Contractual reliance and exporter performance: evidence from Slovenian firm-level data Gabriela Dobrescu∗†, UC B ERKELEY

Helena Schweiger, EBRD‡

(J OB M ARKET PAPER ) November 10, 2008

Abstract We focus on the role of a particular aspect of quality - reliability - in export performance. We argue that ISO certification is a measure of firm reliability. This certification is most relevant in international transactions where contractual problems and information asymmetries are substantial. Our theoretical framework allows for firm-level heterogeneity in productivity and in ’contractual ability’, and for endogenous demand shifters. We show that when ’contractual ability’ is lower in foreign transactions than in domestic ones, there is a threshold ’contractual ability’ below which no level of productivity can compensate for the lack of capacity to contract internationally. This framework predicts that high-productivity exporters that face contractual difficulties seek ISO certification, and that their sales performance improves with certification. We use panel data on Slovenian manufacturing firms from 1995 to 2005 together with firm-level data on ISO certification. We find strong evidence for the self-selection of firms into certification. We estimate a difference-in-differences model controlling for unobserved firm heterogeneity in levels and growth trends and find that ISO certification improves total sales and export sales. The smallest and the largest firms experience the largest gains from certification. The effects of certification on export sales increase with the contractual dependence of the firm’s sector. Because sectoral contractual dependence indicates a sector’s intrinsic reliance on certification, this result strengthens our confidence in the identification of the effect of ISO certification. ∗ I am grateful to my advisors, Professor Pranab Bardhan and Professor G´erard Roland, for their continued guidance and support. I thank Urmila Chatterjee, Ana M. Fernandes, Pierre-Olivier Gourinchas, Robert Johnson, Ann Harrison, Maurice Obstfeld, Steve Tadelis, and participants at the World Bank DRG Trade Unit Lunch Workshop and at the UC Berkeley International Economics Seminar for many valuable suggestions and comments. † University of California Berkeley, 508-1 Evans Hall #3880, Berkeley CA 94720. Email: [email protected]. ‡ European Bank for Reconstruction and Development, Office of the Chief Economist, One Exchange Square, London EC2A 2JN, United Kingdom. Email: [email protected].

1

Introduction

It is a well-established fact in the international trade literature that exporters have to be more competitive than domestic firms. Competitiveness is generally thought of as having lower prices or higher productivity. However, firms that wish to access foreign markets have to be competitive in other dimensions as well. Of these, the literature has recently focused on quality. Quality can generally be thought of as ’anything that increases willingness to pay’ (Sutton (1989)), and has been modeled as a demand shifter. However, this parsimonious representation hides many interesting aspects related to quality. For example, as opposed to prices, quality is unobservable and has a stochastic variance. Because of this, quality can be subject to asymmetric information and contractual problems, which are likely to be most severe in the international trade context. Our contribution is to give a detailed treatment of quality and of its role in export performance. Quality can be decomposed into a quality target level and a quality variance. Both of these elements influence buyers’ ’willingness to pay’. This paper is concerned with the second element of quality, the variance, and we will refer to this as ’reliability’. Reliability means delivering on time and satisfying agreed-upon contract terms. In addition to reliability, trust - the perceived reliability by buyers - matters as well in influencing ’willingness to pay’. The advantage of this perspective is that it highlights the contractual aspects of quality, which are more stringent in international trade. International Standards Organization (henceforth, ISO) certification for quality management systems, ISO 9000, is a good empirical measure for precisely the aspect of quality we focus on, reliability. A management system is a set of procedures used to monitor, control and document production processes, and to track costs and schedules. The quality management system ensures delivery of a consistent product, with low probability of defective items, as well as on-time delivery, and no cost overruns. ISO certification does not guarantee a certain level of product sophistication as it does not certify the firm’s research and development or product innovation activities. This certification is most relevant in the international trade context, where information is scarcer and where being reliable is more challenging. In addition, the ISO 9000 certification is more popular for intermediate input suppliers, where the quality variance matters more1 . For example, consistency and on-time delivery of intermediate inputs are especially relevant to final producers of complex goods, who must coordinate a large number of suppliers. 1

In contrast, the quality level might matter more to final consumers.

2

The key question of our study is how effective ISO 9000 certification is in enhancing firm-level export performance. We are the first to provide firm-level evidence on this issue in an international trade context. We find this evidence to be valuable because ISO certification is most relevant in global contracts, and particularly, for the intermediate inputs sectors, which account for a large portion of international trade. Addressing this question is especially important given the growing popularity2 of ISO certification in Europe, as well as in China and elsewhere. This study also helps us learn about the importance of reliability and information signalling in international trade. Lastly, our research might give some indication of the causes behind underexporting3 by firms from small and/or developing countries. We provide a theoretical framework to formally think about the role of reliability and of ISO certification. We include ’contractual ability’ as a second dimension of heterogeneity alongside productivity4 . ’Contractual ability’ is the ability to be reliable, or, in other words, the effectiveness of investments in the management systems to achieve reliability and trust. Firms endogenously determine their ’reliability’ by choosing how much to invest in management systems. These decisions depend on the firms’ productivity and ’contractual ability’. If ’contractual ability’ is the same for domestic and exporting firms, the model is equivalent to the Melitz (2003) model. It is plausible, however, that ’contractual ability’ is higher in the domestic market than in foreign markets. One reason for this can be that establishing trust is difficult due to more severe asymmetric information problems. Another explanation might be that being reliable is more difficult when serving foreign markets because communication between buyers and sellers is noisy, or because the tastes and requirements of foreign clients are new and diverse, requiring more flexibility on the part of the supplying firm. Therefore, exporting firms need to have more sophisticated management systems in order to establish their reliability abroad. We show that in this case, there is a threshold ’contractual ability’ below which firms will not export, regardless of how productive they are. High ’contractual ability’ can always make up for low productivity, but high productivity cannot compensate for low ’contractual ability’. The intuition for this result is that if a firm has low ’con2 The number of ISO certified firms worldwide has been growing at an average annual rate of around 20% in the past 5 years according to the ISO Survey. 3 Brooks (2006) shows evidence of underexporting by Colombian firms, Dobrescu and Schweiger (2008) document underexporting by Slovenian firms relative to the current benchmark international trade model, Eaton, Kortum, Kramarz (2007) and Yoshino (2007) and Bigsten et al (1999) suggest underexporting by African exporters. 4 We build on the model developed by Melitz (2003), and the extensions by Hallak and Sivadasan (2006) and Johnson (2008) to include endogenous quality choice.

3

tractual ability’, its reliability will be low as well, and the firm cannot capture enough clients in foreign markets to cover the fixed costs of exporting, regardless of how productive it is or how large foreign markets are. This result is a possible explanation for why firm size and productivity are not sufficient to predict export status, and echoes a similar result by Sutton (2007). We allow firms to invest in ISO certification and assume that this certification improves firms’ ’contractual ability’. Our theoretical framework predicts that high-productivity firms that lack ’contractual ability’ abroad select into ISO certification, and that ISO certification increases reliability and trust, and improves firms’ sales performance. In addition, the model predicts that the lower the firm’s initial ’contractual ability’, the larger the benefits from certification are. We explore all these conjectures using Slovenian firm-level data. The novelty in our approach is the use of firm-level information on ISO certification, including the certification date, together with panel data covering nearly the entire population of Slovenian manufacturing firms over the period 1995 to 2005. This data allows us to investigate dynamics in various performance measures before and after firms get certified. We document 4 main findings. First, there is strong evidence of self-selection into certification, ISO exporters performing better than non-ISO exporters across a series of indicators, both in levels and growth rates. Following Bernard and Jensen (1999), we estimate ’premia’ (i.e. ceteris paribus percentage differences) between ISO-certified exporters and non-ISO certified exporters 1, 2 and 3 periods before certification takes place. Before obtaining certification, future ISO exporters have strong premia in export sales (up to 60% higher than non-ISO exporters) and export shares (up to 7% higher than non-ISO exporters). There are also premia for ISO exporters in capital intensity and labor productivity. We also find a higher growth rate of sales for ISO exporters before becoming certified. We take all these selfselection patterns into account in the estimation of the effects of ISO certification. Second, our strictest empirical specification - a differences-in-differences model including firm unobserved heterogeneity in both levels and growth rates to control for self-selection - indicates that total sales increase by around 5%, export sales increase by about 7%, and export shares increase by 1.3%. The effect on domestic sales is significantly positive in some specifications, but it is not robust across all our estimations. Other measures, such as capital intensity, productivity, price-cost margins and wages show either weak or inconsistent results. These results confirm that the main role of ISO certification is to mediate international contracts, increasing sales without necessarily improving

4

operational performance. Third, the strength and role of the signal provided by ISO certification varies by firm size5 . Our theoretical model hypothesizes that small firms should have larger benefits from certification because they are likely to have low ’contractual ability’ for exporting. However, we find that both micro firms and large firms gain most from certification. Large firms might gain from certification because they can better capitalize on this advantage, possibly also complementing it with other reputational devices. The effects for mid-sized firms are not significant or consistent. Forth and last, the effects of ISO certification on export sales and export shares are higher in ’contractually dependent’ sectors. We measure a sector’s ’contractual dependence’ and intrinsic demand for ISO certification and for other similar contractual devices through several indices calculated from the US Input-Output Tables. These indices denote, for example, that the higher the share of intermediate inputs produced in a sector, or the larger the number of intermediate inputs used by a sector’s client industries 6 , the more important timely delivery, reliability or flexibility are. The rate of ISO take-up among Slovenian exporters is higher in sectors with higher indices of ’contractual dependence’. Because firms in contractually dependent sectors generally face more difficulties when exporting and therefore have lower ’contractual ability’, we expect the effect of ISO certification to be larger in these sectors. We indeed find that the strongest effects of ISO certification on export sales and export shares are for firms in the most ’contractually dependent’ sectors. Following Rajan and Zingales (1998), we argue that these findings support our identification of the effect of ISO certification on export performance. The argument is that because ISO certification is related to the contractual aspects captured in these indices, the effects of ISO certification that increase significantly with contractual dependence are more likely to identify ISO certification and not any other unobserved changes in the firm occurring concomitantly with ISO certification. Our results have implications for policies aimed at better assisting exporters. The fact that we find an effect of ISO certification implies that firm-level contractual difficulties play an important role in exporters’ sales performance abroad. Firms from small and/or developing countries7 find 5 We split firms into micro, small, medium and large according to the World Development Indicators definitions of firm size. 6 We use the Herfindahl index of intermediate input usage of a sector to capture this type of ’contractual dependence’, as described by Cowan and Neut (2007) to capture ’contractual dependence’. The argument is that the larger the number of input suppliers a firm uses, and the lower the concentration of the input suppliers, the more important contractual performance is to the firm. 7 The present study is on firms from a stable, open and relatively prosperous emerging market, but are likely to extend to firms from other similar countries. One could argue that the effect of ISO certification could even be larger for firms

5

establishing reputation and visibility in foreign markets especially difficult. These firms might have a lower management capacity8 and might be victims of a broader environment of poor information and contract enforcement. The fact that the effects we find are rather small indicates a scope for policy interventions. Firms could benefit not only from assistance in building better management systems and contractual abilities, but also from being able to credibly relay this signal to foreign buyers. Our results are also relevant to firm managers, helping them understand the benefits of credibly communicating their reliability abroad through international standards. The paper is structured as follows. We start by reviewing the literature and stating the arguments for our interpretation of ISO certification. In Section 3 we present our theoretical framework, gradually introducing heterogeneity in ’contractual ability’, differences between firms’ ’contractual ability’ domestically and abroad and, lastly, the possibility of certification. In Section 4 we describe our data, empirical methodology and discuss our results. A subsection of Section 4 presents the sectoral analysis of the selection and the effects of ISO certification. Section 5 states our conclusions.

2

Related literature

First, we review the international trade literature on information and contract enforcement costs and the recent theoretical literature based on firm-level heterogeneity and export costs. Second, we present the literature on ISO certification, including a discussion on the definition and interpretation of ISO certification.

2.1

The International Trade Literature

The estimation of the effects of trade restrictions related to information and contract enforcement costs on the volume of exports has been done in cross-country, cross-sectoral studies based on the gravity equation model. Information costs have been proxied by either the volume of telephonic traffic, the number of importing country’s banks located in the export’s country (Portes and Ray, 2005), the extent of networks of Chinese immigrants (Rauch and Trindade, 2002) or the frequency of technical service (Evans, 2003). These studies are subject to the usual criticisms of gravity equation models, as surveyed in Anderson and van Wincoop (2004). Moreover, all these proxies, although from poorer countries, where ’contractual ability’ abroad is likely to be more constrained. 8 See Bloom, Dorgan, Dowdy and Van Reenen (2007) for empirical evidence supporting this claim.

6

interesting, are imperfect and fail to address the information asymmetries between exporters and foreign buyers at a detailed micro level . Shepherd (2007) explores whether uniform international standards lower transaction and information costs, thereby promoting international trade. Using a database of product standards in the European Union (EU), Shepherd (2007) shows that a higher proportion of internationally harmonized standards - proxied by the fraction of ISO standards among EU standards - increases the number of varieties exported by EU partner countries. The effect is strongest for poor countries, where presumably the access to information and the quality of management are worse. Shepherd (2007) models international standards as product standards, implying that exporting firms have to pay only one cost in order to access all markets if product standards are harmonized across partner countries. Two recent studies by Nunn (2007) and Levchenko (2007) are noteworthy for addressing the importance of contract enforcement. These papers extend the predictions of the Heckler-Ohlin framework to include contracts as an input in the firm’s production function. Both authors construct sectoral indices of ’contractual intensity’ or ’contractual complexity’. Nunn (2007) uses an index that measures the differentiation of inputs used in the production. Levchenko (2007) uses an index constructed by Cowan and Neut (2007) equal to the Herfindahl index of intermediate inputs usage. The authors interact these indices with country-level institutional indices and find that countries with greater ’contractual endowment’9 export more - either in export volume, export shares, or export varieties - in more contract-intensive sectors. Another strand of the literature focuses on modeling international trade with intra-industry firmlevel heterogeneity. This literature started with the work of Bernard and Jensen (1999, 2004) who carefully document that larger and more productive firms become exporters. The same authors also introduced the idea that firm-level export costs have a significant role in firms’ exporting decisions. These two stylized facts have been elegantly modeled in Melitz (2003). The now popular Melitz model includes export costs and heterogeneity in firm-level productivity to show the self-selection of largest and most productive firms into exporters: only firms with a productivity draw above a certain threshold are able to cover the costs involved in accessing foreign markets. Our theoretical set-up is based on Hallak and Sivadasan (2006) and Johnson (2008) who extend the Melitz model to 9

A country’s ’contractual endowment’ is given by measures of corruption, rule of law or efficiency or the judiciary.

