THE GRAVITY OF SPS RISK
Lesley Ahmed
Peter Herman
Caroline Peters
David Riker
ECONOMICS WORKING PAPER SERIES
Working Paper 2017–11–A
500 E Street SW
Washington, DC 20436
November 2017

Oﬃce of Economics working papers are the result of ongoing professional research of USITC Staﬀ and are solely meant to represent the opinions and professional research of individual authors. These papers are not meant to represent in any way the views of the U.S. International Trade Commission or any of its individual Commissioners. Working papers are circulated to promote the active exchange of ideas between USITC Staﬀ and recognized experts outside the USITC and to promote professional development of Oﬃce Staﬀ by encouraging outside professional critique of staﬀ research.

The Gravity Of SPS Risk
Lesley Ahmed, Peter Herman, Caroline Peters, and David Riker
Oﬃce of Economics Working Paper 2017–11–A

Abstract

As the frequency and magnitude of sanitary and phytosanitary (SPS) issues continues to grow worldwide, understanding the impact that these issues have on agricultural trade is becoming increasingly important. This paper uses a novel, product-centric approach to shed light on this topic. Using characteristics of products such as freshness or preparation method, we classify agricultural products as exhibiting either high or low SPS risk based on their relative likelihood of posing a danger to the human, animal, or plant health concerns that SPS measures address. This classiﬁcation is then used within a collection of gravity trade models to estimate diﬀerences in trade patterns between high and low risk goods. We ﬁnd that high risk goods exhibit higher trade costs based on several measures and that both types of goods face diﬀering impacts from trade agreements providing for customs improvements or regulatory harmonization. Meanwhile, neither category of good experiences signiﬁcant eﬀects from trade facilitation eﬀorts that either primarily lower tariﬀs or address indirectly-related aspects of trade such as intellectual property or services.

JEL Classiﬁcation: F14, Q17

Lesley Ahmed
Oﬃce of Industries
Peter Herman (Corresponding Author)
Oﬃce of Economics
peter.herman@usitc.gov
Caroline Peters
Oﬃce of Economics
David Riker
Oﬃce of Economics

### 1 Introduction

In recent years, non-tariﬀ measures (NTMs) have gained considerable attention in international trade. As tariﬀs worldwide have largely disappeared, NTMs have grown both in the number of measures applied by countries as well as in relative eﬀect as they become, in many cases, the most considerable barrier to trade faced by ﬁrms. This is especially true in the case of sanitary and phytosanitary (SPS) measures, which countries apply to ensure food safety; protect humans, animals, and plants from contaminants, diseases, disease-causing organisms, and pests; and to prevent damage from the entry and spread of pests. The number of SPS measures recorded by the World Trade Organization (2017) has grown by nearly 9.4 percent on average per year from 189 measures in 1995 to 936 measures in 2016.

Given this growth in SPS measures, understanding the eﬀects of these measures on trade it is increasingly important. In this paper, we analyze the relationship between SPS issues and bilateral trade. Using a gravity modeling framework, we show that agricultural products that are especially sensitive to SPS risks exhibit diﬀerent trading patterns than less sensitive agricultural products. High sensitivity goods exhibit higher trade costs with respect to several measures between trading partners. Similarly, high sensitivity products experience statistically diﬀerent impacts from trade facilitating preferential trade agreements (PTA), particularly those that include provisions for regulatory harmonization and customs improvements. Meanwhile, other eﬀorts to liberalize trade, such as those focusing primarily on tariﬀ reductions or indirectly-related aspects of trade such as services or intellectual property provisions, exhibit limited impact on either high or low risk agricultural products, in general or relative to one another.

This paper is a new entry into the growing literature seeking to quantify the eﬀects of non-tariﬀ measures. Much of this literature has relied on gravity models to assess the extent to which NTMs may increase or decrease the cost or level of trade.1 For example, Fontagné et al. (2011) infer the restrictiveness of NTMs using importer ﬁxed eﬀects to compare the relative openness of countries, which they use to calculate a tariﬀ-rate equivalent for each country. Kee et al. (2009) use a slightly diﬀerent approach, employing data on the incidence of certain types of NTMs to identify the trade costs associated with the measures. For a more in-depth survey of diﬀerent types of quantiﬁcation methodologies, see Abbyad and Herman (2017).

Like NTMs more generally, SPS measures have the potential to be either trade-facilitating or trade-diminishing depending upon the nature of the measures themselves and the breadth of their implementation. Nimenya et al. (2012) ﬁnd that SPS measures can be trade catalysts when they reduce information asymmetries in the market, allowing easier comparison of quality attributes across markets subject to diﬀerent public and private SPS requirements. Doing so increases the impact of food product origin as a factor in product diﬀerentiation. Drogué and DeMaria (2012) and Winchester et al. (2012) have found the harmonization of SPS standards to be trade-increasing, although the eﬀect on exports is not uniformly positive as some countries lose their diﬀerentiated edge when standards converge. Crivelli and Groeschl (2016) ﬁnd that if exporters are able to overcome the ﬁxed costs associated with compliance, SPS measures are trade increasing as consumers are reassured of the quality of foreign products.

On the other hand, SPS measures can act as a trade deterrent when they diverge from the general WTO principles of nondiscrimination, scientiﬁc evidence, risk assessment, and least-trade-restrictive alternatives. Exports from developing countries to high-income countries are the most negatively impacted by the implementation of SPS measures, as the costs of compliance are relatively steeper in countries with weaker institutions and less developed agrofood supply chains (see, for example, Disdier et al. (2008), Li and Beghin (2011), Melo et al. (2014), and Henson and Jaﬀee (2008)). With regards to the intensive margin, the multiple sets of standards imposed by various regulatory bodies - large private sector actors in the agrofood supply chain, public sector actors in destination markets, and governing authorities within multilateral organizations - present barriers to market entry for new exporters and make full harmonization of SPS measures diﬃcult overall, as identiﬁed by Henson and Jaﬀee (2008). Moenius (2006) and Tothova and Oehmke (2008) argue that the trade-diminishing impact of SPS measures can also work at the extensive margin, with harmonization eﬀorts restricting the number of varieties available overall, limiting consumer choice and welfare. Work by Fontagné et al. (2015) using ﬁrm level data provides supporting evidence that both the intensive and extensive margins of trade are negatively aﬀected by the presence of SPS measures and that this eﬀect is more severe for smaller ﬁrms.

