Sanitary and phytosanitary regulations and international red meat trade

Abstract

Purpose

The purpose of this paper is to provide an ex post econometric examination of SPS measures and their influences on red meat trade.

Design/methodology/approach

The authors conduct multiple new assessments to further assess the particular effects of specific SPS measures related to animal health, human health and maximum residue limits on red meat trade values. This finer assessment provides updated and more detailed insights into the marginal trade impacts of different SPS measures.

Findings

The current study sheds important light on the determinants of red meat trade. The economic conditions of destination countries and production capability of suppliers are key to determining trade values. Factors including personal income and exporters’ meat supply are identified as trade facilitators. Since the restrictiveness of SPS measures vary across beef and pork sectors, maintaining commodity-specific SPS measures is essential for accurate assessment of trade determinants.

Originality/value

This paper provides multiple contributions to the existing literature and more broadly the authors’ economic understanding on the increasingly contentious issue of global meat trade. Combined, this study yields several implications for food policy, trade negotiators and industry leaders given the growing role and surrounding controversies of trade in meat and livestock markets around the world. The authors further believe the paper would be of notable interest to fellow researchers consistent with the existence of a sizable published literature and ongoing debates in international meat trade.

Citation

Shang, X. and Tonsor, G.T. (2019), "Sanitary and phytosanitary regulations and international red meat trade", British Food Journal, Vol. 121 No. 10, pp. 2309-2321. https://doi.org/10.1108/BFJ-10-2018-0663

Introduction

Global red meat consumption has been gradually increasing for decades, driven by rising incomes and populations as well as productivity growth of meat production (Jones et al., 2013). Emerging markets have growing demand and purchasing power in the global red meat (beef and pork) market. Such escalated demands in international red meat and livestock markets provide meat exporters with substantial opportunities to expand the livestock and meat production. These opportunities also come however with several challenges. Food safety issues have increased both consumer and producer awareness of external effects associated with trade in agricultural products (Schlueter et al., 2009). Various sanitary and phytosanitary (SPS) measures have been widely applied by members of the World Trade Organization (WTO) with the stated purpose to ensure food safety. SPS barriers to trade are growing and become particularly contentious because the barriers have been used to protect favored producers from competition.

Despite the large body of studies analyzing the impacts of SPS measures on agricultural and food trade, little is known about the determinants of red meat international market or the impacts of SPS regulations on meat trade. This analysis contributes by providing useful insights into the trade determinants and the relationship between SPS regulations and red meat trade. The timing of this assessment is also key given worldwide increases in populist notions and associated reductions in support for free-market policies (Economist, 2017). The recent trade disputes between the USA and China on pork and other agricultural products also bring the uncertainties for the global red meat market. In this situation, unbiased assessment of trade factors becomes essential to collective understanding.

Research about SPS measures and meat trade determinants more broadly is essential given the current state of knowledge on trade barriers and growing economic importance of global trade. These SPS measures may be necessary for importers to improve the market benefits of consumer confidence in safer products by providing desired health and safety levels (Wilson and Anton, 2006). On the other hand, red meat trade can be adversely affected by SPS measures by increasing costs for both importers and exporters potentially. Combined, a substantial knowledge gap exists on the specific meat trade impacts of different regulatory instruments.

Although SPS standards and regulations raise the costs of foreign supplies relative to domestic production and conventionally act as “trade barriers” in the literature, the trade impacts of those regulations are not always negative (Anders and Caswell, 2009; Schlueter et al., 2009). Strict food safety restrictions on exports may push meat suppliers to implement higher inspection systems and update production facilities with advanced food safety standards, which increase producer efficiency and maintain consumer confidence in the long run. Under this condition, such safe and healthy efforts could behave like “catalysts” and result in positive trade impacts (Schlueter et al., 2009). It is the balance of these trade cost and catalyst effects that makes the estimations of SPS measures impacts controversial in empirical studies and worthy of focused assessment.

