Trade Barriers and Trade Flows Across Countries and Industries

Board of Governors of the Federal Reserve System

International Finance Discussion Papers Number 476

July 1994

TRADE BARRIERS AND TRADE FLOWS ACROSS COUNTRIES AND INDUSTRIES

Jong-Wha Lee and Phillip Swagel

NOTE: International Finance Discussion Papers are preliminary materials circulated to stimulate discussion and critical comment. References in publications to International Finance Discussion Papers (other than an

acknowledgment that the writer has had access to unpublished material) should be cleared with the author or authors.

ABSTRACT

We use disaggregated data on trade flows, production, and trade barriers for 41 countries in 1988 to examine the political and economic determinants of non-tariff barriers. as well as the impact of protection (both tariff and non-tariff) on trade flows. We use an econometric framework that allows for the simultaneous determination of trade barriers and trade flows. Our results are consistent with political-economy theories of the determinants of protection: even after accounting for industry-specific factors, nations tend to protect industries that are weak, in decline, and threatened by import competition. Countries also give more protection to large industries; these might be thought of as politically important. Nations use tariffs, non-tariff

barriers, and exchange rate controls as complementary instruments of protection.

Trade Barriers and Trade Flows Across Countries and Industries Jong-Wha Lee and Phillip Swagel!

L. [Introduction

Theoretical interest has recently focused on the determinants of nations’ trade barriers. Underlying these theories is the implicit belief that there are common economic and political factors which can explain the structure of protection across countries and industries. This is in contrast to a literature exemplified by Hufbauer and Rosen (1986) which argues instead that at least in the United States, protection is "special" in the sense that it is best explained on a caseby-case or industry-by-industry basis.

The contribution of this paper is entirely empirical. We use disaggregated cross-country, cross-industry data on manufactured goods to examine the political and economic determinants of non-tariff barriers in 1988. As tariff levels have fallen and remained bound by GATT strictures, non-tariff barriers have increasingly become the instrument of choice for protection. The calls for protection from import-competing industries indicate that the pattern of trade is likely to have an effect on the structure of protection. Since protection (both tariff and nontariff) clearly affects trade flows, we use an econometric framework that allows for the

simultaneous determination of trade barriers and trade flows. Unlike previous studies, our

'The authors are respectively: Assistant Professor, Korea University and NBER; and Visiting Assistant Professor, Northwestern University. This paper was written while the second author was a staff economist in the U.S. International Transactions section of the International Finance Division. We thank Judy Hellerstein, Mike Knetter, Kala Krishna, and Andrew Warner for helpful discussions, seminar participants at the 1994 Econometric Society Winter Meetings, the Federal Reserve Board, the New York Fed, and Fletcher for comments, and Refik Erzan and Will Martin for assistance in obtaining data. The opinions expressed in this paper are solely those of the authors, and are not necessarily shared by the Federal Reserve System or its staff.

1

sample includes both developed countries with low barriers, and developing countries with substantial protection across all manufacturing industries. Indeed. a novelty of this paper ts combining disaggregated data on production, trade flows. and trade barriers for a broad range of countries.

Our results are consistent with political-economy theories of the determination of trade protection. We find that non-tariff barriers are determined by more than just industry-specific factors. Nations tend to protect weak industries, as well as industries in decline. Large industries, which we think of as being politically important, also receive protection in the form of non-tariff barriers. Lastly, we find some evidence, although not conclusive, that non-tariff barriers and exchange rate controls were more significant barriers to trade in manufactures than tariffs.

The paper proceeds as follows. The next section discusses political-economy theories of trade protection. In section III, we then specify a model of trade flows based on the monopolistic competition model of trade, along with a model of trade barrier determination. Section IV describes the data, after which we present our empirical results in Section V. Section

VI concludes.

IL. The Political-Economy of Trade Protection

1. Models of Trade Protection

Given the economic consensus regarding the efficiency of free trade, models of trade

barrier determination typically turn to political-economy explanations.* Baldwin (1982) discusses political motivations for trade protection, while Hillman (1989) and Magee, Brock, and Young (1989) survey a variety of models in which political factors influence trade policy. Bhagwati (1982) also contains several interesting models of protection besides those we discuss below.

The key insight is that the structure of the political system can be important in the determination of trade protection when a distinction exists between consumers and producers, or between different groups of consumers and/or producers. The benefits of trade protection typically accrue to a narrow group of stakeholders in the protected industry, while the costs are spread over a much larger number of consumers, each of whom loses only a small amount. This asymmetry means that protection may be politically efficient even if it is inefficient in an economic sense.

Hillman (1982) cites the transactions cost theories of Peltzman and Stigler as an underpinning to this political efficiency of protection. If there are costs to gathering information or voting (hiring a public relations or lobbying firm), then even with majority voting, atomistic consumers will not lobby against the protection sought by a small industry group. Instances of industry-led protection abound. For example, Irwin (1993) provides a fascinating account of the evolution of trade policy in the semiconductor industry, and details the important role played by the Semiconductor Industry Association.

This is formalized in political support models such as Hillman (1982), in which a

?There is of course a large literature on strategic trade policy, but this typically examines trade barriers in a particular industry, rather than the determination of the entire structure of trade protection.

policymaker seeking to ensure reelection balances the welfare of consumer-voters who suffer from protection against the political support (i.e., campaign contributions) provided by an industry seeking protection. Grossman and Helpman (1992) provide a more rigorous theoretical foundation for this literature through an explicit model of the process by which different interest groups bid for protection. With perfect competition in the product market, their model predicts that the structure of protection depends on two factors: the elasticity of import demand, which indicates the degree to which trade barriers distort welfare, and the ratio of imports to domestic output, which reflects the political importance of the domestic industry.

