Heterogeneous firms and homogenising standards in agri-food trade: the Polish meat case

1. Introduction

Technical regulations and standards¹ have gained great importance in international agri-food trade. Governments, particularly of high-income countries, have increasingly implemented tighter and mandatory standards for agri-food products and also demand that agri-food imports comply with them. Since food standards differ across countries, they can potentially restrict market access of exporters. In the context of free trade agreements, the facilitation of trade through dovetailing regulatory frameworks is increasingly important, next to the reduction of traditional trade policy measures such as tariffs and quota. While streamlining, standards can be expected to facilitate bilateral trade, changing from the existing practices to new production processes or new product specifications also involves costs. Producers who wish to participate in and benefit from the deeper integration of markets have to pay an ‘entrance fee’ in the form of investments in compliance with new standards.

In the agri-food sector, the European Union (EU) established common standards that harmonise the diverging standards requirements across the member states to ensure a certain level of food safety and the functioning of the EU single market for agri-food products. EU membership requires the implementation of the EU food standards, and in this regard the most recent wave of eastward enlargement constitutes an interesting case of standards and market integration. In the new member states of Central and Eastern Europe, adopting the EU food standards involved considerable efforts on the side of governments and producers alike. Swinnen and Vandeplas (2007) summarise the changes in agri-food chains in the transition process and how they have affected firms, the organisation of production and trade. While EU membership opened trade opportunities, particularly the small and medium size agri-food firms struggled to meet the more stringent requirements, and the capacity to comply is still a barrier to access the single market.

Some studies on EU enlargement capture the issue of deep integration and trade facilitation through harmonisation in simulation models (Lejour et al., 2001; Michalek et al., 2005). Smith and Venables (1988) first introduced the now standard modelling approach of ‘iceberg tariffs’ in the context of the EU single market. In this approach, non-tariff measures (NTMs) such as standards are regarded as variable trade costs that use resources of exporting firms, and different methods are applied to obtain estimates of respective tariff-equivalents. They range from the direct measurement of compliance costs or careful price comparisons along a given supply chain to cross-country econometric estimations. Ferrantino (2005) provides an overview of recent studies. However, price wedges and ‘iceberg tariffs’ cannot fully represent standards, in particular food standards, as they explicitly lead to fixed (through additional investments) and variable (through additional activities) compliance costs for exporting firms. Standards are thus more than just border measures, and their fixed and variable compliance costs constitute an integral part of the export decision of firms.

In the agri-food sector, like in other sectors, firm size is not uniform but typically shows a skewed pattern, with some large firms and many small firms, and different types of costs will weigh differently across firms. Larger and possibly more productive firms find it easier to cover a given amount of fixed costs than their smaller competitors. Given the fixed costs of exporting, foreign sales therefore tend to be more concentrated on the larger firms. In addition, the empirical trade literature points out that exporting firms typically only export a fraction of their produce; for a review of empirical studies on the characteristics of exporting firms, see, for example, Eaton et al. (2004). These empirical findings have led to advances in trade theory that account for the presence of firm heterogeneity. Heterogeneous firm models such as those developed by Melitz (2003) and Bernard et al. (2003) are particularly useful for analysing standards since different types of compliance costs can be incorporated, while accounting for differences in size and productivity of firms.

The goal of this paper is to explore the market and trade effects of food standards by accounting for their fixed and variable compliance costs in the presence of firm heterogeneity. For our analysis, we develop a partial equilibrium variant of a trade model with heterogeneous firms and apply it to the case of meat trade between Poland (PL) and the old EU member states (EU15). Using simulations, we specifically look at the trade and market structure effects of the financial support to assist Polish firms in complying with the EU food standards. Rau and van Tongeren (2007) account for the fixed and variable costs of compliance in the Polish meat sector in an oligopolistic market framework where the partitioning of firms into exporters and non-exporters is fixed. The current paper extends this analysis by making the decision to export endogenous in the heterogeneous firm modelling framework, and hence provides further insights into the characteristics of firms that engage in exporting.

In the context of EU enlargement, the Polish meat case is particularly illuminating since (i) the Polish meat sector is very fragmented, with a large share of small- and medium-sized firms, (ii) the EU standards are particularly strict for food products of animal origin that can present serious hazards to human health and (iii) the EU standards in the Polish meat sector represent a prototypical instance of a small country adopting standards to gain access to a larger market. Although the paper takes the Polish meat sector as a case study, it contributes to the broader discussion of economic integration. In the heterogeneous firm model, lowering trade costs has the benefit of increasing the number of new varieties and firms, in addition to the usual expansion of trade by existing firms.