The overall objective of this study is to provide an ex post econometric examination of SPS measures and their influences on red meat trade. This research meets this objective using multiple gravity model approaches. First, we evaluate how different SPS regulations imposed to achieve certain SPS goals have diverse trade effects. By disaggregating SPS regulations into specific policy goals, we include finer measurements of effects of SPS regulations related to animal health, human health and maximum residue limits (MRLs) than previously provided in existing studies. This provides valuable information to industry leaders and policy makers that better aligns with the current, complex SPS environment. Second, the potential spillover effects across red meat trade are another focus in this study. It is hypothesized that imposition of SPS measures on pork (or beef) may impact aggregated beef (or pork) trade flows. We conduct a myriad of tests via multiple estimations and detect if corresponding spillover effects exist. Third, since econometric problems such as heteroscedasticity and the presence of zero usually are associated with gravity models, we concentrate on the Poisson pseudo maximum likelihood (PPML) estimators and compare the sensitivity of the related gravity equation parameters across different model quantification scenarios.

The rest of the paper is organized as follows. The next section outlines the background on SPS regulations and red meat trade. The subsequent section describes the specified gravity models and associated econometric issues. Then an overview is provided of the data used along with SPS category classification. Finally, results are presented along with implications and concluding remarks.

SPS background

The basic aims of the SPS is to maintain the sovereign right of WTO members to take SPS measures to protect human, animal or plant life or health. The SPS agreement allows participating countries to adopt their own SPS standards based on the risk assessment and justifications (Grant et al., 2015). In practice, SPS measures are increasingly important to agricultural trade among the list of non-tariff barriers. The imposition of SPS measures on exporters can increase production and trade costs in order to comply with the safety requirements and thus reduce export competitiveness (Otsuki et al., 2001; Disdier et al., 2008; Jayasinghe et al., 2010; Liu and Yue, 2011; Peterson et al., 2013; Grant et al., 2015), especially for the developing countries. Conversely, SPS standards also address externalities including imperfect information and may increase consumer confidence in safer products which hence stimulates consumer demand in the long run (Beghin and Bureau, 2001; Jaffee and Henson, 2004; Anders and Caswell, 2009). The cost of implementing more food safety standards may provide incentives for the modernization of export supply chains and such costs may be offset by benefits, which potentially create new forms of competitive advantage (Henson and Jaffee, 2008).

Although there is an extensive body of literature in agricultural economics related to the effects of SPS regulations on agricultural commodity trade (Grant et al., 2015; Peterson et al., 2013; Xiong and Beghin, 2011; Jayasinghe et al., 2010; Schlueter et al., 2009; Anders and Caswell 2009; Peterson and Orden, 2008; Disdier et al., 2008; Otsuki et al., 2001; Beghin and Bureau, 2001; Koo et al., 1994), only a few studies discussed the relationships between SPS regulation and red meat trade. Most existing studies narrowly focused on either a specific food safety issue like MRLs (Wilson et al., 2003) or an individual animal disease outbreak such as foot-and-mouth disease (FMD) (Yang and Saghaian, 2010). Koo et al. (1994) revised the gravity model for a single agricultural commodity effectively. Their study revealed that trade policies, meat production capacity in countries and long-term agreements are important in determining meat trade flows. However, beyond being dated their analysis has two limitations. First, issues with zero trade flows which may cause biased estimation and heteroscedasticity were not addressed. Second, country-level multilateral resistance terms which are captured by the importer and exporter fixed effects were not included. Schlueter et al. (2009) departed from the impact of specific events or diseases and estimated the SPS regulatory policies in meat trade for the 1996–2007 period. They disaggregated the SPS instruments into classes/groups based on the desired level of SPS health and analyzed the trade effects of different measures that are imposed in the meat sector. In this paper, we follow a similar process and further disaggregate the meat sector or compare potential spillover effects across species.