7

include a second dimension of heterogeneity and endogenous quality choices. Evans (2001) and Rauch and Watson (2003) model firm-level heterogeneity in export costs and relate export costs to contractual and management capacities. Evans (2001) considers heterogeneous firm-level export costs related to setting up distribution networks, establishing brand recognition and tailoring products to fit foreign customers and argues that these costs could differ across firms because of differences in knowledge about foreign markets, differences in productivity of learning about exporting or differences in access to information. Rauch and Watson (2003) consider uncertainty regarding the exporters’ ability to deliver according to contract terms. The authors include heterogeneous export costs and allow foreign buyers to select suppliers based on these costs10 . Rauch and Watson (2003) argue that per period costs differ across suppliers most likely because of differences in the quality of management. Further, they mention mention that suppliers from developing countries ”may not have the requisite managerial capability to learn how to meet international quality and delivery standards,” and that buyers see ISO certification as an indicator of this capability. Eaton, Kortum and Kramarz (2007) generalize the Melitz model to include not only firm-level or market-level heterogeneity in entry costs, but also firm-level or market-level heterogeneity in demand shocks. Two other extensions considered by Arkolakis (2008) and Demidova, Kee and Krishna (2008) are particularly relevant to our theoretical framework. Demidova, Kee and Krishna (2008) introduce firm and market-specific demand shocks to examine the export patterns of Bangladeshi exporters of apparel goods to the US and the EU. They model exogenous demand shocks that incorporate a firm’s market specific endowment of business contacts and networks, or random taste shocks. Arkolakis (2008) models firms’ endogenous choice of the level of market-entry costs to pay. Market-entry costs are considered marketing costs in this framework and the higher the marketing costs a firm pays, the more consumers the firm reaches11 .

2.2

ISO Certification: Definition and Literature

We address firm reputation and reliability empirically by looking at ISO 9000 certification. ISO 9000 is a quality management systems standard developed by the International Standards Organization. ISO 9000 certification is widespread globally, reaching over 800,000 firms worldwide in 2006, and 10

Costs here act somewhat as a signal for the firm’s ability to deliver according to contract terms. Our theoretical framework merges these two models, including firm-level heterogeneity contractual ability and allowing for costs and demand shocks to be endogenous choices. 11

8

growing at an average annual rate of 20% in the past 5 years according to the ISO Survey. ISO 9000 is the most general of ISO standards12 . The certification process involves an initial application from the firm interested in being certified, followed by 1 or 2 audits from an accredited certification agency. If the firm passes the audits, it receives the official certification. This process can take between 6 months to over 1 year. Certification agencies demand application and audit fees that depend on the firm size. Only two studies have looked at ISO certification in the international trade context. Shepherd (2007), mentioned above, looks at bilateral trade flows, and interprets ISO certification as an uniform product standard. Hallak and Sivadasan (2006) use firm-level ISO certification as a proxy for product quality and show that ISO-certified firms command higher prices. However, they do not explore any other patterns related to ISO certification. These authors model minimum quality requirements for exporting to explain the fact that some high-productivity firms do not export. Our model gives an alternative explanation to this observation, without imposing exogenous minimum quality standards as in Hallak and Sivadasan (2006). We depart from the interpretation of ISO certification in the international trade literature so far as a certification of product standards or product quality. Instead, we argue that ISO certification is a proxy for a firm’s better ability to serve contracts and for a firm’s reliability. ISO 9000 is an international standard for quality management systems. A management system represents the monitoring and control processes, such as documentation, tracking costs and schedules and verification of enforcement a firm employs in order to ensure a certain product features. The International Standards Organization website states that ”the ISO 9000 standard provides a [...] framework for [...] managing the organization’s processes so that they consistently turn out products that satisfy customers’ expectations.” The ISO also declares that the objectives of ISO 9000 standard is to provide buyers with ”confidence” about the suppliers’ ”consistency and reliability [...] especially when supplier and client are new to each other, or far removed geographically, as in an export context.” ISO certification acts as a contractual facilitator in international trade. ISO certification is a signalling device that partially corrects the asymmetric information between the domestic supplier and the foreign client, helping to lower contractual terms and costs and increase demand. Hudson and 12 The International Standards Organization has a wide range of management and product standards that are more narrowly specialized within certain industries, such as ISO/TS 16949:2002 for the automotive industry or ISO/IEC 27001:2005 for the information technology industry.

9

Jones (2003) share this view, arguing that ISO certification can ease the asymmetric information frictions13 particularly affecting exporters from small and/or developing countries by lowering transaction and information costs. ISO certification ensures a firm’s ability to deliver on time and according to specified requirements, regardless of whether they are high or low. A firm with ISO certification can have low quality levels, but there is low variance in quality. Quality is a stochastic variable, determined by the desired or targeted quality and the variance in quality or the probability of defective items. ISO guarantees lower variations from a chosen quality target. Aurora and Asundi (1999) further support this view stating that a firm with ISO certification can have low quality, but that a better management lowers costs by ensuring fewer defects, less rework and fewer breaches of contracts. Therefore, reliability regarding established, albeit possibly low, product sophistication implies lower contracting costs and higher demand. Although ISO certification is mainly seen as a signalling device, conversations with firm managers or members of certification agencies reveal a popular view that ISO certification improves productivity and that this is the main reason why firms seek it. Managers surveyed by Aurora and Asundi (1999) also state that ISO guides firms to implement ’sophisticated’ management systems necessary in the competitive international environment when they lack this capacity on their own. We interpret such improvements in ’productivity’ as related to ’contractual productivity’, as opposed to productivity in marginal costs of production. Depending on the initial conditions in a firm, ISO certification can be either a guarantee that the management system is already in place or an actual improvement in quality management. However, whether it is a confirmation of an already in place management system or a guidance towards implementing these, ISO certification lowers contracting costs by offering foreign clients a trusted signal that the firm is reliable, and able to deliver upon specified contract terms. Our view of ISO certification as a contractual reliance tool is further supported by the fact that ISO certification is more prevalent in sectors with higher contractual dependence and in sectors that deliver more intermediate goods. The most popular case is that of exporting firms that deliver parts to automotive producers abroad, where sometimes foreign buyers themselves require certification 13 The asymmetric information problem refers to the fact that foreign buyers do not know about the quality of the product a foreign suppliers offers or about the reliability of that foreign producer to deliver in time and according to pre-specified standards.

10

before trading with suppliers. Reliability in product features and timely delivery are likely to be very important in these cases where the client-industries are demanding, complex industries. Intermediary goods suppliers to these contractually intensive foreign industries also need to show flexibility in meeting client-specific demands. The example of the two branches of a Slovenian firm further illustrate this point: the branch of this firm that has ISO certification is described as: ’our fastest growing production programme, which produces various tools, and high pressure aluminium castings for the highly demanding domestic and foreign automobile, electrical, and mechanical engineering industry.’, while the line of business of this firm that is not ISO-certified is described as: supplies the customers of copper and copper alloy semimanufactures: various rods, wires, forgings and work pieces.’ Specific empirical evidence on the sectoral patterns of ISO take-up are presented in Section 4.4. ISO certification has been studied as a reputation and signalling device outside the international trade field. Banerjee and Duflo (2000) look at the size of reputation effects relative to productivity and other comparative advantages in the Indian software industry. In their model low reputation firms face harsher contract terms regardless of their actual performance. Using detailed information on the contract terms for a small sample of firms in the Indian software industry, Banerjee and Duflo (2000) find firm age to be a significant determinant of the contract type, but they do not find ISO certification to be an effective source of reputation. ISO certification might not be a powerful tool in the software industry because the industry might be too complex or because there are alternative forms of certification specific to this industry. Our empirical work is closest in spirit to that of Terlaak and King (2006) who examine whether ISO certification generates a comparative advantage for US firms. They use panel data on a sample of US firms over 11 years together with information on the certification date. They find that plants tend to grow faster after the adoption of ISO 9000 and this growth is not accounted for by operational improvements. They find evidence supporting a signalling story: the growth effect is greater for firms in sectors with greater reliance on advertising, and therefore sectors where buyers have greater difficulty acquiring information about suppliers. We explore these issues in the international trade context for the first time. In addition, our work uses a near-complete set of firms in the Slovenian manufacturing sector14 , we expand on the empirical methodology used by Terlaak and King (2006) and explore additional sectoral patterns. 14

Terlaak and King (2006) use a dataset comprised of firms reporting to the Toxic Release Inventory.

11

3

Theoretical framework

We include ’contracting ability’ and ISO certification in a theoretical framework in order to formally think about the ways this institution might affect exporting firms. We focus our framework on the case of ISO certification, but it could apply more generally to any investment that a firm might make to improve its ability to serve foreign markets, and it could also apply to other types of certifications, such as certifications for product safety, or for fair trade. We build on the frameworks of Hallak and Sivadasan (2006) and Johnson (2008). We incorporate a second dimension of heterogeneity - ’contractual ability’ - alongside productivity in a Melitz (2003) model with CES demand and firm-level heterogeneity in productivity. ’Contractual ability’ determines the effectiveness of investments in management systems in improving firm reliability or reputation. ’Reliability’ is modeled as a demand-shifter, just like quality more generally is modeled. The firm’s ’contractual ability’ and productivity endogenously determine how much the firm is willing and/or able to pay to improve its ’reliability’ and increase demand. The role of ISO certification is that of an investment that enhances a firm’s contractual ability, allowing it to increase its demand beyond its initial potential. We first take ’contractual ability’ to be the same in the domestic market and in foreign markets. In this case, the model’s predictions are equivalent to that of the Melitz model: firms need a combination of either high marginal costs productivity and low ’contractual ability’ or high ’contractual ability’ and low productivity in order to be able to export. Next, we assume that ’contractual ability’ is lower when the firm is involved foreign transactions rather than in domestic ones. The motivation for this assumption is that contracting internationally is more difficult and requires more sophisticated management systems. Communication between the parts involved in the trade is more difficult, credibly transmitting information is more difficult, and participation of the foreign buyer in the monitoring process is also more difficult. Therefore, a given contractual investment yields a lower reputation, or perceived reliability internationally compared to what it would yield domestically; in order words, contractual ability is lower. In this case we obtain that below a certain level of ’contractual ability’, there is no level of productivity that can make up for the lack of capacity to contract internationally. The intuition for this result is that because low ’contractual ability’ implies low capacity to induce demand, the firm cannot capture a share of the foreign market that is large enough to cover its fixed costs of exporting. 12

The role of ISO certification is to improve a firm’s given ability to contract internationally. The model predicts that firms with low to medium contractual ability but high productivity will seek ISO certification. Firms with high contractual ability do not seek ISO certification as they are able to establish a reputation by themselves, and firms with low productivity or low contractual ability are not able to cover the costs of obtaining certification. The model also predicts that certification increases firms’ exporting performance. The benefits of certification increase with the firm’s initial contractual ability. As a result, we expect sales of small firms and of firms in more contractually demanding sectors to increase most as a result of certification.

3.1

The set-up

Productivity is denoted by z and it is given exogenously, heterogeneous across firms, distributed according to the cumulative distribution function G. Buyers’ utility has a constant elasticity of substitution and demand shifters λ: R θ θ−1 ( i (λ(i) · q(i)) θ di) θ−1 , with θ > 1 and q represents the quantity. The usual interpretation for the demand shifters is product quality. We broaden this definition and assume that λ incorporates either a high quality level, or a low quality variance, or, more generally, reliability and timely delivery. Our model does not focus on the quality levels, but rather on the variance in quality and on the perceived image of the firm regarding its reliability and flexibility. We call λ ’reputation’ to capture these aspects. We make these modifications in order to focus our model on the aspects that ISO-certification applies to. The ’reputation’ λ is endogenously chosen by the firm given its productivity and its ability to establish reputation and to signal reliability. Firms have to incur some fixed costs in order to achieve a certain level of ’reputation.’ In models where λ is taken as product quality, these costs are referred to as research and development investments into quality upgrading. Sutton (1989) mentions that quality can be more generally thought of as anything that increases the consumers’ ’willingness to pay’ and that the fixed costs that endogenously determine the ’willingness to pay’ can be either R&D or advertising-type costs to improve ’product image’. Because we do not focus here on upgrades in the level of quality, but rather on improvements in the variance of quality or improvements in the

13

perceived quality to the client, we consider the costs to be ’contractual costs’. These costs increase with the ’reputation’ target. We call a firm’s ability to establish reputation ’contractual ability’ as it refers to a firm’s capacity to monitor quality, to communicate with and adapt to clients and to enforce specified contract terms. ’Contractual ability’ is denoted by

1 f

and is considered as a second exogenous dimension of firm-

heterogeneity, distributed according to the cumulative distribution H. We can also think of

1 15 f

as entrepreneurial ability, that is, the manager’s ability16 . Contractual costs are investments in the management system that ensure delivery according to specified requirements, as well as investments in communicating their reliability to buyers, i.e. reputation. Contractual costs take the form:

f z

· λα ,

with α > 0, increasing with the ’reputation’ λ a firm wants to achieve, decreasing with the firm’s productivity z, and decreasing with the firm’s contractual ability marginal costs of production:

w z,

1 17 f .

Firms also face the usual

where labor is the only production factor and w represents wages,

and fixed costs of operation in the domestic market F . Lastly, aggregate demand in the domestic market is denoted by E and the aggregate price level is P . For a given firm, profits18 from sales in the domestic market: πd = p−θ · λθ−1 ·

E P

· (p −

w z)

−

f z

· λα

where p represents ’reliability’-adjusted prices. The firm chooses the price and the ’reliability’ that maximize profits, yielding: p= λα0 d =

z f ·α

θ θ−1

·

w z

z (θ−1) θ · ( θ−1 · θ ) · (w)

E P,

where α0 = α − θ + 1. We need to impose here that α0 > 0 (i.e. α > θ − 1) so that a higher f (i.e. lower contractual ability f1 ) is associated with a lower ’reputation’ level λ. A firm with higher productivity or higher contractual ability will endogenously choose a higher ’reputation.’ A firm will sell in the domestic market only if its profit is large enough to cover the fixed costs. The threshold 15 For presentation purposes, we are interested in  = f1 , rather than f itself, such that an increase in this dimension is a positive event. In this case  represents a firm’s contractual ability. 16 Verhoogen (2008) also refers to ’entrepreneurial ability’ as a determinant of quality. 17 We can think of the parameter f as the part of ’contractual ability’ that is independent from productivity. In this perspective our functional form for the management system investments imposes a positive correlation between total contractual ability and productivity. 18 These are gross profits. Net profits are equal to πd − F . Throughout this section we will use π for gross profits.