The paper is structured as follows: we first introduce the case study of Polish meat trade, with a focus on the EU food standards and support programmes to assist firms in their compliance. This is followed by a section on the heterogeneous firm model that we subsequently apply to the case study. We empirically estimate the exogenous model parameters by using a nonlinear least squares method. This estimation approach, including a Monte Carlo simulation to obtain insights into the reliability of parameter estimates, is elaborated in the section on the model application. A number of scenarios are simulated to investigate the impact of the strict EU standards on meat trade between Poland and the EU15. The simulations specifically explore the role of EU support programmes in lowering the costs of market entry for Polish meat firms.

2. The Polish meat case

The integration of Poland into the EU started long before its accession in May 2004. The initial association with the EU and subsequent amendments gradually liberalised agri-food trade between Poland and the EU15 to establish a free trade area with common border protection (custom union). While offering improved trading opportunities, EU membership was conditional on adopting the entirety of the EU legal rules and regulations (acquis communautaire). Poland had to adopt the EU food standards, just like the other new member states, but transitional periods for the time after accession and special safeguard clauses to ensure the functioning of the EU common market were agreed upon; for details see Inglis (2004).

In order to supply meat products on the EU single market, slaughterhouses, cutting plants and processing firms in the member states have to comply with Directive 77/99/EEC and 64/433/EEC. These directives include the EU meat standards and specify some additional provisions regarding product testing, transportation and administrative matters.² In order to meet the EU requirements, Polish meat firms had to modernise production facilities and production methods, which led to fixed costs of compliance. This modernisation potentially promoted the productivity of Polish meat firms, but the necessary changes in processing methods and control systems for food safety/hygiene also increased variable production costs. Variable production costs mainly increased due to the employment of more skilled workers, more frequent checks as well as documentation requirements (Preidel and Rau, 2006). Representative information about the fixed and variable compliance costs in the Polish meat sector is not available. However, Eurostat reports the investments undertaken in the sector, and they most likely contain the fixed costs of compliance.

Figure 1 plots the development of meat trade between Poland and the EU15 against the investments in the Polish meat sector. The volume of Polish meat exports steadily grew in the pre-accession period, and the largest increase is observed after Poland became an EU member country in 2004. To date, the EU member states absorb about 95 per cent of the total Polish meat exports. Almost half of the Polish pork exports go to the other new member states, and with regard to other meat products the EU15 constitute the main export destination: beef (91 per cent), poultry (77 per cent), other types of meat (85 per cent) and meat preparations (65 per cent). While obviously reflecting the liberalisation of meat trade between Poland and the EU15, the trade growth shows a rather strong correlation with the investments undertaken. In the years shortly before accession, investments in the Polish meat sector increased dramatically, and this influx coincides with the EU's pre-accession assistance. The EU Special Accession Programme for Agriculture and Rural Development (SAPARD) in Poland, which was scheduled for the time period 2000–2006, provided funds to the agri-food sector, with the first payments made in 2002. Almost 40 per cent of the total SAPARD funds in Poland were earmarked for supporting agri-food processing firms in their adjustment to the EU sanitary and hygiene standards (Measure 1 ‘Improvement in processing and marketing of agri-food products’). According to the ex-ante evaluation of the Polish SAPARD programme, the meat sector absorbed 48.6 per cent of the financial assistance paid under Measure 1, equalling about EUR 266 million (IERiGZ-PIB, 2007). Mainly large Polish meat firms benefited from the SAPARD funds that were co-financed by the EU15 and Poland and paid up to 50 per cent of the costs of approved investment projects. Considering the total investment in the Polish meat sector, the funds provided were rather modest, and it should be noted that not all investments to comply were subsidised. Relative to total investment expenditures, the SAPARD funds accounted for about 35 per cent of the investment in the Polish meat sector.