Work by Nimenya et al. (2012), Rickard and Lei (2011), and Calvin and Krissoﬀ (1998) has revealed that that the magnitude of welfare gains from removing SPS measures depend largely on the commodity being traded and whether the risk that measures are protecting against is transferable. Calvin and Krissoﬀ (1998) in particular, ﬁnd that unlike with tariﬀs, policymakers must consider both scientiﬁc and economic consequences with the imposition of new SPS measures, weighing the decrease in food safety risk against the potential losses to consumers and producers.

The work presented here builds oﬀ of this research by studying the relationship between bilateral trade determinants and SPS concerns. Unlike most of this research that uses the incidence of SPS measures to identify trade impacts, we take a product-centric approach that identiﬁes goods that are most prone to the types of issues that SPS measures attempt to mitigate. We divide agricultural product categories into two groups depending on whether the product category faces high or low SPS risks. High risk goods are those that are especially susceptible to the health and disease risks covered by SPS measures. Using this grouping, we analyze the trade implications for each group using gravity models to infer diﬀerences in trading behavior. A key beneﬁt of this alternative, product-centric approach is that our results are more closely connected to the health and safety aspects of SPS policies because there is no ex-ante need to determine if a measure is ineﬃciently distortionary as identiﬁcation is not directly based on the policies in place.

The paper proceeds as follows. Section 2 describes our approach to SPS risk and the categorization of agricultural goods. Section 3 describes the data and gravity methodology used for the analysis and presents our ﬁndings. Section 4 concludes.

### 2 SPS Risk

Some products pose greater risk to human, plant, and animal health than other products because of their inherent characteristics and tendency to carry pests, diseases, or organisms. Similarly, the nature of how a product is consumed or used can make it susceptible to SPS risk, such as whether it is for direct human consumption or undergoes heat treatment before it is consumed. In this paper, we deﬁne SPS risk based on whether a product is inherently sensitive to these concerns. Speciﬁcally, a product is considered high SPS risk if it is:

1.
a living plant or animal, hence with greater risk of spreading diseases or pests;
2.
a meat product—whether fresh, frozen or processed—because of risk of spreading human and animal diseases that are not mediated by processing (e.g. foot and mouth disease or bovine spongiform encephalopathy); or
3.
fresh or perishable and intended for human consumption, thereby exhibiting a greater risk to human health.

A few examples of high SPS risk products are dairy products, beef, live plants, and fresh cut ﬂowers. A few examples of lower SPS risk products are vegetable oils, pasta, wool, and dried vegetables. A full list of the HS 4-digit codes used in the analysis and their respective risk categorization can be found in table 5. Of the 210 agricultural product categories, 56 are classiﬁed as high risk; the remaining are considered low risk.

Traditionally, researchers examining the trade eﬀects of SPS measures and technical barriers to trade (TBT) categorize products based on the number of SPS measures or other NTMs they face. These categorizations, while diﬀerent from our inherent risk classiﬁcation, unsurprisingly feature some considerable similarities because products that pose greater SPS risk tend to have more SPS measures applied by importing countries. Prior research on this topic has employed several diﬀerent strategies to identify products subject to SPS measures. Disdier et al. (2008) summarize some common measures to identify potential barriers, using a simple dummy and frequency index along with calculated AVEs to estimate their eﬀect. Fontagné et al. (2005) examine the number countries that had made WTO SPS notiﬁcations for products at the HS 4-digit level. Not surprisingly, the products they label as “sensitive,” with at least 40 countries notifying SPS measures to the WTO, are similar to the group of products we categorize as high risk—including meat products, milk products, ﬂowers, and fresh fruits and vegetables. The products not labeled as “sensitive” exclude many processed food products and are also similar to our SPS risk categorization. Grant and Arita (2017), who examine the incidence and length of speciﬁc trade concerns raised at the WTO SPS committee meetings, ﬁnd the highest incidence of SPS speciﬁc trade concerns were in meat and edible oﬀal, fresh fruits and nuts, live animals, dairy products, and edible vegetables, roots and tubers. They ﬁnd no SPS speciﬁc trade concerns for gums, resins and vegetable saps, preparation of cereals, ﬂour, starches, and pastry; and tobacco and tobacco products. These, again, are in line with our classiﬁcation.

### 3 Gravity Analysis

#### 3.1 Data Description

In addition to the SPS sensitivity data that we produced, several other data sources were combined in order to create the gravity dataset used in the analysis.

Trade data was downloaded from COMTRADE.2 We utilize a cross section consisting of 55 countries for the year 2015. For each country pair, the data includes all reported trade ﬂows belonging to 210 HS2007, 4-digit codes, representing all agriculture products. Any codes not exhibiting positive trade ﬂows between a given country pair is assumed to be an untraded sector and is included as a zero so that the panel is square. As described above, each of these products was designated as being high or low SPS risk using a dummy variable taking the value of one if a product is deemed to be high risk. Recall that a high risk product or sector is one that, because of its inherent nature or general use, poses greater risk to human, animal, and/or plant health. Of the 210 HS 4-digit categories considered, 56 are identiﬁed as having high SPS risk.

#### 3.2 Gravity Models

The objective of this research is to identify the ways in which sensitivity to SPS risks impacts aspects of trade. Given its strength in quantifying the bilateral determinants of trade, the gravity model represents an ideal tool for doing so. To identify these SPS inﬂuences, we employ several gravity speciﬁcations that highlight diﬀerences in trade patterns between high and low SPS risk agriculture products. These speciﬁcations thoroughly explore these relationships by testing the robustness of the identiﬁed diﬀerences with respect to both aggregation and several ﬁxed eﬀect strategies. In total, four speciﬁcations were considered.

Each of the speciﬁcations was estimated using a Poisson Pseudo Maximum Likelihood estimator.6 PPML procedures have become standard in gravity research due to their ability to incorporate zero-value trade ﬂows and superior treatment of heteroskedasticity.

The ﬁrst speciﬁcation follows traditional methods employed in gravity modeling in which trade ﬂows are aggregated at the country pair level. That is, a single observation reﬂects aggregate trade between two partners in the agriculture products considered. In this case, we diﬀerentiate between high and low sensitivity goods by constructing two aggregations. One aggregation includes only those goods listed as being high risk, the other includes only low risk. Once aggregated according to risk, the following gravity speciﬁcation was estimated:

 ${X}_{ij}=\beta {z}_{ij}+{I}_{i}+{I}_{j}$ (i)

Indexes $i$ and $j$ denote exporter and importer respectively. The variable ${X}_{ij}$ denotes the trade value of exports from $j$ to $i$ while ${z}_{ij}$ denotes the following collection of gravity variables: contiguity, common language, distance, colony, and PTA. Identiﬁcation of diﬀerences between high and low risk goods is determined by comparing the respective coeﬃcients for the trade determinants.