14

that determines viable domestic firms is given by: α

(

θ − 1 α·θ 1 z α·(θ−1) z θ−1 E α0 ) α0 · ( ) α0 · ( ) α0 · ·( − α) ≥ F θ w f ·α P θ−1

(3.1)

This equation shows that a combination of high z or low f (high f1 ) - high productivity or high contractual ability - ensures that firms can earn enough revenue to cover their fixed cost in the domestic market. In order to export firms have to incur additional fixed foreign market entry costs fx 19 . For a give firm profits from exporting are given by: −θ · πx = p−θ · λθ−1 · ( E P +τ

E∗ P∗ )

· (p −

w z)

−

f z

· λα ,

where foreign aggregate expenditure is given by E ∗ , foreign aggregate prices are given by P ∗ and τ represents variable costs of exporting20,21 . The optimal price and ’reliability’ are given by: p= λα0 x =

z α·f

θ θ−1

·

w z

z θ−1 θ −θ · · ( θ−1 · (E θ ) · (w) P +τ

E∗ P∗ )

Note that ’reliability’ λ is higher for exporters than for domestic firms because the larger market size gives more incentives to invest in ’reliability.’ A firm will export if the incremental profits from exporting rather than staying domestic are large enough to cover the costs of accessing foreign markets: πx (λx ) − πd (λd ) > fx ⇐⇒ α

(

z α·(θ−1) z θ−1 1 E E∗ α E α0 θ − 1 α·θ ) α0 · ( ) α0 · ( ) α0 · ( − α) · (( + τ −θ · ∗ ) α0 − · ) > fx θ w f ·α θ−1 P P P

Again, a combination of high z and low f (high

1 f)

(3.2)

- high productivity and high contractual

ability - allows firms to export. Equation (3.2) marks the curve in the (z,  = f1 ) space that separates 19

These can be thought of as the simplest costs related to exporting: paying registration fees, establishing distribution α which are not mandatory networks, learning about the foreign market to be accessed. As opposed to contractual costs f ·λ z costs for exporting, the cost fx has to be paid by every firm exporting and it does not vary with the productivity of the firm. It might vary with the number of markets accessed, but we will not explore that route here. 20 Note that we consider here the basic framework developed by Melitz where if firms export, they export to all markets; alternatively, there is only one export market. 21 These variable costs of exporting are known in the literature as ’iceberg trade costs’ and refer to the value of merchandize lost during transportation to the foreign market.

15

exporters from viable domestic firms and viable domestic firms from non-viable ones. This result is graphed in Figure 1 as the curve labeled ’Melitz’22 . Firms above this frontier will export. This curve is labeled ’Melitz’ because this result is equivalent to that in the Melitz model. The variable η = z α+1 f

= z α+1 · summarizes all the firm-level heterogeneity in the model such that the two dimensions

of heterogeneity collapse into one and there is only one threshold for the decisions to be a viable domestic firm or an exporting firm. The same result is obtained by Hallak and Sivadasan (2007) before minimum quality standards are imposed. The interpretation is that any level of productivity can be compensated by high enough ’contractual ability’ and, the reverse, any level of ’contractual ability’ can be compensated by high enough productivity to allow the firm to export. In other words, productivity and ’contractual ability’ are substitutes.

3.2

Differential contractual costs between the domestic and foreign markets

We now assume that a firm’s contractual ability is higher for transactions in the domestic market than for those in foreign markets. This means that any investment in reputation and reliability is less effective in increasing firm demand in foreign markets. This is not a novelty in the international trade literature as contract costs are commonly regarded as being higher for international transactions. This can happen because of frictions in international trade: it is more difficult for foreign buyers to be in contact with the supplier and be part of the monitoring process, information is imperfect and more noisy, contracts are more difficult to enforce and terms are likely to be stricter and more costly. Furthermore, firms need more sophisticated management systems in order to deal with larger markets, a larger variety of tastes or standards, and a larger degree of uncertainty. Let contractual costs be

1 z

· λα for the domestic market23 and

f z

· λα for foreign markets, with f > 1.

The firm’s optimization decision yields the same prices as before, but the ’reliability’ level changes to reflect the different costs. ’Reliability’ for domestic producers is now given by: λα0 d =

z α

z (θ−1) θ · ( θ−1 · θ ) · (w)

E P,

and for exporters by: λα0 x = 22 23

z α·f

z θ−1 θ −θ · · ( θ−1 · (E θ ) · (w) P +τ

E∗ P∗ )

Appendix C details the shapes and properties of the curves graphed in Figure 1. Domestic viability is given by z alone.

16

The frictions in contracting internationally make it harder for exporters to increase the ’willingness to pay’ for their products. However, the larger market size creates stronger incentives for firms to invest in improving their ’reputation’. The overall effect on ’reputation’ λ depends on the level of contractual ability

1 f

relative to relative market sizes

E P ∗ E +τ −θ · E P P∗

24 .

Note here also that we do not al-

low firms to choose different ’reputation’ levels for the domestic market and for the foreign markets. The reason for this is that firms invest in a management system that dictates processes applicable throughout the entire firm, and they could not have different management systems for different markets25 The threshold separating domestic firms from exporting firms is given by: πx (λx ) − πd (λd ) > fx ⇐⇒ (

θ−1 z α·(θ−1) z θ−1 1 E E∗ α E α θ − 1 α·θ ) α0 · ( ) α0 · ( ) α0 · ( − α) · (( + τ −θ · ∗ ) α0 − f α0 · ( ) α0 ) > fx θ w f ·α θ−1 P P P

(3.3)

This inequality (3.3) still say as before (see equation (3.2)) that more productive firms export. Note however, that now with f > 1, the exporting threshold is higher. This happens because the −θ · term (( E P +τ

α E ∗ α0 P∗ )

−f

θ−1 α0

·

E P

α α0

) is lowered by a f > 1, while the rest of the equation is the

same as in (3.2). This higher exporting threshold shows how higher contractual costs can make international trade more difficult. Furthermore, the inequality in (3.3) cannot be satisfied for any z −θ · when the term (( E P +τ

α E ∗ α0 P∗ )

−f

θ−1 α0

·

E P

α α0

) is negative. Therefore, firms export only if:

E 1 P >(E −θ · f + τ P

α

E∗ P∗

) θ−1 = ∗

(3.4)

1 f

has to be high enough for the firm

The interpretation of this condition is that ’contractual ability’

to capture a sales volume in the foreign market at least as large as the sales volume in the domestic market that it would command had it remained solely domestic. The ’contractual ability’ threshold ∗ is always less than the domestic ’contractual ability’ level, 1. Note that a small domestic market size relative to the size of foreign markets makes this threshold of contractual ability lower, while a low elasticity of substitution, θ, increases the contractual ability threshold26 . 24 Johnson (2008) finds that the market size effect dominates, implying that export firms have higher quality. In our model, if the market size effect dominates, export firms chose higher ’reputation.’ 25 One can think of f and λ as firm average ’contractual ability’ and ’reputation’ across both domestic and foreign markets. 26 α > 1. From the earlier condition that α0 = α − θ + 1 > 0 we have that θ−1

17

This result is illustrated in Figure 1 by the curve labeled ’Contractual model’. The frontier separating domestic firms and exporters defined by (3.3) is a hyperbola in z and  =

1 f.

Firms above

this frontier will export. Therefore, below a certain level of contractual ability  - ∗ -, there is no productivity improvement that can make up for low contractual ability. This result is reminiscent of that in Sutton (2007). In a model where firms and countries differ in productivity and quality, Sutton (2007) finds that there is a quality level below which firms cannot export regardless of their marginal costs of production. Here we obtain a similar result, but for contractual ability: there is a level of contractual ability below which firms cannot export regardless of their marginal costs of production. This result explains why firm size is not a perfect predictor of exporting status, as evidenced by the overlap in the distributions of exporters and non-exporters (see Figure 2). Hallak and Sivadasan explain the same empirical fact by imposing exogenous minimum quality standards for exporting.

3.3

Introducing ISO certification

We assume that acquiring certification requires a fee27 fc , that every firm that seeks ISO certification obtains it and we exclude adverse selection effects28 . We also assume that certification improves contractual ability to the point where f = f0 . For simplicity, take f0 = 1; that is, ISO certification makes international contractual ability equal to domestic contractual ability29 . The optimal choice of prices and ’reliability’ under certification are given by expressions similar to the earlier ones: p= λα0 xc =

z α

θ θ−1

·

w z

z θ−1 θ −θ · · ( θ−1 · (E θ ) · (w) P +τ

E∗ P∗ )

Now firms can chose not only between serving domestic markets or foreign markets, but also between serving export markets with or without certification. There are two effects of certification: 1. firms that already are exporters can get certification to improve their sales (the intensive margin) and 2. domestic firms can chose to become exporters with certification (the extensive margin)30 . Because 27

Appendix E discusses alternative functional forms for the certification cost. We assume that the certification process is thorough enough to prevent firms with poor management systems and poor reliability from passing the certification audits. 29 We refer to the firm’s f > 1 before certification as the initial ’contractual ability’. 30 If we extend our framework to include multiple destination markets, the intensive margin would refer to increases in sales volumes per market in already accesses markets, while the extensive margin would refer to increases in the number of export destinations reached. 28

18

our empirical analysis examines only the effects of certification on export sales, and not the effects of certification on firms’ entry into exporting, we focus here only on the decision between exporting with or without certification. Appendix D discusses the case where domestic firms, that previously were unable to export, now chose to become exporters with certification. A firm gets certification if: πxc (λxc ) − πx (λx ) > fc ⇔ (

θ−1 z α·(θ−1) z θ−1 E E∗ α 1 θ − 1 α·θ ) α0 · ( ) α0 · ( ) α0 · (( + τ −θ · ∗ ) α0 ) · ( − α) · (f α0 − 1) > fc θ w f ·α P P θ−1

(3.5)

The first thing to note in equation (3.5) is that only firms with f < 1, or  > 1, would find ISO certification to be profitable. In addition, firms still have to be productive in order to be able to cover the fixed cost of certification. The frontier for ISO certification in the (z,  = f1 ) space is a hyperbola with horizontal asymptote at  = 1. This is represented in Figure 1 by the curve labeled ’ISO frontier’. Firms below the ’ISO’ frontier will seek certification. The model’s prediction is that firms with high productivity and low contractual ability will select into certification. The frontier also shows that the lower the initial ’contractual ability’, the lower the productivity a firm can have and still profit from certification. Thus, the benefit from certification is proportional to the distance between the initial contractual ability and the contractual ability achieved with certification. Based on our model’s results, we make four conjectures that we expect to see in the data. First, we expect firms that get ISO certification to be larger and more productive ones. Second, because we believe contractual issues are likely to be more severe in international transactions, we expect ISO certification to have a stronger effect on export sales and export shares, although the model does not exclude the possibility that ISO certification affects domestic sales as well. Moreover, if we include the extensive margin effect in a multiple export destinations setting, we expect to see firms improving their total export sales after certification because they more markets31 . Third, contractual difficulties vary with industry characteristics. If there are more distortions in f relative to the ideal f = 1 in more contractually demanding sectors, we expect the benefits from certification to be larger in those sectors. Forth, we expect the effect of ISO certification to be larger for small firms because the largest gains of certification are for firms with low initial contractual ability. 31 We do not focus on the extensive margin aspect here, but we plan to use firm-level export destinations data to explore this issue empirically in future work.

19

4

Empirical analysis

We use Slovenian firm-level data from 1995 to 2005 from PASEF - the Data Analysis Service of the Faculty of Economics at the University of Ljubljana - covering all manufacturing sectors and comprising balance sheet and income statement information, including data on export sales. We match these data with data on ISO certification - if and when a firm obtained ISO 9001 quality management certification, collected from the Slovenian Chamber of Commerce and from individual certification agencies. The novelty in our data is the firm-level information on the date when the ISO certification was obtained. This information, together with the panel structure of our dataset, allows us to exploit time dynamics before and after firms obtain the ISO certification. In all our the empirical analysis we follow two sets of performance measures: 1. sales performance measured by total sales, domestic sales, export sales and export shares, and 2. operational performance measured by capital intensity, labor productivity, total factor productivity, price-cost margins and average wages. Export shares are the ratio of export sales to total sales. Capital intensity is the capital32 to employment ratio. Labor productivity is the ratio of gross value added, calculated as total revenue minus the cost of materials and services and other operational costs, to employment. Total factor productivity is estimated using the Levinsohn-Petrin (2003) method33 . The price-cost margin is defined as in Sembenelli and Siotis (2008) equal to: V alue added−W ages V alue added+M aterials Cost

Lastly, average wages are the total wage costs divided by total employment. All these variables are expressed in logs of real 2005 Euros34 , except export shares which are percentages. All the regressions are estimated for a sub-sample of the data where outliers were eliminated. More details about the data, including the data sources and the procedure for eliminating outliers, can be found in Appendix A. Table 1 shows the number of firms in the Slovenian manufacturing industries in the period from 1995 to 2005. There are 6476 firms in the entire manufacturing sector, of which 2520 (39%) are exporters and 686 (11% of all firms and 27% of exporters) are ISO-certified. Table 2 shows summary 32

Capital in our dataset is the book value of all tangible assets. We estimate firm-level total factor productivity at the 2-digit industry level. We use the gross value added as the dependent variable, and using materials and services inputs as proxies for unobservable productivity shocks. Estimated coefficients are available upon request. 34 We used KLEMS price deflators at the 2-digit industry level and the Euro to Slovenian Tolar exchange rate. 33

20

statistics for the main variables used. We use the definition of firm size from the World Development Indicators35 . A firm is ’micro’ if it has less than 10 workers, ’small’ if it has between 10 and 49 workers, ’medium’ if it has between 50 and 249 workers and ’large’ if it has more than 250 workers36 . Table 3 shows basic summary statistics by firm size categories. In our dataset, the majority of firms fall into the micro or small categories. Note also that the fraction of exporters as well as the fraction of exporters with ISO certification are considerably higher among medium and large firms. In our data, micro firms have 3.2 workers on average, while large firms have 643.5 workers on average. The largest firm in our dataset37 has over 6,000 workers and the smallest firms have 1 worker. The differences in sales volumes among firms of different sizes are considerable. Differences in labor productivity and price-cost margins do not seem large among the four firm categories. However, larger firms seem to have higher capital intensity, higher total factor productivity and higher wages on average. Before looking into the dynamics related to ISO certification, we document the contemporaneous advantages enjoyed by non-ISO certified exporters relative to domestic firms and the advantages of ISO-certified exporters relative to non-ISO certified exporters. We split firms into three types: ’domestic firms’ are firms that never export , ’non-ISO exporters’ are firms that export at least once during our our panel period 1995-2005, but do not have ISO certification, and ’ISO-exporters’ are firms that export at least once during our sample period 1995-2005 and have ISO certification38 . Contemporaneous advantages are calculated either as a simple difference: 100 · (M ean

lnXnon− ISO exporters − M ean

100 · (M ean

lnXISO exporters − M ean

lnXdomestic f irms ),

and lnXnon− ISO exporters ),

where X is any of the performance variables listed above39 ; or based on coefficients from a regression of the form:

lnXijkt = β0 + β1 · non ISOexporteri + β2 · ISO exporteri + θ · sizei + κk + ζj + ξt + it 35