With regard to the aforementioned directives, Rau and van Tongeren (2007) elaborate on the state of compliance in the Polish meat sector and the resulting market opportunities for firms in the accession year 2004. Polish meat firms that comply are approved to export to the other EU member states and receive an EU export licence, whereas non-complying firms can sell their product on the domestic market only. The production capacity marks the dividing line between complying and non-complying Polish meat firms. In 2004, almost 70 per cent of the Polish meat firms did not meet the EU requirements, and they were mainly small and medium enterprises with considerably lower production capacities than large firms. Due to their tremendous difficulties to adopt the EU standards, small Polish meat firms have been granted special provisions (mainly relating to the detailed documentation requirements and record keeping) that allow them to continue their production for the domestic market without fully meeting the EU requirements.

The Polish meat sector comprising the production, processing and preserving of meat and meat products can generally be characterised by its very asymmetric structure. With many small firms and only a few large firms, the size distribution of Polish meat firms is extremely skewed. In 2002, 72 per cent of the total of 4,271 Polish meat firms registered by Eurostat employed between one and nine persons, and only 10 per cent had more than 50 employees. By 2004 the number of meat firms had declined to 3,881. Most of the industry exit occurred in the smallest size class whose share accounted for 69 per cent of the Polish meat firms in 2004. Table 1 presents the asymmetric structure in the Polish meat sector by some indicators of inequality estimated for the time period 2000–2004. For comparison, Table 1 also includes the respective indicators for the EU15. Both markets were rather concentrated, with a considerable share of employment, turnover and value-added generated by the larger firms. The Gini-coefficient appears to point to a slightly lower inequality in the structure of the Polish meat sector compared with the EU15. On the other hand, the concentration ratios are higher for Poland, indicating the relative economic dominance of large firms in the sector. According to the C5-ratio, the five largest firms generated 23 per cent of the total turnover in the Polish meat industry; adding the next five largest firms increases the ratio to 27 per cent. In the substantially larger EU15 market, the five largest firms are estimated to account for 18 per cent of the total turnover.

3. The model

The heterogeneous firm model that we derive is a partial equilibrium variant close to the general equilibrium model by Balistreri et al. (2007), which is, in turn, leaning on the seminal theoretical work of Melitz (2003). In our partial equilibrium model capturing firm heterogeneity, factor prices and incomes are exogenous, and another difference to Balistreri et al. concerns the estimation of the model parameters, which shall be highlighted in the application of the model below (see Section 4.2). The basic idea of heterogeneous firm models is that firms differ in their productivity levels and only sufficiently productive firms are able to overcome the hurdles posed by market entry costs. Firms incur fixed costs to enter potential markets where each firm competes with all other suppliers in a large-group monopolistic competition setting. Conceptually, firms pay to participate in a lottery in which they randomly obtain a productivity level, and only those firms that draw productivity levels larger than some threshold level will be able to profitably enter the respective market. The threshold productivity level itself is endogenous and depends on fixed and variable costs as well as on demand characteristics.

Applying the heterogonous firm model approach in the analysis of import standards, we emphasise the importance of productivity for firms to successfully operate on export markets. In any country, only a fraction of firms meets the standards requirements of importing countries and is thus able to export, while the majority of firms serve the domestic market only. That is, import standards affect the firms' fixed and variable costs and hence have an impact on the selection of firms into export markets. As will be shown, cost-increasing requirements tend to raise the productivity threshold for exporters and shift market shares away from less productive firms. Due to compliance costs, standards coincide with a higher observed productivity of exporters, even without the possibility of technological progress induced by more demanding production requirements.

The model features a set of countries/regions R with each region r = 1, 2, … , n_r exporting and importing. For pairs of trading regions, s denotes the exporting or source region and r denotes the importing or destination region; s, r∈R. If s=r products are sold domestically, whereas s≠r implies sales on foreign markets.

Firms set an optimal price p_sr to maximise profits p_srq_sr − (c_sτ_sr/θ_s)q_sr − f_sr on each potential market. From the first order conditions, it is straightforward to derive the optimal pricing rule depending on productivity, cost parameters and the substitution elasticity only: p_sr = (c_sτ_sr/θ_s)[σ/(σ − 1)]. Note that once a firm overcomes the fixed entry costs, they are sunk and hence do not matter in the firm's supply decisions, but as described further below, the fixed entry costs play a role in determining the number of active firms.

4. Application of the model

In our application to the case of meat trade between Poland and the EU15 (two countries/regions case), the model consists of a total of 15 parameters that have to be determined. The 2 × 2 Pareto shape and location parameters are estimated outside the model (see Appendix A1). This leaves 11 parameters to be determined: c_s, f_sr, τ_sr and σ. Since τ_ss = 1, the number of parameters to be estimated reduces to n_r × (1 + 2 × n_r) − 1 = 9. Before elaborating on the estimation of the nine parameters, information on the data used in the model application is provided first. The model parameters estimated are applied in the simulation exercise, and the simulation scenarios and results are reported subsequently.