As in Kasa (1991), protection for declining industries can be explained as an attempt to mitigate the adjustment costs incurred in factor reallocation. In Cassing and Hillman (1986), an industry’s slow decline can turn to sudden collapse once the industry shrinks below the threshold where it is large enough to gain the ear of politicians. Mayer (1984), on the other hand, shows that small industries might be more likely to garner protection, since the welfare loss from the protection will be small and thus unlikely to raise opposition. Stole and Zame (1993) allow for the possibility of foreign direct investment, and show that domestic firms might reduce demands for trade relief in order to avoid more intense direct competition from transplant industries, particularly in expanding industries.

Cassing and Hillman (1985) show that political considerations will also apply to the choice of protectionist instrument--that it is sometimes advantageous to use a quota instead of a tariff despite the loss of quota rents. In practice, giving up quota rents to foreigners might be used to "buy off" the affected firms in exporting countries in order to forestall protectionist

retaliation (Marvel and Ray (1987)).

2. Empirical Studies of Trade Protection

The papers by Marvel and Ray (1983. 1987. 1981a, 1981b, 1985) examine various aspects of the implication of the theoretical literature that the structure of protection across industries depends on the particular political and economic characteristics of each industry. Ray (1981ib) estimates equations for the simultaneous determination of imports and trade barriers (both tariff and non-tariff) in the United States. The import equation is based loosely on the Heckscher-Ohlin model, while trade barriers are determined by industry characteristics such as a measure of capital intensity. the proportion of skilled labor, the domestic supply elasticity, and the concentration ratio. He finds that non-tariff barriers in the United States fell mainly on capital-intensive, low-skill industries. Ray (1981a) estimates trade and protection equations for both the U.S. and for an aggregate of foreign countries. While he finds that tariffs and NTB’s were used as complements, he finds no effect of trade protection on U.S. imports. More recently. Trefler (1993) estimates trade and NTB equations for the United States and: shows the importance of taking into account the simultaneous determination of imports and trade protection. Trefler also finds that political factors and proxies for industrial structure, such as measures of union density and industry concentration, have the expected positive impact on the level of protection in the United States.

Marvel and Ray (1983) estimate equations for the determination of U.S. tariffs and NTB’s alone. They find that protection was given to politically important industries, and industries under threat, while healthy industries received less protection. They ascribe this to the Peltzman-Becker theory of regulation, which suggests that policymakers will seek to share

an industry's good fortune with weaker sectors. Finally, Ray and Marvel (1985) estimate tariff

and NTB equations alone for the U.S.. Canada. Japan, and the EC as a whole. They find broad similarities in the structure of protection in these countries. Although tariff rates were generally low, they find that NTB’s were used to undercut this apparent liberality. particularly in the EC. Dick (1994) also finds that NTB’s were used to compensate industries affected by reduced tariffs.

Like Ray (198la, 1981b) and Trefler (1993), we combine the literatures on the determination of trade barriers and trade flows, and attempt to control for their simultaneous determination. One advantage of our analysis is that our sample includes both developed and developing countries, and thus encompasses substantially more variation in the structure of trade

flows and trade barriers.

III. Models of Trade Barriers and Trade Flows 1. The Monopolistic Competition Model of Trade

Following Krugman and Helpman (1985), we assume a monopolistic competition model of trade, in which goods are imperfectly substitutable and differentiated by country of origin. With identical homothetic preferences for consumers, each country consumes identical proportions of each product. Since production of each variety of a product occurs in only one

country, the model gives a prediction of the volume of trade as follows:

(1) IM;, = s; (Q - Qi)

where: IM; = import of good i by country j

= production of good i in country j total world production of good i share of country j in world income

OL no

n II

Equation (1) states that country j’s import of good i is Proportional to the amount of good i produced outside country j, and provides a basic framework to estimate the volume of trade.

For example, Lawrence ( 1987) estimates a logarithmic variant: (2) log(IM,/DU,) = constant + a log(Q;\/Q) + B log(Distance,) + Uy

Domestic use, DU; equals production plus imports minus exports, while a trade-weighted measure of distance between the capital of each country and the capitals of its trading partners is used to proxy for transportation costs, as in Bergstrand (1989).

The monopolistic competition model gives a Prediction of the volume of trade in the absence of trade barriers. When the Helpman-Krugman model is extended to include trade policies, as in Flam and Helpman (1987), it gives ambiguous predictions about the effects of protection on welfare, production, and trade flows. In the simple framework above, however, the model unambiguously predicts that the presence of trade barriers, such as tariffs, non-tariff barriers, and exchange controls, will diminish the volume of trade. As trade barriers increase the prices of foreign goods, consumption of imports falls while consumption of domestic goods rises.

Lawrence does not have measures of trade barriers, but instead identifies this with the residual; that is, he attributes any deviation of actual imports from predicted imports to the effects of protection. Because we have measures of trade barriers, we do not have to make this assumption, but can instead directly examine the impact of trade barriers on trade flows. We

extend Lawrence’s specification to consider the effects on the volume of trade of distortions such

as tariffs, non-tariff barriers, and exchange rate controls. This is similar to Harrigan (1993), who estimates the effects of tariff and non-tariff barriers (which are taken as exogenous) on bilateral trade flows in manufactures in OECD countries.