(4.1)

World Development Indicators (2007), The World Bank. Size is a time-invariant variable given by the median firm size during a firm’s existence in our panel. 37 These statistics are performed on the subset of the original dataset where outliers have been excluded. 38 Note that these firm definitions are time-invariant. 39 This summary statistic is inspired from those presented by Bernard, Eaton, Kortum, Kramartz (2003) for US exporters. 36

21

where k indexes regions, j indexes sectors and t indexes years. The advantages are equal to 100 · (expβi − 1), where i=1,2. These advantages are calculated both for the levels of the performance variables as well as for their growth rates. Table 4 shows exporters with ISO certification are better in all dimensions considered than exporters without ISO certification and exporters without ISO certification are better than domestic firms. Table 4(a) shows that ISO-certified exporters have both higher domestic sales and higher export sales by a large margin (60-90%), and their export shares are higher by about 3%. These firms also enjoy higher capital intensity, higher productivity and higher wages relative to non-ISO exporters, but no significant advantage in price-cost margins. Furthermore, Table 4(b) shows that ISO-exporters have higher growth rates in both domestic and export sales, as well as in productivity measures relative to non-ISO exporters, while non-ISO exporters have higher growth in total sales40 and in capital intensity. The hierarchy of ISO-exporters being better than non-ISO exporters, and these being better than domestic firms is illustrated also in the distributions graphed in Figure 2. Next, we describe how we can disentangle the self-selection of better firms into ISO certification from the effect of ISO certification. Our theoretical framework predicts that there should be self-selection of firms into ISO certification and that ISO certification should improve a firm’s sales performance. First, firms with high productivity and mid-level contractual ability should select into certification. Because cannot observe a firm’s contractual ability, we rely on observable variables - sales, capital intensity, productivity, price-cost margins and average wages -, to document self-selection patterns. Second, ISO certification should have an effect on firm sales. The effect should occur especially on export sales because contractual constraints are more severe in international contracts. Our theoretical model also indicates that the effect of ISO certification varies with the initial contractual ability of the firm. We reinterpret this as the effect of ISO certification varies by firm size. Lastly, the effect of ISO certification should be larger in sectors where contractual difficulties are more prevalent, and therefore where initial contractual abilities of firms would be lower. We test all these hypotheses in turn. We use methodologies borrowed from the empirical international trade literature. One of the aspects this literature has been concerned in the past few years is disentangling the selection into exporting from the learning from exporting hypotheses, as surveyed by Wagner (2007). The standard 40

This is almost automatic since non-ISO exporters have foreign sales which domestic firms do not have.

22

approach has been to calculate ’exporter premia’ from OLS regressions of log performance in periods before or after firms start to export on an exporting status dummy and other controls. We conduct a similar analysis examining instead various measures of performance before and after a firm’s being granted ISO certification. First, we describe the self-selection patterns of exporters with ISO certification. Second, we document the premia for exporters after they obtain certification, relative to those exporters that do not have ISO certification. Lastly, in order to take into account the self-selection of exporters into certification, we estimate a series of differences-in-differences models controlling for either firm fixed effects, firm growth trends or both. For all of these questions, we consider two sets of variables: sales performance (looking separately at domestic sales and export sales) and operational performance, measured by capital intensity, measures of productivity, the price-cost margin and the average wage. Lastly, we explore the effects of ISO certification by firm size and by sector. Throughout our empirical analysis we use the following two time-variant variables to keep track of when a firm obtains the ISO certification: 0 ISO0 is equal to 0 in all periods before the firm became ISO certified, and equal to 1 in all periods afterwards; and 0 ISOstart0 is equal to 1 only in the year when the firm has obtained the certification. We denote the period when the firm obtained the ISO certification for the first time by TiISO .

4.1

Self-selection

We investigate self-selection by looking at various measures of firm performance 1, 2 and 3 years before the firm becomes certified, controlling for size and other firm characteristics, as well as region, sector and time fixed effects:

lnXijk,t−n = β0 + β1 · ISO startit + β2 · lnZi,t−n + θ · sizei + κk + ζj + τt + ξit ,

(4.2)

where i indexes firms, j indexes sectors, k indexes regions and t indexes years, κk represents region fixed effects, ζj represents sector fixed effects, τt represents time fixed effects, the vector Zit represents lags of firm characteristics, such as sales, capital intensity, productivity, price-cost margins, wages. ξit is the error term. n takes the values 1, 2 or 3. Each ISO exporter is included only once when t = TiISO . Selection premia can be calculated for growth rates as well, by changing the dependent variable in equation (4.2) to lnXijk,t − lnXijk,t−n . Premia are calculated based on the formula: 100 · (expβ1 − 1).

23

The pre-ISO premia presented in Table 5(a) show that ISO-certified exporters have 13% higher total sales and up to 60% higher export sales relative to non-ISO exporters. ISO exporters also have higher capital intensity and total factor productivity relative to non-ISO exporters. Note that these premia increase closer to the time of certification. Total sales are higher by about 13% for ISO exporters, but export sales are higher by up to 60% relative to non-ISO exporters. The capital intensity is consistently about 30% higher for ISO exporters and total factor productivity is about 7% higher. Capital intensity might just be a proxy for firm size or for the skill intensity used in production, or it might indicate that more capital intensive firms can standardize more because they rely on automated processes of production. Therefore, some firms could have low skill and high capital intensity and produce goods at low quality, but with low quality variance. There does not seem to be any precertification differences in terms of price-cost margins or average wages. Table 5(b) shows that ISO exporters also have higher growth rates of total sales, and again, especially of export sales. Export sales are growing by an additional 16% more for ISO-exporters in the period right before certification. This might be due however to delays in recording the formal certification or to the fact that firms announce that they have ISO certification while still in the process of obtaining it. Another aspect going in the firm’s self-selection into certification is the timing during a firm’s experience as an exporter when it finds certification worthwhile. Table 6 shows that the largest proportion of ISO-certified firms obtain this certification in the first 4 years of exporting experience. The probabilities of continuing to export conditional on a certain export experience indicate that after 4 periods of exporting it becomes very probable that the firm is a successful exporter and will continue exporting. Therefore, obtaining ISO certification is not attractive after the firm learns that it is successful at exporting.

4.2

Post-ISO Premia

We investigate changes after ISO certification by calculating ’post-ISO’ premia and then by differencesin-differences estimation. The ’post-ISO’ premia are calculated from the regression:

lnXijk,t+n = β0 + β1 · ISO startit + β2 · lnZi,t + θ · sizei + κk + ζj + τt + ξt where the premia is equal to 100 · (expβ1 − 1).

24

(4.3)

The first indication of the changes after ISO certification can be seen in Figure 3. Those figures graph the distributions of the performance variables, as well as those of two other interesting variables: the number and the popularity of export destinations41 . The distributions are plotted for ISO exporters only, and cover up to 3 periods before the certification was obtained and up to 3 periods after. The figure shows a clear shift of the distributions in most cases. In addition, Figures 3(e) and 3(f)42 show that ISO exporters serve more markets and less popular markets that are harder to access. The post-ISO premia are presented in Tables 7(a) and 7(b). In 7(a) shows that after certification, ISO exporters have between 60 and 80% higher export sales than their non-certified counterparts, and have higher export shares by about 4-5%. Capital intensity continues to be significantly larger, but the size of the difference is now around 4% compared to the 30% difference with non-ISO exporters before certification. Domestic sales, for which there were no pre-ISO advantage, are higher after certification. Both labor productivity and total factor productivity are higher after certification by between 8 and 16%. There is still no significant change in price-cost margins, but we do observe higher average wages in certified firms after they obtain the certification. The fact that we do not see a significant effect in price-cost margins can indicate that both the signalling effect of ISO certification and the real changes effect of ISO certification are at play. Price-cost margins might not change because although prices might increase, there could be increases in costs associated with the implementation of ISO standards and of a new management system43 . Table 7(b) shows significantly higher growth rates post-certification not only in total sales and export sales, but also in capital intensity, productivity and average wages.

4.3

Differences-in-differences estimation

In order to take into account for self-selection into certification, we estimate several differences-indifferences models (Ashenfelter (1978), Ashenfelter and Card (1985), Card (1992), Card and Krueger (1994)). The basic differences-in-differences model estimates the change in the outcome for ’treated’ (ISO exporters) after the ’treatment’ occurred (ISO certification obtained) relative to a constant, timeinvariant, treatment-independent difference in the outcome variable between ’treated’ and ’non41

The popularity of an export destination is given by the fraction of Slovenian exporters that serve it. The data source for firm-level export destinations is the Statistical Office of the Republic of Slovenia. 43 Quality control, for example, can involve both fixed and variable costs. At the same time, costs might decrease with certification. Yet another effect might be that ISO certification is prevalent in more competitive industries where price-cost margins are low. 42

25

treated’. We refer to this as the Basic OLS model:

lnXijk,t = β0 + β1 · ISOit + β2 · ISO exporteri + β3 · lnZi,t−1 + θ · sizei + κk + ζj + τt + ξit

(4.4)

This model includes, as before, sector, region and year fixed effects, as well as lagged firm characteristics, and in addition, θi , which represents firm size fixed effects. To control further for any unobservable time-invariant firm characteristics, we estimate this model using firm fixed effects44 . Therefore, we have the Basic FE model :

lnXijk,t = β0 + β1 · ISOit + β2 · lnZi,t−1 + γi + τt + ξit

(4.5)

where γi represents firm fixed effects. These specifications control for the fact that exporters that obtain ISO certification have higher levels of sales, capital intensity, productivity etc., but do not control for the fact that ISO exporters might also have higher growth rates in any of these performance variables. Our first attempt to control for higher growth rates of exporters that select into ISO certification is to include a firm-specific growth trend. We calculate the average growth rate in variable X for firm i before the certification was obtained, gX,i = meant
avoid any confounding effects of the firm announcing its ISO certificate before the formal certification. Therefore, we estimate Trend Models in OLS:

lnXijk,t = β0 + β1 · ISOit + β2 · ISO exporteri + β3 · lnZi,t−1 + θ · sizei + gX,i · t + κk + ζj + τt + ξit (4.6)

and with firm FE:

lnXijk,t = β0 + β1 · ISOit + β3 · lnZi,t−1 + gX,i · t + γi + τt + ξit

(4.7)

Lastly, if we believe there is firm-level unobserved heterogeneity in growth rates, then the model would look like:

lnXijk,t = β0 + β1 · ISOit + β2 · lnZi,t−1 + γi + γX,i · t + +τt + ξit 44

In this specification, firm fixed effects replace the dummy for ’treated’, the ’ISO exporter’ dummy.

26

(4.8)

where now γX,i are firm-specific parameters to be estimated. We can estimate this model by differencing it, and call it the Differenced Model. Therefore, we estimate this model without any lagged changes firm controls ∆Zit :

lnXijk,t − lnXijk,t−1 = β1 · ∆ · ISOit + γX,i + τt + ∆ξit

(4.9)

and with lag changes of firm controls:

lnXijk,t − lnXijk,t−1 = β1 · ∆ · ISOit + β2 · ∆lnZi,t−1 + γX,i + τt + ∆ξit

(4.10)

These models identify a causal effect of ISO certification on performance if there are no other unobserved changes at the firm-level occurring at the same time as certification. Table 8 supports the evidence illustrated by the post-ISO premia through more rigorous differences-in-differences estimations of the effects of ISO certification on exporter performance. The first thing to note is that ISO certification has a significantly positive effect on total sales, export sales and export shares almost throughout the panel of specifications. Note that the effects of ISO certification are diluted as more controls are included in the regressions, and as firm-level unobserved heterogeneity and growth trends are taken into account. However, in the strictest specifications the FE Trends model and the Differenced model with FE and lags - we still observe effects of 4.6% increase in total sales, 7% increase in export sales, and 0.7% increase in export shares due to certification. Firm unobserved heterogeneity accounts for a large portion of the differences between ISO-certified firms and non-ISO firms, as seen by comparing the results from the OLS and the firm fixed effects specifications. The OLS Trends indicates increases by up to 35% in export sales, increases of 7% in total sales, and export shares higher by about 4%. There is no consistent effect on capital intensity or price-cost margins. There is only a weak indication of increased productivity and average wages after certification. We should carefully emphasize that these estimates do not cleanly identify a causal effect of ISO certification. For the coefficients estimated in the models described above to be causal effects of ISO certification we have to assume that there is no unobservable change at the firm-level occurring at the same time as ISO certification. The fact that ISO certification can not only verify that management systems are in place, but also induce changes in the management systems used by firms, 27

might also confound our results. In particular, the effects we observe could be due to real changes in firm’s management, rather than due to the signalling and reputational role of ISO certification. By considering the two different sets of firm performance - sales and operations - we believe that it is possible to tell which of these effects is at play. Although the econometric models do not disentangle from pure contractual innovations and the signalling effect of ISO, both effects are related to contractual improvements, so our results suggest that contractual improvements matter for exporter performance. Furthermore, regardless of whether the effects we observe are due to the signalling role of ISO certification or due to real changes in the firm occurring with ISO certification, both of these events are due to the certification. Therefore, although the exact mechanism might remain uncertain, we can tell whether ISO certification has any role in exporter performance and this is what matters for policy purposes.

