WPS5407 Policy Research Working Paper 5407 Export Entrepreneurs Evidence from Peru Caroline Freund Martha Denisse Pierola The World Bank Development Research Group Trade and Integration Team August 2010 Policy Research Working Paper 5407 Abstract This paper examines firm entry and survival in exporting, discovered. The results imply that high sunk costs of and in products and markets not previously served by entry are of concern for product discovery, especially any domestic exporters. The authors use data on the for products that are not consumed domestically. In nontraditional agriculture sector in Peru, which grew contrast, the tremendous entry and exit in exporting seven-fold from 1994 to 2007. They find tremendous and in new markets suggests that initial sunk costs are firm entry and exit in the export sector, with exits more relatively low. The authors develop a model that explains likely after one year and among firms that start small. how entrepreneurs decide to export and to develop new There is also significant entry and exit in new markets. In export products and markets when there are sunk costs of contrast, such trial and error in new products is rare. New discovery and uncertainty about idiosyncratic costs. The products are typically discovered by large experienced model explains many features of the data. exporters and there is increased entry after products are This paper--a product of the Trade and Integration Team, Development Research Group--is part of a larger effort in the department to understand export growth. Policy Research Working Papers are also posted on the Web at http://econ. worldbank.org. The authors may be contacted at cfreund@worldbank.org and mpierola@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 Export Entrepreneurs: Evidence from Peru* Caroline Freund World Bank Martha Denisse Pierola World Bank * Contact information: cfreund@worldbank.org, mpierola@worldbank.org. We are very grateful to Richard Baldwin for extensive discussions and comments on an earlier draft, and to the private sector representatives who kindly agreed to discuss the development of the industry. We are also grateful to Andrew Bernard, Ana Paula Cusolito, Leonardo Iacovone, Marc Melitz, Peter Neary, Emanuel Ornelas, Nina Pavcnik, and seminar participants at the American Economic Association Meetings (2010), European Workshop on International Trade (2010), London School of Economics, Oxford University, the World Bank, and the Globalization Conference sponsored by Southern Methodist University and the Dallas Federal Reserve Bank. This paper received financial support from the governments of Finland, Norway, Sweden and the United Kingdom through the Multidonor Trust Fund for Trade and Development. The views presented here represent the views of the authors and do not represent the views of the Board of the World Bank. I. Introduction Recent empirical work highlights high rates of entry and exit into exporting, and explores the way in which exporters access foreign markets.1 While the standard heterogeneous-firm model (Melitz 2003) does an excellent job of explaining empirical findings on exporters' characteristics,2 it is less equipped to explain these entry and exit patterns and how foreign markets are accessed. It uses comparative statics to explain entry and exit, and as a result, it cannot explain why a firm would enter and then immediately exit exporting with no change in trade costs, as is observed frequently. It also cannot explain how firms develop new products or enter new markets. Studying these factors is important, especially in dynamic sectors, where these entries account for a large share of export growth. In this paper, we explore the role of idiosyncratic uncertainty and sunk entry costs in explaining why many firms enter the export sector and then exit almost immediately. We also examine entry and exit into new products and markets, and describe characteristics of the pioneers. While some of our results are consistent with earlier work, our contribution is to uncover precisely why entry and exit in exporting is so 1 Eaton, Kortum and Kramarz (2004) examine French data and find that most firms sell to only one market, typically the most popular one; while some firms that export widely serve the less popular markets. Eaton et al. (2008) examine data from Colombia and find extremely high entry and exit rates into exporting--total entrants in a given year exceed the number of continuing firms and most entrants exit after one year. Alvarez and Lopez (2008) use data from Chile and also find high rates of entry and exit. Volpe Martincus and Carballo (2008) examine exports from Peru from 2001 to 2005 and find that large firms export more products to more markets. Iacovone and Javorcik (2010) examine data from Mexico from 1994-2003 and find that new exporters tend to start small and that there is a lot of churning of products within firms. 2 A number of studies find that exporting firms are bigger, more productive, pay higher wages and offer better working conditions than otherwise similar import-competing firms. Bernard and Jensen (1995) report detailed statistics for the United States. A number of papers followed their approach and find similar results in both developing and developed economies. Shank, Schnabel and Wagner (2007) provide a summary of these papers, and offer similar evidence for Germany. Bernard, Jensen, Redding, and Schott (2008) also provide a summary. 2 common and identify stylized facts about new product and market development within a dynamic industry. Using exporter-level international transactions data, we focus on the nontraditional agricultural sector in Peru, which grew more than 700 percent from 1994 to 2007 (compared with 450 percent for traditional agriculture).3 Entry of new firms and expansion into new products and markets was an important part of the story. Specifically, exporters that began exporting after 1994 account for nearly three-quarters of total exports in 2007. Similarly, exports of products new to the country and entry into new markets for existing products together account for almost a third of total exports at the end of the sample period. We have three main findings about the way firms access markets abroad: (i) Firms start exporting with small trials and there is significant entry, exit, and reentry, implying that entry into exporting does not entail large initial fixed costs. (ii) Entry into new markets (for an existing export product) is more cumbersome, but the amount of trial and error suggests that entry costs are still not so large as to prohibit new markets from being discovered. And (iii) there are fewer trials in new goods, pioneers are typically relatively large exporters that are more successful than followers, and there is herding following product discovery. This suggests that finding new export products is more costly and that many new (and profitable) products may not be discovered because of high sunk costs. Interestingly, many of the new export products are not consumed domestically-- they are produced only for foreign consumption. This is a common pattern in developing 3 Official numbers from the Central Bank of Peru. 3 countries,4 which cannot be explained by firms exporting products that they are most efficient at producing for the home market. Exporting such products must involve discovery. We therefore also explore the distinguishing characteristics of firms that develop these "untasted" products. We develop a model that generates entry and exit as a form of trial and error. We extend the model to the case of new products and markets, where discovery costs are likely to be relatively large. Specifically, entrepreneurs first decide whether to enter the export sector, and then whether to continue exporting, and finally whether to develop new products that have not been exported previously by any firm (or similarly to access markets new to a specific product). Prior to entry, each exporter faces uncertainty about their cost of exporting a particular product, and once they export the cost is revealed. The uncertainty generates significant entry and exit--some entrepreneurs with a negative expected value of entry will attempt to export, and if their cost draw is bad they will exit. The intuition is that there is a lifetime value of getting a good cost draw and only a one- period negative shock from a bad draw. This implies that the present value from attempting exports can be positive even if the one-period expected gain is negative. It also means that with sunk costs of entry, there can still be significant entry and exit. We show that if small trials are possible, the range of firms which attempt exporting expands. In addition to entering existing markets and products, exporters can also start product lines that are new to the country (or enter markets that are new to the product line). Such development is relatively costly because the firm must develop a new product 4 For example, cut flowers in Kenya, coffee in Rwanda, semiconductors in Costa Rica, and flat screen TVs in China. Countries are increasingly setting up special programs such as export processing zones to encourage such production. There are now 60 million people working in 3,500 export processing zones spanning 130 countries producing clothes, shoes, sneakers, electronics, and toys for export (Boyenge 2007). 