Adding the measures of trade barriers, the empirical specification in equation (2)

becomes:

(3) logIM;/DU,) = a, + a, log(Q;/Q;) + a, log(Distance,) + a; log(1 +tariff,) + a, log(l+NTB,) + a; log(l1+BMP)) + uj;

The variables for tariff and non-tariff barriers (NTB) measure the intensity of trade barriers on good i in country j, while the black market premium (BMP) is meant to capture the distortionary effects of exchange controls that might hinder imports. These are described in Section IV. Since production of each good is determined simultaneously with trade flows, we follow Harrigan (1992b) and use factor endowments to instrument for sectoral production. We use the economy-wide factor endowments of each nation’s capital stock, labor force, human capital, and land area. These are also described in Section IV. Unfortunately, data on sector-specific inputs are not available for our wide range of countries. An immediate implication of this is that we cannot compare our results with those from a Heckscher-Ohlin model like the one Harkness (1978) estimates for the United States, in which factor-endowments determine the pattern of

trade.

2. Endogenous Determination of Trade Barriers

As discussed in Section II, the structure of production is probably best thought of as

endogenously determined by both economic and political factors. To take into account the notion that the political power of the industry is likely to be important. we use two proxies for a sector’s political influence: the size of the industry as measured by its share of value-added within a country. as well as the industry’s share of labor. Of course. these are likely to be imperfect indicators of political importance, since small industries may be seen as crucial to national security. or in many countries, might be directly or indirectly owned by policymakers.

There may also be political pressure to protect "weak" industries, such as those with low productivity. To examine this, we use value-added per worker as a measure of industry productivity in our model of trade barrier determination. Of course, it may be impossible to make inferences about causality here, since protection could lead to a lazy industry with low productivity rather than a weak industry receiving protection. We need some instrument for productivity to better make this distinction. To examine the tendency of declining industries to receive protection, we include the five-year change in wage per worker (from 1982 to 1987) as an explanatory variable. If there is profit-sharing in an industry, declining wages would indicate declining rents, and thus shifting comparative advantage.

Not surprisingly, there are other determinants of trade protection for which we could not obtain data. For example, Grossman and Helpman (1992) show that trade barriers are more likely to exist the lower the own price elasticity of demand for an industry’s product, since this entails a smaller deadweight loss to consumers. Similarly, the higher the foreign price elasticity of supply, the more effective will be a given trade barrier in changing the pattern of trade. Since we could not obtain cross-country, cross-industry data on elasticities and industry

characteristics such as concentration ratios, we must instead rely on the inclusion of industry

fixed-effects to account for any omitted industry-specific effects. This will werk to the extent that these omitted factors are constant across countries.

We take tariff rates as exogenous; while not strictly correct. this is probably not too bad an assumption relative to non-tariff barriers. since tariff rates in many countries are under GATT strictures. For the U.S., Ray (1981b) finds no feedback from NTB’s to tariffs.

In our basic specification, then, non-tariff barriers in each industry of each country are

expected to respond to sectoral imports and other economic and political factors:

(4) logi+NTB)) = c, + c¢, logdIM,/DU;) + c, log(VA,/L,) + c¢; log(WA,/VA) +

Q

4 Alog(W;,/L;) + cs log(1+tariff;,) + c,log(1+BMP;) + uy;

where: VA,/L; = labor productivity (value-added per worker) VA,/VA; = sectoral share of value-added A(W,,/L;,) = five year change of real wage per worker

As discussed above, the labor productivity represents each sector’s competitive position, while the sectoral share of value-added (or alternatively, the sectoral share of labor, L;/L,) is meant as a proxy of political power. The change of real wages captures the evolution of each industry. Since declining industries typically call for protection, we would expect this to be negatively associated with non-tariff barriers. In addition, tariff rates and the black market premium are included in the regression to examine whether the different varieties of trade restrictions tend to be used as substitutes or in tandem. We also added several additional variables to this base specification: these are discussed in Section IV below.

The production share and distance measure are used as instruments for imports and thus

10

do not appear in the NTB equation, while the "political-economy" variables instrument for the level of protection, and do not enter into the import volume equation. This latter identifying restricticn is clearly more troubling, since trade flows might have a direct effect on productivity, the industry share of value-added (or workers), and the evolution of wages.

As detailed in the next section, our measure of non-tariff barriers is a coverage ratio

which is bounded from below at zero. To take this censoring into account, we specify the NTB

equation (4) as a Tobit:

log(1+NTB’,) = c, + c, log(IM;/DU;) + c, log(VA;/L;) + c¢; log(Lj/L) + c, Alog(W;/L;) + cs log(1 +tariff;) + c, log(1+BMP;) + uy; (5) NTB,; = NTB, if NTB; > 0

= 0 otherwise

We assume that the error terms u, and uy are distributed with a bivariate normal, and estimate equations (3) and (5) jointly using the simultaneous equations Tobit methodology of Nelson and Olson (1978). We estimate the equations both with and without industry fixed effects. Finally, we also estimate the equations using the corresponding single-equation methodologies (OLS and

Tobit), and calculate a Hausman test of the null hypothesis of no simultaneity bias.

IV. Data 1. Trade and Protection Data

To measure the degree of trade barriers across industry and country, we use the dataset of

11

trade control measures (TCM) compiled by UNCTAD (1991). This provides information on both tariffs and additional charges on imports in 1988, as well as information on the coverage of non-tariff measures (NTM’s) at the most detailed level of the Customs Co-operation Council Nomenclature (CCCN)--four digits plus up to two alphabetic codes.

