a¤ect the productivity of domestic …rms endogenously through the spillover channels. 53 on Bangladeshi garment exports. Such policies may signi…cantly alter the market structure and factor prices of the …rms. To accommodate this, I allow the polynomial function to di¤er between the pre- and post-EBA period. In other words, I proxy the unobserved …rm productivity with 4 di¤erent polynomial functions — domestic …rms in period 1999-2000; FDI …rms in period 1999- 2000; domestic …rms in period 2001-2003; FDI …rms in period 2001-2003. The coe¢ cients of these polynomial functions are free to be di¤erent to re‡ect the di¤erent market conditions. Thus the …rst stage estimation involves using a polynomial function hF DI;t (kit ; ait ; iit ) to control for ! it in order to estimate the coe¢ cients on the variable inputs, lit and mit ; which are decided after ! it are observed. yit = t + L lit + M mit + K kit + A ait + F F DIi + hF DI;t (kit ; ait ; iit ) + it = L lit + M mit + F DI;t (kit ; ait ; iit ) + it ; where F DI;t (kit ; ait ; iit ) = t + K kit + A ait + F F DIi + ! it ; combines t; K kit , A ait and F F DIi with hF DI;t (:) : Provided that hF DI;t (:) is successful in controlling for ! it ; the least squares estimates for L and M are consistent, and I denote them as ^ L and ^ M . To estimate K and A; one needs to control for the propensity to exit to address the endogenous exiting which is a¤ected by the size and age of the …rms. For each …rm i; in order to maximize the present discounted value of current and future pro…ts, the optimal exit rule having observed ! it is 8 < 1 (continue) it = if ! it ? ! F DI;t (kit ; ait ) ; (32) : 0 (exit) where ! t is the cuto¤ productivity to continue exporting. Thus, the probability for …rm i to survive in year t + 1 given information set in year t; Jt ; is Pr it+1 = 1jJt = Pr (! it+1 > ! F DI;t+1 (kit+1 ; ait+1 ) jJt ) ~ t (! it ; ! F DI;t+1 (kit+1 ; ait+1 )) = ' e ~ F DI;t (! it ; kit+1 ; ait+1 ) = ' = 'F DI;t (kit ; ait ; iit ) = Pit+1 (33) 54 where the …rst equality holds because of the exit rule (32) ; the second and third equalities hold due to the assumption of the exogenous Markov process of ! it ; and the last equality holds because the investment function iit = iF DI;t (kit ; ait ; ! it ) is a bijection in ! it conditional on (kit ; ait ) ; and kit+1 and ajt+1 can be inferred from kit ; iit and ait ; from their laws of motion, Kit+1 = Kit (1 ) + Iit+1 ; and Ait+1 = Ait + 1: (34) In other words, in the second stage, I can estimate the survival probability in t+1 non-parametrically using a period speci…c polynomial function of (kit ; ait ; iit ) in a probit regression. This would allow factors like the existence of the EBA to a¤ect exit decisions. I denote the estimated survival ^ it+1 : probability in t + 1 as P According to (30) ; the expected value of output net of in‡uence of labor and material in t + 1; given the information set in t and survival in t + 1 is E yit+1 L lit+1 M mit+1 jJit ; it+1 =1 = t+1 + A ait+1 + F F DIi + K kit+1 + E ! it+1 jJit ; it+1 =1 = t+1 + A ait+1 + F F DIi + K kit+1 + g (! it ; Pit+1 ) = t+1 + A ait+1 + F F DIi + K kit+1 + g0 ( t t K kit A ait ; F F DIi ; Pit+1 )(35) where the …rst equality holds because ait+1 and kit+1 are known in t due to (34) : Given the assumption of the Markov process, ! it+1 only depends on ! it and the probability of surviving in t + 1 is given in (33). Equation (35) suggests that I run the following nonlinear estimation in the third stage with g0 ( t t K kit A ait F F DIi ; Pit+1 ) being approximated parsimoniously with a polyno- mial function, to obtain t; A; F and K; yit+1 ^ L lit+1 ^ M mit+1 = ( L ^ L ) lit+1 + ( M ^ M ) mit+1 + t+1 + A ait+1 + F F DIi + K kit+1 +g 0 ^t t K kit A ait ^ F F DIi ; Pit+1 + it + it ; (36) where, by construction, E it + it jJit ; it+1 = 1 = 0; and ^ L ; ^ M and ^t are obtained from 55 ^ it+1 is from the second stage probit regression. The the …rst stage least squares regression and P results of the industry speci…c regressions are presented in Table 11. Overall the augmented OP procedure works well in correcting for input endogeneity and selection bias. Varible inputs have no explanatory power in the second stage, as emphasized in Ackerberg, Benkard, Berry and Pakes (2007). Firm productivity is constructed based on the following results which forms the basis of the empirical exercise,38 Non-Woven: ln TFP_AOPit = yit 0:156mit 0:283lit 0:303kit ; (37) Woven: ln TFP_AOPit = yit 0:549mit 0:357lit 0:122kit : (38) Note that, since the production functions are estimated separately for the two industries, I restrict the empirical exercises only to within …rm comparisons of productivity over time, in order to avoid questionable cross-regression comparisons.39 For comparison, without any correction, the TFP estimates from OLS are constructed as the following: Non-Woven : ln TFP_OLSit = yit 0:177mit 0:416lit 0:121kit (39) Woven : ln TFP_OLSit = yit 0:524mit 0:396lit + 0:013kit : (40) References [1] Ackerberg, Daniel, Benkard, C. Lanier, Berry, Steven, and Ariel Pakes (2007). “Econometric Tools for Analyzing Market Outcomes”. In J.J. Heckman (Ed.) The Handbook of Econometrics, Volume 6. 