4 or meet new market requirements. The model shows that the quality of the pioneers in new products (and markets) is increasing in the cost of discovery. As a result, these entrepreneurs are less likely than followers to cease exporting these products after entry. Several other recent papers focus on related issues. Segura-Cayuela and Vilarrubia (2008) and Eaton et al. (2009) incorporate uncertainty that is alleviated as firms learn about a market. In Eaton et al. the uncertainty is firm specific while in Segura-Cayuela and Vilarrubia uncertainty about a market is reduced as more firms enter. In these models entry is suboptimally slow, in contrast, in our model greater uncertainty leads to more entry and exit by firms, except in the case of new products where the discovery cost is large. Like ours, the model of Albornoz et al. (2010) has uncertainty about the profitability of a particular market that is revealed when a producer enters a market. However, their focus is on the sequence of entry into new markets and not on entry and exit in existing markets and the development of products and markets that are new to the country. Hausmann and Rodrik (2003) offer a model of self-discovery, with uncertainty and high costs of starting a new product. Their model is similar is spirit to the entry into new products that we discuss. However, in their model the threat of imitation discourages firms from innovating and that leads to suboptimal discovery. In contrast, we show that discovery costs alone generate similar effects. The literature on multi-product firms also explores some of these issues, but it focuses on products or markets new to the firm and not to the country.5 5 The multiproduct firm models do a nice job explaining the efficient use of resources within a firm and how a trade shock alters within firm resource allocation, but they cannot explain the discovery of a completely new export product or market, one of the things we want to model. (See Nocke and Yeaple (2006), Eckel and Neary (2008), Bernard Redding and Schott (2010)). 5 Our theoretical framework is also related to the literature on hysteresis and trade flows, which shows that with sunk costs and uncertainty about market conditions, positive shocks that lead to entry may not produce exit when they are reversed.6 In these models, only bad market conditions induce exit and hence entry and exit will not be positively correlated. Our departure from these models is that we assume there is uncertainty about the firm's potential in a market. Specifically, export costs are revealed only if the firm enters, and the firm can exit if the cost is high. This generates a strong positive correlation between entry and exit, a feature confirmed in the data. In sum, our work builds on previous theoretical and empirical developments in the literature of exports at the firm-level, but instead of focusing on equilibrium effects, we focus on the dynamics of a growing sector. In particular, the patterns of entry and exit of firms in exporting and the discovery of new products and new markets. The paper is organized as follows. The next section develops the model. Section III examines the predictions from the model using transactions level data from customs. Section IV offers background information on the nontraditional agriculture exports in Peru that supports the findings from the previous section. Finally, Section V concludes. II. Model Before developing the model, we use an example to highlight the issues that we wish to address. Consider three entrepreneurs that want to access foreign markets. Sr. Lopez wants to start an export business but he does not know whether the cost of exporting will be prohibitive. He must gather information on regulations and paperwork 6 See Baldwin (1988), Baldwin and Krugman (1989), and Dixit (1989). Roberts and Tybout (1997) also use this framework and find evidence that sunk costs are important in explaining entry into exporting by Colombian firms. 6 required for his products to be shipped abroad. After paying this entry cost, his success will depend on the overhead cost of exporting he faces. Sr. Martinez, an entrepreneur already established in foreign markets, wants to break into new markets. He needs to find the right distributor in the new destination and market his products so that they will be appealing to his new customers. Subsequent to this investment, he also faces uncertainty in delivery costs that will determine profitability and survival. Finally, Sra. Nuñez is considering how to develop an export product nobody in the country has ever sold abroad. Her decision about whether to invest in product development depends on the magnitude of discovery costs and whether another firm has already taken the lead and she can save time and money on establishing new production techniques. This model is about the collective experience of the many entrepreneurs like these three in their attempts to break into foreign markets. There are several important features of exporting that we want to capture in the model. First, there are heterogeneous entrepreneurs in terms of ability. The ability of the entrepreneur is related to management skills and technical knowledge. Second, there is idiosyncratic uncertainty--a firm does not know how costly it will be to export a particular product to a given market until the firm tries. Third, there is a sunk cost of entry into exporting, reflecting changes to the product, required paperwork, and the gathering of market information that must be completed before exporting. The model is meant to be illustrative and highlight the way entrepreneurs behave; it does not take into account general equilibrium effects.7 We first describe the basic model then we discuss 7 We abstract from the precise production function in terms of labor and capital because when we go to the data, we will only observe exports. 7 how the model changes if small trials are possible. Finally, we discuss how the model can be adapted to describe entry into new products and new goods. i. Basic Model We start with an entrepreneur, of type i, where ranges from 0 to 1, and a higher represents a more productive entrepreneur. It is the amount of product the entrepreneur can produce and it is known by him from the beginning. In this model, there are two different markets: foreign and domestic. If a product is sold in foreign market k (k denotes the product-market combination), the entrepreneur receives price Pk, which is known. For example, an entrepreneur can observe the price of a specific product in a specific market and knows how much he can produce, thus he has a very good estimate of potential revenues from that product-market combination. If the product is sold domestically, the entrepreneur charges a price PD. Foreign and domestic markets entail distinct costs. An entrepreneur serving the foreign market pays a sunk entry cost and a fixed per-period cost of exporting (i.e. a fixed overhead cost). An entrepreneur selling to the domestic market pays only a fixed per- period cost. Specifically: Ck is the overhead cost that a firm pays to export to foreign market k. This cost is associated with bureaucracy and logistics. This cost is unknown to the entrepreneur before exporting, and it is not revealed until he exports. The entrepreneur has an expectation of what this cost will be before trying to export. Specifically, with probability q he gets a low cost draw, CkL, and with probability (1-q) he receives a high cost draw, CkH. F is a sunk cost of entry into a foreign market. This is the cost that the entrepreneur has to incur to adapt his factory or his land to produce a particular product for export. CD is the overhead cost that the entrepreneur pays to serve the domestic market. 8 We assume that the overhead cost of exporting, Ck, is larger than the cost in the domestic market CD. The intuition is that exporting requires the producer to get the product through local distribution to the ports as well as through foreign distribution. In addition, we assume that for export goods the price in the foreign market, Pk, is larger than the price in the domestic sector, PD.8 Given the higher costs of accessing the foreign market, Pk must be greater for the entrepreneur to have incentive to export to that product-market. The sequence of decisions to be made by the entrepreneur is the following. First, the entrepreneur faces the decision of whether to enter the export sector or the domestic sector. If the entrepreneur goes to the domestic sector he earns i PD and pays CD. He receives profits (i PD - CD ) for life, discounted at the rate . If the entrepreneur enters the export sector, he earns i Pk and pays the realization of the overhead cost of exporting, Ck, plus the sunk cost F in the first period. As noted above, there are two possibilities for the cost of exporting: with probability q, the exporter will obtain a low cost, CkL, and with probability (1-q), he will obtain a high cost, CkH. To concentrate on the trade-off that is important in the data, we impose a number of regularity conditions on the parameters. First, we assume that Pk-CH>PD-CD, so that exporting is always more interesting than domestic sales on a period-by-period basis for a firm with the highest quality. Second, we assume that the sunk cost, F, is small enough such that some entrepreneurs attempt exporting even if they may exit ex-post. 8 For Peruvian agricultural exports considered here this is a sensible assumption, in part because Peru exports many products when the US and Europe and other northern hemisphere markets are not producing. Still, to check we identified four nearly identical products included in the Peruvian and US CPI, oranges, tomatoes, (delicious) apples and bananas. These products are all exported from Peru. We compared monthly prices in $US for one kilogram of each product. On average the US price was 4.5 times higher in 2008 and 2009, and ranged from 2.8 to 5.3. 9 Specifically, the condition is that there exists an i, such that expected lifetime profits given the entry cost are positive, but given a high overhead cost the firm prefers to exit i Pk E (C k ) P CD ( F i D & i Pk C H i PD C D ) , where E(Ck) is the expected 1 1 overhead cost of exporting. Later, in a sub-section (II.iii), we discuss the situation when entry costs are large enough to preclude an enter-exit strategy. Now, we can solve the model backward. We examine what happens in the second period to a firm that entered the export sector in the first period. The decision is whether to stay in or exit the foreign market given the realization of Ck. This will depend on the profits from staying versus shifting to the domestic sector. Subsequent to entry, the profits from staying in the export sector are Profitstay = 1 ( i Pk C ik ), and the profits from exit are 1 Profitexit = 1 ( i PD C D ). 