The tariff provided by UNCTAD is the ad valorem rate for total import charges: this includes all duties and customs fees collected at national borders. The measure of non-tariff barriers reports the coverage ratio for "core" NTM‘’s; this includes essentially all non-tariff restrictions applied at the border, including quantitative restrictions (QRs), Voluntary Export Restraints (VERs), and advance payment requirements. Note however, that the measures of both tariffs and NTB’s do not include restrictions which apply inside national borders, such as consumption taxes in countries with no domestic production. See UNCTAD (1987, 1991) and Laird and Yeats (1990) for details.

The coverage ratio indicates the extent to which the tariff lines within a CCCN category are affected by core NTM’s. For instance, the index equals zero for a particular 4 or 5 digit CCCN category if no NTM’s apply to any of the products which make up that category. The CCCN category for autos might include tariff lines for both small and medium-sized products. If a country has an NTM on small but not medium-sized cars, then the coverage ratio for that CCCN category would equal 0.5 regardless of the composition of auto imports. The NTM coverage ratio thus captures only the frequency of the non-tariff restrictions, but provides no information on the severity of the distortions or the distribution of the resulting quota rents.

The data set also provides the value of trade flows taken from the United Nations

COMTRADE database. The trade flow data is at the level of disaggregation needed to match

12

the tariff and NTM data (either 4 or 5 digits of SITC Revision 2). so that the UNCTAD database facilitates combining the trade barrier data with import values.

Unfortunately. reliable cross-country. cross-industry production data is available only at the 3 digit level of the ISIC classification, so we must aggregate up the data on trade barriers and trade flows by weighting them by the country’s import value. Weighting by the own import-values has the well-known problem that a high level of protection typically results in a low leve. of imports, and thus a low weight. As a check on this bias, Table 1 reports tariff and non-tariff barriers aggregated from the 3 digit level to a single value for each country (what might be thought of as the "0 digit" level), using both import and production weights. Just as import weights will understate trade barriers, production weights will overstate them, since a high trace barrier will result in larger domestic production than would occur in the absence of all barriers. At this "0" digit level, the simple correlation between the two weighting schemes is 0.959 for both tariffs and NTB’s, while the rank correlations are 0.935 for tariffs and 0.945 for NTB’s. These high values provide some hope that using import weights to go from 4 or 5

digits to 3 digits will not introduce too much bias into the measures of trade protection.

2. Production, Labor and Wage Data

Data on gross-output and value-added, as well as industry wages and employment at the 3 digit level of the ISIC classification system are from the United Nations Industrial Statistics Yearbook Volume 1, as found in the BESD database of the World Bank. The data on wages includes wages, salaries, and supplements. This wage data, along with the data for gross-output

and value-added are in home-country currency. To match the trade flow data which is reported

13

in US dollars, we convert the currencies using the exchange rate series from the Summers and Heston dataset. Implicit GDP deflators from the Summers and Heston dataset are used to deflate lagged wages to obtain the change in wage per worker in 1988 dollars.

The data for the black market premium, distance, and the factor endowments of land. area, human capital, and labor force are from Barro and Lee (1993). The measure of distance is the import-weighted distance between a nation’s capital and the capitals of its trading partners. The black market premium on foreign exchange is measured as an average of the period from

1980 to 1984.

3. Features of Trade Barriers

Table 1 gives a summary of protection by country, while Table 2 summarizes protection across industries. Our sample is limited by the availability of both the trade data and the production data. For example, the UNCTAD trade barrier database does not include non-EC developed countries such as Austria, Switzerland, Australia, and New Zealand. Similarly, reliable production data are not available for many countries, particularly developing countries.

Construction of the import- and production-weighted tariffs and NTB’s are discussed above; the standard deviations shown in Tables 1 and 2 are of the unweighted tariff rates and NTM coverage ratios. As expected, import-weighting typically results in smaller measures of protection than production-weighting, although again, the two are highly-correlated.

Even a brief glance at Table 1 reveals the dramatically higher levels of protection in developing countries than in developed (the table is sorted by IMF country codes, so the

developed countries appear at the top). On the other hand, while tariff rates are quite lov in

14

most developed countries, these nations employ a notably higher level of non-tariff barrier protection. This is consistent with the findings of Marvel and Ray (1983) and Dick (1994) that NTB’s were used to offset the diminished tariffs negotiated in the various GATT rounds. The higher taritf levels in developing countries probably also reflect the greater importance of these relatively easily collected revenues in government finance.

Table 2 shows that protection, particularly non-tariff barriers, is concentrated in certain industries, notably food, clothing, steel, and transport equipment. Of course, these are for the most part explained by industry-specific managed trade arrangements, such as the Multi-Fibre Arrangement for textiles and clothing, and the web of bilateral quantitative restraints that govern trade in steel and automobiles. This suggests that the pattern of protection Hufbauer and Rosen (1986) ascribe to the United States might apply to the rest of the world as well--that protection is "special" in that it is industry-specific. If this were the case, then including industry fixedeffects in the basic regression specification of equation (5) would eliminate the statistical significance of political-economic determinants such as the political importance or competitive position of an industry in each particular country.

We should note that outlying observations in Tables 1 and 2 are the effect of particularly large protection by specific countries in certain industries. For example, despite otherwise moderate barriers, Venezuela protected its furniture industry in 1988 with an 85% tariff and 93% non-tariff barrier coverage, accounting for the large standard deviations of Venezuela's protection found in Table 1. Similarly, Egypt protected its beverage industry with a tariff rate

of 2200% [sic], accounting for the large standard deviation of Egypt’s tariff rates in Table | as

15

well as the large standard deviation of the beverage industry in Table 2.° Finally, we should note that the United Nations concordance we use to go from the detailed CCCN classification to the 3 digit ISIC classification assigns no products to ISIC category 356, Plastic Products, but

instead places these in various other categories.