4.4

Results by firm size

The effects of ISO certification might differ by firm-type. Our theoretical framework indicates that the benefit of certification is higher the lower the firm’s initial ’contractual ability’. This might indicate that smaller firms should benefit more from certification, as they are the more prone to information and contractual problems. On the other hand, one might argue that larger firms are better able to seize the benefits of ISO certification as they have more resources to complement ISO certification and build their reputation. We include interactions between our ISO dummies and firm size in our estimation of post-ISO premia, as well as in the estimation of the Trend model and the Differenced model with lagged firm controls and firm FE to test for these differential effects. Tables 9 and 10 indicate that the largest effects of ISO certification are for the smallest and for the largest firms. The premia indicate that total sales and domestic sales increase most for micro firms, but export sales increase most for medium and small firms. The premia for wages are highest for micro firms. The premia for capital intensity or productivity do not give a consistent picture. The results from the differences-in-differences estimations in Table 10 also give suggestive evidence that micro and large firms benefit most from the certification. Our theoretical model indicates that the largest benefit should be for micro firms, which presumably are productive but have a low contractual ability. The fact that we observe large effects of certification for large firms indicates that perhaps the costs of obtaining certification are lower for these firms and that they are most able to capitalize

28

on the benefits from this reputational tool by complementing it with other related actions. The effects by firm size are not robust across different specifications though. In the Trends FE estimation, export sales increase for micro firms by about 58% after certification. This effects are diluted in the Differenced FE w/ Lags model, where a 15% increase in export sales is suggested, but is not statistically significant. On the other hand, there is almost no effect of certification for large firms in the Trends FE model, but we find a 22% increase in export sales for large firms in the more demanding Differenced FE w/ Lags estimation. The effects are significantly different across different firm sizes only for the Trends FE model45 .

4.5

Sectoral analysis

ISO certification is a signal of firm contractual ability and reliability, and this signal should be more relevant in ’contractually intensive’ sectors and in intermediary goods sectors. Therefore we expect ISO take-up rates to be higher in these sectors and the role of ISO certification in performance to be greater for firms in these sectors. Furthermore, evidence that the effects of ISO certification increases with sectoral contractual intensity strengthens our confidence that we identify effects related to changes in contractual ability or perception, and not from changes of other nature. This empirical analysis is related to the identification used by Rajan and Zingales (1998): firms in sectors that rely more on ISO certification due to the inherent nature of the production process in those sectors, are likely to benefit most when ISO certification is adopted. Along the same argument, in addition to several sectoral indices of contractual intensity, we also look at sectoral rates of ISO take-up among US exporters, which indicate an inherent demand for ISO certification at the sector level. For this analysis use the US Input-Output Benchmark tables to construct five sectoral indices: capital intensity, contractual complexity, client industries’ contractual complexity, final consumption share in output and the US rate of ISO certification, each hypothesized to indicate a differing role of ISO certification across sectors. The first two indices are calculated based on the formulas proposed by Cowan and Neut (2007): the KL, capital intensity index is taken from Cowan and Neut and the industry Herfindal index of intermediate use is recalculated46 as described in Appendix B. The latter index indicates the contractual dependence of a sector: the lower the index, the higher the contrac45

Note that both sample sizes, as well as the fraction of exporters and the fraction of ISO-certified exporters vary greatly among the four firm size groups. 46 We recalculated these indices from the Benchmark US Input-Output table for 1997, at the 6-digit NAICS level. We used industry concordances to obtain indices for the Nace Rev. 1 codes.

29

tual complexity as the producer relies on more intermediate suppliers. Firms in sectors with high contractual complexity are concerned with the timely delivery of the intermediary inputs, as well as with the variance in intermediary product features47 . For our purposes here we are actually interested in whether the firm is a supplier to an industry with higher contractual complexity. Therefore we construct a ’supplier index’ as the average contractual complexity of the industries an industry supplies to, weighted by the share of total output that is supplied to each client industry. Lastly, we construct the share of an industry’s total output that goes to final consumption and the ISO certification rate among exporters for US industries. Each of these indices relate to a sector’s intrinsic demand for ISO certification or other similar contractual or reputational devices. More details on these indices and the data sources used to construct them are found in Appendix B. Table 11 indicates that ISO certification is more relevant in sectors where contracts are more important. This table present averages of these indices for each 2-digit manufacturing industry, together with the rate of certification among exporters for each industry48 . The first pattern to note is that the correlations between the rate of ISO certification among Slovenian exporters and each of these indices are as expected: sectors with higher capital intensity, with higher contractual complexity49 , sectors that supply to industries with higher contractual complexity, sectors that supply more intermediate goods, and sectors where the certification rate is higher among US exporters all have higher rates of certification among Slovenian exporters. We examine post-ISO changes depending on sector characteristics by including interactions between the ISO dummies and the sectoral indices in the regressions described in the previous section. We first estimate post-ISO premia in Table 12 and then the differences-in-differences Trend FE and Differenced FE w/ lag changes models in Table 13. Premia are calculated based on the coefficients from regressions of the form: lnXijk,t+n = β0 + β1 · ISO startit + βI · ISO startit · Indexk + β2 · lnZi,t + θ · sizei + κk + ζj + ξt , where region, sector, year and size controls are included, as well as firm characteristics at time t. We calculate premia for the least and most contractually dependent sectors based on formulas: 47

The argument behind both of these statements can be that firms in high contractual complexity sectors need to coordinate inputs from many suppliers and therefore any lapse in timely delivery or pre-established input features are more costly. 48 While we calculated and used in all the regressions indices at the 4-digit level, for presentation purposes we show only averages at the 2-digit level in the table. 49 Note that a lower Herfindal index or a lower supplier index represents higher contractual complexity.

30

100 · (exp(β1 +βI ·mink Indexk −1) ), and respectively, 100 · (exp(β1 +βI ·maxk Indexk −1) ) Table 12 presents for each contractual index the ranges of post-ISO premia between the sector with the lowest contractual dependence and the sector with the highest contractual dependence. Although there is a strong correlation between ISO certification and the industry’s capital intensity, we do not observe any significant differences in the effects of ISO according to this sectoral index. This may be due to the fact that the capital intensity index does not capture well the nature of the certification as a contractual device. For the indices for which we do have a strong inclination to believe they are good measures of the relevance of contractual tools such as ISO certification - the Herfindahl index of intermediate input usage, the suppliers index and the final consumption index -, Table 12 shows that indeed the strongest post-ISO premia in export sales and export shares are for the most contractually intensive sectors. The premia by the US ISO certification rate also indicate that export shares improve mostly in the sectors most reliant (by nature, or by the intrinsic nature of the sector’s technology) on ISO certification and other similar institutions. An interesting and consistent result in the post-ISO premia table is the fact that the price-cost margins increase for ISO-certified firms in the most contractually intensive sectors (and decrease in the least contractually intensive ones). This finding may be due to the fact that in the most contractually intensive sectors the effect that dominates is the fact that firms can charge higher prices after proving their reliability, whereas in least contractually intensive sectors the effect that dominates is the increased costs due to implementing the certification. We also see that the strongest effect on domestic sales is found in the sectors with the lowest contractual intensity. Lastly, the strongest increases in wages occur in sectors with lower contractual intensity. The results on export shares and export sales, together with the latter result for the price-cost margins confirm that 1. that ISO certification is most relevant in international contracts, and 2. that ISO certification allows firms to credibly inform buyers about their reliability and therefore capture larger sales and higher margins. Lastly, the results for the differences-in-differences models presented in Table 13 confirm the fact that ISO certification enhances mainly export sales and that the effect of certification is highest in the 31

sectors where certification is hypothesized to matter most. The most important results are for the Herfindahl index and for the supplier index. We find a strong, significant negative coefficient on the interaction term between certification and these indices, indicating that the higher a sector’s contractual dependence, and the higher a sector’s client industries’ contractual dependence, the larger the effects of certification50 . This result is robust in our strictest econometric model, the Differenced FE model with lag firm characteristics. The results for the share of output going to final consumption and for the US ISO certification rate also indicate that the effect of ISO certification is higher in sectors where this certification is most relevant and required. Lastly, the results for the capital intensity index suggest that the effect of ISO certification is highest in sectors with lower contractual intensity. This fact might occur because standardization is more difficult in sectors with lower contractual intensity, and therefore when standardization occurs, the rewards are larger.

5

Conclusion

We find strong self-selection of exporters into ISO certification. After controlling for observed selfselection, as well as unobserved time-invariant firm fixed effects and firm time trends, we find an effect of ISO certification on total sales and export sales. Because we do not find a consistent or robust effect for any of the operational measures, we incline to believe that the effect of ISO certification comes mostly through its signalling role. We find that the smallest and the largest firms benefit most from certification. We also find that firms in more contractually dependent sectors gain the largest increases in export sales after ISO certification. These varying effects of certification by firm type, together with the relative small average aggregate effects of certification indicate that ISO certification is not always a clear recipe for success. At the same time, we have seen that the effects can be considerably large in particular types of firms. We explored the situations that make ISO certification more valuable, but further work in this area might be worthwhile, especially for policy purposes. The limitation of our empirical work is the inability to control for unobserved firm-level changes that might occur concomitantly with certification. In further work we plan to use firm-level data on export destinations to construct an instrumental variable that would improve our control for unobserved sales shocks or news. Specifically, the average growth on a firm’s export destinations’ in the firm’s sector could control for unobserved firm-level information regarding future contracts. 50

A higher Herfindahl Index and supplier index signify lower contractual dependence.

32

Also, we want to use firm-level export destinations data to explore the extensive margin effect of ISO certification. Going back to the motivation for our study, our results suggest that reliability and signalling are important in international trade. Some firms are able to surpass these hurdles and establish their reputation abroad. However, many firms are not, and for some even private-sector institutions such as the ISO certification are not effective. This indicates a scope for government action to correct for these frictions and asymmetries in international trade, that affect mostly small and mediumsize firms. Lastly, our study complements current research showing that contractual problems can significantly alter theoretical predictions regarding international trade.

33

Bibliography [1] Anderson, James and Eric van Wincoop (2004). ”Trade Costs.” Journal of Economic Literature, Volume 42, Number 3, pp. 691-751. [2] Arkolakis, Costas (2008). ”Market Penetration Costs and the New Consumers Margin in International Trade.” NBER Working Paper 14214. [3] Arora and Asundi (1999). ”Quality Certification and the Economics of Contract Software Development: A Study of the Indian Software Industry.” NBER WP, July 1999. [4] Ashenfelter, O. (1978). ”Estimating the Effect of Training Programs on Earnings.” Review of Economics and Statistics, Volume 60, pp. 4757. [5] Ashenfelter, O. and Card, D. (1985). ”Using the Longitudinal Structure of Earnings to Estimate the Effects of Training Programs.” Review of Economics and Statistics,Volume 67, pp. 648660. [6] Banerjee, Abhijit V. and Esther Duflo (2000). ”Reputation Effects and the Limits of Contracting: A Study of the Indian Software Industry.” The Quarterly Journal of Economics, Volume 115, Number 3, pp. 9891017. [7] Bernard, Andrew and J. Bradford Jensen (1999). ”Exceptional Exporter Performance: Cause, Effect or Both?” Journal of International Economics, Volume 47, pp. 1-25 [8] Bernard, Andrew and J. Bradford Jensen (2004). ”Why Some Firms Export.” Review of Economics and Statistics, Volume 86, Number 2, pp. 561569. [9] Bernard, Andrew, Jonathan Eaton, J. Bradford Jensen and Samuel Kortum (2003). ”Plants and Productivity in International Trade.” The American Economic Review, Volume 93, Number 4, pp. 1268-1290. [10] Bigsten, Arne, Paul Collier, Stefan Dercon, Marcel Fafcharnps, Bernard Gauthier, Jan Willern Gunning, Jean Habarurema, Anders Isaksson, Abena Oduro, Remco Oostendorp, , Mans Soderborn, Francis Teal, and Albert Zeufack (2006). ”Exports of African Manufacturers: Macroeconomic Policies and Firm Behavior.” Journal of International Trade and Economic Development, Volume 8, Number 1, pp. 5371. [11] Bloom, Nick, Stephen Dorgan, John Dowdy and John Van Reenen (2007). ”Management Practices across Europe, the US and Asia.” Stanford University mimeo. [12] Brooks, Eileen (2006). ”Why Don’t Firms Export More?”. Journal of Development Economics, Volume 80, Issue 1, pp. 160-178. [13] Bureau of Economic Analysis (1997). ”Benchmark Input-Output Accounts of the United States, 1992.” [14] Card, David (1992). ”Do Minimum Wages Reduce Employment? A Case Study of California, 19871989.” Industrial and Labor Relations Review, Volume 46, pp. 3854. [15] Card, D. and Krueger, A. B. (1994). ”Minimum Wages and Employment: A Case Study of the Fast-Food Industry in New Jersey and Pennsylvania.” American Economic Review, Volume 84, pp. 772793. [16] Cowan, Kevin and Alejandro Neut (2007). ”Intermediary Goods, Institutions and Output per Worker.” Bank of Chile Working Paper 420. [17] Demidova, Svetlana, Hiau Looi Kee and Kala Krishna (2008). ”Do Trade Policy Differences Induce Sorting? Theory and Evidence from Bangladeshi Apparel Exporters.” mimeo. [18] Dobrescu, Gabriela and Helena Schweiger (2008). ”Underexporting Puzzles.” University of California at Berkeley, mimeo. [19] Eaton, Jonathan, Samuel Kortun and Francis Kramarz (2007). ”An Anatomy of International Trade: Evidence from French Firms.” University of Minnesota, mimeo. [20] Evans, Carolyn (2001). ”Home Bias in Trade: Location or Foreign-ness?”. Federal Reserve Bank of New York Working Paper.