1 The threshold , above which firms choose to stay in the export market (stay), can be calculated from comparing exporter profits if he stays in the foreign market forever (Profitstay) and profits if he exits the foreign market after one period and goes to the domestic sector (Profitexit). Profitstay must be larger than or equal to Profitexit for the entrepreneur to continue exporting. This implies that the threshold for staying in the export market is (1) C ik C D stay (C ik ) , Pk PD 10 Where Cik CkL , CkH . Given the regularity conditions mentioned above, we know that stay is positive. All entrepreneurs with an i equal to or above this threshold, given the realization of their overhead cost, will continue exporting. Now, having solved for the cutoff stay in the second period, we go back to the first period and solve for the threshold level of for the entrepreneur to enter the export sector. In order for an entrepreneur to enter the export sector, it must be the case that the value of entry exceeds the value of going to the domestic sector. There are two possibilities for entry. In the first case, an entrepreneur enters and stays in the foreign market irrespective of the cost draw. This is the case for highly productive entrepreneurs, those with always above stay in Equation 1. This yields the value function of entry (2) 1 1 V Xistay ( , Pk , C kL , C kH , F , PD , C D ) q ( i Pk C kL ) (1 q ) ( i Pk C kH ) F . 1 1 In the second case, an entrepreneur enters the export sector and stays only if he receives a low cost draw --he exits the foreign market if the cost is high. This is the case for firms with above stay(CL) but below stay(CH).9 The value function in this case is (3) V Xiexit ( , Pk , C kL , C kH , F , PD , C D ) q 1 ( i Pk C kL ) (1 q ) ( i Pk C kH ) ( i PD C D ) F . 1 1 For firms to choose to enter the export sector, the expected value of attempting export (Equation 2 or 3, depending on ) must be larger than the value of producing for the domestic sector. The value of selling domestically, VDi, is 1 (4) VDi ( i PD C D ). 1 9 Note that an entrepreneur will never enter and then exit if the cost draw is low. If the value of entry (where cost is unknown) is greater than being in the domestic sector then it must be the case that the value of staying with a low cost draw is better than being in the domestic sector since CLVwait, which yields the additional condition: . (14) (1 ) (qC kL (1 q )C kH ) (1 ) F pioneer ((1 2 p ) D (1 ) R) . PK (1 2 2 p ) Pk This implies that is higher for a new product also because there is a value to waiting. When the discovery cost, D, is big then this condition is hard to satisfy. No individual firm wants to expend D to find a new product, even if once found it is profitable for all exporters. Similarly, if the opportunity cost of investing in a new product, R, is large then is greater. Provided D is not too small, the cutoff is increasing in the probability because as p goes up the benefit from waiting expands. Finally, the cutoff rate is decreasing in the profitability of the product. The greater is Pk and the lower is the average Ck, the more likely is discovery because the gain from starting the product is high. (2008). The main insight from this literature is that there will be a suboptimal rate of discovery (or technology adoption) because firms would rather wait to invest. 18 After products are discovered and the cost of discovery is no longer relevant there will be increased entry, and the cutoff falls to equation (5). Because firms are waiting for others to expend discovery costs, the rate of discovery is suboptimal. There can be products that can be exported competitively but which are not exported because of the high discovery cost a single firm must face. This is the standard problem of innovation. If the gains are relatively greater for followers then there is little incentive to innovate. This is true even if imitation does not erode the pioneers products, but simply because imitation is less costly. In sum, developing new products requires a much larger entry cost, because the production process is very different for these products. Similarly, this may be the case for entering distant markets where new standards must be met. This implies that firms that start new products or new markets are likely to be the better firms, and these firms are likely to have a lower exit rate than later entrants, all else equal. If fixed costs of discovery are large, after successful products (or markets) are found, there will be herding into those markets as such costs fade. Several testable predictions come out of the model: 1. Size and quality. There is self selection into exporting with high and medium productivity entrepreneurs exporting (they are on average more productive than the average in the industry). The highest productivity exporters will enter and survive in more products and markets on average, and export more to each product-market. 19 2. Entry and exit patterns. This model naturally generates entry and exit by the same firm. Exit is especially likely after the first attempt. This yields a positive correlation between entry and exit. Weaker entrepreneurs (small entrants) are more likely to exit. Weaker firms will enter into exporting with small trials in order to avoid high entry costs if they receive a bad cost draw. As a result of entry and exit, in the first year of a given cohort, there will be more different quality types of firms in the export sector. This implies that the variation with respect to the mean of exporters' size should decrease with age of the firm. Many of the lower quality exporters will exit, while some that receive a good cost draw will expand. 3. New products and new markets. Firms that pioneer new products or new markets tend to be high productivity (large) firms, and have a lower exit rate than later entrants. After a product or market is discovered and discovery costs disappear, there is more entry (herding). III. Empirical Evidence from Transactions Data In this section, we examine whether the predictions of the model are consistent with the Peruvian experience in the nontraditional agricultural sector. This is a particularly dynamic sector (Figure 3), which began a period of rapid growth in the mid to late 1990s. The product that mainly explains this surge is asparagus, but there is also considerable growth in the exports of other nontraditional crops (in particular, prepared/preserved artichokes, avocados, paprika, grapes and mangos) in recent years (Figure 4). 20 Although several factors contributed to the surge,13 firm entry into exporting and the discovery of new products and markets was an important part of the story. Thus, the study of this sector allows us to examine the transition to equilibrium, and in particular the export decisions by firms as was explored in the model in the previous section. We start by describing the data and then we proceed to explore the predictions from the theoretical framework. i. Description of the Data We use transaction data on Peruvian export flows included within Chapter 7 (Edible vegetables and certain roots and tubers), Chapter 8 (Edible fruit and nuts; peel of citrus fruit or melons), Chapter 9 (but only the lines related to the exports of paprika) and Chapter 20 (Preparations of vegetables, fruit, nuts or other parts of plants) of the HS Code. Although we have daily information on all shipments between years 1994 and 2007, for much of the analysis, we report annual results. The dataset allows the identification of the exporter (information on firms' names and corresponding Tax ID number), the destination market for each trade flow, the custom port from which the merchandise is shipped, the description of the item exported (at 10-digit) and the FOB value of each shipment. The values exported by year/date of the different products under analysis in this study (i.e. asparagus, prepared/preserved artichokes, avocados, mangoes, paprika, grapes, 13 Two important factors that led to large scale investment were land privatization in 1993, which removed constraints on the size of plots, and the capture of the head of the Shining Path in 1992, which greatly improved investor confidence in rural areas in Peru. Some additional conditions accelerated this investment. The introduction of the drip irrigation system (imported from Israel) in the late 1980s was completed in the late 1990s. The Andean Trade Preferences Act (then extended under the ATPDEA) eliminated the tariffs for the Peruvian exports of asparagus to the United States from 1993. Currently, these preferences have been included in the FTA signed between Peru and the United States. 21 etc.) include all the relevant lines and items of the HS code. In the definition of each product, we included all those lines related to the exports of each product in its different forms/presentations.14 After collapsing the information by firm, year, market and product we obtain 16,053 observations. The summary statistics of the data (by firms, products and markets) can be found in Annex 1. The details of all the lines or items included in the definition of each product can be found in Annex 2.15 To analyze the model's predictions, we split the presentation of the evidence up into three parts. The first part focuses on exporting firm characteristics: correlations between the number of markets and products and size of firms. The second part focuses on entry and exit of firms into exporting. The third part focuses on entry into new products and markets. ii. Characteristics of the Exporters This section explores the main characteristics of the exporters in our sample relating size across products and markets. The model suggests that higher quality entrepreneurs export more to a given product-market, export to more markets, and export more products. Since we cannot directly estimate firm quality, we examine whether 14 We also proceeded this way to avoid problems associated with changes in the product classification in the HS Code in 1996 and 2002. 