Unconstrained Trade: The Impact of EU Cage Bans on Exports of Poultry-Keeping Equipment

This study evaluates the impact of conventional cage bans for laying hens in the EU on exports of poultry-keeping equipment. Using detailed data on international trade in poultry-keeping equipment combined with an event study regression approach yields several new ﬁndings. The results suggest that the cage bans were associated with an increase in intra-EU trade, and also an increase in exports of poultry equipment from EU member states to non-EU countries where conventional cages are still permitted. The results suggest that some banned cages were likely exported to countries outside the EU to be used in egg production.


Introduction
Consumers and citizens in many countries continue to demand higher standards for the welfare of farm animals (Eurobarometer 2016). Governments have thus responded by passing legislation with the objective to improve farm animal welfare. In the case of laying hens, the banning of conventional cages has been implemented in many countries.
A potential concern, however, is that more stringent rules in one jurisdiction will lead to higher costs, farm exits (Harvey and Hubbard 2013), and also a shift in animal production to other jurisdictions where animal welfare regulations are less stringent. This phenomenon, known as policy leakage, is important because it can partially undermine the intent of the regulations. It is thus important to understand the scope of policy leakage with respect to animal welfare regulations. The export of banned production equipment is one way that animal welfare policies in one region can lead to leakage to "low animal welfare havens." This study presents the results of an ex-post evaluation of the impact of conventional cage bans for laying hens in the EU on exports of poultry-keeping equipment, which includes cages. The analysis focuses primarily on the impact of the 2012 EU conventional cage ban, but also includes EU countries that implemented bans before 2012. The analysis includes both intra-EU trade and exports of poultry cages from EU member states to non-EU countries where conventional cages are still permitted.
Studying the impact of stricter animal welfare regulations on international trade in animal-keeping equipment is new in the literature. In contrast, other studies in this literature have focused on the impact on production and trade in animal-based food products.
Using detailed annual trade data at the country-pair-product level and an event study regression approach, the analysis finds that EU countries' exports of poultrykeeping equipment increased around the time of the bans. These results suggest that some egg producers sold their conventional cage systems to egg producers in other countries outside the EU. In addition to a statistically significant impact on EU exports 2 of poultry-keeping equipment, a large and temporary increase in intra-EU trade in poultry-keeping equipment is observed around the time of the bans. The increase in intra-EU trade around the time of the conventional cage bans is a natural consequence of procuring new poultry-keeping equipment within the EU single market. EU exports of poultry-keeping equipment may, however, indicate that used cages were sold to egg producers outside the EU.
This study builds on a small and recent empirical literature evaluating ex post the impact of stricter animal welfare regulations on economic outcomes in general, and trade in particular. The most relevant studies of the trade effects of animal welfare regulations are based on the 2015 California cage ban. Mullally and Lusk (2018) find that egg production and the number of egg-laying hens were about 35 percent lower due to the law, but imports from other U.S. states compensated for this decrease in production. Carter et al. (2021) study the impact of inter-state trade in more detail, and find that imports to California were characterized by higher firm-level concentration. Carter et al. (2021) also find prices in other states rose in response to California's law, resulting in a loss of consumer surplus and retailer surplus at the national level. 1 The potential for an increase in cage exports in response to a cage ban relates to the phenomenon of policy leakage via input markets, where the input in this case is cages. The leakage mechanism studied here is thus conceptually distinct from the earlier work by Malone and Lusk (2016), Mullally andLusk (2018), andCarter et al. (2021), which focus on leakage via output markets (eggs). This study is the first to analyze international trade in animal-keeping equipment affecting farm animal welfare and its subsequent response to a ban.
The rest of the study proceeds as follows. In section 2 I characterize the market for poultry equipment in the EU and explain the historical details of conventional cage bans at the EU and national levels. Section 3 describes the international trade data used in the analysis. In section 4 I specify the event study methodology used in the analysis, and in section 5 the results of the event study are described and discussed.
Conclusions follow in section 6.