The remaining speciﬁcations utilize disaggregated data in an eﬀort to better identify the nuances present at the product or sector level. For this purpose, we consider three diﬀerent speciﬁcations. The ﬁrst of these speciﬁcations does not attempt to control for sector-level variation, including only importer and exporter ﬁxed eﬀects. The second speciﬁcation includes an additional set of product level ﬁxed eﬀects corresponding to each of the HS 4-digit codes included in our sample. Speciﬁcally, these two speciﬁcations take the following forms, respectively:

 ${X}_{ijs}=\beta {z}_{ij}+{\gamma }_{0}sps\text{_}ris{k}_{s}+\gamma \left(sps\text{_}ris{k}_{s}\ast {z}_{ij}\right)+{I}_{i}+{I}_{j}$ (ii)
 ${X}_{ijs}=\beta {z}_{ij}+{\gamma }_{0}sps\text{_}ris{k}_{s}+\gamma \left(sps\text{_}ris{k}_{s}\ast {z}_{ij}\right)+{I}_{i}+{I}_{j}+{I}_{s}$ (iii)

The additional index $s$ denotes the product and $sps\text{_}ris{k}_{s}$ denotes the dummy variable reﬂecting SPS risk for product $s$. In these two cases, the eﬀects of SPS sensitivity are identiﬁed through the inclusion of the dummy for SPS risk that takes the value of one if a product is high risk and its interaction with the other bilateral determinants.

The fourth speciﬁcation controls for importer-sector and exporter-sector variation by estimating a separate model for each of the products considered. By estimating each $s$ separately, the ﬁxed eﬀects are allowed to vary across sectors.7 Because products are estimated separately, the identiﬁcation of diﬀerences between high and low risk products relies on non-parametric methods. Speciﬁcally, we report Kolmogorov-Smirnov statistics for each variable that determines if the distribution of coeﬃcients, conditional on risk type, diﬀers in a statistically signiﬁcant way.

#### 3.3 Results

##### 3.3.1 Baseline Gravity Variables

The results for speciﬁcations (i)-(iii) are presented in table 1. A summary of the results for speciﬁcation (iv) over all sectors is presented in table 2. Additionally, kernel-density plots of the estimated coeﬃcients for each sector are provided in ﬁgure 1.

Table 1: Gravity estimates for speciﬁcation (i)-(iii)
 (i) (i) (ii) (iii) (High SPS Risk) (Low SPS Risk) Contiguity 0.452*** (0.120) 0.349*** (0.102) 0.394*** (0.094) 0.394*** (0.094) Common Language 0.442*** (0.108) 0.187 (0.111) 0.200 (0.104) 0.200 (0.104) Log Distance -0.868*** (0.066) -0.821*** (0.054) -0.798*** (0.052) -0.798*** (0.052) Colony -0.045 (0.160) 0.239 (0.131) 0.193 (0.145) 0.193 (0.145) PTA 0.767*** (0.105) 0.230** (0.084) 0.260** (0.086) 0.260** (0.086) SPS Risk 0.236 (0.522) 5.792*** (0.697) Contiguity * SPS Risk -0.049 (0.150) -0.049 (0.150) Language * SPS Risk 0.120 (0.128) 0.120 (0.128) Distance * SPS Risk -0.056 (0.059) -0.056 (0.059) Colony * SPS Risk -0.044 (0.203) -0.044 (0.203) PTA * SPS Risk 0.547*** (0.142) 0.547*** (0.142) Constant 5.065*** (0.837) 8.797*** (0.562) 3.295*** (0.550) -2.691*** (0.649) Importer, Exporter F.E. yes yes yes yes Sector F.E. no no no yes N 2886 2886 606060 606060

Standard errors clustered at the country pair level in parentheses. *** $p<0.01$, ** $p<0.05$, * $p<0.1$.

Table 2: Summary of the estimation results for individual sectors (iv)
 Variable Low Risk High Risk K-S test stat. K-S p-value Mean $\beta$ S.D. Mean $\beta$ S.D. Distance (log) -1.223 (0.855) -1.596 (1.451) 0.274 0.003 Contiguity 0.332 (0.908) 0.244 (0.797) 0.115 0.605 Common Language 0.205 (0.827) 0.554 (0.835) 0.164 0.197 Colony -0.101 (1.303) 0.131 (0.638) 0.170 0.164 PTA 0.518 (1.296) 0.972 (1.170) 0.305 0.001

In each speciﬁcation, the results are generally consistent with previous literature. As is typically expected, contiguity, common language, and PTAs are trade facilitating while geographic distance is trade deterring. With regards to the relationships between trade determinants and SPS risk, we ﬁnd several consistent results. First, when accounting for sector-speciﬁc ﬁxed eﬀects in speciﬁcation (iii), we ﬁnd that high SPS risk goods trade more on average than low SPS risk goods. There are several possible explanations for this trend. By their nature, high risk goods tend to be fresh and seasonal, relying to a higher degree on seasonal trade for year-round availability than other agricultural products. Similarly, the fresh nature of these goods likely implies that they are generally of a higher quality and more prone to trade than their preserved or otherwise prepared, low risk counterparts in the vein of the so-called “Washington Apples” eﬀect. Finally, the SPS risks of the goods may themselves be a reﬂection of the fact that they trade more; the high risk goods may be characterized as high risk because of past SPS incidences encountered during trade.

Second, speciﬁcation (iv) ﬁnds that high risk goods are particularly sensitive to distance. Increased distance, which is often interpreted as being representative of transport time and costs, are especially impactful to high risk goods. This is likely a reﬂection of the relatively greater risk of rot and and other quality degradations inherent in the higher SPS risk products.

Finally, we ﬁnd that under all four speciﬁcations, PTAs are signiﬁcantly more inﬂuential in increasing trade among high risk goods. This ﬁnding suggests that policies put in place by countries that restrict trade are especially onerous to goods that face higher sensitivity to SPS issues. Because the existence of a PTA between countries is a rather blunt measure of trade facilitation, there are several possible explanations for why these agreements have a stronger inﬂuence on high risk goods. Because of the added complexity of trading high risk goods, eﬀorts to reduce barriers to trade likely aﬀect these goods more signiﬁcantly because there is a larger potential gain from doing so. Additionally, many trade agreements—particularly in recent years—focus increasingly on NTMs, which will aﬀect SPS sensitive goods more signiﬁcantly.8 An outcome of this trend is a general reduction in the burden of NTMs, a mitigation on the delays faced at borders, and an improvement in the eﬃciency of achieving the desired SPS safety measures.