Gravity model and SPS regulations

The gravity model approach is often applied as an effective trade model for empirical trade analysis. Tinbergen (1962) first performed the gravity model on international trade independently. Then the gravity model has inspired many researchers to explain bilateral trade flows. The basic idea behind the gravity model of international trade is that the bilateral trade volumes from one country to another can be explained by factors that capture the potential of a country to export and import goods. In general, the gravity equation posits that bilateral trade can be represented by a function of trade factors such as importer’s demand, exporter’s supply and various bilateral trade costs. Through decades of development, the work of Anderson (1979), Bergstrand (1985), Helpman and Krugman (1985), Anderson and van Wincoop (2003) and Baldwin and Taglioni (2006) established a solid theoretical foundation for the gravity model application. Among a great number of agricultural trade studies, the research by Koo et al. (1994), Dascal et al. (2002), Peterson and Orden (2008), Schlueter et al. (2009), Villoria, Sun and Reed (2010), Yang and Saghaian (2010), Jayasinghe et al. (2010), Xiong and Beghin (2011), Peterson et al. (2013) and Grant et al. (2015) discuss and apply the gravity model on trade of agricultural commodities to provide evidence on the trade impacts of tariff and non-tariff measures.

Empirical issues of gravity model

In the process of model estimation, two econometric issues related to the gravity equation model and trade data are widely discussed in previous literature: sample selection bias and heteroscedasticity. A common feature of sample selection bias in bilateral trade data is that zero trade volumes are frequent across country pairs and products[1]. Consequently, the existence of zero trade observations creates problems for the use of the common log-linear form of the gravity equation. Furthermore, zeros cannot be simply omitted because it may delete important information on trade patterns and cause biased estimates and inconsistency. Heteroscedasticity is caused by the varying data levels because the data sample usually consists of trade volumes from different sources and countries with diverse productivities (Xiong and Chen, 2014). Therefore, the ordinary least squares (OLS) and non-linear least squares (NLS) estimators cannot be efficient, as they require the conditional variance to be constant. All estimators of logarithm form gravity models are generally inconsistent because of Jensen’s inequality (Silva and Tenreyro, 2006).

To accommodate these potential problems, a PPML estimator[2] which was originally introduced by Silva and Tenreyro (2006) has been widely applied in recent research (Schlueter et al., 2009; Sun and Reed, 2010; Jayasinghe et al., 2010; Xiong and Beghin, 2011; Peterson et al., 2013; Grant et al., 2015). The PPML estimator is relatively robust to heteroscedasticity and well behaved over other estimators including OLS, Tobit and NLS). Even when the dependent variable has a large proportion of zeros and the data suffers from over-dispersion, the PPML estimator still performs robustly in empirical studies (Silva and Tenreyro, 2011). Fally (2015) further emphasized that estimating gravity equations using PPML with fixed effects automatically satisfies the “adding-up” constraints and it is consistent with the introduction of multilateral resistance as in Anderson and van Wincoop (2003).

Model specification with PPML estimator[3]

Based on the modeling frameworks of Anderson and van Wincoop (2003) and Baldwin and Taglioni (2006), multiple product-level gravity models are conducted for estimating the effects of SPS measures on red meat trade. The associated gravity equation in this study with a PPML estimator is specified as follows:

(1) T i j t k = exp { β 0 k + β 1 k ln ( G D P i t ) + β 2 k ln ( I P r o d i t k ) + β 3 k ln ( E P r o d j t k ) + β 4 k D F t a i j t + ∑ n λ n k Ln ( S P S i j t k ) + π i j + π t } ε i j t ,

where “exp” refers to the exponential function; T i j t k indicates the trade value (in current US dollars) of red meat k (beef and pork) imported from exporter j to importer i in year t; GDP_it is the GDP per capita (in current US dollars) of importer i in year t; I P r o d i t k and E P r o d j t k are the total production of red meat k in importing country i and exporting country j in year t, respectively. DFta_ijt is a dummy variable for free trade agreements (FTA) between importer i and exporter j starting from year t; S P S i j t k represents a series of SPS measures with different arrangements imposed by importer i to exporter j on meat k in time t. Multiple SPS measures with different policy goals can be examined separately using this approach. π_ij and π_t are importer–exporter (country-pair) fixed effects and year fixed effects which represent the multilateral resistance terms following the suggestions of Anderson and van Wincoop (2003). The country-pair dummy π_ij absorbs all time-invariant determinants of bilateral trade costs such as common language and distance (Baldwin and Taglioni, 2006). ε_ijt is the error term.