The EU Conventional Cage Ban
The EU Council Directive 1999/74/EC stipulated a ban of conventional laying hen cages starting January 2012. 2 This directive was passed in 1999, which gave EU member states several years to comply. This directive forced egg producers to replace their conventional cages with so-called "furnished" or "enriched" cages. The directive also permitted cage-free systems, which includes production in barns, free-range, and organic production. The directive stipulated that furnished cages must provide at least 750 cm 2 of area per hen, of which 600 cm 2 is at least 45 cm high. Furnished cages must also provide a nest, a littered area, a sufficiently large perch and food trough, and a claw shortening device. 3 A few EU countries banned conventional cages prior to the EU-wide ban. The official date of these national bans usually did not correspond to the timing of full compliance, as phase-out periods were granted. In Sweden, for example, the final day of banning conventional cages according to the legislation was January 1st, 1999, but extensions were granted for most producers until 2004 (Berg and Yngvesson 2006 Some industry stakeholders were initially concerned about the prospect of increased egg imports due to the EU cage ban. 9 The EU has not banned the import of eggs produced with conventional cage systems, although a majority of Members of the European Parliament supported a Resolution on the European Citizens' Initiative 'End the cage age' (2021/2633(RSP)) on 10 June 2021. Banning imports of eggs produced using conventional battery cages would constitute a barrier to trade, which could lead to a trade dispute at the World Trade Organisation (WTO). The EU-Mercosur Trade Agreement is expected to include a requirement that eggs imported from Mercosur countries must comply with EU standards for laying hen production (Hagemejer et al. 2021). The figure suggests that there was a temporary increase in egg imports in 2012, but egg imports then fell after 2012. The EU is generally self-sufficient in egg production, with the value of annual imports ranging between EUR 20 million and EUR 40 million most years since 2004, which corresponds to import volumes of between 20 thousand and 40 thousand tonnes egg equivalent. In contrast, the EU produced a total of 7 million metric tonnes (mmt) of eggs i 2020, with the top 5 largest egg producers in being France (979 mmt), Germany (967 mmt), Spain (918 mmt), Italy (806 mmt) and the Netherlands (703 mmt).

Data and Descriptive Statistics
The analysis uses bilateral trade flow data from EU member states to each export destination, both within and outside the EU. This data is available at the 6-digit Harmonized System (HS) level. The source of the international trade data is CEPII's BACI 9 See https://www.fwi.co.uk/news/farm-policy/deluge-of-egg-imports-predicted-after-eu-cage-ban 6 database (Gaulier and Zignago 2010). 10 The data is converted from USD to EUR using exchange rate data from the Penn World Table (Feenstra et al. 2015). 11 Trade value data is converted to constant 2015 prices using data from the Organisation for Economic Cooperation and Development (OECD). 12 The analysis uses bilateral trade flow data for the period 1995-2019, where 2019 is the latest available year available for the international trade data. This timespan provides a long span of pre-treatment and post-treatment periods in the event study.  Table version 10.0 is available at https://www.rug.nl/ggdc/productivity/pwt/. 12 Consumer Price Index, All Items for the Euro Area, Index 2015=100, Annual 13 HS 8432: Agricultural, horticultural or forestry machinery for soil preparation or cultivation; lawn or sports-ground rollers. HS 8433: Harvesting or threshing machinery, including straw or fodder balers; Grass or hay mowers; Machines for cleaning, sorting or grading eggs, fruit or other agricultural produce, other than machinery of heading 8437. HS 8434: Milking machines and dairy machinery. HS 8435: Presses, crushers similar machinery used in the manufacture of wine, cider, fruit juices or similar beverages. HS 8437: Machines for cleaning, sorting or grading seed, grain or dried leguminous vegetables; Machinery used in the milling industry or for the working or cereals or dried leguminous vegetables, other than farm-type machinery.
supply as poultry cages, making them a suitable control group. The control group product is referred to as "Agricultural machinery" throughout the rest of the analysis.
According to manufacturing production data available via the EU PRODCOM database, Germany, Italy, Spain and the Netherlands are the largest producers of "Poultry-keeping machinery (excluding poultry incubators and brooders)" (PROD-COM code 28308500). 14 At the same time, several EU members states report that they do not produce any such products. Replacing conventional cages with enriched cages thus requires EU member states to import equipment in many cases. However, exports of poultry-keeping equipment rose during this time.
The total export value of "Other poultry-keeping machinery" to non-EU countries over the period 1995-2019, reported by EU member state, is provided in Table A.1 in the appendix. The largest producers of eggs and poultry-keeping equipment feature prominently among the EU exporters. Germany, Italy, Spain, and the Netherlands were the largest exporters of "Other poultry-keeping machinery" to countries outside the EU. advance of the 2012 deadline as well as delays in compliance imply that sales of used 14 Prodcom data is available at https://ec.europa.eu/eurostat/web/prodcom/data/ excel-files-nace-rev.2. Since HS and CN product classifications are identical at the 6-digit level, it is possible to find the corresponding Prodcom product code using Eurostat's Prodcom-CN concordances available at https://ec.europa.eu/eurostat/ramon/relations/index.cfm?TargetUrl=LST_REL. 8 cages could be detected several years before and after the 2012 deadline.
The top-20 destinations for EU exports of "Other poultry-keeping machinery" over the period 1995-2019 are reported in Table A.2 in the appendix. Russia was the largest export destination during this period, followed by USA, Japan, and Ukraine. The