##### 3.3.2 Expanded Trade Agreement Variables

The gravity results corresponding to this second collection of variables are presented in table 3 for speciﬁcations (i)-(iii). Summary statistics for the results of the individual sector estimates of speciﬁcation (iv) are presented in table 4 and as kernel density plots in ﬁgures 2 and 3.9 As before, the standard gravity components exhibit the expected eﬀects. Of the added PTA variables, several exhibit interesting relationships with agricultural goods in general and high or low risk goods in particular.

Table 3: Gravity estimates for speciﬁcation (i)-(iii) with added PTA variables
 (i) (i) (ii) (iii) (High SPS Risk) (Low SPS Risk) Contiguity 0.494*** (0.113) 0.338*** (0.098) 0.393*** (0.092) 0.393*** (0.092) Common Language 0.436*** (0.107) 0.261* (0.107) 0.221* (0.104) 0.221* (0.104) Log Distance -0.770*** (0.064) -0.710*** (0.055) -0.753*** (0.056) -0.753*** (0.056) Colony 0.062 (0.148) 0.373** (0.122) 0.335* (0.134) 0.335* (0.134) Agricultural Provisions 0.533 (0.470) -0.633 (0.359) -0.478 (0.292) -0.478 (0.292) Customs Provisions 0.110 (0.232) 0.910** (0.329) 0.664** (0.230) 0.664** (0.230) SPS Provisions 0.387* (0.193) -0.194 (0.201) -0.073 (0.155) -0.073 (0.155) TRIPS 0.096 (0.168) -0.010 (0.190) -0.045 (0.156) -0.045 (0.156) GATS -0.724** (0.230) 0.376 (0.225) 0.269 (0.214) 0.269 (0.214) Customs Union 0.901* (0.377) 0.493 (0.262) 0.475 (0.250) 0.475 (0.250) EIA 0.096 (0.211) 0.027 (0.218) -0.037 (0.195) -0.037 (0.195) FTA 0.261 (0.363) -0.152 (0.207) -0.038 (0.189) -0.038 (0.189) High SPS Risk -1.575* (0.667) 3.981*** (0.780) Contiguity * High Risk 0.011 (0.145) 0.011 (0.145) Language * High Risk 0.244 (0.131) 0.244 (0.131) Distance * High Risk 0.138 (0.077) 0.138 (0.077) Colony * High Risk -0.112 (0.187) -0.112 (0.187) Ag. Prov. * High Risk 1.456** (0.562) 1.456** (0.562) Cust. Prov. * High Risk -0.406 (0.360) -0.406 (0.360) SPS Prov. * High Risk 0.369 (0.246) 0.369 (0.246) TRIPS * High Risk 0.489* (0.229) 0.489* (0.229) GATS * High Risk -0.762** (0.270) -0.762** (0.270) CU * High Risk -0.181 (0.414) -0.181 (0.414) EIA * High Risk 0.190 (0.209) 0.190 (0.209) FTA * High Risk -0.601 (0.374) -0.601 (0.374) Constant 8.121*** (0.706) 8.172*** (0.646) 3.806*** (0.617) -2.662*** (0.749) Importer, Exporter F.E. yes yes yes yes Sector F.E. no no no yes N 2886 2886 606060 606060

Standard errors clustered at the country pair level in parentheses. *** $p<0.01$, ** $p<0.05$, * $p<0.1$.

Table 4: Summary of the estimation results for individual sectors using expanded trade agreement data (iv)
 Variable Low Risk High Risk K-S test stat. K-S p-value Mean $\beta$ S.D. Mean $\beta$ S.D. Distance (log) -1.140 (1.027) -1.531 (1.462) 0.244 0.013 Contiguity 0.354 (0.855) 0.274 (0.738) 0.125 0.502 Common Language 0.233 (0.961) 0.523 (0.858) 0.143 0.339 Colony -0.052 (1.400) 0.215 (0.673) 0.213 0.042 Agriculture Provision -0.504 (7.817) -1.282 (5.716) 0.138 0.380 Customs Provision 0.649 (6.401) 0.017 (2.343) 0.185 0.106 SPS Provision -0.589 (5.967) 0.240 (2.125) 0.185 0.106 TRIPS 0.496 (4.419) 0.113 (1.343) 0.091 0.855 GATS 0.328 (2.708) 0.025 (2.769) 0.096 0.810 Customs Union 0.892 (5.466) 3.107 (5.963) 0.192 0.086 EIA 0.025 (4.835) -0.165 (2.128) 0.149 0.289 FTA -0.039 (5.396) 1.864 (5.255) 0.162 0.206

Of the three types of special provisions included in the model speciﬁcation—agriculture, customs, and SPS—each exhibits some speciﬁc inﬂuence on agricultural trade.

First, agricultural provisions facilitate trade in high risk agricultural products signiﬁcantly more than low risk products within the frameworks of speciﬁcations (ii) and (iii).10 However, when controlling for country-sector speciﬁc eﬀects under speciﬁcation (iv), there is limited statistical diﬀerence between high and low risk goods. Agricultural provisions, which reﬂect both tariﬀ and non-tariﬀ liberalizations on agricultural products, represent a slightly complicated case to analyze due to how common they are. According to Hofmann et al. (2017), agricultural provisions are present in 99.6 percent of trade agreements covered by the World Bank’s database. Thus, it is likely that the variable is only identifying general trade facilitation eﬀects, which are also being picked up in other variables, potentially explaining the inconsistencies between speciﬁcations (i)-(iii) and (iv).

By comparison, customs provisions appear to be eﬀective at increasing the trade of agricultural goods in general but do not exhibit a statistical diﬀerence in their eﬀects on high or low risk goods under most speciﬁcations. However, speciﬁcation (iv) suggests that customs provisions are relatively more trade facilitating for low risk goods at close to a ten percent signiﬁcance level. This relatively stronger impact for low risk goods may suggest that high risk goods, for which long clearance times tend to result is substantial depreciation of value such as rot, already experience special customs consideration outside of those identiﬁed by the variable. Thus, the broad provisions for customs transparency and clearance improvements controlled for here may largely represent improvements for non-fresh agricultural products not already covered.

Finally, SPS provisions, which reﬂect the harmonization of measures targeting the types of risks inherent to the studied products, exhibit limited inﬂuence on trade in agricultural goods generally or goods of a speciﬁc risk. The exception to this is speciﬁcation (iv), which suggests that SPS provisions are more inﬂuential to high risk goods, again at a nearly ten percent conﬁdence level. This provides some support for the idea that NTM harmonization is trade facilitating.