38 How di¤erent are these estimates compared to Demidova, Kee and Krishna (2012), when market speci…c demand shocks are controlled for instead of the FDI status of the …rms? While the point estimates of L ; M; and K are slightly di¤erent between the two versions, simple t-tests reveal that the di¤erences are not statistically signi…cant with 95% con…dence level. 39 There may be a concern that the non-woven industry appears to have decreasing returns to scale, based on the point estimates of Equation (37) ; ^ M + ^ L + ^ K = 0:8: I tested for the following null hypothesis of constant returns to scale: H0 : M + L + K = 1: Based on the bootstrapped standard error of 0.33, the t-statistic is -0.61, which is not statistically di¤erent from 0. Thus, the constant returns to scale hypothesis is not rejected. 56 [2] Blomstrom, Magnus and Hakan Persson (1983). “Foreign Investment and Spillover E¢ ciency in an Underdeveloped Economy: Evidence from the Mexican Manufacturing Industry,” World Development 11(6), 493-501. [3] Blomstrom, Magnus and Edward N. Wol¤ (1994), “Multinational Corporations and Productiv- ity Convergence in Mexico," in Baumol, William J., Richard R. Nelson, and Edward N. Wol¤ (eds.), Convergence of Productivity: Cross-National Studies and Historical Evidence. Oxford: Oxford University Press, 263-84. [4] Caves, Richard E. (1974). “Multinational Firms, Competition and Productivity in Host-Country Markets,” Economica 41(162), 176-93. [5] De Loecker, Jan (2007). “Do Exports Generate Higher Productivity? Evidence from Slovenia,” Journal of International Economics 73, 69–98. [6] Demidova, Svetlana, Hiau Looi Kee and Kala Krishna (2012). “Do Trade Policy Di¤erences Induce Sorting? Theory and Evidence From Bangladeshi Apparel Exporters,” Journal of Inter- national Economics 87(2), 247-261. [7] Djankov, Simeon and Bernard Hoekman (2000). “Foreign Investment and Productivity Growth in Czech Enterprises.” World Bank Economic Review 14(1), 49-64. [8] Eslava, Marcela, John Haltiwanger, Adriana Kugler, and Maurice Kugler (2004). “The E¤ect of Structural Reforms on Productivity and Pro…tability Enhancing Reallocation: Evidence from Colombia,” Journal of Development Economics 75(2), 333-371. [9] Findlay, Ronald (1978). “Backwardness, Direct Foreign Investment, and the Transfer of Tech- nology: A Simple Dynamic Model,” Quarterly Journal of Economics 92(1), 1-16. [10] Haddad, Mona and Ann E. Harrison (1993). “Are There Positive Spillover from Direct Foreign Investment? Evidence from Panel Data for Morocco." Journal of Development Economics 42(1), 51-74. [11] Haskel, Jonathan, Sonia Pereira and Matthew Slaughter (2007). “Does Inward Foreign Direct Investment Boost the Productivity of Domestic Firms?” Review of Economics and Statistics, 89(3), 482-496. 57 [12] Konings, Jozef (2001). “The E¤ects of Foreign Direct Investment on Domestic Firms.” Eco- nomics of Transition 9(3), 619-633. [13] Markusen, James R. and Anthony J. Venables (1999). “Foreign Direct Investment as a Catalyst for Industrial Development,” European Economic Review 43, 335-356. [14] Olley, G. Steven and Ariel Pakes (1996). “The Dynamics of Productivity in the Telecommu- nications Equipment Industry,” Econometrica 64(6), 1263-1297. [15] Poole, Jennifer P. (2013). “Knowledge Transfers from Multinational to Domestic Firms: Evi- dence from Worker Mobility,” Review of Economics and Statistics 95 (2), pp. 393-406. [16] Van Biesebroeck, J. (2006). “Exporting Raise Productivity in sub-Saharan African Manufac- turing Firms.” Journal of International Economics 67(2), 373-391. 58 Table 11: Dependent variable: Log of output (1) (2) (3) (4) (5) (6) Industry Non-woven Non-woven Non-woven Woven Woven Woven Dependent Variable yit yit yit+1 0:156mit+1 0:283lit+1 yijt yit yit+1 0:549mit+1 0:357lit+1 Materials 0.177*** 0.156*** -0.004 0.524*** 0.549*** 0.013 (0.051) (0.046) (0.040) (0.044) (0.045) (0.028) Labor 0.416*** 0.283*** -0.019 0.396*** 0.357*** -0.012 (0.086) (0.099) (0.085) (0.076) (0.085) (0.056) Capital 0.121*** 0.303*** -0.013 0.122*** (0.048) (0.081) (0.032) (0.053) Age -0.085 -0.226 59 (0.281) (0.162) FDI -0.370 -0.421 (0.555) (0.305) Endogeneity correction1 No Yes Yes No Yes Yes Selectivity correction 2 No No Yes No No Yes Year …xed e¤ects No Yes Yes No Yes Yes Observations 387 346 254 826 729 558 Notes: Heteroscedasticity corrected white robust standard errors in parentheses. 1 A 2rd order polynomial function of age, capital, investment, FDI status are included. 2 A 3rd order polynomial function of propensity to stay in business and the …tted output net of labor and capital are included. Columns (2) and (5) are for observations with positive investments. Columns (3) and (6) lose one year of observations due to the lead variables. *, **, *** indicate statistical signi…cance at 90%, 95% and 99% con…dence levels.