Event Study Methodology
The analysis applies an event study regression methodology to study the impact of conventional cage bans. As noted earlier, the timing of the ban was 2012 for all EU countries except for Sweden, Austria, and Germany, which had national conventional cage bans in 2004, 2009 and 2010 respectively. As mentioned earlier, the treatment group product is "Other poultry-keeping machinery", while the control group is "Agricultural machinery".
The event study regressions employ a Poisson pseudo-likelihood regression with multiple levels of fixed effects (Correia et al. 2019(Correia et al. , 2020. Poisson regressions have the advantage of allowing for zeros in the trade flow data, and have recently gained popularity in the international trade literature, starting with work by Silva and Tenreyro (2006). The estimation follows the standard event study approach, and includes seven pre-treatment and post-treatment periods. Pre-and post-treatment effects greater than seven years are included in the regression, but not reported.
The event study regression model is a special version of the Difference-in-differences model, employing multiple time periods and including lag and lead treatment terms. 16 The event study model used in this analysis takes the following form: where Y ijkt is the value of exports from EU member state i to destination j of good k at time t. α ijk and α ijt are origin-destination-product and origin-destination-year fixed effects respectively. D e ikt is an indicator for member state i being e periods away from initial treatment at time t. D e ikt always takes a value of zero for the control good. ijkt is the error term. Indicators greater than seven years before or after treatment are binned. t − 5 is used as the baseline event-time, for reasons explained below.
Fixed effects are used to control for other factors that affect international trade in the treatment and control goods. Origin-destination-year fixed effects control for all explanatory factors such as trade agreements that affect both poultry-keeping machinery (the treatment group) and other agricultural machinery (the control group). Origindestination-product fixed effects control for any time-constant explanatory factors that are specific to either poultry-keeping machinery or agricultural machinery. These fixed effects also control for the standard "gravity model of trade" variables such as GDP, distance, and the price indices. The point estimates are clustered at the origin country and destination country level, which is the most conservative clustering choice.
The event study analysis is divided into two parts: an analysis of intra-EU trade due to the cage bans, and an analysis of exports from EU member states to non-EU countries. Estimating the impact of the bans on intra-EU trade is interesting in its own right, but is also useful for detecting how many years before the ban and after the ban 16 The main advantage of an event study regression methodology is that it allows for estimation of pre-and post-treatment effects. In contrast, a simple difference-in-difference specification with a post-ban indicator would only allow for comparison for the period before versus after the ban in each country. that egg producers began replacing their old conventional cages with furnished cages, or with cage-free systems. If the bans on conventional cages is associated with increases in exports form EU countries to non-EU countries, this could be indicative of sales of used cages, implying that policy leakage occurred.
Egg producers could begin adapting to the cage ban well in advance of the deadline, and some egg producers did not comply with the cage bans on time. One can thus expect that trade in cages will be affected several years before and after the formal cage ban. The base year used as the benchmark for determining statistical significant effects is thus somewhat ambiguous. A standard event study would set event-time t − 1 as the base year, but since trade in cages may have preceded several years in advance of the ban, t − 5 is used as the baseline event-time. We use seven pre-treatment and post-treatment periods so that the event study results for cohort affected by the 2012 EU-wide ban does not overlap with the 10 countries that joined the EU in 2004 or the withdrawal of the United Kingdom from the EU in 2020.