The two measures representing membership to the TRIPS and GATS largely reinforce the initial observations that agricultural goods beneﬁt from general trade facilitation. Both agreements represent eﬀorts to improve trade by reducing costs, alleviating barriers, and addressing other frictions faced by IP sensitive goods and cross-border services. TRIPS, which targets intellectual property issues in trade, is relatively more facilitating for high risk goods than low risk goods under speciﬁcations (ii) and (iii). This relationship may be reﬂective of more sophisticated and proprietary IP embedded in the high risk goods, which tend to be fresh and of higher quality. Meanwhile, the GATS appears to be more facilitating for low risk goods under speciﬁcations (ii) and (iii). This observation may be related to that made above with regards to the eﬀects of customs provisions. It may well be the case that the general services considerations have limited additional impact on high risk goods because they already exhibit special considerations due to the sensitivity of the goods. However, under speciﬁcation (iv), neither TRIPS nor GATS exhibits signiﬁcant diﬀerences across risk types when sectoral eﬀects are permitted to vary across importers and exporters. This suggests that when controlling for variations in the types of products countries import and export, such as technology intensity in the case of TRIPS, these indirectly related trade agreements no longer disproportionately beneﬁt high or low SPS risk goods. This observation provides further evidence that it is the speciﬁc provisions addressing aspects of certain high or low risk goods rather than general or indirect trade facilitation because the interaction between the country ﬁxed eﬀects and the sectors will have diluted the varying role of technology or services highlighted above.

The remaining three variables characterizing the type of agreement that both partners belong to—CU, EIA, and FTA—exhibit slightly curious results. Under most speciﬁcations, none of these types of agreements appear to be signiﬁcant determinants of trade or exhibit large diﬀerences between high and low risk goods. This observation suggests that much of the trade facilitating behavior identiﬁed using the broadly deﬁned PTA variable from the initial collection of gravity variables, which included agreements in all three categories, is speciﬁc to certain types of provisions rather than general trade facilitation. That is, the positive eﬀects of trade agreements are more closely tied to speciﬁc cases in which provisions address the goods being traded rather than spillovers resulting from increased economic cooperation. This observation also provides support for the ﬁrst of the two eﬀects described above with respect to the TRIPS and GATS. If it is generally the case that speciﬁc types of provisions are what cause increases in trade ﬂows, it is likely that the positive eﬀect that these two agreements have is reﬂective of their provisions for IP and services rather than more general, indirect liberalization. The one exception to this observation is that under speciﬁcation (iv), customs unions are signiﬁcantly more trade facilitating for high risk goods. This observation is consistent with much of the other results in that a customs union reﬂects general regulatory cooperation in addition to tariﬀ reductions. As a result, it is not surprising that customs unions exhibit stronger eﬀects on regulation-sensitive, high SPS risk goods when estimated at the sector level and controlling for country eﬀects at that level.

These results, all taken together, provide strong evidence for several aspects of trade and SPS risk. High risk goods appear to be consistently aﬀected by aspects of trade that slow the ﬂow of goods from exporter to importer. The shortening of distances between trading partners and the introduction of policies that mitigate regulatory burdens through transparency and cooperation tend to increase the trade of high risk goods signiﬁcantly more than low risk goods. These results suggest that policymakers with speciﬁc industries in mind ought to consider intrinsic characteristics of their goods when they design trade policy as not all goods are aﬀected equally. Furthermore, the observed evidence suggests that increasing trade in speciﬁc types of products requires the introduction of agreements and provisions that directly target the products in question. Broad agreements representing general and indirect liberalizations do not appear to consistently increase trade in a speciﬁc subsets of products, namely high and low SPS risk agricultural goods.

Curiously, many of the results are sensitive to the model speciﬁcation. In light of this, it is important to reassess these speciﬁcations in terms of model selection. Due to the important inﬂuences of sector speciﬁc ﬁxed eﬀects within the model, speciﬁcation (iv) is our preferred speciﬁcation. Of the four speciﬁcations, (iv) is the one the best controls for sector speciﬁc idiosyncrasies. There are many reasons for which a particular country may not import or export a particular HS 4-digit category that are unrelated to either SPS risk or trade costs. For example, a country in a temperate climate will likely never be a large exporter of certain agricultural products, regardless of their trade agreements or other trade costs. Speciﬁcation (iv) with the additional trade agreement variables provides for the best control of this type of eﬀect and the strongest identiﬁcation strategy.

As such, our preferred interpretation of the results follows those reﬂecting the speciﬁcation (iv) estimates. In summary, distance and colonial ties, which both represent proxies for trade costs between partners, indicate that high risk goods beneﬁt from cost reductions relatively more than low risk goods. Similarly, trade facilitation eﬀorts in the form of trade agreements generally aﬀect high risk goods more when they address issues that directly aﬀect aspects of the goods being traded. More broadly deﬁned, indirectly-related trade facilitation measures, or those that primarily lower tariﬀs rather than address regulatory harmonization, do not exhibit statistical diﬀerences in their eﬀects when comparing high and low risk goods.

In addition to the speciﬁc results regarding the trade determinants of high and low SPS risk agricultural goods, these ﬁndings provide interesting insight into the ways in which economic research studies SPS issues and NTMs more broadly. As discussed in the introduction, most research assesses the eﬀects of NTMs on trade by studying the eﬀects of the measures themselves rather than by directly studying the product characteristics that have inspired the measures. Our ﬁndings provide evidence that this alternative, product-focused approach can return economically interesting policy implications while foregoing any need to ex-ante diﬀerentiate between non-tariﬀ measures that address socially desirable issues and those that are either domestic protection eﬀorts veiled as NTMs or unnecessarily burdensome due ineﬃciently designed measures (i.e. NTBs). Assuming that the trade agreements that countries enter into and the provisions included within tighten regulatory eﬃciency rather than concede socially desirable measures, the policy eﬀects identiﬁed by our results are related only to protectionism or the ineﬃcient application of NTMs. As such, this product-focused approach represents a promising strategy for future NTM research.

### 4 Conclusion

SPS concerns and the measures put in place to address them signiﬁcantly inﬂuence trading patterns. Using a novel approach of identifying SPS issues by noting the agricultural goods for which SPS human, plant, and animal health risks are highest, we have observed several consistent trends. The estimates from several gravity model speciﬁcations suggest that high risk goods experience higher standards costs to trade as measured by distance between partners and, under the preferred speciﬁcation, colonial relationships. Similarly, high risk goods tend to beneﬁt more from trade facilitation in the form of preferential trade agreements broadly or speciﬁc types of facilitation such as customs unions or SPS provisions speciﬁcally. Likewise, high risk goods tend to be less aﬀected by speciﬁc customs provisions than low risk goods. Neither category of goods appears to exhibit signiﬁcant diﬀerences with respect to other trade facilitation eﬀorts that primarily lower tariﬀs or address indirectly-related aspects of trade, such as the TRIPS or GATS.