Data description

The data for red meat trade values utilized in this study were obtained from the Global Trade Atlas (GTA) of Global Trade Information Services from 1997 through 2013. The GTA data set, which is commonly used by government officials and organizations (e.g. USDA, USITC and US Meat Export Federation), allows users to track imports and exports of products grouped by Harmonized System (HS) codes from the most general to the most detailed levels. HS four-digit products include HS 0201 (meat of bovine animals, fresh or chilled), HS 0202 (meat of bovine animals, frozen) and HS 0203 (meat of swine, fresh, chilled or frozen) have been considered to represent “red meat” in this study. Furthermore, HS 0201 and HS 0202 were merged to create a single beef series. In total, 15 countries/regions which account for over 80 percent of red meat trade worldwide (Argentina, Australia, Brazil, Canada, Chile, China, EU-15, Hong Kong, Japan, South Korea, Mexico, New Zealand, Russia, Taiwan and the USA) are included over the sample period in the current study[4]. To accurately reflect the real trade flows between country pairs, we select the annual trade values (import values) reported by importers[5] and include the zero trade flows.

The SPS data is extracted from the WTO SPS information Management System (SPS-IMS)[6]. We manually search and gather the information of SPS regulations on the red meat sector grouped by HS four-digit code. The variables of SPS measures in our study are arranged into aggregated SPS, SPS with policy goals (to protect animal health, human health and/or MRLs), SPS related to BSE (bovine spongiform encephalopathy) and SPS related to FMD. The effects of the SPS variables are estimated separately in our study which helps to understand the relationship between trade flows and SPS measures in more detail.

The additional variables in gravity equations are obtained from multiple sources. The income impact on consumer demand for red meat is captured by the GDP per capita expressed in current US dollars, which was collected from the World Bank Databank except for Taiwan. Data about Taiwan was obtained from the International Macroeconomics Data Set of USDA ERS. Annual production/supply quantities of beef and pork for both importers and exporters are extracted from the Food and Agriculture Organization of the United Nations Statistics Division (FAOSTAT) and included to capture production capacity. The dummy variable for bilateral or regional FTA applied in this study is based on those agreements notified to the WTO and obtained from the Regional Trade Agreements Information System[7]. Table I presents the summary statistics for selected variables of the data sets used.

Results and further discussions

This section reports the results of the estimated gravity models for beef and pork under different SPS quantification scenarios and trade data formats with the frequency method. Overall, Table II presents the results of eight estimated product-level gravity equations[8]. Columns (1) and (2) report the results of gravity equations with SPS measures included in an aggregated level for beef and pork. Columns (3) and (4) present the parameter estimates of models with SPS measures specific to three different policy goals. Models with BSE focused SPS measures are reported in Columns (5) and (6) while Columns (7) and (8) present model results when SPS measures with special focus on FMD. In each subsection of estimated models, robustness checks to extensive zeros and tests of misspecification are employed. Following the suggestion of Silva and Tenreyro (2006), all models and estimators are tested by the Ramsey’s regression equation specification error test (RESET), which helps to detect whether the gravity equations are correctly specified[9].

Model results

Table II shows the estimated beef and pork trade effects for the data set whose percentages of zeros are 47 and 49 percent, respectively. The results of RESET tests reject the hypothesis that the coefficients on the test variables are zero for both meat products with four SPS variable specifications, which suggest that the PPML estimators are appropriate in this case. The importer–exporter fixed effects are also included to capture multilateral resistance. The parameter estimates of the gravity equations under the four SPS specifications are similar and more broadly results are consistent with expectations. Increases in GDP per capita and production by exporters foster additional trade while domestic production of importers reduces trade. Comparing meats, larger effects of exporter production on beef and larger effects of GDP and domestic importer production on pork trade are identified. For example, a 1 percent increase in exporter production increases beef trade by about 2.3 percent and increases pork trade by 1.5 percent. Conversely, larger importer, domestic production reduces beef trade by about 0.7 percent and pork trade by over 0.9 percent. The impacts of aggregated SPS regulations (Columns (1) and (2)) are negative and significant on both beef and pork trade. A 1 percent increase is found to decrease beef trade by 0.07 percent and pork trade by 0.12 percent. Considering SPS regulations separately by policy goals reveals a finer set of trade impacts. Beef trade is reduced by the SPS measures focusing on animal health (Column 3) while pork trade is reduced by SPS maximum residue focused measures. When SPS measures specific to BSE or FMD are considered (Columns (5)–(8)) the only significant effect identified is a positive impact of BSE focused SPS measures on pork trade. This finding motivates the focused cross-species spillover assessment discussed later.