Intra-EU Imports
The analysis begins with the event study results focusing on intra-EU imports around the time of the cage bans, including all 25 EU member states in the same regression.
The point estimates and 95 percent confidence intervals estimating the impact of the cage bans on imports of "Other poultry-keeping machinery" from other EU countries, relative to other agricultural machinery are illustrated in figure 4. Using year t − 5 as the baseline year is a reasonable choice, as intra-EU trade in "Other poultry-keeping machinery" was steady five years before the ban and earlier.
The results suggest that imports of poultry equipment from other EU-countries increased rapidly during the three years before the ban, reaching its peak in the year of the ban in each country. Intra-EU trade then declined and fell back to a statistically insignificant level three years after the bans. The regression coefficient on the treatment indicator at time t = 0 is 1.01, which implies that intra-EU trade rose by (exp(1.01) − 1) × 100 = 174 percent compared to 5 years prior to the cage ban deadline.
In sum, the results presented in in figure 4 suggest that intra-EU imports of "Other poultry-keeping machinery" clearly increased around the time of the ban. This trade likely included shipments of furnished cages meant to replace the banned conventional cages.

Exports to non-EU destinations
The next part of the event study analysis focuses on exports from EU member states to non-EU destinations, again including all 25 EU member states in the same regression.
The results of this analysis are illustrated in figure 5. The results indicate that exports of "Other poultry-keeping machinery" was arguably steady until two years before the ban in each country. The point estimate is positive and statistically significant one year before the ban and during the year of the ban. The regression coefficient on the treatment indicator at time t = 0 is 0.36, which implies that exports to non-EU countries rose by (exp(0.36) − 1) × 100 = 43 percent compared to 5 years prior to the cage ban deadline. The banning of conventional cages thus corresponded to a large percentage increase in the exports of poultry-keeping machinery to non-EU destinations. The point estimate is also positive at time t = 5, which I explore in more detail when studying the results by treatment cohort.
It is important to note that one cannot rule out that the exports of "Other poultrykeeping machinery" around the time of the conventional cage bans may be exports of furnished cages or other equipment used in cage-free systems. This limitation is due to the fact that data on international trade in used cages is not available. It is unlikely, however, that exports of poultry equipment would increase at exactly the same time as domestic demand for equipment to replace conventional cages is high due to the bans. The results are thus highly suggestive that such exports to destinations without 12 conventional cage bans could indeed be used conventional cages. Such trade may be replacing old cages in other non-EU countries, or it could be purchased by foreign egg producers that were aiming to increase egg production.

Additional Robustness
One threat to identification is that exports may be affected by other confounders that occur at the country-pair-product-year level. The most obvious confounder is that countries export more poultry equipment to destinations that have a comparative advantage in egg production. I use the bilateral imports of eggs, interacted with the treatment product dummy variable, in order to control for the impact of egg production on imports of poultry equipment such as battery cages. I use both contemporaneous and lagged egg imports as an additional control in the event study. The point estimates for egg imports are not statistically significant, and the event study results, illustrated in figure A.2 in the appendix, are robust to including these controls.
In the main analysis of exports from EU member states to non-EU countries I included the U.S. as an export destination. Even though the U.S. has not implemented a national conventional cage ban, including the U.S. may be slightly problematic since California implemented a cage ban during the study period. As a robustness check I drop exports to the U.S. from the analysis. The event study results for EU exports are presented in figure A.3 in the appendix. The main effect of excluding the U.S. is that there is no longer a statistically significant point estimate in the year t + 5.
As a final robustness check I drop exports from Greece and Italy from the analysis, since these two countries were the slowest to comply with the EU-wide ban. The event study results for EU exports are presented in figure A.4 in the appendix. Again, the main effect of excluding Greece and Italy is that there is no longer a statistically significant point estimate in the year t + 5.