34

[21] Evans, Carolyn (2003). ”The Economic Significance of National Border Effects.” The American Economic Review, Volume 93, Number 4, pp. 1291-1312. [22] EU KLEMS Dataset (2008). (http://www.euklems.net/). [23] Hallak, Juan Carlos and Jagadeesh Sivadasan (2006). ”Productivity, Quality and Exporting Behavior under Minimum Quality Requirements.” University of Michigan, mimeo. [24] Hudson, John and Philip Jones (2003). ”International Trade in ”Quality Goods”: Signalling Problems for Developing Countries.” Journal of International Development, Volume 15, Issue 8, pp. 999-1013. [25] International Standards Organization (2007). ”2006 ISO Survey.” [26] Johnson, Robert (2008). ”Trade and Prices with Heterogeneous Firms.” UC Berkeley, mimeo. [27] Levchenko, Andrei A. (2007). ”Institutional Quality and International Trade.” Review of Economic Studies, Volume 74, pp. 791-819. [28] Levinsohn, James and Amil Petrin (2003). ”Estimating Production Functions Using Inputs to Control for Unobservables.” Review of Economic Studies, Volume 70(2), Number 243, pp. 317-342. [29] Melitz, Marc (2003). ”The impact of Trade on Intra-Industry Reallocations and Aggregate Industry Productivity.” Econometrica, Volume 71, Issue 6, pp. 1695-1725. [30] Nunn, Nathan (2007). ”Relationship-Specificity, Incomplete Contracts and the Pattern of Trade.” The Quarterly Journal of Economics, Volume 122, Number 2, pp. 569-600. [31] Portes, Richard and Helene Rey (2005). ”The Determinants of Cross-Border Equity Flows.” Journal of International Economics, Volume 65, pp. 269296. [32] Rajan, Raghuram G. and Luigi Zingales (1998). ”Financial Dependence and Growth.” The American Economic Review, Volume 88, Number 3, pp. 559-586. [33] Rauch, James E. and Joel Trindade (2002). ”Ethnic Chinese Networks in International Trade.” The Review of Economics and Statistics, Volume 84, Number 1, pp. 116-130. [34] Rauch, James E. and Joel Watson (2003). ”Starting Small in an Unfamiliar Environment,” International Journal of Industrial Organization, Volume 21. [35] Sembenelli, Alessandro and Georges Siotis (2008). ”Foreign Direct Investment and mark-up dynamics: Evidence from Spanish firms.” Journal of International Economics, Volume 76, pp. 107-115. [36] Shepherd, Ben (2007). ”Product Standards, Harmonization and Trade: Evidence from the Extensive Margin.” World Bank Policy Research Working Paper, Number 4390. [37] Sutton, John (1989). ”Endogenous Sunk Costs and the Structure of Advertising Intensive Industries.” European Economic Review, Volume 33, pp. 335-344. [38] Sutton, John (2007). ”Quality, Trade and the Moving Windows: Competitiveness and the Globalization Process.” London School of Economics, mimeo. [39] Terlaak, Ann and Andrew A. King (2006). ”The Effect of Certification with the ISO 9000 Quality Management Standard: A signalling Approach.” The Journal of Economic Behavior and Organization, Volume 60, Number 40, pp. 579-602. [40] United Nations Statistics Division (2008). ”UN Classifications Registry.” [41] US Census Bureau (1992). ”1992 Census of Manufacturers” [42] US Census Bureau (2002). ”2002 Economic Census: Manufacturing” [43] Verhoogen, Erik A. (2008). ”Trade, Quality Upgrading and Wage Inequality in the Mexican Manufacturing Sector.” Quarterly Journal of Economics, Volume 123, Number 2, pp. 489-530. [44] Wagner, Joachim (2007). ”Exports and Productivity: A Survey of the Evidence from Firm-Level Data.” The World Economy, pp. 60-82. [45] Yoshino, Yutaka (2007). ”Domestic Constraints, Firm Characteristics, and Geographical Diversification of Firm-level Manufacturing Exports in Africa.” The World Bank, mimeo. [46] ”World Development Indicators.” The World Bank, 2007.

35

Appendix A

Data description and cleaning

The data used in this paper is from two main sources. The balance sheet and income statement data is from PASEF/AJPES. It covers all the officially registered firms in Slovenia over the period 1995 to 200551 . The data on ISO certification is from the Slovenian Chamber of Commerce and from the respective certification agencies: Slovenian Institute of Quality and Metrology (SIQ), Bureau Veritas (BV) office in Ljubljana, ¨ TUVRheinland (TUV) office in Maribor and Slovenian Quality (SQ). The Slovenian Chamber of Commerce supplied us with a list of all ISO 9001 certified firms as of April 2006. We obtained data on the certification date from the certification agencies. In addition, we requested this information ourselves by email or telephone for several firms. Data on the region a firm belongs to, the birth year and foreign ownership are from the Business Register. The dataset also includes the industry code, based on the NACE rev. 1 classification, at the 5 digit level. We consider for our analysis only firms with the modal industry code for the period of 1995 to 2005 in the manufacturing industries (as listed in Table 1)52 The original data is reported in thousands local currency. We use KLEMS 53 output and value added price indices at the 2 digit sector level to transform the values for sales and value added in real values. We use consumer goods and capital goods price indices from the Slovenian Statistical Office to transform wages and capital into real values. We then use the exchange rate for the Slovenian Tolar to the Euro for 2005 to transform into thousands of constant 2005 Euros units. The PASEF data contains information on capital54 , labor, sales - including export sales separated from domestic sales - and firm costs from which we calculate all other variables of interest. Employment is number of workers calculated based on hours worked55 . This labor measure does not include the owners who are not employed as managers in the firm, student and contract labor, or overtime hours. Due to this definition we have observations with zero labor in the dataset56 . We drop firms that report zero labor in the middle of their existence in our dataset and keep firms that have zero labor in the first two years or in the last two years. For these firms we drop only those observations where labor is zero, and therefore cannot calculate either capital intensity of labor productivity). We calculate labor productivity as the ratio of gross value added (total sales minus costs of materials and services and other operational costs) over labor. We estimated total factor productivity using the LevinsohnPetrin (2003) method, using materials and services inputs as proxies for unobservable productivity shocks, V alue added−W ages and gross value added as the dependent variable. The price-cost margin is defined as V alue added+M aterials Cost . Average wages are calculated as the total wage costs over the number of employers. Firm size is defined as in World Development Indicators: micro firms have less than 10 workers, small firms have between 10 and 49 workers, medium firms have between 50 and 249 workers and large firms have more than 250 workers. From the initial dataset containing 71,236 observations and 9757 firms we eliminated outliers according to the following procedure. We first drop all the firms where labor is zero somewhere in the middle of the panel; then we drop the observations where firms have zero labor in either the first or last two periods. This reduces our dataset to 6,800 firms, and around 50,000 observations. We then remove outliers as follows: for the log capital intensity and the log labor productivity intensity we create thresholds equal to the average +/3 standard deviations, by year and by sectors. We categorize an observation as outlier if the capital intensity and the labor productivity values are outliers, or if either the capital intensity or the labor productivity values are outliers and the sum of the ratios of deviations from means, calculated in standard deviations, of capital intensity relative to labor productivity and the inverse of this are larger than 3. Any firm that has at least 1 observation considered outlier in any year in the panel is considered outlier firm and thus eliminated from our sample. After this process we are left with 45,246 observations 6089 firms, so about 10% of the original dataset was eliminated as outliers. 51

All firms are legally obliged to report to AJPES. However, each year there is a small fraction of firms that do not report and instead pay a penalty for not doing so. Note also that sole proprietors are not included here. 52 2 sectors were too small and we merged them with the most similar larger sector at the 2 digit level. Sector ”manufacturing of tobacco products” was merged with ”manufacturing of textiles”. Sector ”manufacture of coke, refined petroleum products and nuclear fuel” was merged with ”manufacture of chemicals and chemical products”. 53 http://www.euklems.net/index.html 54 Capital here is the book value of physical/tangible assets. 55 Given a certain number of working weeks in a year, and usually 8 hours of work in a day for each full time worker. 56 Because firms always report to AJPES they sometimes report income statements with no sales, or sometimes even with no labor as well because they are legally still registered.

36

We obtained a list of the Slovenian certified companies from the Slovenian Chamber of Commerce as of April 2007 and the certification date for each of these from the respective certification agencies, or from the firms themselves. There are still a few firms for which we were unable to obtain the certification date. Because we know all the firms that are certified, we dropped the firms for which we do not have the certification date so as to not bias our results.

Appendix B

Industry-level indices

We construct 4 industry-level indices inspired from the work by Cowan and Neut (2007). Cowan and Neut (2007) calculate measures of sectoral contractual dependence from the 1992 Benchmark Input-Output table for the US. We use their capital intensity index: the average capital intensity of firms in the respective industry. We use their methodology to calculate the Herfindal index of intermediate goods usage at the 6-digit level NAICS codes. This latter index takes into account not only the number of intermediates needed, but also the share of each intermediate used in the production. We re-calculated these indices from the original IO use tables and translated them from SIC industry codes into NACE 4 digit codes using concordances from the UN57 . We thank Andrei Levchenko for providing us with data and codes for calculating these indices. The regressions use the indices at the 4 digit level, while Table 11 reports the averages for the 2 digit level industries. The indices constructed by Cowan and Neut (2007) indicate how reliant on contracts is the production of goods in a sector. For our purposes here we are interested in whether the industry supplies to industries that are contractually dependent. Contractually dependent industries care more about product quality and features and are more likely to have specific product requirements. Therefore, for industries that supply more to more contractually dependent industries, ISO certification is presumably more relevant, and P further, the role of certification might be different. The precise formula for our supplier index is: Si = j fij · Herfj , where i indexes the respective industry and j indexes all the industries that industry i supplies to, and fij is the fraction of industry i’s output that is supplied to industry j. We also construct from the US IO Table the share of an industry’s output that goes directly to final consumption. Lastly we gathered data to calculate the US certification rate among exporters. We thank Ann Terlaak for providing us with data on the US firms that are ISO 9001 certified as of 2002. The data source for US ISO 9000 certified companies is the McGraw-Hill Directory. We obtained the total number of establishment in the US and the total number of exporting establishments by industry from the 1992 Census of manufacturers. We calculated the fraction of exporting establishments. We obtained the number of firms for 2002 from the 2002 Census of Businesses, and applies the fraction of exporting establishments to obtain an approximate of the number of exporting firms in 2002. We then calculated the fraction of ISO firms among exporters by dividing the number of ISO firms to the number of exporting firms in 2002. This number is a good proxy for the US ISO rate as long as the fraction of exporting establishments is a good proxy for the fraction of exporting firms by industry.

Appendix C

Graphing the model

Figure 1 represents three curves in the (z,  = f1 ) space. The first curve is labeled as ’Melitz model exporting frontier’ and graphs equation (3.2), which can be rewritten as: z a · b = k1 where a, b and k1 are constants given by the parameters of the model. This equation represents a standard hyperbola. The second curve is labeled as ’Contractual model exporting frontier’ and graphs equation (3.3), which can be rewritten as: z a · (c1 · b − c0 ) = k2 57

http://unstats.un.org/unsd/cr/registry/regct.asp?Lg=1

37

where, again, a, b, c1 , c0 and k2 are constants given by the parameters of the model. This equation represents a hyperbola with a vertical asymptote at 0 and a horizontal asymptote at cc10 , which is in fact equal to ∗ . The third curve is labeled as ’ISO frontier’ and graphs equation (3.5), which can be rewritten as: z a · (1 − b ) = k3 where a, b and k3 are constants given by the parameters of the model. This equation, as opposed to the two pervious ones that described an inverse relationship between z and , represents a positive relationship between z and . In addition, this hyperbola has a horizontal hyperbola at  = 1.

Appendix D

The extensive margin from ISO certification

In addition to exporting firms obtaining certification, we might have domestic firms, that previously were unable to export, that decide to become exporters with certification. Firms in this situation are described by πx (λx ) < πd (λd ) < πxc (λxc ). The first part of this sequence of inequalities is described by the ’Contractual model exporting frontier’: firms below this curve have πx (λx ) < πd (λd ) and remain domestic firms in the absence of certification. If in addition these firms satisfy πd (λd ) < πxc (λxc ), they will become exporters with certification. This last condition is equivalent to: ( θ−1 θ )

α·θ α0

· ( wz )

α·(θ−1) α0

· ( αz )

θ−1 α0

∗

α

α

1 E α0 α0 − ( · ( θ−1 − α) · (τ −θ · ( E P∗ ) P ) ) > fx + fc

This condition is satisfied for firms with productivity above a certain threshold z > z ∗ if the foreign markets are large enough. To summarize, high productivity firms with low contractual ability that were previously unable to become exporters, can now access foreign markets by obtaining certification.

Appendix E

Alternative cost of certification functional forms

In section 3.3 we modeled the cost of certification in the simplest form, as a fixed fee. However, the model’s properties are robust to other specifications that might be more realistic. [1] The cost of certification can take the form fzc , decreasing with higher firm productivity. In this case the certification condition given by equation (3.5) remains the same except the coefficient on z changes. Therefore, the qualitative implications remain. [2] The cost of certification can take the form ffc , decreasing with higher ’contractual ability’. In this case the certification condition given by equation (3.5) remains the same except the coefficient on f changes. Therefore, the qualitative implications remain. [3] The cost of certification can be proportional with firm size, therefore becoming part of marginal costs. In this case the ’ISO frontier’ curve asymptotes to 1 − k · fc (instead of 1), where k is a constant given by the parameters of the model. In this case, firms with contractual ability above a certain level do not get ISO certification.

38

39

Notes: The table shows average annual numbers over the period 1995 to 2005.

Table 1: General description of the Slovenian Manufacturing Industries, 1995-2005 Industry (2-digit NACE rev. 1) Number Number of Number of of firms exporters ISO-exporters Manufacture of food products and beverages 398 103 40 Manufacture of textiles and tobacco 236 105 28 Manufacture of wearing apparel, dressing and dying of fur 332 91 9 Leather tanning, dressing, luggage, saddlery, shoes etc. manufacture 86 39 6 Manufacture of wood, cork, straw, plaiting materials products 507 208 18 Manufacture of pulp, paper and paper products 96 45 19 Publishing, printing and reproduction of recorded media 861 192 24 Manufacture of chemicals and chemical products; manufacture of coke, re- 165 89 46 fined petroleum products and nuclear fuel Manufacture of rubber and plastic products 454 209 66 Manufacture of other non-metallic mineral products 218 86 27 Manufacture of basic metals 81 51 27 Manufacture of fabricated metal products, except machinery and equipment 1169 445 114 Manufacture of machinery and equipment not elsewhere classified 481 272 86 Manufacture of office machinery and computers 99 27 9 Manufacture of electrical machinery and apparatus nec 308 137 52 Manufacture of radio, television, communication equipment and apparatus 141 66 27 Manufacture of medical, precision, optical instruments, watches and clocks 220 104 37 Manufacture of motor vehicles, trailers, semi-trailers 81 51 20 Manufacture of other transport equipment 48 24 3 Manufacture of furniture, manufacturing nec. 495 176 28 Total 6476 2520 686

Table 2: Summary Statistics of the Main Variables Used Variable

Mean

Median

Employment Total sales Domestic sales Export sales Export shares (%) Capital intensity Labor productivity Total factor productivity Price-cost margin Average wages

1.857 5.252 4.955 4.881 37.607 2.310 2.633 0.911 -1.160 1.719

1.386 5.201 4.977 4.896 27.818 2.497 2.674 0.918 -1.145 1.786

Standard Deviation 1.722 2.302 2.163 2.821 33.905 1.571 0.832 0.648 0.931 0.652

Notes: All variables except export shares are in logs of real values in constant 2005 Euros. The statistics for export shares and export sales are calculated conditional on firms exporting. Export shares are the fraction of export sales in total sales and are reported as percentages. Capital intensity is the capital to labor ratio. Labor productivity is the ratio of gross value added to labor. Total factor productivity is calculated by the Levinsohn-Petrin method. The price-cost margin is proxied by the ratio of gross value added minus labor costs to gross value added plus the cost of materials. Average wages are total wages divided by the total number of workers. All the summary statistics are reported as averages over the entire panel period 1995 to 2005.