V. Empirical Results

Equations (3) and (5) are estimated jointly using a simultaneous equations Tobit estimator, where the import equation (3) is the usual linear model, and the NTB equation (5) is censored at zero. The results are found in Tables 3 to 5. As discussed above, all exogenous variables are used as instruments, with the political-economy variables excluded trom the import equation, and the production share and distance measure are excluded from the NTB equation. In order to examine the degree to which industry-specific factors determine protection. we estimate the models both with and without industry fixed effects. As noted before, factor endowments are used to instrument for the production share; this first stage regression has an

adjusted-R* of 0.83 without industry fixed effects and 0.93 with fixed effects.

1. Determinants of Manufactured Imports

Table 3 contains estimates of four specifications of the import equation (3). Columns (1) and (3) are for the base case, without and with industry fixed effects; these correspond to first column of Tables 4 and 5, which contain the results for the NTB equation. Cclumns (2) and

(4) in Table 3 correspond to the third column in Tables 4 and 5, in which we add the Summers-

‘Dropping this observation from the sample does not affect our results.

16

Heston measure of openness, defined as the sum of imports and exports as a fraction of GDP. as an additional explanatory variable in the NTB equation (that is, as an additional instrument in the import equation). The reasons for considering openness are discussed below in the context of the results for the NTB equation (5).

The negative and highly significant coefficient on the output share in all four specifications indicates that the monopolistic competition model seems to work well. On the other hand, the low R°’s of columns (1) and (2) without fixed effects show that the simple version of the model we use does not explain most of the variation in world trade flows.* Adding :ndustry fixed effects improves the fit of the equation; this most likely indicates that the products of certain industries are simply more frequently traded than others, or that global trade in certain industries is comprised to a larger degree of two-way shipments of differentiated products of the type best characterized by the monopolistic competition model. As expected, the coefficients on distance and the black market premium are always significant and negative.

The specifications in Table 3 give mixed evidence on the extent to which tariffs and nontariff barriers reduce imports: as openness and industry fixed effects are added in going from column (1) to column (4), the coefficients on tariffs and NTB’s change in both sign and significance. When industry fixed effects are included in columns (3) and (4), non-tariff barriers appear to be more substantial barriers to imports than tariffs, which have a statistically insignificant effect on imports. This result, though not incredibly robust, is the opposite of

Harrigan (1993), who finds that "tariffs . . . were a more substantial barrier to trade in

‘Leamer (1984, 1992) discusses many issues involved with the estimation of trade models.

17

5

manufactures between developed countries than were non-tariff barriers."~ The values of the coefficients on NTB’s in columns (3) and (4) imply that an increase in the coverage ratio of one percentage point leads to a two to three percentage point drop in import penetration.

The Hausman tests at the bottom of Table 3 show that we cannot reject the null hypothesis of no simultaneity bias in the import equation. particularly in the estimates with industry fixed effects. This implies that our finding of a significant response of iraports to trade barriers is not simply the result of endogenizing NTB’s. This is borne out in the top half of Table 6, which shows the results for the coefficient on NTB’s in the import equation for both single equation OLS and the linear equation of the simultaneous equations Tobit. In all four single-equation specifications, we find a negative and statistically significant effect of NTB’s on import penetration. Allowing for feedback from imports to NTB’s gives an estimate of at most only slightly more than twice the magnitude (-2.918 and -2.327 versus -1.130 and -1.165 for the equations with fixed effects) as when this endogeneity is ignored. This contrasts with Trefler’s results for the US, in which he strongly rejects the null hypothesis of no simultaneity bias, and finds that taking into account the endogeneity of trade barriers gives an estimate of the effect of NTB’s on import penetration ten times larger than when the simultaneity is ignored. While we also find a statistically significant effect of NTB’s on import penetration, the fairly small impact which comes solely from endogenizing trade barriers points to the importance of not generalizing

US-specific results to other nations.

5Of course, we use a different dataset and a more diverse set of countries. Our specification is also substantially different, since Harrigan estimates bilateral trade flows. whereas our data is limited to each nation’s overall trade in each industry.

18

2. Determinants of Non-Tariff Barriers

Table 4 contains the results for the non-tariff barrier equation without fixed effects, which are then added in Table 5. In general, the results are well-reconciled with political-economy theories of protection. In both tables, the coefficient on the share of imports is always significant and positive, showing that "threatened" industries receive protection. In contrast to the import equation, the large statistics for the Hausman tests on nearly all specifications reject the null hypothesis of no simultaneity bias for the NTB equation, indicating that the simultaneity of trade barriers and trade flows matters for the determinants of trade barriers. This can be seen clearly :n the bottom half of Table 6, which shows the coefficients on import penetration in the NTB equation for both the single-equation and the simultaneous equations Tobit estimators. In the single-equation estimates, where the effect of NTB’s on imports is neglected, we find either an insignificant or negative effect of imports on NTB’s: this no doubt reflects the usual importreducing effects of NTB’s. But this is strongly reversed in the simultaneous equations estimates, in which we find a positive effect of import penetration on NTB’s. In other words, our results indicate that it is not simply that industries with high import penetration receive protection, but rather that industries receive more protection to the extent that they have high import penetration after taking into account the level of NTB’s. And in results not shown, we find that adding the three-, four-, or five-year change in import penetration to the NTB equation results in all cases in an insignificant coefficient on this new variable. These results are again the opposite as Trefler, who finds that the level of import penetration has no effect on the level of U.S. trade barriers, while the change in penetration has a highly significant positive effect.