15 As part of the data cleaning process, we eliminated trade flows registered under the name of individuals that showed erratic patterns (i.e., exports of tiny amounts for one or few years not consecutively registered to an individual). These individuals are 579 of a total of 2,676 exporters (see also Annex 1), and on average, they represented 1.5% of the yearly total amount exported during 1993 and 2007. Market participants informed us that these are individuals, so-called "gatherers," who buy from small farms and sell on an agricultural exchange. If we include them in our sample, none of the results change dramatically, except that the one-year exits are more extreme. In order to avoid the inclusion of export flows that could be related to the export of product samples (we observed many erratic flows in very small values), we set a threshold as a filter. Specifically, we excluded exports flows that were less than US$ 1,000 a year, after collapsing by firm, product, market and year. We checked the robustness of all of our findings under different scenarios for this threshold (US$50, US$100, US$200 and US$500) and found that the substance of all of our results holds. 22 larger exporters export more to a given product-market, export to more markets, and/or export more products. Figure (5a) plots average exports in the product-markets to which a firm exports against average size for the beginning and end of the sample period (1994 and 2007). It shows that the largest firms export more on average to each product-market. In addition, we observe that larger firms not only tend to export more of a product to a given market, but they also export to more product-markets--if they only export to one product-market and grow the picture would be the 45o line. Figure (5b) confirms that larger firms export more products. While this fact holds throughout the period, it appears to strengthen over time. For instance in 2007, we observe relatively more firms exporting a large number of products (above 5) and most of these firms are in the upper half of the distribution of firms by size. Similarly, larger firms export to more markets, especially in the last years, where we observe that most of the firms exporting to more than 10 markets are located in the upper fourth of the distribution by size (Figure 5c).16 iii. Entry and Exit in Exporting This section examines the pattern of entry and exit into exporting across the firms in our sample. The model suggests that we should observe a large number of entries and exits, and entries and exits will be positively correlated. In addition, exit is more likely in the first year and among firms that start small (relative to other entrants). 16 In all comparisons, we have evaluated the pattern for each of the years included in the sample and we observe the same: larger firms export more products and to more markets and this trend accentuates with time. However, we only report the results for years 1994 and 2007 for simplicity in the presentation of the results. 23 Figure 6 shows firm entry and exit by year. Entry and exit is common. The number of entries and exits has increased throughout the period; however, the entries have remained higher than the exits for most of the period analyzed. Another striking result is the correlation between entries and exits (0.87).17 As we will see below, this can be explained by the large number of exits after the first year, thus when entry increases, we expect to see exit increase the next year. Entries are very important in terms of the development of the industry. Figure 7 shows the cumulative market share in 2007 by cohort in the traditional and nontraditional agricultural sectors.18 Firms that enter during the period under analysis in the nontraditional products make up nearly three quarters of exports by 2007. This differs from what is observed in traditional products where entries are important but to a lesser degree, making up just over 50 percent of exports by the end of the period. Although entries' importance in both sectors is similar during the first years in our sample, after 1998, importance of entries in traditional products begins to lag behind relative to nontraditional products. This is consistent with the take-off observed in the nontraditional sector, which begins in 1998, as shown in Figure 3. In fact, many of the entries that occur in the nontraditional products correspond to large and growing firms. In particular, the strongest entries happened in 1998, 1999, and 2001 with firms that combined 17 An observation is considered an entry if a firm was not exporting in the previous year. It is considered an exit if it disappears in the next year. 18 More complete set of statistics by cohort (in terms of the number of firms, the total and average value they represent) can be found in Annex 3. The products (and their respective HS codes) considered within the group of "traditional" exports are (as classified by the Central Bank of Peru): - Cotton (5201/5202) - Coffee (0901) - Sugar and molasses (1701/1703) - Wool (5101/5102/5103/5104) - Raw hides (pieles) (4101/4102/4103) - Coca leaves (1211300000) 24 concentrated one-third of the market in 2007. These strong entries corresponded to Sociedad Agricola Drokasa in 1998, Camposol in 1999 ­two of the largest exporters- and a Consortium of fruit producers in 2001.19 Figure 8 presents the average number of exits according to the age of the exporting spell. We observe a drastic decrease in the average number of exits after the first year of exporting. In particular, we observe that in 667 exporting attempts, exporters cease to export after their first year of operation. Then, for spells that lasted at least two years, on average, only 271 came to an end after their second year of operations. If we translate the exits into the fail rate by age group (Figure 8b), we observe that the decrease in the fail rate remains, although it is less abrupt. For instance, a one-year old spell has a 34% probability of failure (exiting the market), a two-year old spell has a 27% chance of failure. This declining trend continues as the attempts last longer. Who are these exits? A large part of them are occasional exporters that try with only one shipment. Figure 9 shows the distribution of the annual number of shipments exported by firms that lasted only one year. Fifty-six percent of these single-year firms exported only one shipment.20 In addition, the model suggests that lower quality entrepreneurs are more likely to exit. To examine this hypothesis, we develop a binary variable for the entrants that is one if the firm exited after one year. We expect exits to occur more frequently among low 19 According to the export transaction data from SUNAT, Camposol exported for the first time using that name in 2002. However, based on information obtained from the company's website and during an interview with a representative of the company, we observed that Camposol started to export in 1999 under the name of Sol Produce (and a different id number), previous name of the company and one of the brand names that the company uses today for its exports of packed asparagus. We took note of that fact and we combined the export transaction data from both companies and treated them as one under the name of Camposol. 20 We made a similar calculation for the group of individuals with single-year entries and obtained an even larger percent: 60% of the individuals that lasted one year exported only one shipment. 25 quality entrepreneurs. Low quality entrepreneurs are also likely to start with smaller exports, so as to extend a small share of the fixed entry cost. In Table 1, we report results from a Probit regression of exit after one year on the log value of exports during the initial entry, controlling for crop, market and year (Column 1) ­results from a similar regression using OLS are reported in Column 2. We find a robust negative relationship, indicating that a ten percent larger entry is associated with about a 1 percent lower likelihood of exit. However, we know that many of these exits occur after the initial shipment. Therefore, firms may all start with similar size shipments, with some firms exiting after one shipment while others continue. This would generate a negative relationship between size and exit in the annual data, but only because firms that exit have fewer shipments. To control for this possibility, we also regress exit on the log value of the initial shipment exported by each firm (Column 3) ­results from a similar regression using OLS are reported in Column 4.21 We find that a ten percent larger initial shipment is associated with a 0.3 percent lower chance of exiting the market after the first year. The smaller coefficient suggests that part of what is driving the coefficient at the annual level is variation in the number of shipments. Frequent entry and exit imply that the sunk costs to entry into exporting are not large. In addition to the high number of entries and exits observed in the data, additional evidence of the presence of small sunk costs (for the entry into exporting) is the observed pattern of re-entry of some firms in our sample. Not all firms enter and exit exporting only once. There are 194 firms (almost 10% of the total number of firms, excluding individuals, in our sample) that reenter after a few years (see also Annex 1). This is not consistent with very high sunk costs on entry. 21 We also tried Logit and results are similar, not reported. 