40

Table 3: Statistics of Slovenian Manufacturing firms by firm size, 1995-2005 (a)

Number of firms

Number of firms Number of exporters Number of exporters with ISO Fraction of exporters Fraction of exporters with ISO (b)

Micro 3829 2082 78

Small 1250 1024 177

Medium 698 651 240

Large 213 211 123

0.544 0.037

0.819 0.173

0.933 0.369

0.991 0.583

Summary statistics of the main variables of interest

Employment Total sales Domestic sales Export sales Export shares (%) Capital intensity Labor productivity Total factor productivity Price-cost margin Average wages

Micro 3.212 4.903 4.729 3.202 0.279 2.291 2.671 0.826

Small 22.0 6.830 6.363 5.034 0.339 2.548 2.814 1.012

Medium 113.8 8.389 7.562 7.125 0.467 2.725 2.712 1.147

Large 643.5 10.182 9.055 9.283 0.578 3.127 2.867 1.449

-1.251 1.682

-1.364 1.922

-1.537 1.944

-1.576 2.004

Notes: All variables except export shares and employment are in logs of real values in constant 2005 Euros. Export shares are reported as percentages. All the summary statistics are reported as averages over the entire panel period 1995 to 2005. Export sales and export sales summary statistics are calculated conditional on exporting. Employment is the total number of workers employed and is reported in levels. Firm size is according to the definition provided in World Development Indicators (2007): micro firms have less than 9 workers, small firms have between 10 and 49 workers, medium firms have between 50 and 249 workers and large firms have more than 250 workers.

41

42

(b)

Growth

54.4*** 33.2*** 5.5*** -3.2*** 16.6***

7.1 0.8 1.1 -1.1 -0.9

2.7*** -0.4 0.8 1.3 -0.8

Non-ISO exporters vs. Domestic firms Simple difference Premia 3.7 1.5** 3.0 1.0

53.1 34.5 13.3 -23.7 23.4

ISO exporters vs. Non-ISO exporters Simple Difference Premia 6.2 5.9*** 3.7 4.3*** 6.7 7.6*** 2.1 3.6** 3.2 1.3 2.2 1.4* 1.7 1.2* 0.1 1.5 -1.6 0.5

ISO exporters vs. Non-ISO exporters Simple Difference Premia 277.9 72.1*** 232.1 67.6*** 299.2 93.4*** 19.9 3.3*** 70.1 60.9*** 29.7 33.2*** 34.5 14.6*** -25.2 5.9*** 32.0 12.9***

Notes: All numbers represent percentages. Simple differences are the difference between the mean log X for Non-ISO exporters or ISO exporters respectively and the mean log X for domestic firms: 100 · (M ean lnXISO exporters − M ean lnXdomestic f irms ) and 100 · (M ean lnXnon−ISO exporters − M ean lnXdomestic f irms ), where X is any of the variables on the left-most column. This summary statistic follows Bernard, Eaton, Jensen, Kortum (2003). The ’Premia’ column reports coefficients on ISO exporters and non-ISO exporters from regressions of the form: lnXijkt = β0 + β1 · non ISOexporteri + β2 · ISO exporteri + θ · sizei + κk + ζj + ξt + it . Premia are calculated from the formula 100 · (expβ − 1). * significant at 10%; ** significant at 5%; *** significant at 1%. Robust standard errors are computed. The number of observations range between 32,000 and 17,000.

Total sales Domestic sales Export sales Export shares Capital intensity Labor productivity Total factor productivity Price-cost margin Average wages

Total sales Domestic sales Export sales Export shares Capital Intensity Labor Productivity Total Factor Productivity Price-cost margin Average wages

Levels

Non-ISO exporters vs. Domestic firms Simple difference Premia 184.6 131.4*** 147.6 74.5***

(a)

Table 4: Contemporaneous advantages

Table 5: Are ISO firms better before they obtain certification? (a)

Total sales Domestic sales Export sales Export shares Capital Intensity Labor Productivity Total Factor Productivity Price-cost margin Average wages

Premia at T ISO − 1 13.0*** 1.5 59.8*** 5.8*** 28.1*** 2.6* 7.1*** -0.8 -0.4 (b)

Total sales Domestic sales Export sales Export shares Capital Intensity Labor Productivity Total Factor Productivity Price-cost margin Average wages

Levels Premia (%) Premia at T ISO − 2 12.0*** -2.0 37.9*** 4.7*** 29.3*** 3.7** 8.0*** -1.9 -1.4

Premia at T ISO − 3 8.0** 2.0 33.9*** 4.6*** 0*** 2.6 7.5*** 0 -0.8

Growth Premia (%)

Premia at T ISO − 1 5.1* 6.7* 16.6** 5.3 -4.0 1.1 1.1 0.0 0.9

Premia at T ISO − 2 5.5*** 3.0 8.2** 1.3 0.0 3.0** 0.8 0.3 1.6*

Notes: Premia are calculated based on coefficients from regressions of the form: lnXijk,t−n = β0 + β1 · ISO startit + β2 · lnZi,t−n + θ · sizei + κk + ζj + τt + ξit , where region, sector, year and size controls are included, as well as firm characteristics at time t − k, with k = 1, 2, or 3. ISO startit is equal to 1 only once for ISO certified exporters, in the first year they become certified. All numbers represent percentages. Premia calculated as percentages based on formula 100 · (expβ1 − 1), except the premia for export shares where we report the actual coefficient times 100. * significant at 10%; ** significant at 5%; *** significant at 1%. Robust standard errors are computed. Number of observations range between 22,000 and 16,000, depending on the variable and time lags (k) used.

43

Table 6: Age at certification Export Age 1 2 3 4 5 6 7 8 9 10

Probability continuing exporting, conditional on export age 50.1 64.3 76.1 79.2 83.1 87.2 88.4 93.8 95.8 97.4

Fraction of ISO starters conditional on export age 15.8 9.3 11.2 14.4 6.5 7.0 7.4 2.8 3.3 2.8

Notes: The export age is the number of periods a firm exported before and including the current period. We excluded exporting firms in 1995 for which we do not know the exporting status in the previous year. The probability of survival represents the fraction of exporters at that respective export age that continue to export an additional period. The fraction of ISO starters conditional on export age is the fraction of exporters that become certified at that specific export age over the total number of ISO-certified exporters.

44

Table 7: Post-ISO Premia (a)

Total sales Domestic sales Export sales Export shares Capital Intensity Labor Productivity Total Factor Productivity Price-cost margin Average wages

Premia at T ISO + 1 10.1*** 10.2** 79.8*** 4.2*** 4.7** 8.2*** 11.5*** 2.2 2.7*** (b)

Total sales Domestic sales Export sales Export shares Capital Intensity Labor Productivity Total Factor Productivity Price-cost margin Average wages

Levels Premia (%) Premia at T ISO + 2 19.8*** 17.1*** 62.4*** 4.6*** 2.8 9.1*** 14.1*** 1.6 2.7***

Premia at T ISO + 3 24.1*** 19.6*** 74.0*** 5.6*** 5.2 12.7*** 16.4*** 4.2 4.1***

Growth Premia (%)

Premia at T ISO + 1 10.1*** 9.0*** 11.1*** -0.1 4.7** 8.2*** 4.9*** 2.2 2.7***

Premia at T ISO + 2 9.5*** 7.2*** 5.5* 0.2 1.4 4.4*** 3.1*** 0.8 1.4***

premia at T ISO + 3 7.4*** 5.2*** 2.9 0.3 1.7 4.0*** 2.8*** 1.4 1.4***

Notes: All numbers represent percentages. Premia are calculated based on coefficients from regressions of the form: lnXijk,t+n = β0 + β1 · ISO startit + β2 · lnZi,t + θ · sizei + κk + ζj + ξt , where region, sector, year and size controls are included as well as the following firm characteristics lnZit at time t: total sales, capital intensity, labor productivity, price-cost margins and average wages. ISO startit is equal to 1 only once for ISO certified exporters, in the first year they become certified. Premia are calculated based on formula 100 · (expβ1 − 1), except the premia for export shares where we report the actual coefficient times 100. * significant at 10%; ** significant at 5%; *** significant at 1%. Robust standard errors are computed. Number of observations are between 22,000 and 16,000 depending on the respective independent variable and the time lags used.

45

46 X

Basic OLS lnX 0.114*** [0.018] -0.024* [0.013] 0.406*** [0.053] 0.067*** [0.007] -0.027** [0.014] 0.010 [0.010] 0.008 [0.007] -0.009 [0.012] 0.007 [0.005] X X X X X X X X

Basic FE lnX 0.149*** [0.016] 0.049** [0.019] 0.169*** [0.045] 0.026*** [0.005] 0.008 [0.023] 0.063*** [0.015] 0.024* [0.013] 0.010 [0.019] 0.010 [0.09]

X

X

Basic w/ Trend OLS lnX 0.069*** [0.019] -0.024* [0.014] 0.352*** [0.052] 0.041*** [0.008] -0.053*** [0.015] -0.015 [0.010] -0.002 [0.008] -0.021 [0.013] -0.007 [0.005] X X X X X

X

X X X

Basic w/ Trend FE lnX 0.090*** [0.015] 0.050*** [0.019] 0.070 [0.043] 0.025*** [0.005] -0.057** [0.023] 0.049*** [0.015] 0.028** [0.013] 0.006 [0.019] 0.008 [0.008]

X X

Differenced FE ∆lnX 0.121*** [0.023] 0.116*** [0.026] 0.120** [0.047] 0.007 [0.006] 0.050* [0.026] 0.006 [0.016] -0.001 [0.015] -0.018 [0.019] 0.046*** [0.008]

X

X X

Differenced FE w/ lags ∆lnX 0.046** [0.021] 0.026 [0.026] 0.071 [0.053] 0.013** [0.006] 0.053** [0.025] 0.009 [0.016] 0.015 [0.015] -0.009 [0.019] 0.016** [0.008]

Notes: * significant at 10%; ** significant at 5%; *** significant at 1%. Standard errors are reported in brackets. Robust standard errors are computed. Number of observations is between 32,000 and 26,000, and the number of firms is between 4000 and 3650, depending on the empirical model and the dependent variable used.

ISO dummy Sector FE Region FE Size FE Year FE Firm FE Lag firm controls Lag ∆ firm controls Pre-ISO Growth trend in X

Average wages

Price-cost margins

Total Factor Productivity

Labor Productivity

Capital Intensity

Export Shares

Export Sales

Domestic Sales

Dependent Variable Total Sales

Model

Table 8: Differences in Differences estimation of the effect of ISO certification

Table 9: Post-ISO Premia by firm size Premia at T+1 Total sales Domestic sales Export sales Export shares Capital Intensity Labor Productivity Total Factor Productivity Price-cost margin Average wages

28.4*** 50.3*** 44.1 -3.6 4.7 11.7** 9.5* 5.0 7.7**

Total sales Domestic sales Export sales Export shares Capital Intensity Labor Productivity Total Factor Productivity Price-cost margin Average wages

11.0*** -1.1 86.8*** 5.3* 11.6** 10.1*** 9.4*** 4.9 3.0**

Total sales Domestic sales Export sales Export shares Capital Intensity Labor Productivity Total Factor Productivity Price-cost margin Average wages

5.1*** 7.5 97.9*** 6*** 3.4 7.6*** 11.9*** 0.1 2.5*

Total sales Domestic sales Export sales Export shares Capital Intensity Labor Productivity Total Factor Productivity Price-cost margin Average wages

3.8* 11.2 98.5*** 4.4 0.3 4.4* 16.5*** 2.1 0.1

Premia at T+2 Micro Firms 65.2*** 76.6** 73.4* -3.5 -12.2 -1.4 0.8 -19.8 8.0* Small Firms 21.2*** 8.9 47.4* 4.4 22.2*** 13.2*** 11.6*** 3.8 4.7** Medium Firms 10.4*** 11.1 61.2*** 5.2** -1.1 9.9*** 15.9*** 3.4 2.0* Large Firms 7.5* 13.7 51.1*** 4.8 0.4 6.8 18.6*** 4.6 -0.2

Premia at T+3 70.5*** 102.1*** 76.2 -4.5 -3.4 18.8 11.4 -14.7 15.0** 33.6*** 15.2 44.4 5.1 16.7 20.3*** 16.5*** 13.0** 7.0*** 14.2*** 12.7 81.3*** 6.7*** 0.1 9.4*** 13.6*** -2.0 3.0** -2.6 8.1 53.2*** 4.4 15.3* 9.5* 18.8*** 14.6** 1.0

Notes: All numbers represent percentages. Firm size definitions are from World Development Indicators (2007): micro firms have less than 10 workers, small firms have between 10 and 49 workers, medium firms have between 50 and 249 workers and large firms have more than 250 workers. * significant at 10%; ** significant at 5%; *** significant at 1%. Robust standard errors are computed. Number of observations is around 22,000.

47

48

Small Firms Trend, FE Differenced, FE w/ lags 0.095*** 0.048 [0.026] [0.037] 0.065** -0.008 [0.032] [0.054] 0.123 -0.014 [0.076] [0.106] 0.031*** 0.016 [0.009] [0.010] 0.022 0.012 [0.039] [0.047] 0.018 0.011 [0.025] [0.028] 0.02 0.005 [0.022] [0.026] 0.044 -0.027 [0.032] [0.040] 0.023 0.008 [0.015] [0.014]

Medium Firms Trend, FE Differenced, FE w/ lags 0.046** 0.022 [0.023] [0.028] -0.006 0.039 [0.028] [0.031] -0.041 0.058 [0.062] [0.072] 0.022*** 0.005 [0.008] [0.007] -0.158*** 0.038 [0.034] [0.030] 0.069*** 0.018 [0.022] [0.024] 0.056*** 0.027 [0.019] [0.024] 0.009 0.004 [0.028] [0.022] -0.01 0.024*** [0.013] [0.009]

Large Firms Trend, FE Differenced, FE w/ lags 0.113*** 0.074 [0.038] [0.067] 0.116** 0.027 [0.047] [0.071] 0.036 0.227* [0.099] [0.137] 0 0.021* [0.012] [0.011] 0.037 0.032 [0.056] [0.069] 0.112*** -0.026 [0.037] [0.038] 0 0.036 [0.032] [0.038] -0.03 -0.014 [0.047] [0.044] 0.01 -0.004 [0.021] [0.019]

Notes: Firm size definitions are from World Development Indicators (2007): micro firms have less than 10 workers, small firms have between 10 and 49 workers, medium firms have between 50 and 249 workers and large firms have more than 250 workers. Coefficients in the Trends, FE model are statistically different among the 4 firm sizes for all variables except for price-cost margins and average wages, according to F-tests. Coefficients in the bf Differenced, FE w/ lags model are not statistically different from eachother for different firm sizes. * significant at 10%; ** significant at 5%; *** significant at 1%. Number of observations is around 22,000.