To summarize, Table II suggests that there is little variation in the effects of trade factors within beef and pork models no matter which SPS scenarios were applied. However, variations of effects are notable across commodities. For example, the extensive beef production of exporters has stronger positive effect on beef exports than the pork supply on pork exports. Also, pork trade seems to be more easily influenced by the change of GDP per capita than beef. This could indicate that pork trade is more sensitive toward personal income for importers. Within each individual model, the impact of GDP per capita is relatively large among all trade factors, suggesting that the personal incomes in destination countries are a crucial meat trade determinant The FTA is not a significant trading facilitator within these top meat trading countries. It may be the reason that some markets with huge meat demand (e.g. China) do not establish FTA with major exporters. Table II also suggests that SPS measures at aggregated levels on pork appear to be more trade restricting than for beef.

Further discussions: spillover effects

Across all models presented so far, BSE focused SPS measures are found to increase pork trade. This motivates direct assessment of cross-species spillover effects. Empirical models for detecting product-level spillover effects are developed from the previous gravity equations using the PPML estimator. Specifically, two separate gravity equations are estimated individually with additional regressors to capture effects from the competing industry. By including variables of beef (or pork) production along with the SPS measures on beef (or pork) products in the pork (or beef) model we can identify how beef (pork) industry adjustments impact pork (or beef) trade. A joint test is made to examine if issues of production and SPS in a competing industry affect trade[10].

Two model specifications have been identified. One of two gravity equations (e.g. pork) with PPML estimator is specified as follows:

(2) T i j t pork = exp { β 0 pork + β 1 pork ln ( G D P i t ) + β 2 pork ln ( I P r o d i t pork ) + β 3 pork ln ( E P r o d j t pork ) + β 4 pork ln ( D i s t i j ) + β 5 pork D F t a i j + β 6 pork D L a n i j + λ 1 pork L n ( S P S i j t pork ) + β 7 pork ln ( I P r o d i t beef ) + β 8 pork ln ( E P r o d j t beef ) + λ 2 pork L n ( S P S i j t beef ) + π i + π j + π t } ε i j t ,

where the coefficients of β 7 pork , β 8 pork and λ 2 pork capture spillover effects individually and jointly from the beef industry on pork trade. DLan_ij indicates the common language dummy between country i and j and Dist_ij represents the distance. A parallel beef trade model considering pork industry impacts is also specified.

Table III indicates model results with spillover effects for beef trade. The estimated coefficients from the beef gravity equation indicate pork-related SPS measures in an aggregated level have no significant effects on beef trade. However, the coefficients indicate that 1 percent increase in importer pork production reduces beef import by 1.2 percent, which suggest that domestically produced pork is a substitute to imported beef. This has clear implications on future trade expectations. For instance, consider renewed discussions of US beef entering China – this finding suggests some moderation in projected beef trade given China’s substantial pork production. Furthermore, importer pork production has a larger negative impact than their beef production on beef trade flows. The joint test of pork-related factors (pork production of importers and exporters and pork-related SPS measures) suggests that factors in the pork industry collectively influence beef trade.

Table IV shows parallel spillover model results for pork trade. The individual factors of exporter beef production and beef-related SPS measures are insignificant in the pork gravity equation. Importers’ domestic beef production has a positive spillover effect on pork trade, which suggests that domestically produced beef is a complement to imported pork. However, the joint test of beef-related factors (beef production of importers and exporters and beef-related SPS measures) suggests beef products do not collectively impact pork trade. Combined this suggests beef industry stakeholders have a burden of monitoring pork industry conditions to understand beef trade that is not reciprocally shared.

Conclusion and implication

As the international meat trade is growing rapidly, the trade barriers in export of meat products need to be addressed. SPS measures imposed by importers are an important example of non-tariff measures faced by red meat exporters. Using a series of product-level gravity models, this study analyzes the trade effects of diverse SPS measures that are imposed in the meat industry to further discover the determinants of global red meat trade. Red meat products were classified by HS four-digit code and the trade volume of top 15 meat trade countries were used with different portion of zero trade flows to further assess performance of the PPML estimator. Econometrically, the empirical results largely support the theory of Silva and Tenreyro (2006, 2011) and indicate that PPML estimator is generally well behaved.