26 Finally, we examine the distribution of the size of firms as they age. In Figure 10 we present the residuals from a regression of size on age, controlling for main product exported and year. In the Figure 10a we observe that the variation in the residuals among firms declines significantly as they age. Also, we analyze the distribution of these residuals in two different ages (Figure 10b) and we confirm that there is less dispersion among firms as they age from their first year to their ninth. This is consistent with weak firms with high cost draws exiting and weak firms with low cost draws expanding. In sum, we observe considerable entry and exit of exporters each year; they are positively correlated; exit is especially likely after the first year and among firms that start small; there is less variation in terms of size as firms grow older; entry is important, accounting for two-thirds of total exports in 2007. All of these findings are consistent with the model, where entry and exit are a form of trial and error, and initial sunk costs are not very high. iv. Innovation: The Discovery of New Products and New Markets This section examines which exporters (by size and experience) are the first to enter new products and new product-markets (defined at the country level). Once they enter new products and markets, we also examine the development of the industry. The model suggests that, when sunk discovery costs are high, larger exporters will be more likely to start new products (or markets), that they will be more successful in surviving in the export of these new products (or new markets) and that there will be herding after successful products (or markets) are discovered. 27 A product is defined as "new" in our sample if the product was not exported from Peru in 1994 (the first year of our sample) and was later exported for at least four years consecutively at any time within our sample.22 A product is defined as "old" if it was exported for at least for two years consecutively starting in 1994. All cases not covered by these definitions are either intermittent products or products that were exported only once in our sample. In these cases, we dub these products "trials", unless exports are either left or right censored. New markets are defined at the product level, in a similar fashion. Specifically, a product-market combination is "new" if it was not served in 1994 and then was later covered for at least four consecutive years. A product-market combination is defined as "old" if it was covered consecutively for at least two years starting in 1994. And cases not covered by the types described above are either intermittent product-market combinations or product-markets that have been covered only once according to our sample. In these cases, we define product-markets as "trials", with the exception of product-markets whose coverage is left or right censored. Using these definitions, exports of new products made up 12 percent of the value of exports in 2007, and exports of old products to new markets made up 16 percent of exports. Thus, without these discoveries, growth would have been significantly slower. Table 2 shows the distribution of the various types of products and product- markets over the sample period. New products make up 20 percent of the total number of products that are exported and new product-market combinations are 25 percent of the total number of product-markets served. An important difference between products and 22 We excluded from this group the products that never exceeded a total amount exported of US$10,000 in any of the years included in the sample. The only products excluded for this reason are "carrots, turnips & other edible roots, frozen or chilled" (all grouped under HS codes 0706). 28 markets is the amount of trials. Market trials are commonplace, with 496 new market attempts in specific products, or 62 percent of the total number of product-market combinations served in the sample period being trials, i.e. unsuccessful. In contrast, in products there are 10 trials, which amount to only 17 percent of the observations. Thus, the data indicate trials are much more likely in markets, suggestive of lower fixed entry costs. One caveat is that this could be related to the number of possible trials. Specifically, the universe of potential new products (the number of HS 6-digit lines) is significantly smaller than the universe of potential new product-markets (the number of HS 6-digit lines multiplied by the total number of markets/countries). As an alternative, which does not suffer from this potential bias, we compare the success rate of trials in both products and markets. As the model shows, a higher entry cost should yield a higher success rate, as only really good firms will try. We find that the success rate in products (55 percent) is significantly higher than the rate observed for markets (29 percent), which also points toward entry costs into new markets for existing products being relatively low. Table 3 presents the statistics on the characteristics of entrants23 in all the products that can be considered "new" in the sample from 1994 to 2007. If entry costs are high, we expect pioneers to be the better (larger) and more experienced firms. For pioneers in goods that are produced for export but are not consumed locally ("untasted" in the country), we expect stronger results as these are likely to involve the largest discovery costs. 23 Entrants are defined as all firms exporting in the first three years of the lifecycle of a specific product. 29 Column (1) shows the total number of firms that started to export each new product. Column (2) presents the exporters with previous experience as a share of the total number of entrants. In 68 percent of the cases, entrants are exporters with previous experience. The entrants in products belonging to the primarily for export (X) category are on average, more experienced than the entrants in goods that are produced for domestic consumption (D)--specifically, 76 percent of X exporters have experience, as compared with 56 percent of D exporters.24 Column (3) shows average value of the exports in the main product in the previous period relative to the exports of the average exporter in that product (mainly to observe the size of the entrants relative to their main competitors). Exporters that start new products tend to be larger on average by about 24 percent. However, if we decompose this in terms of the products that belong to each categories (D and X), we observe that the entrants in the X category are on average 59 percent larger than their main competitors, while entrants in the D group are not necessarily the larger exporters in the year previous to their entry. This last observation could be due to the fact that the exporters in this group compete mainly with producers in the domestic sector, which are likely to be smaller but are not taken into account in the average calculated for exporters since we do not have information on their size. Column (4) shows the average ratio of the count of products exported by the entrants over the average number of products exported by all the firms whose main product exported was the same as the entrant's main product during the year of entry into the new product. The 24 The D category refers to the new export goods for which there is a demand in the domestic market. The X category includes new export goods that are mainly produced for exportation. The grouping of new goods in terms of these two categories was based on information on domestic demand obtained from the Ministry of Agriculture (mainly for fresh produce), the National Institute of Statistics and the website of a supermarket (E.Wong). 30 ratios above one show that, on average, the entrants export more products than their main competitors. In all cases, the entrants export more products than their average competitor; however, again, this ratio is larger for the entrants in products of the X category. In addition, differences are observed between fresh produce and processed produce. The new products with more entrants are mostly products in the segment of fresh produce (avocado, passion fruit, piquillo pepper, etc.), suggesting that discovery may be easier in these products. Overall, these results imply that exporters of new products tend to be bigger, export more products, and be experienced exporters, and these results are particularly important for products that are "untasted" in the country, where discovery is likely to be more costly. Given that in the case of the pioneers in product-markets combinations, we have a larger sample (1,767 observations), we use regression analysis to complement the analysis of the characteristics of the entrants into new markets presented above.25 We create a dependent variable that is one if a firm is a pioneer in a product-market and zero if the firm is a late entrant into that product-market. We regress that on the size in the year before entry, experience in the product, and experience in the market (an indicator that is one if the firm served that product or market previously). Table 4 presents the results. Using Probit, we find that entrants into new product-market combinations are relatively large exporters (the coefficient of the size of the exporter in t-1 is positive and significant). Experience exporting the same product is positively and significantly 25 Again entrants are defined in terms of the firms that started to export during the first three years in the lifecycle of a particular new product-market combination. Here we make no distinction between products that are consumed domestically and products that are mainly exported because the focus in this part is on the discovery of markets "new" to the exporters in a particular product regardless of the type of product involved in the transaction. 