Average wages

Price-cost margin

Total Factor Productivity

Labor Productivity

Capital Intensity

Export shares

Export sales

Domestic sales

Total sales

Model

Micro Firms Trend, FE Differenced, FE w/ lags 0.216*** 0.107 [0.046] [0.074] 0.137** 0.082 [0.056] [0.089] 0.587*** 0.154 [0.145] [0.182] 0.053*** 0.026 [0.015] [0.029] -0.036 0.266** [0.068] [0.108] -0.033 0.015 [0.044] [0.057] -0.017 -0.034 [0.038] [0.034] -0.074 0.004 [0.057] [0.065] 0.029 0.04 [0.025] [0.037]

Table 10: Differences in Differences estimation by firm size

49

Herf Index 0.157 0.152 0.136 0.165 0.217 0.260 0.113 0.125 0.115 0.074 0.104 0.112 0.062 0.096 0.090 0.131 0.062 0.096 0.093 0.099 -0.123

K:L Ratio 1.050 0.902 0.698 0.847 0.898 1.091 0.833 1.190 0.959 1.127 1.175 0.956 0.963 0.931 0.961 0.942 0.912 0.978 1.000 0.882 0.762***

0.112 0.081 0.079 0.113 0.084 0.098 0.095 0.111 0.095 0.105 0.082 0.100 -0.220

Supplier Index 0.135 0.138 0.133 0.164 0.138 0.159 0.124 0.130 0.241 0.329 0.003 0.139 0.075 0.074 0.069 0.140 0.140 0.169 0.295 0.504 -0.476**

Final Cons. 0.679 0.337 1.220 1.471 0.113 0.079 0.310 0.242 7.3 5.5 8.4 7.5 11.3 12.0 17.6 19.2 18.0 14.4 17.7 2.9 0.377*

US ISO Rate 0.9 5.5 1.5 4.9 0.7 0.2 2.5 12.7

45.9 39.5 63.0 38.0 56.5 27.1 44.4 46.7 47.2 62.9 49.8 35.6 0.495**

SL. Export Rate 25.8 44.5 27.4 45.2 41.1 46.6 22.3 53.9

Notes: Industries are based on 2-digit Nace Rev. 1. Indices are explained in Section 4.4 as well as in Appendix B. All indices have been calculated at the 4-digit level. For presentation purposes, this table presents averages of each respective index at the 2-digit level. All regressions use indices at the 4-digit level. ISO rates are calculated as the ratio of ISO firms to exporting firms. The Slovenian Export rate is the fraction of exporting firms to the total number of firms in each industry. *** significant at 1%, ** significant at 5%, * significant at 10%.

Manufacture of food products and beverages Manufacture of textiles and tobacco Manufacture of wearing apparel, dressing and dying of fur Leather tanning, dressing, luggage, saddlery, shoes etc. manufacture Manufacture of wood, cork, straw, and plaiting materials products Manufacture of pulp, paper and paper products Publishing, printing and reproduction of recorded media Manufacture of chemicals and chemical products; manufacture of coke, refined petroleum products and nuclear fuel Manufacture of rubber and plastic products Manufacture of other non-metallic mineral products Manufacture of basic metals Manufacture of fabricated metal products, except machinery and equipment Manufacture of machinery and equipment n.e.c. Manufacture of office machinery and computers Manufacture of electrical machinery and apparatus nec Manufacture of radio, television, communication equipment and apparatus Manufacture of medical, precision, optical instruments, watches and clocks Manufacture of motor vehicles, trailers, semi-trailers Manufacture of other transport equipment Manufacture of furniture, manufacturing nec Correlation with Slovenian ISO rate

Industry

Table 11: Industry Contractual Indices

Table 12: Range of Post-ISO premia, by sector contract intensity Premia at T ISO + 1 Min. Max. Total sales Domestic sales Export sales Export shares Capital Intensity Labor Productivity Total Factor Productivity Price-cost margin Average wages

11.1 0.2 79.2 5.5 11.2 4.0 -45.9*** 3.7 1.7

Total sales Domestic sales Export sales Export shares Capital Intensity Labor Productivity Total Factor Productivity Price-cost margin Average wages

14.2 96.9*** -4.0*** -13.2*** 8.7 6.3 17.9 -16.9*** 11.9***

Total sales Domestic sales Export sales Export shares Capital Intensity Labor Productivity Total Factor Productivity Price-cost margin Average wages

13.4 63.3*** 43.5 -7.3*** 10.1 6.3 25.9 -12.5*** 8.6

Total sales 9.3 Domestic sales 268.3*** Export sales -16.5 Export shares -31.6*** Capital Intensity 19.9 Labor Productivity 10.7 Total Factor Productivity -9.8* Price-cost margin -6.7 Average wages 6.8 Continued on Next Page. . .

Premia at T ISO + 2 Premia at T ISO + 3 Min. Max. Min. Max. Capital Intensity Index 9.6 25.4 14.9 21.9** 22.8** 8.8 9.2 12.9 11.5 14.8 115.3 48.3 88.1 32.6 124.0 6.7 5.9 6.3 5.6 8.2 0.3 14.7 -3.8 17.2 -1.0 13.8 1.9* 18.6* 5.1 23.9 88.0*** -47.4*** 101.5*** -48.6*** 111.4*** 1.2 -4.9 7.4 -5.8 12.7 6.6 4.6 5.2 7.5 7.0 Herfindal Index 9.1 23.8 17.9 53.2* 13.3* -11.9*** 103.2*** -6.1*** 191.0*** -15.9*** 166.7*** -21.2*** 124.0*** -25.4*** 148.1*** 12.8*** -12.7*** 12.5*** -15.9*** 15.6*** 4.9 17.4 1.4 13.5 6.9 9.9 6.6 12.1 15.2 15.0 12.9 19.5 17.5 27.8 16.5 6.8*** -19.4*** 7.4*** -14.2 8.7 2.0*** 11.1*** 2.7*** 11.9 5.3 Supplier Index 8.4 22.9 16.9 46.2* 10.2* -13.4*** 68.3*** -7.3*** 131.7*** -21.3*** 138.4 17.8 101.3 24.0 114.0 12.8*** -6.9*** 12.5*** -8.7*** 15.6*** 2.9 21.3* -3.6* 22.6 0.2 10.9 9.4 11.6 17.1 14.0 7.6 29.3 11.4 31.7 12.1 9.4*** -15.2*** 10.8*** -12.3** 13.0** 2.2 10.6** 1.6** 12.3* 4.0* Final Consumption Index 9.8 17.0 18.0 41.4 17.3 -13.3*** 290.1*** -8.3*** 393.4*** -10.7*** 130.4 -31.8* 93.8* -30.8* 104.5* 13.5*** -33.3*** 13.7*** -34.5*** 15.2*** 2.0 46.7** -3.2** 17.9 2.8 9.0 25.3* 8.3* 20.6 13.8 18.9* -4.2 22.1 -15.4*** 26.5*** 3.9 -1.3 2.5 -6.6 6.5 3.5 13.2** 2.7** 7.3 5.7

50

Table 12 – Continued Total sales Domestic sales Export sales Export shares Capital Intensity Labor Productivity Total Factor Productivity Price-cost margin Average wages

9.3 29.4** 81.0 -2.2*** 3.2 7.7 7.8 -1.6 4.0

11.4 -8.0** 77.6 11.6*** 7.2 9.7 16.3 6.1 2.1

US ISO certification rate 21.8 17.3 42.7*** -6.0*** 54.3 69.7 -2.2*** 13.8*** 5.1 -0.6 13.8* 3.5* 13.3 15.5 -0.9 4.8 4.9 1.0

28.4 51.3*** 69.3 -3.1*** 2.5 16.2 16.0 1.4 5.3

19.0 -9.3*** 79.2 17.1*** 8.6 9.7 18.3 7.4 3.3

Notes: All numbers represent percentages. For each index, the ’min.’ column shows the premia in the least contractually intensive industry and the ’max.’ column shows the premia in the most contractually intensive industry. * significant at 10%; ** significant at 5%; *** significant at 1%. Robust standard errors are computed. Number of observations are around 22,000.

Table 13: Effects of ISO certification, by sector contract intensity Trends, FE ISO ISO · Index Total sales Domestic sales Export sales Export shares Capital Intensity Labor Productivity Total Factor Productivity Price-cost margin Average wages

Differenced, FE w/ lags ISO ISO · Index

Capital Intensity Index -0.215 0.303 [0.161] [0.249] -0.361* 0.222 [0.199] [0.280] -1.328*** 0.254 [0.442] [0.634] -0.111** 0.019 [0.053] [0.052] 0.096 0.037 [0.240] [0.224] 0.149 -0.184 [0.157] [0.209] -0.067 -0.401* [0.135] [0.219] -0.237 -0.044 [0.199] [0.208] 0.105 0.000 [0.089] [0.082]

0.301* [0.159] 0.404** [0.196] 1.379*** [0.438] 0.134** [0.052] -0.152 [0.237] -0.098 [0.155] 0.094 [0.133] 0.238 [0.196] -0.095 [0.088]

Continued on Next Page. . .

51

-0.262 [0.255] -0.200 [0.282] -0.186 [0.650] -0.006 [0.052] 0.017 [0.225] 0.197 [0.215] 0.425* [0.227] 0.036 [0.212] 0.016 [0.083]

Table 13 – Continued

Total sales Domestic sales Export sales Export shares Capital Intensity Labor Productivity Total Factor Productivity Price-cost margin Average wages

Total sales Domestic sales Export sales Export shares Capital Intensity Labor Productivity Total Factor Productivity Price-cost margin Average wages

Trends, FE Differenced, FE w/ lags ISO ISO · Index ISO ISO · Index Herfindahl Index 0.162*** -0.634*** 0.079* -0.285 [0.030] [0.232] [0.044] [0.356] 0.065* -0.486* 0.055 -0.251 [0.038] [0.288] [0.052] [0.421] 0.275*** -1.346** 0.296*** -1.993** [0.088] [0.672] [0.113] [0.968] 0.026*** -0.055 0.025** -0.108 [0.010] [0.076] [0.012] [0.078] -0.017 0.137 0.029 0.214 [0.046] [0.354] [0.048] [0.332] 0.081*** -0.244 0.001 0.067 [0.030] [0.229] [0.034] [0.304] 0.048* -0.288 0.012 0.024 [0.026] [0.196] [0.032] [0.305] 0.047 -0.316 -0.024 0.134 [0.038] [0.291] [0.047] [0.400] -0.002 0.073 0.011 0.045 [0.017] [0.130] [0.015] [0.100] Supplier Index 0.256*** -1.509*** 0.082 -0.321 [0.050] [0.434] [0.080] [0.701] 0.158** -1.342** 0.074 -0.434 [0.062] [0.538] [0.094] [0.847] 0.438*** -2.857** 0.454** -3.455* [0.143] [1.230] [0.196] [1.769] 0.038** -0.163 0.024 -0.097 [0.016] [0.142] [0.015] [0.116] -0.023 0.199 0.012 0.374 [0.077] [0.661] [0.075] [0.617] 0.086* -0.294 -0.075 0.752 [0.050] [0.429] [0.057] [0.530] 0.068 -0.474 -0.066 0.729 [0.042] [0.366] [0.061] [0.582] 0.05 -0.359 -0.098 0.807 [0.063] [0.544] [0.083] [0.759] 0.003 0.024 -0.001 0.159 [0.028] [0.243] [0.022] [0.182]

Continued on Next Page. . .

52

Table 13 – Continued

Total sales Domestic sales Export sales Export shares Capital Intensity Labor Productivity Total Factor Productivity Price-cost margin Average wages

Total sales Domestic sales Export sales Export shares Capital Intensity Labor Productivity Total Factor Productivity Price-cost margin Average wages

Trends, FE Differenced, FE w/ lags ISO ISO · Index ISO ISO · Index Final Consumption Share 0.125*** -0.198*** 0.054** -0.043 [0.019] [0.061] [0.026] [0.101] 0.014 -0.024 0.021 0.028 [0.023] [0.075] [0.031] [0.102] 0.181*** -0.335* 0.089 -0.103 [0.053] [0.171] [0.063] [0.209] 0.029*** -0.055*** 0.017** -0.024 [0.006] [0.020] [0.007] [0.016] 0.001 -0.015 0.031 0.124 [0.028] [0.093] [0.032] [0.087] 0.083*** -0.170*** 0.016 -0.038 [0.018] [0.060] [0.021] [0.068] 0.040** -0.137*** 0.02 -0.032 [0.016] [0.051] [0.020] [0.066] 0.040* -0.167** -0.027 0.105 [0.024] [0.081] [0.024] [0.074] 0.016 -0.058* 0.016* 0 [0.010] [0.034] [0.009] [0.026] US ISO rate 0.004 0.009*** 0.035 0.054 [0.029] [0.003] [0.031] [0.104] -0.021 0.008** 0.007 0.09 [0.035] [0.003] [0.036] [0.116] -0.154* 0.024*** 0.072 -0.006 [0.081] [0.007] [0.072] [0.232] 0.006 0.002** 0.014* -0.003 [0.009] [0.001] [0.008] [0.018] -0.042 -0.002 0.055 -0.005 [0.043] [0.004] [0.036] [0.082] 0.032 0.002 -0.005 0.065 [0.028] [0.003] [0.022] [0.056] 0.018 0.001 -0.009 0.112** [0.024] [0.002] [0.019] [0.049] -0.038 0.005 -0.034 0.119 [0.035] [0.003] [0.028] [0.085] -0.007 0.002 0.017* -0.005 [0.016] [0.001] [0.011] [0.032]

Notes: * significant at 10%; ** significant at 5%; *** significant at 1%. Number of observations is between 29,000 and 16,000.

53

Figure 1: Exporting and ISO certification frontiers

54

Figure 2: Advantages of ISO exporters over non-ISO exporters

(a) Total sales advantages

(b) Domestic sales advantages

(c) Export sales advantages

(d) Export shares advantages

(e) Capital intensity advantages

(f) Labor productivity advantages

(g) TFP advantages

(h) Price-cost margins advantages

55

(i) Average wages advantages

Figure 3: Changes in the distributions of ISO exporters, before and after ISO certification

(a) Total sales Changes

(b) Domestic sales Changes

(c) Export sales Changes

(d) Export shares Changes

(e) Number of destination markets Changes

(f) Popularity of destination markets Changes

(g) Capital intensity Changes

(h) Labor Productivity Changes

(i) Total Factor Productivity Changes

(j) Price-cost Margins Changes

(k) Average wages Changes

56