The current study sheds important light on the determinants of red meat trade. Similar to previous agricultural trade studies, the economic conditions of destination countries and production capability of suppliers are key to determining trade values (Grant et al., 2015; Peterson et al., 2013; Xiong and Beghin, 2011; Jayasinghe et al., 2010; Schlueter et al., 2009; Anders and Caswell 2009; Peterson and Orden, 2008; Disdier et al., 2008; Otsuki et al., 2001; Beghin and Bureau, 2001). Factors including personal income and exporters’ meat supply are identified as trade facilitators. Since the restrictiveness of SPS measures vary across beef and pork sectors, maintaining commodity-specific SPS measures is essential for accurate assessment of trade determinants. Diverse scenarios were considered with classification of SPS impacts being based on specific topic areas and policy goals. Among multiple product-level gravity equations, the sum of all counts of SPS measures for a particular country present significant and negative effects on both beef and pork trade. The further disaggregated SPS measures with policy goals shows that there are specific SPS measures that have a substantial positive impact and others with a significant negative impact on either frozen beef or pork. These effects can offset each other within a class. All combined, the empirical results are mostly consistent with existing literature (Schlueter et al., 2009), in which SPS appear to be more of a trade barrier than a catalyst to agricultural trade, including read meat.

Detecting spillover effects across and industries on trade is another contribution of this study. Pork-related SPS measures do not have significant effects (or spillover) on beef while importer pork production decreases beef trade flow. Furthermore, the joint test of pork-related factors suggests that pork industry factors significantly impact beef trade. Accordingly, the beef industry should monitor pork industry developments to better assess their own industry’s trade situation and future prospects.

While results presented here are new, timely and informative, additional future considerations are encouraged. Future research could extend the SPS data from WTO SPS-IMS and search for additional, further detailed information. The four-digit HS code data could also be further disaggregated into six-digit HS code level data to add analysis on more specific products. Given the ongoing expectation of growing global meat demand and the current environment questioning the benefits of free-market, open-trade policies, any future research examining additional or more specified meat can enhance the understanding of meat trade effects and by extension improve policy surrounding SPS concerns.

Table I

Summary statistics of selected variables

Data types	Beef (HS 0201 and HS 0202)		Pork (HS 0203)
Variables	Mean	SD	Mean	SD
Trade values	465.82	1,687.24	413.78	1,635.2
GDP per capita i	21,852.4	14,693.4	21,801.57	14,567.36
Importer production	31.23	34.96	72.71	119.7
Exporter production	19.48	32.12	38.85	86.61
FTA dummy	0.17	0.38	0.13	0.34
Aggregated SPS’s	1.18	2.43	0.86	2.21
SPS for policy goals
SPS (animal health)	0.21	0.59	0.07	0.27
SPS (human health)	1.09	2.37	0.81	2.18
SPS (residue)	0.79	2.26	0.73	2.15
SPS for diseases
SPS (BSE)	0.15	0.49	0.02	0.14
SPS (FMD)	0.03	0.19	0.02	0.17
No. of observation	4,267		3,587
Percentage of zero	47		49

Notes: Trade values are expressed as import values in $100,000. GDP per capita of importers is expressed in current US dollars. Quantities of importer and exporter production are in 100,000 metric tons. It is possible that the mean of importers’ production is relatively larger than the mean of exporters’ production. This is because we have fifteen countries as importers and some importers are both larger meat producers and importers