31 correlated with pioneering new product-markets. Past experience in the same market is never significant. These results hold using OLS as well. This suggests that firms that pioneer new markets have experience exporting the product to other markets and they tend to be larger than other firms. Another feature suggested by the theory is that the pioneering firms must be higher quality if fixed costs of discovery are large. This suggests that pioneering firms are more likely to survive than followers. Table (5) shows the average one-year survival rates in new products and new product-markets for the group of pioneers and followers. On average, pioneers always survive longer than followers. The difference between the survival rate of pioneers and followers is larger in the case of the products that are mainly exported. Again, this implies that these products are the most difficult to develop. There is also a difference in survival rates of pioneers in new products and in new markets. The difference between the survival rate of pioneers in new products and of pioneers in new markets is positive. In contrast, the followers in both types have similar rates of survival. This offers additional evidence that the discovery costs of new products are higher than the discovery costs of finding new markets for an existing export product. To enter into new products, exporters need to be of very high productivity and thus are quite likely to survive, while to enter into new product-markets, the cutoff productivity level is somewhat lower. Finally, if discovery costs are large, we should observe herding after successful entry, when other firms can imitate this success without paying large sunk discovery costs. We now examine the pattern of imitation. Figure (11a) shows the mean and median of entry in new products (D and X) over the lifecycle of the new products. We 32 observe increasing entry a few years after discovery in the case of new products (both type D and X). Figure 11b shows a similar picture for new product-markets. Herding is less obvious: while the mean of entries increases over time, the median remains almost flat from the fourth year onwards and the scale is much smaller. Again, this suggests that the entry costs to new product-markets are not as high as the costs of discovering of new products, therefore, the role of the pioneers in new markets is not as strong as it is in the case of new products. The imitation that takes place in the case of new products could be the result of a product becoming more attractive--i.e. an increasing foreign price. Figure 12 shows the mean and median of unit values in the products. Peru appears to be largely a price taker. Thus, the increased entry appears to be the result of following the pioneers into the product rather than expanding foreign demand. In sum, we find limited evidence that sunk costs of discovering new markets discourage entry, and strong evidence that sunk costs of discovering new products discourage entry (especially in the case of products that are produced mainly for exportation). Both entrants in new products and new markets tend to be relatively large and more experienced. However, while trials are very common in new product-markets, they are very rare in new products. Also, the rate of success of entrants relative to followers is greater for products as compared with new product-markets, however, within products, this difference is greater in the case of entrants in products not consumed in Peru. We observe herding after entry in new products but not new markets. These results, taken with the results from the previous section imply that if there is a role for policy to stimulate entry, it is in new products where entry is rare. And it is especially 33 important in the case of products produced for exportation. The role for policy is much less for entry into exporting or into new markets. In the next section, we discuss anecdotal evidence on the discovery of new products that confirms that entry was in fact costly into many new products. IV. Anecdotal Evidence on Product Discovery in Peru The empirical work above suggests that discovering new products is costly, but once products are discovered imitation is relatively straightforward. Below, we describe briefly the story behind the development of the asparagus industry--the main Peruvian nontraditional crop--and then explain the discovery of other new crops. This anecdotal evidence offers further support for the presence of high discovery costs of new products. i. The Development of the Asparagus Industry Asparagus is the most important nontraditional crop. It was not consumed locally when it was first developed, and its exports began with direct market intervention. The production of asparagus started in the 1950's in the valleys of the North coast of Peru, with exports of canned white asparagus. The expansion into fresh asparagus was due to an experiment in the south of Lima, involving both the private sector and assistance from the U.S. Agency for International Development (USAID). The Ica Farmers' Association decided to explore options to replace traditional crops with export crops. With funding from USAID, many products were studied for this purpose (melons, paprika, green beans and asparagus); the one with the most profit potential was asparagus (Shimizu 2006). As a result, a new variety of seed designed for Peru (UC-157, created by an expert from the 34 University of California, Davis) was introduced successfully. USAID also provided funding for experts who advised on crop management, packing, and exporting. Fresh asparagus was first exported at the end of the 1980s, and in 2002 exports of fresh asparagus surpassed exports of canned asparagus. This highlights the potential role for intervention in finding new products. Next we turn to other more recent discoveries. ii. The Development of Other New Crops Recent and rapidly growing export crops, include among others, preserved artichoke and paprika. Like asparagus, both of these are not consumed domestically. The case of artichokes is especially interesting and provides evidence on the importance of sunk costs of discovery, and how networks and coordination help firms to overcome them. Artichoke exports were first attempted by the large asparagus firms. Several trial plots for artichokes were developed independently­according to different sector participants. However, the trials were costly and the farmers ultimately gave up. A seed distributor (Mr. Fumagalli) heard of these trials, studied the market for artichoke seeds and invited the exporters to present this information. As a result of this meeting, many of the attendees decided to conduct a large coordinated effort. The advantage was that they could try many seed varieties, climates, and irrigation techniques and share information on what was most efficient. This culminated in the takeoff of the exports of preserved artichokes; the trials revealed that the climate was inappropriate for fresh artichokes (Klinger 2007). The case of paprika is a case of pure private entrepreneurship. It was the initiative of a seed distributor (Mr. Chepote) who learned of paprika through a friend in Chile and 35 decided to try it in Peru. He formed a company that produced and exported paprika. They were successful on a small scale and with the help of Spanish investment expanded significantly. After the expansion was complete, a virus destroyed the whole crop and Mr. Chepote had to close down. However, due to the original success of paprika, Mr. Chepote marketed his knowledge to other producers and the exports of paprika took off (Klinger 2007). These stories show that the way Peruvian exporters decide to try new varieties is typically based on extensive research and development and in some cases market intervention or coordination. This evidence on what are now some of the biggest crops in Peru in combination with the empirical results above imply that discovering new products involves sizeable sunk costs. This suggests that there is a role for facilitating coordination among producers and subsidizing research. V. Conclusion We examine the development of nontraditional agriculture exports in Peru. Our theoretical framework assumes that there is idiosyncratic uncertainty about the profitability of exporting and that there are sunk costs of entry--this leads to a process of trial and error (observed in the industry), with a high share of exits after one year. Many firms start with small trials and increase their exports over time, in this way avoiding losses from potentially uncompetitive products. Entrepreneurs in large firms export more to a given product-market pair on average, enter more markets and more products, and enter new markets and products earlier. Because they are relatively high quality they survive in new products and new markets at a higher rate than subsequent entrants. These predictions are confirmed in the data. 36 The results highlight significant differences between entry into exporting, entry into new markets, and entry into new product lines. The large amount of entry and exit in exporting that we uncover, suggests that entry costs are not so large as to deter entry. As the model shows, this is true provided firms can enter small and sunk costs are not too large relative to lifetime gains. This appears to be the case for firm entry into existing products and existing markets. Firms entering new markets with old products face somewhat higher costs of entry, but still, the large amount of trial and error suggests that they are not excessive. However, completely new products are different. They are costly to introduce, which deters entry, especially since followers do not have to pay discovery costs. Firms that discover new products are larger and more likely to succeed than followers. There are few new product trials and there is herding after products are discovered. We also examine separately export discoveries of products new to the country that are not consumed locally. By definition these are products in which local producers have no initial expertise. There is no home market effect, and more productive domestic firms do not become exporters. Rather these are export products that entrepreneurs invest both time and money to develop. Theory suggests that only the highest quality entrepreneurs will discover such goods. Indeed, we find that entrepreneurs in these products are better (larger, more experienced, higher survival rate) than other types of pioneers in new goods, and that entry rates into such products are lower. In addition, there is herding after product discovery, implying that the cost of imitation is lower than of discovery. Our results imply that the rate of new product discovery is likely to be suboptimal, and therefore there is a role for government policy targeting the discovery of 37 new products. In Peru, in the early stages of the development of nontraditional exports, one form of government assistance was subsidizing producer-exporter associations (Diaz 2007). For example, IPEH, which promotes exports and competitiveness in asparagus and other nontraditional vegetables, was formed with government assistance (O'Brien and Diaz 2004). An important indication of its success is that it is now funded entirely by the private sector. Similarly, Volpe Martincus and Carballo (2008) find evidence that the main export promotion agency, PROMPEX, has helped stimulate exports more generally. More research into how to assist export discovery in developing countries, without introducing costly distortions, is warranted. 