The significant negative coefficient on the change in wage per worker indicates that

19

declining industries receive more protection: this too is robust to different specifications for the number and range of years of the change in wages. In results not shown, we add the five-vear change in the sectoral share of labor, but this additional measure of industry decline has no explanatory power beyond the change in wages.

The significant positive coefficient on the industry share of value-added indicates that large industries, which we interpret as politically important. also tend to receive more protection. As seen in column (2), this resuli (as well as those in all other rows and columns) holds when the industry share of labor is used instead as the measure of political importance--this might better reflect the actual number of voters and thus the raw political importance of the sector rather than the economic importance indicated by the share of value-added. As expected, the coefficient on value-added per worker is negative and significant, showing that nations give more protection to "weak" industries. The positive and significant coefficients on tariffs and the black market premium indicate that other trade measures and exchange rate controls are used in conjunction with non-tariff barriers; this matches the findings of Marvel and Ray.

In column (3), we add country openness to the regression. This is to examine the possibility that the outward orientation of each nation’s political system has explanatory power for the inter-industry structure of trade protection. For example, political entities such as Singapore and Hong Kong which by their institutional nature are "open" may eschew trade barriers regardless of sectoral conditions, while "closed" countries such as India may erect across-the-board barriers as a matter of course. The significant negative coefficient on openness seems to bear this out. In results not shown, we find that including a complete set of country

fixed effects instead of just the single openness variable gives the odd result that a higher level

20

of NTB’s leads to strongly significant greater import penetration, though it does not greatly change cur other results. While the industry fixed effects are meant to capture omitted industry characte-istics, a similar rationale does not exist for country fixed effects. Since the openness variable is meant to capture a crucial dimension along which countries differ, this can be thought of as a parsimonious alternative to including country fixed effects.

In column (4), we add the share of industry output which is exported, and find that this has a significant negative relationship with the level of protection. This accords with the idea that nations refrain from protecting industries for which exports are important out of fear that their trading partners will retaliate for any import restraints.

Column (5) replaces the measure of productivity, value-added per worker, with wages per worker. As before, the coefficient is significant and negative, indicating that high-wage sectors receive less protection. The significant negative coefficient on the labor share of valueadded in column (6) indicates that less protection is given to labor-intensive industries, which probably correspond to low-skill industries. This matches Ray’s finding for the U.S. that protection tends to be given to capital-intensive and skill-intensive industries. Finally, column (7) includes both wages per worker and value-added per worker. The coefficient on wages remains negative and significant, indicating as before that nations give less protection to industries with labor rents--industries in which wages are high after controlling for productivity. And this negative coefficient somewhat allays our fears regarding the endogeneity of our righthand-sicle variables, since we would usually expect wages to rise if there were affect by trade barriers.

Table 5 adds industry fixed effects to the specifications of Table 4. What is most

21

encouraging is that all of the sector-specific political-economy variables remain significant. even after taking into account the possibility that certain industries get protection across countries, As in Table 4, the significant negative coefficients on value-added per worker and the change in wage per worker indicate that nations protect weak industries and industries in decline, while the positive coefficients on the shares of workers and valued-added show that large, politically important, industries receive more protection. The coefficient on country openness remains negative and statistically significant. And again, all columns of Table 5 indicate that tariffs and

exchange rate controls are used in conjunction with non-tariff barriers.

VI._Conclusion

Our results indicate that protection is not specific to particular countries and industries, but instead that the structure of non-tariff barriers across countries and industries can be explained by sectoral conditions. This is consistent with political-economy explanations of trade protection. Of course, we have not tested a specific model of protection, but rather examined some of its general determinants.

Also, we have only a single cross-section of data on tariffs and NTB’s, and are thus not able to look at the effects of changes in protection over time. This is clearly a concern for our identifying restrictions, which use industry conditions such as labor productivity and wages per worker to instrument for the level of non-tariff barriers. If protective measures are longstanding, the causality might be the reverse; that is, the existence of barriers could influence industry conditions, rather than policymakers responding to industry-specific calls for protection.

However, the significant negative coefficients we find on wages per worker and changes in

22

wages per worker somewhat mitigate this concern, since we would expect wages to rise rather than fall in response to protection.

Further work is needed to examine the impact of variables such as demand and supply elasticitizs which figure prominently in the theoretical literature but for which we were unable to obtain data. Also, our measures of an industry’s political importance--industry shares of labor and value-added--are probably far from ideal. For some countries, unionization data might be an important measure, while for others, the extent of ownership by the ruling party or family would no doubt be closely tied to the structure of trade protection. And as mentioned before, disaggregated data on factor endowments would allow us to compare the results of a Heckscher- Ohlin model with those from our monopolistic competition model of trade.

The limitations imposed by data availability notwithstanding, we obtain remarkably robust results that sectoral factors are important determinants of the structure of trade protection, even after taking into account industry-specific fixed effects. Our results thus provide encouraging

support “or the burgeoning literature on the political-economy of trade protection.

23

REFERENCES

Baldwin. Robert E., 1982. “The Political Economy of Protectionism.” in Jagdish Bhagwati, ed., Import Competition and Response, pp. 263-286.

Barro. Robert J., and Jong-Wha Lee. 1993. "Data Set tor a Panel of 138 Countries. 1960- 1985," Harvard mimeo, February.

Bhagwati, Jagdish, ed., 1982. Import Competition and Response, Chicago: University of Chicago Press, pp. 263-286.