Results by treatment cohort and quantifying the effects
Recent studies have shown that the coefficient of interest in a two way fixed effects specification is not guaranteed to recover an interpretable causal parameter if there is variation in treatment timing and heterogeneous treatment effects (de Chaisemartin and D'Haultfoeuille 2020, Goodman-Bacon 2021, Callaway and Sant'Anna 2021. In an effort to deal with this potential concern, I perform the event study analysis separately for each treatment cohort. There are four treatment cohorts in this study: Sweden (2004) The results of the event study estimation for intra-EU imports and exports from EU member states to non-EU countries for each treatment cohort are presented in figures 6 and 7 respectively. The results in figure 6 suggest that the imports of other poultry equipment to Austria from other EU member states were not affected by the Austrian conventional cage ban, but that imports of equipment to Sweden, Germany and other EU member states affected by the 2012 EU-wide ban did respond. The point estimates for the affected cohorts are at most around 1, which is nearly identical to the event study results including all cohorts in the same regression. Specifically, these point estimates imply that intra-EU imports rose by 199 percent, 194 percent, and 184 percent compared to 5 years prior to the cage ban deadline in Sweden, Germany, and the 2012 cohort respectively.
In order to quantify the effects of the cage bans, I convert the point estimates to monetary values by using the value of exports five years prior to the bans as a baseline.
In the case of Sweden, intra-EU imports five years prior to the ban were valued at The event study results for Germany in year t + 5 and for the 2012 cohort in year t + 3 also suggest that there was an increase in exports of "Other poultry-keeping machinery" relative to the control group that occurred in 2015. As exports to the U.S. are already dropped from the analysis, further inspection suggests that this temporary increase is driven by exports to Russia. A comparison of EU exports of "Other poultrykeeping machinery" and "Ag. machinery" to Russia is illustrated in figure A.5 in the appendix. The pattern of trade in the figure suggests that EU exports of "Other poultry-keeping machinery" temporarily diverged from the control group in 2015. This temporary divergence may have been driven by Russian import bans related to the conflict between Russia and Ukraine, which began in 2014.

Poultry equipment exports and egg imports
Given that some industry stakeholders were initially concerned about the prospect of increased egg imports due to the EU cage ban, it is worth exploring whether exports of poultry equipment is correlated with imports of eggs. Table A.3 in the appendix reports the results of a simple panel regression using poultry equipment exports from the EU and its lag as the independent variables and egg imports to the EU as the dependent variable. This analysis is purely descriptive in nature, and the results should only be interpreted as correlation, not causation.
The estimations reported in Table A.3 are performed at four levels of aggregation.
Column (1) reports the results using annual bilateral trade flow data. Column (2) reports the results aggregating each EU country's total trade with the rest of the (non-EU) world. Column (3) reports the results aggregating over all EU trade with each non-EU country. Finally, column (4) reports the results aggregating total trade between the EU as a whole and the Rest of the World as a whole. Panel and year fixed effects are included in columns (1)-(3), and a time trend is included in column (4).
The results reported in column (1) of Table A.3 suggest that there is no significant relationship between an EU country's bilateral exports of poultry equipment and its imports of eggs from the same country. Aggregating to over all non-EU destinations does not change this result, and aggregating total EU trade implies a statistically significant negative relationship between the trade flows. Only when aggregating to total RoW-EU trade in column (4) is the point estimate for poultry equipment exports positive and statistically significant. Thus, there appears to be a positive relationship between poultry equipment exports and egg imports overall, but its effect cannot be 16 attributed to a particular origin or destination.

Conclusion
This study presents the results of an ex-post evaluation of the impact of conventional cage bans in the EU on international trade in poultry-keeping equipment. The results suggest that the cage bans were associated with an increase in intra-EU trade and exports of poultry equipment, such as cages, from EU member states to non-EU countries where conventional cages are still permitted. Although it is not possible to provide direct evidence that the increase in exports was in fact used conventional cages, the results are highly suggestive that some banned cages were exported to countries outside the EU to be used in egg production where conventional cages are still allowed.
The conventional cage bans implemented in EU countries did not include any policy to ensure that used cages were not sold to egg producers outside the EU. In order to avoid the risk of exporting banned cages in the future, the EU may want to implement measures to avoid such "leakage" of cages to other countries. This issue may become important as EU countries consider banning furnished cages for laying hens, and as other countries implement bans on conventional cages.
One important potential limitation of the analysis is that it relies on trade data for a broader category of poultry-keeping equipment, as data focusing specifically on trade in cages, especially used cages, is not available. Although the goal of this study is to show that it is possible to detect trade in cages using the data that is presently available, including new cages and used cages as a separate product codes in international trade data would greatly simplify the tracking of cross-border trade in cages. For example, including product codes for these items in new versions of the EU Combined Nomenclature (CN) system of classifying goods would ease future work in this topic in the European context.
This study does not analyze the animal welfare implications of an export ban for animal welfare in other countries, and it is important to emphasize that the welfare implications are potentially ambiguous. For example, exports of conventional cages from the EU may be displacing worse cage systems in other countries. I leave the study of the animal welfare implications of international trade in cages for future research.
In sum, this study highlights that the complex issue of cage exports requires more attention from researchers and policymakers.      Notes: Based on observations from Figure 5. Source: BACI database.  Robust standard errors in parentheses *** p<0.01, ** p<0.05, * p<0.1