4.1 Data

The application of the model relies on publicly available data for Poland and the EU15. To estimate the parameters of the Pareto distribution of productivities, we use the information on the number of firms, production and sales as well as distributional information that the Eurostat Business Statistics provide as grouped data per firm size class; see Appendix A1. The Eurostat data refer to meat production and processing (NACE DA151), and given the available data, firm productivity is approximated by the turnover per person employed, which is admittedly inferior to some measure of total factor productivity. Trade statistics come from UN COMTRADE (HS code 0201-2010: beef, pork, poultry and other meat types; HS code 1601-1602: meat preparations).

For the estimation of the parameters, short time series of the years 2002–2004 are used, while the simulation are performed with the data for the accession year 2004 as the baseline. For the simulation scenarios, information on the SAPARD programme is obtained from the ex-ante evaluation report, IERiGŻ-PIB (2007).

4.2 Parameter estimation using nonlinear least squares

The estimation method is inspired by Balistreri et al. (2007) who propose a nonlinear least squares method to estimate some of the unobservable trade cost parameters by imposing exogenous estimates of the variety substitution elasticity, the Pareto location parameter and the fixed entry costs on the domestic market. They also assume the productivity distribution to be the same in all regions. For the estimation, Balistreri et al. use cross-section data on bilateral trade, breaking global trade down into nine regions and seven sectors, whereas we use (short) time series and two trading partner countries/regions. Another difference to Balistreri et al. is that we estimate both the Pareto shape and location parameters separately for Poland and the EU15 using OLS on grouped data. We also estimate the fixed entry costs on domestic and foreign markets and the variety substitution elasticity consistently with the rest of parameters and the model equations.

The estimation strategy is to minimise the sum of squared differences between the trade values observed and those generated by the model, taking the full set of model equations as restrictions and taking the Pareto parameters as given. This yields model-consistent estimates of all remaining parameters. In order to obtain insights into the reliability of these estimates, Monte Carlo simulations are used to generate a synthetic data set that is then used to re-estimate the parameters of the model. The distribution of these estimates is investigated by using Kernel density estimation.

Table 3 presents the estimates obtained by this method as well as those of the parameters of the Pareto productivity distribution that have been obtained by an OLS estimation outside the model as already mentioned.

Our estimates of both the variable and fixed trade costs are higher for firms in the EU15 than for their Polish counterparts. It appears to be far less costly for Polish firms to enter the EU market, and once they entered, the variable trade costs are also smaller. A Polish meat firm incurs a domestic fixed cost of almost EUR 23,000, and when exporting to the EU15 it faces an additional EUR 15,000 of fixed market entry costs. For EU15 meat firms, the domestic market entry costs amount to EUR 585,000, and exporting to the Polish market adds EUR 1.2 million to the domestic market entry costs. The very high estimates of the EU15 firms' fixed costs to enter the Polish market may be a result of the fact that EU15 exports to Poland constitute only a very small fraction of production, and only the very productive firms are exporters in the model. In other words, the model requires large fixed entry costs for EU15 firms to replicate the small export volume from the EU15 to Poland. The estimate may, therefore, be biased upward. Another explanation may be the quality differences that are not captured in the model, with the EU15 meat exports being of a relatively higher quality level. Concerning the considerably lower estimates of the fixed costs for Polish meat firms, it should be noted that the effects of subsidies to comply with standards are implicitly taken into account, and we may hence underestimate the true costs. The finding that domestic set-up costs are also much smaller in Poland than in the EU15 is consistent with the size distribution of firms in Poland being dominated by many small firms that produce less capital-intensively but with low productivity. On average, EU15 firms are larger and even the smallest firms in the EU15 have a productivity that is 4.5 times greater than the productivity of the smallest Polish firms. The estimated unit price for variable inputs is slightly smaller in Poland, but the difference is far less pronounced than the extraordinary difference in the other cost parameter estimates. Finally, the estimate of the elasticity of substitution between varieties (σ) is found to be on its upper bound, dictated by the value of the Pareto shape parameter for Poland.⁵

Unfortunately, the reliability of the parameter estimates cannot be judged on the basis of standard test statistics. First, it is not obvious how to obtain covariance matrices of the parameters and, second, we do not know the distributions that govern the parameters. The latter implies that we cannot rely on test statistics that assume normality. Some insights into the distribution of the parameters can be gained by Monte Carlo simulation.