Bergstrand, Jeffrey, 1989. "The Generalized Gravity Equation, Monopolistic Competition. and the Factor-Proportion Theory in International Trade," The Review of Economics and

Statistics, V.71, no.1, pp. 143-153.

Cassing, James H., and Arye L. Hillman, 1985. "Political Influence Motives and the Choice Between Tariffs and Quotas,” Journal of International Economics, vol. 19, pp. 279-290.

-------- , 1986. "Shifting Comparative Advantage and Senescent Industry Collapse," American Economic Review, Vol. 76, No. 3, pp. 516-523.

Dick, Andrew, 1994. "Explaining Managed Trade: Non-Cooperative Politics or Cooperative Economics?" UCLA mimeo, January.

Flam, Harry, and Elhanan Helpman, 1987. "Industrial Policy under Monopolistic Competition,” Journal of International Economics, 22, pp.79-102.

Grossman, Gene, and Elhanan Helpman, 1992. "Protection for Sale," NBER Working Paper number 4149, August.

Harkness, Jon, 1978. "Factor Abundance and Comparative Advantage,” American Economic Reviewn, December, pp. 784-800.

Harrigan, James, 1992a. "Openness to Trade in Manufactures in the OECD,” mimeo, July.

-------- , 1992b. "Factor Endowments & the International Location of Production: Microeconometric Evidence for the OECD, 1970-1985," mimeo, November.

a------- . 1993. "OECD Imports and Trade Barriers in 1983," Journal of Internaticnal Economics, vol. 35, no. 1/2, pp. 91-112.

Helpman, Elhanan, and Paul Krugman, 1985. Market Structure and Foreign Trade: Increasing Returns, Imperfect Competition and the international Economy. Cambridge: MIT Press.

24

Hillman, Arye, 1982. "Declining Industries and Political-Support Protectionist Motives," Avnerican Economic Review, Vol. 72, No. 5, December, pp. 1180-1187

Hufbauer, Gary C., and Howard F. Rosen, 1986. Trade Policy for Troubled Industries, Washington, D.C.: Institute for International Economics.

Irwin, Douglas A., 1993. "Trade Politics and the Semiconductor Industry," mimeo, November. Kasa, Kenneth, 1991. "Declining Industries and Protectionist Lobbying,” mimeo, June.

Laird, Sam, and Alexander Yeats, 1990. Quantitative Methods for Trade Barrier Analysis, New York: New York University Press.

Lawrence, Robert Z., 1987. "Does Japan Import Too Little: Closed Markets or Minds?," Brookings Papers in Economic Activity, no. 2, pp. 517-54.

Leamer, Edward, 1984. Sources of International Comparative Advantage: Theory and Evidence, Cambridge: MIT Press.

wonenan- , 1992. "Testing Trade Theory," NBER Working Paper number 3957, January.

Magee, Stephen, William Brock, and Leslie Young, 1989. Black Hole Tariffs and Endogenous Policy Theory. Cambridge: Cambridge University Press.

Marvel, Howard P., and Edward Ray, 1983. "The Kennedy Round: Evidence on the Regulation of International Trade in the United States," American Economic Review, Vol. 73, No. 1, pp. 190-197.

-------- , [987. "Intraindustry Trade: Sources and Effects on Protection," Journal of Political Economy, vol. 95, December, pp. 1278-1291.

Mayer, Wolfgang, 1984. "Endogenous Tariff Formation," American Economic Review, vol. 74, Liecember, pp. 970-985.

Nelson I*., and L. Olson, 1978. "Specification and Estimation of a Simultaneous Equation Model with Limited Dependent Variables," International Economic Review, Vol. 19, pp. 695-710.

Ray, Edward, 1981a. "Tariff and Non-tariff Barriers to Trade in the United States and Abroad," The Review of Economics and Statistics, May, pp. 161-168.

on------ , 1981b. "The Determinants of Tariffs and Nontariff Trade Restrictions in the U.S.," Journal of Political Economy, Vol 89, February, pp. 105-121.

25

on------ . and Howard Marvel. 1985. "The Pattern of Protection in the Industrialized World." The Review of Economics and Statistics, pp. 452-458.

Stole, Lars, and Ed Zame, 1993. "Foreign Direct Investment and the Political Economy of Protection,” Chicago GSB mimeo, August.

Trefler, Daniel, 1993. "Trade Liberalization and the Theory of Endogenous Protection: An Econometric Study of US Import Policy," Journal of Political Economy, vol. 101, no. 1, pp. 138-160.

UNCTAD, Handbook of Trade Control Measures of Developing Countries, primary volume and Supplement, 1987.

UNCTAD, UNCTAD Micro TCM System, 1991.

26

Table 1: Measures of Protection by Country

Taritf Rate Non-Taruf Barriers

Country Seetors impot- | prodn- std. import- | prodnweighted weighted ‘ weighted Weighted United States United Kingdom Belgium Denmark France Germany. West Italy Netherlands Swecen Canada Japan

Greece Ireland Portugal Spain Turk zy Chile Colombia Ecuador Guatemala Venezuela Barbados Cyprus Jordan Syria

Egypt Bangladesh Sri Lanka Honz Kong India Indonesia Korea Malaysia Pakistan Philippines Singapore Thailand Kenya Mauritius Zim dabwe

Tun.sia

27

a a oe

ro

x

> x

pays w

SAAT g, PPLIP | -UON

oO

pays -odut

ey

(car

cro,

“ w vet oo

ta a

gL Ly

re

cl

ir, val

al

6

oO

an

LOC O11

pays

SPY [PAV

Noa

-tpoid

9

oO 9 0

rm Ol

Ir,

oc,

ol

g

~™

9 9

y 9 9 0

6

Palys tar - ody

oat

on

OF

SALNUNO SD)