38 References Albornoz, Calvo Pardo, Corcos and Ornelas (2010) "Sequential Exporting" London School of Economics, CEP Discussion Paper 0974. Alvarez, R. and R. Lopez (2008) "Entry and Exit in International Markets: Evidence from Chilean Data," Review of International Economics 16(4): 692-708. Baldwin, R. E. (1988) "Hysteresis in Import Prices: The Beachhead Effect." American Economic Review, September 1988, 78(4): 773-85. Baldwin, R. E. and P. Krugman (1989) "Persistent Trade Effects of Large Exchange Rate Shocks." Quarterly Journal of Economics, November 1989, 704(4): 635-54. Banerjee, A. (1992) "A Simple Model of Herd Behavior" Quarterly Journal of Economics 107(3): 797-817. Bernard, A. and B. Jensen, (1995) "Exporters, Jobs, and Wages in U.S. Manufacturing: 1976­1987" Brookings Papers on Economic Activity. Microeconomics, pp. 67­119. Bernard, A. Jensen, B, Redding, S. and Schott (2007) "Firms in International Trade" Journal of Economic Perspectives, 21(3): 105-130. Bernard, A. S. Redding and P. Schott (2010) "Multi-product Firms and Product Switching" American Economic Review 100(1): 7097. Boyenge, S. (2007) "ILO Database on Export Processing Zones" ILO working Paper 251. Diaz, L. (2007) "Agro-industries Characterization and Appraisal: Asparagus in Peru" FAO, Agricultural Management, Marketing and Finance, Working document No. 23. Dixit, A. (1989) "Entry and Exit Decisions Under Uncertainty." Journal of Political Economy, June 1989a, 97(3): 620-38. Eaton, J. , M. Eslava, M. Kugler and J. Tybout (2008) "Export Dynamics in Colombia: Transactions Level Evidence" NBER Working Paper 13531. Eckel, C. and J.P. Neary (2008) "Multi-Product Firms and Flexible Manufacturing in the Global Economy" Oxford University. Mimeo. Eaton, J., S. Kortum, and F. Kramarz (2004) "Dissecting Trade: Firms, Industries, and Export Destinations" American Economic Review (94) 2: 150-154. Hausman, R. and D. Rodrik (2003) "Economic Development as Self-Discovery" Journal of Development Economics 72: 603-633. 39 Iacovone, L. and B. Javorcik (2010) "Multi-Product Exporters: Product Churning, Uncertainty and Export Discoveries". The Economic Journal 120: 481-499. Klinger, B. (2007) "Uncertainty in the Search for New Exports" Center for International Development at Harvard University, Working Paper No.16. Melitz, M. (2003) "The Impact of Trade on Intra-Industry Reallocations and Aggregate Industry Productivity" Econometrica, 71: 1695-1725. O'Brien, T. and A. Diaz (2004) "Improving Competitiveness and Market Access for Agricultural Exports through the Development and Application of Food Safety and quality Standards. The Example of Peruvian Asparagus" IICA, http://www.bountyfresh.com/Peru-case-english.pdf Nocke, V. and S. Yeaple (2006) "Globalization and Endogenous Firm Scope".NBER Working Paper No. 12322. Rauch, J. and J. Watson (2003) "Starting Small in an Unfamiliar Environment". International Journal of Industrial Organization, Vol. 21: 1021-1042. Rob, R. (1991), "Learning and Capacity Expansion Under Demand Uncertainty", Review of Economic Studies, 58: 655­675. Roberts, M. and J. Tybout (1997) "The Decision to Export in Colombia: An Empirical Model of Entry with Sunk Costs" American Economic Review, 87(4): 545-564. Segura-Cayuela, R. and J. Vilarrubia (2008) "Uncertainty and Entry into Export Markets" Banco de España Working Paper No. 0811. Shimizu, T. (2006) "Expansion of Asparagus Production and Export in Peru" Institute of Developing Economies. http://www.ide.go.jp/English/Publish/Dp/Abstract/073.html. Volpe Martincus, C. and J. Carballo (2008) "Is Export Promotion Effective in Developing Countries? Firm-level Evidence on the Intensive and the Extensive Margins of Exports" Journal of International Economics 76: 89-106. Zeira, J. (1987) "Investment as a Process of Search" Journal of Political Economy 95(1): 204-210. 40 Figure 1: The Value of Exporting and the Type of Entrepreneur Value Value of enter and stay, VXistay Value of enter and exit if CkH, VXiexit Value of remaining in domestic sector, VDi 0 CD * entry stay 1 qC L C (1 q)(CkH D ) F k 1 1 1 ( qC kL (1 q )C kH ) F 1 Figure 2: Entry and Exit with High Fixed Entry Costs Value Value of enter and stay, VXistay Value of enter and exit if CH, VXiexit Value of remaining in domestic sector, VDi 0 CD * enter&stay 1 qCkL C (1 q)(CkH D ) F 1 1 1 ( qCkL (1 q )C kH ) F 1 41 Figure 3: Surge in the nontraditional agricultural exports in Peru US$ Billions 1.6 1.4 1.2 1 0.8 0.6 0.4 0.2 0 1980 1981 1982 1983 1984 1985 1986 1987 1988 1989 1990 1991 1992 1993 1994 1995 1996 1997 1998 1999 2000 2001 2002 2003 2004 2005 2006 2007 Traditional Non Traditional Source: WITS Figure 4: Main export products 300 US$ Millions 250 200 150 Fresh Asparagus Canned Artichoke 100 Canned Asparagus Paprika Grapes 50 Mangos Avocados 0 1994 1995 1996 1997 1998 1999 2000 2001 2002 2003 2004 2005 2006 2007 50 Source: SUNAT 42 Figure 5: Size versus products and markets a) Average exports by product-market 1994 2007 20 20 15 15 lnaveXpdtmkt lnaveXpdtmkt 10 10 5 5 5 10 15 20 5 10 15 20 lnvalue lnvalue b) Number of products 1994 2007 15 8 6 10 nr_pdt nr_pdt 4 5 2 0 0 8 10 12 14 16 5 10 15 20 lnvalue lnvalue c) Number of markets 1994 2007 20 20 15 15 nr_mkt nr_mkt 10 10 5 5 0 0 8 10 12 14 16 5 10 15 20 lnvalue lnvalue 43 Figure 6: Entry and exit of firms into exporting (by year) 250 200 Correlation: 0.87 150 100 50 0 1995 1996 1997 1998 1999 2000 2001 2002 2003 2004 2005 2006 2007 Entries Exits Figure 7: Cumulative market shares in 2007 by cohorts 80% 70% 73% 60% 50% 40% 52.4% 30% 20% 10% 0% 1994 1995 1996 1997 1998 1999 2000 2001 2002 2003 2004 2005 2006 2007 Traditional NonTraditional 44 Figure 8: Exits of firms by age a) Average number of exits by age of the b) Percent of exit according by age of the spell spell 800 40% 700 35% 667 34% 600 30% 27% 500 25% 23% 400 20% 17% 300 15% 15% 271 12% 200 10% 8% 9% 8% 141 100 5% 5% 67 4% 3% 43 26 2% 0 13 10 7 3 2 1 1 0% 1 2 3 4 5 6 7 8 9 10 11 12 13 1 2 3 4 5 6 7 8 9 10 11 12 13 45 Figure 9: Distribution of the number of shipments exported within the single-year entry firms 600 494 500 Number of singleyear entries 400 300 200 151 100 73 41 24 15 13 15 26 9 3 5 6 4 4 0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 >15 Number of shipments by year Figure 10: Distribution of the size of the firms by age a) Residuals vs. age b) Distribution of residuals .3 5 .2 kdensity res Residuals 0 .1 0 -5 -5 0 5 x 0 5 10 15 Age 1 Age 9 age 46 Figure 11: Entries after Discovery a) In new products 7 7 6 6 5 5 4 4 3 3 2 2 1 1 0 0 Y1 Y2 Y3 Y4 Y5 Y6 Y7 Y8 Y9 Y1 Y2 Y3 Y4 Y5 Y6 Y7 Y8 Y9 Mean (X) Median (X) Mean (D) Median (D) b) In new product-markets 4 3.5 3 2.5 2 1.5 1 0.5 0 Y1 Y2 Y3 Y4 Y5 Y6 Y7 Mean Median Figure 12: Unit Values in New Products (US$/Kilogram) 1.8 1.7 1.6 1.5 1.4 1.3 1.2 1.1 1 0.9 0.8 Y1 Y2 Y3 Y4 Y5 Y6 Y7 Y8 Y9 Y10 Mean Median 47 Table 1: Probit Regression on Probability of Exit Dependent variable: Exit Probit OLS Probit OLS (1) (2) (3) (4) ln(initial exports) -0.14*** -0.12*** [0.010] [0.01] ln (first shipment value) -0.03*** -0.03*** [0.010] [0.01] Observations 1370 1397 1370 1397 Product Yes Yes Yes Yes Market Yes Yes Yes Yes Year Yes Yes Yes Yes Robust standard errors in brackets *** p<0.01, ** p<0.05, * p<0.1 Table 2: Trade flows by the type, role of trials Type Markets Products Number % Number % New (a) 201 24.94 12 20.00 Old (b) 109 13.52 38 63.33 Trials (c) 496 61.54 10 16.67 Total 806 100.00 60 100.00 Entries (a) + (c) 697 22 Success (a)/ (a) + (c) 29% 55% Table 3: Characteristics of exporters that start exporting completely new products Average(Firm i's number of products Average (Firm i's main product's Exporters with previous experience exported in t/ average number of Total Entrants exports in t1/ average exports of all Product (%) products exported by all firms with firms with same main product in t1) same main product in t) (1) (2) (3) (4) Domestically consumed products (D) Avocadoes 19 37% 0.78 1.30 Guanabana Juice 2 100% 1.56 1.47 Mango Juice 7 71% 0.64 1.04 Passion Fruit 7 71% 0.81 2.11 Pinneaples 2 0% 0.00 2.33 Average (D) 7 56% 0.76 1.65 Products mainly for exportation (X) Prepared/preserved Artichoke 7 100% 3.03 1.71 Prepared/preserved Mango 4 100% 1.42 2.30 Prepared/preserved Nuts 2 50% 0.02 1.00 Prepared/preserved Papaya 2 100% 1.16 3.38 Prepared/preserved Sweet Corn 4 50% 3.44 2.28 Papaya Juice 1 100% 0.92 1.00 Piquillo Pepper 9 33% 1.14 1.54 Average (X) 4 76% 1.59 1.89 Average (total) 68% 1.24 1.79 48 Table 4: Characteristics of exporters that start exporting to new product-market combinations Dependent variable: Entry during first three years in each productmarket Probit OLS (1) (2) (3) (4) (5) (6) (7) (8) ln (total exports in t1) 0.029* 0.0243* 0.0113*** 0.008* (0.02) (0.01) (0.00) (0.00) Past experience in same product 0.099*** 0.152** 0.059*** 0.077*** (0.03) (0.07) (0.02) (0.03) Past experience in same market 0.03 0.02 0.028 0.020 (0.03) (0.05) (0.02) (0.03) Observations 734 1,152 1,152 734 1,128 1,767 1,767 1,128 Rsquared 0.65 0.64 0.64 0.66 Robust standard errors in brackets *** p<0.01, ** p<0.05, * p<0.1 Product, market and year fixed effects are controlled for in all regressions. Table 5: Average survival rates after one year, in new products and new product- markets