4.2.1 Monte Carlo simulation

Given the parameter estimates and observations on exogenous variables, we generate a new data set of trade values by adding random disturbances to the trade values which the model simulates: ⁠. The random disturbances u are assumed to be iid N(0, s²), and the standard deviation s is derived from the residuals of the original least squares estimation. Subsequently, the least squares estimation is repeated with assuming the role of x^o. The Monte Carlo procedure thus first generates synthetic data through a completely controlled process and then re-estimates the model parameters using this synthetic data set. After repeating this process 100 times, a reasonably large sample of parameter estimates is obtained.⁶ This sample of estimates reveals that the parameters are certainly not normally distributed, as seen, for example, by comparing the mean and median values presented in Table 4, and hence the usual normality-based test statistics do not apply.

4.2.2 Kernel density estimation

The non-parametric kernel density estimation allows us to empirically construct a density function of estimated parameters on the basis of the sample of observations generated through the Monte Carlo simulations (Silverman, 1986). The results of the kernel density estimates are presented in Figure 2. As shown in the figure, the estimates of the substitution elasticity (σ) are very much clustered near its theoretical upper bound, while some estimates also yield lower values. In comparison to other estimates, the unit price of variable inputs (c_s) is most symmetric around the mean. The kernel density estimates suggest that there is indeed a difference in c_s, with both the median and mean values for the EU15 being larger than for Poland. In contrast, the estimated distributions of trade costs are all extremely left-skewed. The median value of the variable trade costs (τ_sr) for Poland is below the EU15 median, and this appears to indicate a consistent difference in variable trade costs. The densities of the fixed costs of supplying the domestic market (f_ss) are sufficiently far apart, and we can conclude that there is a significant difference between the two markets. Likewise, for Polish and EU15 firms, the fixed costs of exporting (f_sr) differ considerably although the density estimates reveal a similar shape.

The results presented in this section indicate that a consistent estimation of the parameters for the heterogeneous firm model is possible with limited information, but the estimates cannot be assumed to be normally distributed. We have been able to estimate the unobservable trade costs by imposing a huge theoretical structure on the estimation problem. The highly non-linear characteristics of the model in combination with theoretical bounds on parameters and the very pronounced size asymmetry between the two markets makes the parameter estimation a perilous task.

4.3 Scenarios and simulation results

In our simulations, we explore the effects of subsidising compliance costs, like the EU SAPARD programme did in the Polish meat sector. With the calibration, the simulations take the accession year 2004 as the base, and our parameter estimates implicitly include the effects of both the strict EU standards and the SAPARD funds in Poland. We therefore ‘backcast’ the model in the first set of simulations that consequently reflect scenarios where standards are in place but the subsidies to comply with them are reduced. This allows us to investigate the effects of supporting firms in their compliance with standards.

In the first simulation experiment (S1), we increase the fixed entry costs for Polish firms on both the domestic and the EU15 market (f_PL,PL and f_PL,EU15) in order to reflect the fact that the SAPARD programme was not specifically targeted at exporting firms. Given the information about the SAPARD programme, the size of the shock amounts to 25 per cent and approximates the full removal of the subsidy paid to the meat firms in Poland (see Section 2). In the second simulation (S2), we refer to the considerable testing and documentation requirements for exporting to the EU market, and thus increase the variable ‘iceberg trade costs’ for Polish meat firms (τ_PL,EU15). Without representative information, we also apply a 25 per cent shock. The third simulation (S3) is a combination of S1 and S2. The fourth simulation (S4) relates to the upgrading of the Polish meat sector to the EU standards that may lead to more dynamic productivity gains. Whether standards bring about an overall increase in productivity through technological improvements is first and foremost an empirical question. However, since the Polish meat sector can generally be considered as rather traditional, we assume that the investments undertaken to meet the EU requirements had a positive effect on the level of productivity of Polish meat firms. In the simulation, we mimic this by raising the minimum productivity level determined by the Pareto location parameter β_PL.