This work provides evidence that the risks underlying SPS measures exhibit speciﬁc nuances that ought to be considered when designing trade policies and, more speciﬁcally, SPS measures. In general, it appears that high risk SPS agricultural goods rely to a greater degree on trade facilitation policies than do lower risk goods. Furthermore, increasing trade in these high risk goods tends to require the introduction of policies that target speciﬁc issues related to SPS sensitivity rather than spillovers from broader liberalization eﬀorts. Eﬀorts to facilitate (or deter) trade in speciﬁc types of agricultural products ought to be composed of these types of attributes in order to be most eﬀective.

In addition to providing insight into the trading patterns around SPS issues, this work also shows that a product-centric approach to the analysis on NTM issues can generate useful perspective on the ways in which NTMs prohibit trade. Unlike most research, which studies these issues through the lens of applied measures that may exist with multiple underlying motives, the presence of SPS risk in a product is not subject to politicking, rent-seeking, protectionism, or administrative ineﬃciency. The trends identiﬁed here are those associated only with actual health SPS concerns, bypassing any need to pass judgment on the scientiﬁc validity of any measures. For this reason, analysis of trade determinants using similar product-centric approaches can provide valuable information for designing measures that eﬀectively provide the desired social protections while avoiding the introduction of ineﬃcient barriers to trade.

This work opens the door for several promising avenues for further research extending the ﬁndings here. First, the incorporation of a panel dataset covering more years would provide better identiﬁcation of the eﬀects of preferential trade relationships by taking advantage of variation in membership overtime. A longer time series would also help answer lingering questions about the eﬀects of longer-lived trade agreements. Second, the estimates could be used to parameterize general equilibrium experiments, such as those described in Baier and Bergstrand (2009), in order to simulate the eﬀects of policy changes. Such work could, for example, better describe or predict the impacts of signing ”deeper” trade agreements on speciﬁc markets by incorporating their SPS risk characteristics into the assessment.

### References

Abbyad, N. and P. Herman (2017). How Much Do Non-Tariﬀ Measures Cost? A Survey of Quantiﬁcation Methods.

Baier, S. L. and J. H. Bergstrand (2009). Bonus vetus OLS: A simple method for approximating international trade-cost eﬀects using the gravity equation. Journal of International Economics 77, 77–85.

Calvin, L. and B. Krissoﬀ (1998). Technical barriers to trade: A case study of phytosanitary barriers and U.S.-Japanese apple trade. Journal of Agricultural and Resource Economics 23(2), 351–366.

Crivelli, P. and J. Groeschl (2016). The Impact of Sanitary and Phytosanitary Measures on Market Entry and Trade Flows. World Economy 39(3), 444–473.

Disdier, A. C., L. Fontagné, and M. Mimouni (2008). The impact of regulations on agricultural trade: Evidence from the SPS and TBT agreements. American Journal of Agricultural Economics 90(2), 336–350.

Drogué, S. and F. DeMaria (2012). Pesticide residues and trade, the apple of discord? Food Policy 37(6), 641–649.

Fontagné, L., A. Guillin, and C. Mitaritonna (2011). Estimations of Tariﬀ Equivalents for the Services Sectors.

Fontagné, L., M. Mimouni, and J.-M. Pasteels (2005). Estimating the Impact of Environmental SPS and TBT on International Trade. Integration and Trade Journal 9(22(1)), 7–37.

Fontagné, L., G. Oreﬁce, R. Piermartini, and N. Rocha (2015). Product standards and margins of trade: Firm-level evidence. Journal of International Economics 97(1), 29–44.

Grant, J. and S. Arita (2017). Sanitary and Phyto-Sanitary Measures : Assessment , Measurement , and Impact.

Gurevich, T., P. Herman, S. Shikher, and R. Ubee (2017). Extending the CEPII Gravity Data Set.

Henson, S. and S. Jaﬀee (2008). Understanding developing country strategic responses to the enhancement of food safety standards. World Economy 31(4), 548–568.

Hofmann, C., A. Osnago, and M. Ruta (2017). Horizontal Depth: A New Database on the Content of Preferential Trade Agreements.

Kee, H. L., A. Nicita, and M. Olarreaga (2009). Estimating Trade Restrictiveness Indices. The Economic Journal 119(534), 172–199.

Li, Y. and J. C. Beghin (2011). A meta-analysis of estimates of technical barriers to trade. Journal of Policy Modeling (34), 497–511.

Melo, O., A. Engler, L. Nahuehual, G. Cofre, and J. Barrena (2014). Do Sanitary, phytosanitary, and quality-related standards aﬀect international trade? Evidence from Chilean fruit exports. World Development 54, 350–359.

Moenius, J. (2006). The Good , the Bad and the Ambiguous : Standards and Trade in Agricultural Products. In IARTC Summer Symposium.

Neufeld, N. (2014). Trade Facilitation Provisions in Regional Trade Agreements: Traits and Trends.

Nimenya, N., P. F. Ndimira, and B. H. De Frahan (2012). Tariﬀ equivalents of nontariﬀ measures: The case of European horticultural and ﬁsh imports from African countries. Agricultural Economics 43, 635–653.

Rickard, B. J. and L. Lei (2011). How important are tariﬀs and nontariﬀ barriers in international markets for fresh fruit? Agricultural Economics 42(supplement), 19–32.

Santos Silva, J. M. C. and S. Tenreyro (2006). The Log of Gravity. The Review of Economics and Statistics 88 (4)(4), 641–658.

Tothova, M. and J. F. Oehmke (2008). To Regulate Or Not? The Trade-Oﬀ Between Food Safety and Consumer Choice. Journal of International Agricultural Trade and Development 4(1), 63–82.

Winchester, N., M.-L. Rau, C. Goetz, B. Larue, T. Otsuki, K. Shutes, C. Wieck, H. L. Burnquist, M. J. Pinto de Souza, and R. Nunes de Faria (2012). The Impact of Regulatory Heterogeneity on Agri-food Trade. World Economy 35(8), 973–993.

World Trade Organization (2017). Sanitary and Phytosanitary Information Management System.