combinations, only entrants and later entrants Product (11) Surv. Rate 1year Pioneers Surv. Rate 1year Followers Average products 56.42 32.38 Avera ge (D) 54.65 34.91 Avera ge (X) 57.90 30.27 Productmarkets (109) Surv. Rate 1year Pioneers Surv. Rate 1year Followers Average productmarkets 47.89 30.00 Pioneers (Product Market): Followers (Product Market): Difference in Survival Rate: 8.53 2.38 49 Annex 1: Summary statistics of the data a) By year Year Nr. Observations Exports 1994 567 140,069,787 1995 620 169,345,874 1996 613 202,059,329 1997 652 208,153,040 1998 583 198,863,592 1999 782 263,332,652 2000 1,064 250,119,272 2001 1,050 304,436,362 2002 1,183 371,673,187 2003 1,297 438,881,446 2004 1,530 562,062,708 2005 1,744 726,224,496 2006 2,046 884,797,308 2007 2,322 1,101,137,051 Total 16,053 b) By type of exporter Non singleyear exp. Singleyear exporters Total Non individuals 1,272 825 2,097 Individuals 239 340 579 Total 1,511 1,165 2,676 Exporters nonrentry Exporters with reentry Total Non individuals 1,903 194 2,097 Individuals 536 43 579 Total 2,439 237 2,676 50 c) By firms Average value Number of Year exported by S.D. Min Max firms firm 1994 210 666,999 1,311,441 1,265 8,686,215 1995 221 766,271 1,581,295 1,018 11,918,041 1996 239 845,437 1,933,473 1,000 17,818,464 1997 225 925,125 2,325,943 1,006 24,140,284 1998 203 979,624 2,593,879 1,920 24,164,342 1999 267 986,265 2,659,069 1,013 23,044,694 2000 307 814,721 2,317,083 1,048 23,325,820 2001 351 867,340 2,406,457 1,000 23,349,340 2002 392 948,146 2,774,388 1,004 29,665,694 2003 432 1,015,929 3,453,340 1,058 47,235,336 2004 468 1,200,989 4,099,953 1,001 61,607,304 2005 540 1,344,860 4,758,673 1,008 76,113,736 2006 595 1,487,054 5,999,261 1,207 97,699,096 2007 643 1,712,499 6,832,192 1,015 110,384,024 d) By products Average value Number of Year exported by S.D. Min Max products product 1994 42 3,334,995 9,959,187 1,042 61,421,740 1995 48 3,528,039 11,752,034 1,050 77,926,088 1996 44 4,592,258 14,655,546 4,098 93,610,584 1997 40 5,203,826 14,749,356 2,212 88,928,112 1998 39 5,099,067 13,837,446 6,150 79,323,688 1999 47 5,602,823 14,825,945 1,254 87,683,368 2000 47 5,321,687 14,115,093 1,100 80,498,160 2001 50 6,088,727 14,907,061 1,040 80,892,736 2002 46 8,079,852 19,334,110 1,137 99,071,856 2003 50 8,777,629 21,354,464 1,210 123,434,096 2004 48 11,700,000 26,273,140 1,017 156,307,728 2005 49 14,800,000 31,484,366 2,883 179,588,880 2006 53 16,700,000 35,586,412 1,227 212,422,752 2007 58 19,000,000 42,798,808 1,114 259,384,112 51 e) By markets Average value Number of Year exported by S.D. Min Max markets market 1994 45 3,112,662 13,800,000 1,265 89,960,888 1995 47 3,603,104 16,500,000 13,650 108,818,136 1996 48 4,209,570 19,800,000 2,100 131,603,600 1997 52 4,002,943 19,200,000 7,313 131,696,648 1998 41 4,850,332 19,300,000 1,605 110,707,712 1999 50 5,266,653 23,100,000 1,013 137,202,016 2000 54 4,631,839 20,800,000 3,500 132,834,520 2001 48 6,342,424 26,900,000 2,880 151,131,312 2002 59 6,299,546 29,900,000 3,831 181,650,688 2003 59 7,438,669 36,000,000 1,016 226,051,696 2004 67 8,388,996 42,900,000 1,238 268,860,672 2005 74 9,813,845 51,400,000 1,227 321,110,624 2006 78 11,300,000 61,900,000 1,904 398,428,704 2007 76 14,500,000 76,400,000 4,120 511,210,848 52 Annex 2: Product classification 1. Vegetables (Chapter 7) 2. Fruits (Chapter 8) & Paprika 3. Processed Food (Chapter 20) Potatoes: Coconuts and nuts: Prepared/preserved Cucumbers: 0701100000 (0801100000-0801190000) 2001100000 0701900000 Nuts: Prepared/preserved Onions and 0710100000 (0801200000-0802900000) Garlic: Tomatoes: Bananas: 2001200000 0702000000 (0803000000-0803002000) Canned Olives: Onions, Garlic and other Avocados: 2001901000 alliaceous vegetables: 0804400000 2005700000 (0703100000-0703900000) Pineapples: Prepared/preserved Tomatoes: 0712200000 0804300000 (2002100000-2002900000) 0712901000 Guayabana: Canned Fungi 0711100000 0804500010 (2003100000-2003900000) 0712100000 0804501000 Prepared/preserved Potatoes: Cauliflower, Cabbage and Mangos: 2004100000 Broccoli: 0804500020 2005200000 (0704100000-0704900000) 0804502000 Canned Asparagus: Lettuce: 0811909100 2005600000 (0705110000-0705290000) Citrus fruits: Prepared/preserved Artichoke Carrots: (0805100000-0805900000) 2005991000 (0706100000-0706900000) (0814000000-0814009000) 2005901000 Cucumbers: Grapes: Prepared/preserved Piquillo 0707000000 (0806100000-0806200000) Pepper: 0711400000 Melons: 2005992000 Legumes Shelled or Unshelled: 0807100010 Prepared/preserved Legumes (0708200000-0708900000) 0807190000 Shelled or Unshelled: (0710220000-0710290000) Watermelons: (2005510000-2005590000) (0713209000-0713909000 ) 0807100020 Canned Peas: Peas: 0807110000 2005400000 0708100000 Papaya: Canned Sweet Corn: 0713101000 0807200000 2005800000 0713109010 0811909600 Other Prepared/preserved 0713109020 Chirimoya: vegetables: 0710210000 0810900000 (2001909000-2001909090) Asparagus: 0810902000 (2004900000-2005100000) 0709200000 Passion Fruit: 2005300000 0710801000 0810901000 (2005909000-2005910000) Fungi: 0811909400 (2005999000-2006000000) (0709510000-0709590000) Camu Camu: Jams, fruit jellies, marmalades: (0711510000-0711590000) 0811909200 (2007100000-2007999200) (0712300000-0712390000) Lucuma: Prepared/preserved Palm: Spinach: 0811909300 2008910000 0709700000 Guanabana: Prepared/preserved Mango: 0710300000 0811909500 2008993000 Piquillo Pepper: Other Fruits: Canned Peanut: 0709600000 (0804100000-0804200000) (2008111000-2008119000) Olives: (0808100000-0810500000) Canned Nuts: 0709902000 (0810903000-0811909000) (2008191000-2008199000) 0711200000 (0811909900-0813500000) Prepared/preserved Pineapples: 0709900010 Paprika: (2008200000-2008209000) Sweet Corn: 0904200000 Prepared/preserved Citrus Fruits: 0709901000 0904201010 2008300000 0710400000 0904201020 Prepared/preserved Papaya: 0712902000 0904201030 2008992000 53 Artichokes: 0904209000 2008999100 0709903000 Other Prepared/preserved Fruits: 0709100000 (2008400000-2008809000) Other roots and tubers: (2008920000-2008991000) (0714100000-0714909000) 2008999000 Other Vegetables: (2008999200-2008999900) 0709300000 Mango Juice: 0709400000 2009801400 0709900090 Pineapple Juice: 0709909000 (2009400000-2009490000) 0710800000 Tomato Juice: 0710809000 2009500000 0710900000 Guanabana Juice: 0711900000 2009801300 0712909000 Passion Fruit Juice: 2009801200 2009801910 Camu Camu Juice: 2009801500 Papaya Juice: 2009801100 Other Juices: (2009110000-2009399000) 2009801900 (2009690000-2009790000) (2009801990-2009900000) 54 Annex 3: Summary statistics of exports and entries by cohort Exports of Non - Traditional Agricultural products by cohorts (number of firms) Cohort1994 Cohort1995 Cohort1996 Cohort1997 Cohort1998 Cohort1999 Cohort2000 Cohort2001 Cohort2002 Cohort2003 Cohort2004 Cohort2005 Cohort2006 Cohort2007 Total 1994 210 214 1995 133 88 220 1996 108 47 84 237 85 1997 29 44 67 218 64 1998 18 24 38 59 198 60 1999 18 23 34 42 90 260 53 2000 13 23 30 31 52 105 288 45 2001 12 13 24 27 42 57 131 324 44 2002 12 15 17 21 35 41 85 122 367 39 2003 13 16 15 19 28 33 61 66 142 399 38 2004 13 15 13 15 19 32 47 46 88 142 429 38 2005 12 9 15 15 16 28 48 41 63 85 170 464 36 2006 11 16 11 10 18 25 43 30 46 57 95 197 539 36 2007 10 12 8 8 18 23 42 27 42 39 80 106 192 593 Exports of Non - Traditional Agricultural products by cohorts (total values) Cohort1994 Cohort1995 Cohort1996 Cohort1997 Cohort1998 Cohort1999 Cohort2000 Cohort2001 Cohort2002 Cohort2003 Cohort2004 Cohort2005 Cohort2006 Cohort2007 Total 1994 140,100,000 140,100,000 1995 159,900,000 9,426,866 169,274,226 1996 172,500,000 15,552,607 14,056,445 201,974,201 1997 159,500,000 10,834,199 21,587,571 16,273,075 206,203,057 1998 148,500,000 8,930,792 10,056,849 17,152,844 14,222,079 197,966,429 1999 178,400,000 8,838,484 12,977,285 20,885,127 26,528,712 15,656,570 260,408,235 2000 141,700,000 7,447,453 13,075,186 16,950,109 27,906,561 28,077,101 14,956,075 244,266,281 2001 132,800,000 8,286,807 13,955,155 13,445,579 39,147,358 47,488,972 23,251,201 26,050,653 288,865,162 2002 142,800,000 8,984,371 18,310,625 12,548,193 51,868,733 52,043,031 23,072,951 39,325,798 22,748,088 352,305,389 2003 148,100,000 10,445,214 20,777,208 12,753,798 54,739,189 70,272,676 25,328,759 45,433,146 28,756,004 22,257,925 416,093,798 2004 167,300,000 11,482,384 24,536,344 11,923,248 57,495,428 88,773,745 35,677,086 62,503,635 39,772,636 41,634,382 20,983,038 511,921,373 2005 196,700,000 15,699,955 28,494,756 13,700,013 80,100,001 107,200,000 50,322,894 78,511,288 40,180,376 47,358,004 38,982,921 28,984,573 630,712,166 2006 256,200,000 17,504,631 29,462,398 13,803,093 89,073,786 134,500,000 52,399,298 86,320,778 42,943,540 46,239,579 33,174,894 47,672,319 35,426,928 812,001,938 2007 295,000,000 16,614,316 29,169,730 18,109,317 94,163,165 161,100,000 58,690,569 103,900,000 56,546,712 57,807,309 42,251,032 56,689,385 56,457,480 54,595,331 1,001,945,518 Annex 4: Mathematical Appendix, Profits at alpha entry q (C kL C kH ) q (C kL C D ) (C kH C D ) F (1 q )((q (C kL (1 q )C kH ) C D ) 1 1 1 Pr ofits ( )F 1 (1 q ) (1 q ) 1 1 q (C kL ) F (q )((q (C kL (1 q )C kH )) 1 ( 1 1 )F 1 (1 q ) (1 q ) 1 1 q (1 q )(C kL C kH ) F 1 1 F 1 (1 q ) 1 q (1 q )( (C kL C kH ) F ) 1 1 q The regularity condition on price ensures the denominator in Equation (6) is positive. The regularity conditions on fixed costs i Pk E (C k ) P CD ( F i D & i Pk C H i PD C D ) 1 1 ensures the numerator is positive. To see this, multiply both sides of the second condition by 1/(1-) and subtract the left side from the left side of the first condition and the right side from the right side of the C H E (C k ) F 0. second condition. This yields 1 ). Substituting qCL+(1-q)CH for E(C) q(C L C H ) F 0 this is the same as 1 . Note that some firms with expected lifetime profits from exporting (net of fixed cost) below zero will chose to enter because of the option of exit. The expected present value of net profits is 1 Pr ofits ( i ( Pk PD ) qC kL (1 q)C kH C D ) F , At entry this is 1 56 q (1 q )( (C kL C kH ) F ) Pr ofits 1 , 1 q which is negative given the regularity condition. 57