Focusing on the perspective of Polish meat firms, Table 5 summarizes key simulation results. The 25 per cent rise in fixed entry costs following an assumed withdrawal of the SAPARD programme leads to an increase of the average productivity of Polish meat exporters (10.6 per cent). This is not due to technological improvements, but the result of less productive firms being driven out of the EU15 export market. In other words, without the financial assistance, less productive firms cannot bear the higher market entry costs. Only those firms that have a productivity that is at least 10 per cent higher can enter the EU15 market without the subsidy, and the number of exporters consequently falls if the subsidy is removed. The change in the number of firms is an indication for a shrinking extensive trade margin: in the model, each firm is producing one product variety and hence the fewer firms there are, the fewer varieties are exported. The intensive margin, on the other hand, relates to the amount of exports by existing firms to existing destinations. In the simulation of higher fixed costs of exporting, the extensive margin shrinks (fewer varieties) while the intensive margin grows (expansion of exports of surviving firms).

The results are different when the variable ‘iceberg trade costs’ are increased (S2). Although the average productivity of exporters increases, the number of exporters considerably falls and export values decline. The reason for the different effects between fixed and variable trade costs is that the former are sunk and are not taken into account once the firm has entered the market. In contrast, the variable trade costs play a role in the firm's pricing decisions. At any given level of market entry costs, higher variable trade costs lead to higher prices for exports to the EU15, reducing export demand for existing firms (the intensive margin shrinks) and driving firms out of the EU15 export market (the extensive margin shrinks). Note that the quantity supplied by the average exporting firm does not change, even though the average productivity of exporters increases by about 25 per cent. The reason is that the change in the average productivity (and in the cut-off productivity) is as large as the change in variable trade costs (25 per cent) under the current parameterisation of the model, so that the two effects cancel out in the optimal pricing equation (equation 2). The price set by the average firm does not change, and hence export demand remains constant. Combining higher fixed trade costs with higher variable trade costs reinforces the previous effects: decrease of the number of exporting firms and a higher average productivity of exporters, covering the fixed and variable trade costs.

Table 5 also reports the simulation results for the consumer utility index. In the case of higher fixed trade costs (S1), Polish consumers are affected by higher prices, but this does not significantly spill over to EU15 consumers. Polish meat has a very small market share on the EU15 market, and consequently the supply changes do not have big market effects in the EU15. In contrast, the higher variable trade costs in S2 negatively affect EU15 consumers' welfare since higher prices and less variety reduce welfare. Overall, the welfare of consumers in the EU15 decreases in all scenarios, and thus the subsidy for compliance is beneficial from the consumers' consumption point of view.

As described, the model allows the disentangling of the intensive and extensive margin effects of higher trade costs due to standards, and this constitutes a main difference between this model and the model by Rau and van Tongeren (2007). Although the fixed and variable compliance costs are accounted for as drivers for change in meat trade between Poland and the EU15, the latter model depicts homogenous goods such that it is impossible to distinguish the effects of changes in the number of varieties (extensive margin) from the effects of changed export supply (intensive margin). However, it should be noted that the results for quantity and welfare generated by the homogeneous product model appear to be qualitatively similar to those presented here, but the market structure implications are far less elaborate. With the heterogeneous firm model, the firms' decision to enter foreign markets and export in the face of standards is modelled endogenously. In addition, the model yields insights into the relationship between productivity and exporting that requires compliance with the standards imposed by importing countries.

Producing according to standards can be expected to lead to productivity upgrades, which will be reinforced if compliance costs are subsidised. To capture this more dynamic effect, the last column in Table 5 shows the effects of a 10 per cent increase in the minimum productivity level in the Polish meat sector (β_PL). This is a very modest increase considering that the economy-wide labour productivity growth in Poland between 2002 and 2004 amounted to about 4 per cent per year (Eurostat). The change in the minimum productivity shifts the lower end of the productivity distribution to the right, but it does not affect the export market entry decisions. While Polish meat firms that were able to export before the productivity upgrades continue to do so, some of the firms that previously only produced for the domestic Polish market reach the unchanged cut-off productivity for exporting and thus start supplying the EU15 market. The simulation results reveal that such a productivity increase would have been sufficient to compensate for the export loss caused by increased fixed costs of compliance without the subsidy.