 Table 5: List of agricultural, HS 4-digit codes and risk classiﬁcation High HS Product Description Risk 4-digit x 0101 Live horses, asses, mules and hinnies. x 0102 Live bovine animals. x 0103 Live swine. x 0104 Live sheep and goats. x 0105 Live poultry x 0106 Other live animals. x 0201 Meat of bovine animals, fresh. x 0202 Meat of bovine animals, frozen. x 0203 Meat of swine. x 0204 Meat of sheep or goats. x 0205 Meat of horses, asses, mules or hinnies. x 0206 Edible oﬀal of animals. x 0207 Meat and edible oﬀal of the poultry. x 0208 Other meat and edible meat oﬀal. x 0209 Pig fat. x 0210 Meat and oﬀal, salted, in brine, dried or smoked; edible ﬂours and meals of meat or meat oﬀal. x 0401 Milk and cream, not concentrated or sweetened. x 0402 Milk and cream, concentrated orsweetened. x 0403 Buttermilk, curdled milk and cream, yogurt, kephir and other fermented or acidiﬁed milk and cream. x 0404 Whey and products consisting of natural milk constituents. x 0405 Butter and other fats and oils derived from milk; dairy spreads. x 0406 Cheese and curd. x 0407 Birds’ eggs, in shell. x 0408 Birds’ eggs, not in shell, and egg yolk. 0409 Natural honey. x 0410 Edible products of animal origin, n.e.s. 0501 Human hair. 0502 Pigs’, hogs’ or boars’ bristles and hair; badger hair and other brush making hair. x 0504 Guts, bladders and stomachs of animals. 0505 Skins and other parts of birds. 0506 Bones and horn-cores. 0507 Ivory, tortoise-shell, whalebone and whalebone hair, horns, antlers, hooves, nails, claws and beaks. 0510 Ambergris, castoreum, civet and musk; cantharides; bile, whether or not dried; glands and other animal products for pharmaceuticals. 0511 Animal productsn.e.s; dead animals, unﬁt for human consumption. x 0601 Bulbs, tubers, tuberous roots, corms, crowns and rhizomes; chicory plants and roots. x 0602 Other live plants (including their roots), cuttings and slips; mushroom spawn. x 0603 Cut ﬂowers and ﬂower buds for ornamental purposes. x 0604 Foliage, branches and other parts of plantsfor ornamental purposes. x 0701 Potatoes, fresh or chilled. x 0702 Tomatoes, fresh or chilled. x 0703 Alliaceous vegetables, fresh or chilled. x 0704 Edible brassicas, fresh or chilled. x 0705 Lettuce and chicory, fresh or chilled. x 0706 Carrots, turnips, and similar edible roots, fresh or chilled. x 0707 Cucumbers and gherkins, fresh or chilled. x 0708 Leguminous vegetables, fresh or chilled. x 0709 Other vegetables, fresh or chilled. 0710 Vegetables, frozen. 0711 Vegetables provisionally preserved but unsuitable for immediate consumption. 0712 Dried vegetables. 0713 Dried leguminous vegetables. 0714 Roots and tubers with high starch or inulin content. x 0801 Coconuts, Brazil nuts and cashew nuts, fresh or dried. x 0802 Other nuts, fresh or dried. x 0803 Bananas, including plantains, fresh or dried. x 0804 Dates, ﬁgs, pineapples, avocados, guavas, mangoes and mangosteens, fresh or dried. x 0805 Citrus fruit, fresh or dried. x 0806 Grapes, fresh or dried. x 0807 Melons and papaws (papayas), fresh. x 0808 Apples, pears and quinces, fresh. x 0809 Apricots, cherries, peaches, nectarines, plums and sloes, fresh. x 0810 Other fruit, fresh. x 0811 Fruit and nuts, frozen. 0812 Fruit and nuts, provisionally preserved but unsuitable for immediate consumption. 0813 Fruit, dried; mixtures of nuts or dried fruits 0814 Peel of citrus fruit or melons. 0901 Coﬀee. 0902 Tea. 0903 Mat. 0904 Pepper of the genus Piper;fruits of the genus Capsicum or of the genus Pimenta. 0905 Vanilla. 0906 Cinnamon and cinnamon-tree ﬂowers. 0907 Cloves (whole fruit, cloves and stems). 0908 Nutmeg, mace and cardamoms. 0909 Seeds of anise, badian, fennel, coriander, cumin or caraway; juniper berries. 0910 Ginger, saﬀron, turmeric (curcuma), thyme, bay leaves, curry and other spices. 1001 Wheat and meslin. 1002 Rye. 1003 Barley. 1004 Oats. 1005 Maize (corn). 1006 Rice. 1007 Grain sorghum. 1008 Buckwheat, millet and canary seeds; other cereals. 1101 Wheat or meslin ﬂour. 1102 Cereal ﬂours other than of wheat or meslin. 1103 Cereal groats, meal and pellets. 1104 Cereal grains otherwise worked; germ of cereal. 1105 Flour, meal, powder, ﬂakes, granules and pellets of potatoes. 1106 Flour, meal and powder of dried leguminous vegetables, of sago, or of roots or tubers. 1107 Malt. 1108 Starches; inulin. 1109 Wheat gluten. 1201 Soya beans. 1202 Ground-nuts, not roasted or otherwise cooked. 1203 Copra. 1204 Linseed. 1205 Rape or colza seeds. 1206 Sunﬂower seeds. 1207 Other oil seeds and oleaginous fruits. 1208 Flours and meals of oil seeds or oleaginous fruits. 1209 Seeds, fruit and spores, of a kind used for sowing. 1210 Hop cones; lupulin. 1211 Plants and parts of plants used primarily in perfumery, in pharmacy or for insecticidal, fungicidal or similar purposes. 1212 Locust beans, seaweeds and other algae, sugar beet and sugar cane; fruit stones and kernels and other vegetable products 1213 Cereal straw and husks. 1214 Swedes, mangolds, fodder roots, hay, lucerne (alfalfa), clover, sainfoin, forage kale, lupines, vetches and similar forage products. 1301 Lac; natural gums, resins, gum-resins and oleoresins. 1302 Vegetable saps and extracts; pectic substances, pectinates and pectates; agar-agar and other mucilages and thickeners. 1401 Vegetable materials of a kind used primarily for plaiting. 1404 Vegetable products n.e.s. 1501 Pig fat and poultry fat. 1502 Fats of bovine animals, sheep, or goats. 1503 Lard stearin, lard oil, oleostearin, oleo-oil and tallow oil. 1504 Fats and oils and their fractions, of ﬁsh or marine mammals. 1505 Wool grease and fatty substances. 1506 Other animal fats and oils and their fractions. 1507 Soya-bean oil and its fractions. 1508 Ground-nut oil and its fraction. 1509 Olive oil and its fractions. 1510 Other oils and their fractions, obtained solely from olives. 1511 Palm oil and its fractions. 1512 Sunﬂower-seed, saﬄower or cotton-seed oil and fractions thereof. 1513 Coconut (copra), palm kernel or babassu oil. 1514 Rape, colza or mustard oil and fractions thereof. 1515 Other ﬁxed vegetable fats and oils. 1516 Animal or vegetable fats and oils and their fractions, partly or wholly hydrogenated. 