Table II

Parameter estimates of gravity equations – the frequency method

SPS specifications	Aggregated SPS		SPS policy goals		SPS BSE		SPS FMD
Column	(1)	(2)	(3)	(4)	(5)	(6)	(7)	(8)
Variable	Beef (HS 0201 and HS 0202)	Pork (HS 0203)	Beef (HS 0201 and HS 0202)	Pork (HS 0203)	Beef (HS 0201 and HS 0202)	Pork (HS 0203)	Beef (HS 0201 and HS 0202)	Pork (HS 0203)
GDP per capita	0.856*** (0.132)	1.106*** (0.121)	0.915*** (0.134)	1.050*** (0.122)	0.900*** (0.135)	1.250*** (0.112)	0.893*** (0.135)	1.245*** (0.112)
Importer production	−0.666*** (0.205)	−0.901*** (0.226)	−0.697*** (0.204)	−0.938*** (0.219)	−0.675*** (0.207)	−0.982*** (0.223)	−0.633*** (0.206)	−0.963*** (0.223)
Exporter production	2.280*** (0.200)	1.492*** (0.298)	2.274*** (0.201)	1.463*** (0.299)	2.287*** (0.200)	1.545*** (0.296)	2.277*** (0.199)	1.531*** (0.297)
Free trade agreement	−0.153 (0.109)	0.019 (0.093)	−0.132 (0.109)	0.0171 (0.095)	−0.123 (0.110)	−0.019 (0.093)	−0.107 (0.108)	−0.013 (0.093)
Aggregated SPS	−0.065* (0.0361)	−0.123*** (0.032)
SPS_Animal Health			−0.149* (0.080)	−0.045 (0.087)
SPS_Human Health			−0.035 (0.066)	0.105 (0.108)
SPS_Maximum Residue			0.013 (0.078)	−0.271** (0.116)
SPS_BSE					−0.093 (0.0836)	0.340*** (0.076)
SPS_FMD							0.007 (0.181)	0.124 (0.089)
Constant	−14.90*** (1.759)	−8.100*** (1.371)	−15.27*** (1.762)	−7.512*** (1.384)	−15.27*** (1.782)	−9.345*** (1.338)	−15.32*** (1.781)	−9.289*** (1.341)
R2	0.866	0.950	0.867	0.951	0.866	0.948	0.866	0.947
Country pair fixed effects	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes
Year fixed effects	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes
Ramsey RESET	Accept	Accept	Accept	Accept	Accept	Accept	Accept	Accept
Observations	4,267	3,587	4,267	3,587	4,267	3,587	4,267	3,587
Percentage of zeros	47	49	47	49	47	49	47	49

Notes: Yes denotes that the fixed effects have been included in estimation. *,**,***Significant at 1, 5 and 10 percent levels, respectively

Table III

Estimated coefficients of beef gravity equation with spillover effects (PPML)

Variables	Coeff.	SE
Log GDP per capita	1.329***	(0.232)
Log Importer Production_Beef	−0.605**	(0.302)
Log Exporter Production_Beef	2.598***	(0.311)
Log Importer Production_Pork	−1.150***	(0.449)
Log Exporter Production_Pork	−0.166	(0.284)
Log Aggregate SPS_Beef	−0.065	(0.125)
Log Aggregate SPS_Pork	0.035	(0.139)
Log Distance	−1.051***	(0.055)
Free trade agreement	1.045***	(0.119)
Common language	−0.263***	(0.130)
Joint test of pork-related factors (p-value)	0.038
Fixed effects	Yes
No. of observations/% of zeros	12,257/82.21

Notes: *,**,***Significant at 1, 5 and 10 percent levels, respectively

Table IV

Estimated coefficients of pork gravity equation with spillover effects (PPML)

Variables	Coeff.	SE
Log GDP per capita	1.109***	(0.168)
Log Importer Production_Pork	−0.825**	(0.406)
Log Exporter Production_Pork	0.719	(0.483)
Log Importer Production_Beef	0.418*	(0.231)
Log Exporter Production_Beef	0.425	(0.410)
Log Aggregate SPS_Pork	−0.123	(0.097)
Log Aggregate SPS_Beef	−0.020	(0.082)
Log Distance	−0.671***	(0.078)
Free trade agreement	1.362***	(0.144)
Common language	0.774***	(0.172)
Joint test of beef-related factors (p-value)	0.172
Fixed effects	Yes
No. of observations/% of zeros	12,257/85.14