4.4 Sensitivity analysis with respect to the variety substitution elasticity (σ)

It has been shown above that there is some uncertainty regarding the estimates of model parameters. A sensitivity analysis can shed some light on the importance of that uncertainty. Since we have already perturbed the fixed and variable trade cost parameters in the simulations, we here concentrate on one remaining key parameter in the model: the elasticity of substitution between varieties (σ). Among others, Chaney (2008) elaborates on the important impact of σ on trade effects in a model with heterogeneous firms. In the sensitivity analysis, we take the estimated median value and vary it within its estimated 96 per cent confidence interval (see Table 4). The simulations reported in the previous section are subsequently repeated three times with σ taking a low, median and high value.⁷ For exports from Poland into the EU15, Figure 3 shows the trade effects following the variations in market entry parameters and minimum productivity.

As illustrated, a low elasticity of substitution coincides with greater effects of increasing trade costs: the lower the elasticity of substitution, the more exports from Poland to the EU15 contract. Not shown in Figure 3 are imports into Poland that tend to expand more under lower values of the substitution elasticity. This finding contradicts the usual model results in which a lower σ implies that cost changes are not directly translated into changes in market shares. In standard trade models, a higher σ means more competition between varieties and more pronounced effects of changes in trade costs. Without firm heterogeneity, all firms expand their exports following a decline of trade costs: the intensive margin increases and this effect is more pronounced the higher the elasticity of substitution. In contrast, in the heterogeneous firm model, a low elasticity of substitution means that increasing (falling) trade costs drive out (attract) relatively productive and large firms. In other words, the lower the σ, the more concentrated the exports among the more productive and larger firms. Consequently, higher trade costs will lead to a more pronounced decrease in the quantity supplied as relatively large firms are driven out of the market. As Chaney (2008) formulates, the extensive margin is more sensitive to changes in trade costs for lower values of σ. The simulation results of the Polish meat case presented indicate that the extensive margin dominates: the lower the elasticity of substitution the larger the effect on trade.

5. Conclusion

The heterogeneous firm model that we develop recognises the asymmetric situation, where firms complying with standards and non-complying firms co-exist but supply different markets. The model features are characterised by firm-level monopolistic competition with firm heterogeneity introduced through a distribution of productivities. For our application to the case of meat trade between Poland and the EU15, we estimate the model parameters by using limited data and a nonlinear least squares method to match observed patterns of trade. The estimates of the size and productivity distribution of Polish and EU15 meat firms and the approximated compliance costs reflect the heterogeneity in the sector. Our analysis examines the effects of standards on indicators of market structure as well as on trade.

Given heterogeneous firms with asymmetrically distributed productivities, homogenising standards, which leads to compliance costs for producers, tend to increase the concentration of production and exports among the more productive and larger firms. This effect is more pronounced if product varieties are less substitutable in the eyes of consumers.

While we specifically consider the financial assistance that helped Polish meat firms to adjust to the tight EU food standards in the simulations, some more general conclusions about the effects of subsidising compliance can be drawn. We find that subsidies to comply with the standards required by importing countries lower the firms' productivity threshold to qualify for exporting. Since smaller and less productive firms that already operate in the industry can continue to exist structural changes in the industry of the exporting country tend to be dampened. In order to present a more complete picture, we combine the direct cost-increasing effects of standards with productivity upgrades in the analysis. Although data limitations do not allow us to exactly pin down the size of the productivity gains, a modest 10 per cent increase of the minimum productivity level in the Polish meat sector would be more than sufficient to compensate for the export loss following a simulated 25 per cent increase in the firms' fixed costs of compliance.

Acknowledgements

The authors thank Abdelhakim Hammoudi, Ruben Hoffmann and Yves Surry for their valuable comments on earlier versions of the manuscript. In addition, the comments provided by three anonymous referees have been immensely helpful to improve the paper. The views expressed are those of authors and do not reflect the official view of the OECD or of the governments of its member countries.

References

Appendix A1: Estimating the Pareto productivity distribution with grouped data

The estimation results are presented in Table A1. According to the standard errors and the proportion of variation explained (R²), the estimated equation appears to fit the data very well. The smaller value of the shape parameter α indicates that the productivity in the Polish meat sector is more unequally distributed than in the EU15. At the same time, the average productivity is more than three times higher in the EU15 than in Poland. This corresponds

to an average turnover of EUR 156,000 per person employed in the EU15 versus an average turnover of EUR 46,000 per person employed in Poland.

Appendix A2: Estimation of indicators of market concentration

With the assumed Pareto size distribution, the expression for the Ck index becomes Ck = (k/n)^(α−1)/α where α refers to the shape parameter of the Pareto distribution.

Author notes