1517 Margarine; edible mixtures of animal or vegetable fats or oils. 1518 Animal or vegetable fats and oils and their fractions, chemically modiﬁed. 1520 Glycerol, crude; glycerol waters and glycerol lyes. 1521 Vegetable waxes, beeswax, other insect waxes and spermaceti. 1522 Degras; residues resulting from fatty substances or animal or vegetable waxes. x 1601 Sausages and similar meat products. x 1602 Other prepared or preserved meat products. 1701 Cane or beet sugar and chemically pure sucrose, in solid form. 1702 Other sugars, in solid form; sugar syrups; artiﬁcial honey; caramel. 1703 Molasses. 1704 Sugar confectionery. 1801 Cocoa bean. 1802 Cocoa shells, husks, skins and other cocoa waste. 1803 Cocoa paste. 1804 Cocoa butter, fat and oil. 1805 Cocoa powder, not containing added sweetening. 1806 Chocolate and other food preparations containing cocoa. 1901 Malt extract; food preparations of ﬂour, groats, meal, starch or malt extract, not containing minimal or no cocoa. 1902 Pasta. 1903 Tapioca and substitutes therefor prepared from starch. 1904 Foods prepared with cereals or cereal products . 1905 Bread, pastry, cakes, biscuits and other bakers’ wares. 2001 Vegetables, fruit, nuts and other edible parts of plants, prepared or preserved by vinegar or acetic acid. 2002 Tomatoes prepared or preserved otherwise than by vinegar or acetic acid. 2003 Mushrooms and truﬄes, prepared or preserved otherwise than by vinegar or acetic acid. 2004 Other vegetables prepared or preserved otherwise than by vinegar or acetic acid, frozen. 2005 Other vegetables prepared or preserved otherwise than by vinegar or acetic acid, not frozen. 2006 Vegetables, fruit, nuts, fruit-peel and other parts of plants, preserved by sugar. 2007 Jams, fruit jellies, marmalades, fruit or nut pure and fruit or nut pastes. 2008 Fruit, nuts and other edible parts of plants, otherwise prepared or preserved n.e.s. 2009 Fruit juices (including grape must) and vegetable juices. 2101 Extracts, essences and concentrates, of coﬀee, tea or mat. 2102 Yeasts ; other single-cell micro-organisms; prepared baking powders. 2103 Sauces; mixed condiments and seasonings; mustard ﬂour and meal and prepared mustard. 2104 Soups and broths and preparations therefor; homogenised composite food preparations. x 2105 Ice cream and other edible ice. 2106 Food preparations not elsewhere speciﬁed or included. 2201 Waters, not containing sweetening nor ﬂavoured; ice and snow. 2202 Waters, containing sweetening, and other non-alcoholic beverages, not including fruit or vegetable juices. 2203 Beer made from malt. 2204 Wine of fresh grapes. 2205 Vermouth and other wine of fresh grapes ﬂavoured with plants or aromatic substances. 2206 Other fermented beverages. 2207 Undenatured, 80∖% or higher ethyl alcohol; ethyl alcohol and other spirits, denatured, of any strength. 2208 Undenatured, 80∖% or less ethyl alcohol; spirits, liqueurs and other spirituous beverages. 2209 Vinegar and substitutes for vinegar obtained from acetic acid. 2301 Flours, meals and pellets, of meat or meat oﬀal, of ﬁsh or other aquatic animals. 2302 Bran, sharps and other residues, derived from the working of cereals or of leguminous plants. 2303 Residues of starch manufacture and similar residues, beet-pulp, bagasse and other waste of sugar manufacture, brewing or distilling dregs and waste. 2304 Oil-cake and other solid residues resulting from the extraction of soyabean oil. 2305 Oil-cake and other solid residues resulting from the extraction of ground-nut oil. 2306 Oil-cake and other solid residues resulting from the extraction of vegetable fats or oils. 2307 Wine lees; argol. 2308 Vegetable materials and vegetable waste, used in animal feeding, n.e.s. 2309 Preparations of a kind used in animal feeding. 2401 Unmanufactured tobacco; tobacco refuse. 2402 Cigars, cheroots, cigarillos and cigarettes, of tobacco or of tobacco substitutes. 2403 Other manufactured tobacco and manufactured tobacco substitutes. 2905 Acyclic alcohols and their halogenated, sulphonated, nitrated or nitrosated derivatives. 3301 Essential oils (terpeneless or not). 3501 Casein, caseinates and other casein derivatives; casein glues. 3502 Albumins. 3503 Gelatin. 3504 Peptones and other protein substances n.e.s.; hide powder, whether or not chromed. 3505 Dextrins and other modiﬁed starches; glues based on starches, dextrins, or other modiﬁed starches. 3809 Finishing agents and accelerating dye carriers 3823 Industrial monocarboxylic fatty acids; acid oils from reﬁning; industrial fatty alcohols. 3824 Prepared binders for foundry moulds or cores; chemical products and preparations of the chemical or allied industries (including those consisting of mixtures of natural products), n.e.s. x 4101 Raw hides and skins of bovine or equine animals. x 4102 Raw skins of sheep or lambs. 4103 Other raw hides and skins. x 4301 Raw furskins. 5001 Silk-worm cocoons suitable for reeling. 5002 Raw silk (not thrown). 5003 Silk waste (including cocoons unsuitable for reeling, yarn waste and garnetted stock). 5101 Wool, not carded or combed. 5102 Fine or coarse animal hair, not carded or combed. 5103 Waste of wool or of ﬁne or coarse animal hair, including yarn waste but excluding garnetted stock. 5201 Cotton, not carded or combed. 5202 Cotton waste (including yarn waste and garnetted stock). 5203 Cotton, carded or combed. 5301 Flax, raw or processed but not spun; ﬂax tow and waste (including yarn waste and garnetted stock). 5302 True hemp (Cannabis sativa L.), raw or processed but not spun; tow and waste of true hemp.

1In the past, NTMs have often been referred to as non-tariﬀ barriers (NTBs). Although subtle, there is an important distinction between the terms NTM and NTB in modern convention. A measure can be put in place for a wide variety of reasons. Some of these reasons target important social objectives such as ensuring food safety or environmental protections. Other measures may be put in place for purely protectionist purposes similar to tariﬀs. A measure falling onto this latter category is generally considered an NTB. By comparison, the term NTM makes no such judgment regarding the protective nature of the measure and may be either protective or intended to provide a legitimate social beneﬁt.