Notes: *,**,***Significant at 1, 5 and 10 percent levels, respectively

Table AI

List of bilateral and regional FTA and the effective dates

ASEAN – Australia – New Zealand	January 1, 2010	EU – Central America	August 1, 2013
ASEAN – China	January 1, 2005	EU – Chile	February 1, 2003
ASEAN – India	January 1, 2010	EU – South Korea	July 1, 2011
ASEAN – Japan	December 1, 2008	EU – Mexico	July 1, 2000
ASEAN – South Korea	January 1, 2010	EU – Norway	July 1, 1973
Australia – Chile	March 6, 2009	EU – Serbia	February 1, 2010
Australia – New Zealand (ANZCERTA)	January 1, 1983	Hong Kong, China – New Zealand	January 1, 2011
Canada – Chile	July 5, 1997	India – Japan	August 1, 2011
Canada – Colombia	August 15, 2011	South Korea – Chile	April 1, 2004
Canada – Costa Rica	November 1, 2002	South Korea – USA	March 15, 2012
Chile – China	October 1, 2006	New Zealand – Chinese Taipei	December 1, 2013
Chile – Colombia	May 8, 2009	Nicaragua – Chinese Taipei	January 1, 2008
Chile – Costa Rica (Chile – Central America)	February 15, 2002	North American Free Trade Agreement (NAFTA)	January 1, 1994
Chile – Japan	September 3, 2007	Russian Federation – Belarus	April 20, 1993
Chile – Mexico	August 1, 1999	Russian Federation – Kazakhstan	June 7, 1993
China – Costa Rica	August 1, 2011	Russian Federation – Republic of Moldova	March 30, 1993
China – Hong Kong, China	June 29, 2003	Russian Federation – Serbia	June 3, 2006
China – New Zealand	October 1, 2008	Thailand – Australia	January 1, 2005
Colombia – Mexico	January 1, 1995	Thailand – New Zealand	July 1, 2005
EFTA – Canada	July 1, 2009	Ukraine – Russian Federation	February 21, 1994
EFTA – Chile	December 1, 2004	USA – Australia	January 1, 2005
EFTA – Colombia	July 1, 2011	USA – Chile	January 1, 2004
EFTA – South Korea	September 1, 2006	USA – Colombia	May 15, 2012

Notes

1. One reason is that some country pairs simply do not trade in certain products. This problem is more likely to occur when small countries are taken into consideration or when less aggregated products are examined (Silva and Tenreyro, 2006). The other reason is potential exporters may find it unprofitable given costs of trade in some specific markets (Helpman et al., 2008). Rounding errors may also be an additional source of zeros if trade flow is measured in big units (like thousands of dollars) (Silva and Tenreyro, 2006).

2. The estimations of PPML do not actually require the data to follow a Poisson distribution, which is the reason why the estimation is a “pseudo maximum likelihood” (Silva and Tenreyro, 2006).

3. The advantages and disadvantages of other estimators (OLS and Heckman Selection) for the gravity equation have also been summarized in the Table AI. The estimation results of OLS and Heckman Selection (Heckman, 1979) are available by request.

4. Since the major destinations of Indian beef are the developing regions like Southeast Asia, the Middle East and North Africa, these 15 countries do not have significant demands for beef from other exporters.

5. Because of asymmetric information following issues such as meat smuggling, transportation loss and perhaps political motives to over-state exports, the import values reported by importers and the export values reported by exporters do not exactly match.

6. SPS information Management System Data is available online at http://spsims.wto.org/

7. See Table AI for the list of free trade agreements. The free trade agreement dummy variables in this study do not distinguish specific commodities.

8. Eight gravity equation=1 estimator (PPML)×4 SPS specification approaches (aggregated SPS, SPS with Policy Goals, SPS related to BSE and SPS related to FMD)×2 meat products (beef and pork).

9. The RESET tests the null hypothesis that additional regressors (xβ)² and (xβ)³ cannot help to further explain the dependent variables by running auxiliary regressions (Silva and Tenreyro, 2006; Schlueter et al., 2009).

10. One limitation of this “spillover” specification is that it cannot detect those effects across particular country pairs. The importer and exporter fixed effects controls for the aggregated impacts across country pairs which say nothing about specific spillover effects among countries. For instance, this model cannot specifically explain how Australian red meat trade values change when Asian countries imposed SPS measures on US red meat exports. But it will say how global pork trade is impacted by beef SPS measures in an aggregated level.