SPOON CRE OVANVAL TELL O WOOL OLILEN TOS Y TVNOISS TO Ud TNGINdEAOT EYOdSNV ULL

ORLOV AQANIH OVA

TWORLLOEP EE 1d LONE AGEINIELOVIN SEONdUOUd TW ETA CELE V ORT Vel SIVAN, SOOURAENON

TELLS GNV NOU

OUd WWE INTIAL OUT TV.ESIACNON SAHLLO SLONdOUd GNV SSWI1D “RIVAANGIELL VEE V NERO AUT LLOd SLONGOUd WAAAY

WOO UNV WOAFTOULAd OSIW SHIMANE WA TOU d

SIVOINSHO YAH.LO

SIVOINAHO TVIALSAGNI

ONITSPTEAd (NV ONEENTYd SLONGOUd GNV WldVd

TWIAW Ld tOXd AMINA SPOLLING Da Ld OX SLONGOUd COOM YO WAGGA LdHOXT YVAMLOOA SLONGOUd YSULVAT

WVAMLOOA dd OXA TANVddV ONIAVAM SATLLXAL

OOOVAOL

SAOVUAATE

SLONGOUd GOOA

Aaisnpu]

Caysnpuyp <q WorojOd gp JO SIANSLITAl

TOME,

28

Table 3: Determinants of Manufactured Imports

Dependent Variable: imports/domestic use 1031 Observations

No Fixed Effects with Industry Independent Variables Fixed Effects

(1) (2) B) (ay

output share -0.279 (-11.50)

distance -0.668 (-6.70)

| + tariff rate -1.766 (-5.36)

1 + NTB coverage ratio 0.884 (1.61)

| + black market premium -3.129 (-6.82)

openness in NTB equation? no

adjusted R?

Hausman test statistic, x” (significance level)

degrees of freedom for test

Notes: |. linear equation in simultaneous equations Tobit estimation

2. {-Statistics in parentheses, except for Hausman test

3. all variables in logs

4. results for constant omitted

5. a low significance level for the Hausman test means that the null hypothesis of no simultaneity bias cannot be rejected.

29

Table 4: Determinants of Non-Tariff Barriers, No Fixed Effects

Dependent Variable: | ~ Non-Tariff Barrier Coverage Ratio 103} observations

Independent Variables

Import Share

imports. domestic use 0.172 (13.39)

Sector’s Comparative Advantage

labor productivity (value-added/worker)

wage per worker

labor share of valueadded

Sector's Political Importance

share of value-added 0.059 (11.29)

share of workers 0.055 i (9.18)

Demand for. Protection

change in wage per -0.120 | -0.111 worker : (-4.19) 3 : (-4.48)

share of exports in -0.018

gross output : (-5.66)

Other Trade Influences

| + tariff rate 0.516 | 0.537 } 0.586 : 0.503 | / 0.572 5 0.465 (12.94) : (11.93) : (24.89) } (13.47) } : (13.97) } (11.66)

| + black market 0.702 0.732, 0.771 | 0840 | | 0.746 (0.640

premium (13.75) : (12.61) } (28.77) } (12.33) : : (12.87) : (12.39)

Country Openness -0.251 : (-15.64)

Hausman test statistic 24.368 | 183.811 | 37.829 | (significance level) : (0.99) : (0.99) : (0.99)

degrees of freedom 7 8 8

Tobit equation in simultaneous equations Tobit estimation t-Statistics in parentheses. except for Hausman test . all variables in logs . results for constant omitted _ a low significance level for the Hausman test means that the null hypothesis of no simultaneity bias cannot be rejected

Notes:

a em wha

30

Table 5: Determinants of Non-Tariff Barriers, with Industry Fixed Effects

Dependent Variable: | ~ Non-Tariff Barrier Coverage Ratio 1031 observations Independent Variables

import Share

imports. domestic use

Sector's Comparative Advantage

labor preductivity (value-acided’ worker)

wage per worker

labor share of valueadded

Sector's Political Importar.ce

share of value-added share of workers

Demand for Protection

change in wage per worker

share of exports in gross output

Other Trade Influences

| ~ tariff rate 0.442 | 0.454 | | 0.501; 0.419 (10.40) : (10.25) } : (13.61) : (10.46)

| + black market 0.719 0.723 | 0.791 | 0.694 premium (11.70) : (11.50) } : (13.56) } (11.99)

Country Openness

Hausman test statistic 43.611 (significance level) i (0.90)

degrees of freedom

Tobit equation in simultaneous equations Tobit estimation

. t-statistics in parentheses. except tor Hausman test

all variables in logs

results for constant omitted

a low significance level for the Hausman test means that the null hypothesis of no simultaneity bias cannot be rejected.

Notes:

Tn he od

31

Table 6: Coefficients on NTB’s and Import Penetration

Import Equation: Effect of Non-Tariff Barriers on Import Penetration

No Fixed Effects with Industry Fixed Effects openness j included?

Single Equation Simultaneous Equations

NTB equation: Effect of Import Penetration on Non-Tariff Barriers

No Fixed Effects with Industry Fixed Effects (1) 2 (2) (3) 2 (4) openness 2 included?

Single Equation Simultaneous Equations

Note: _ t-statistics in parentheses

32

IFDP Number

+76

475

474

473

472

47]

470

469

468

467

466

465

464

International Finance Discussion Papers

Titles

1994

Trade Barriers and Trade Flows Across Countries and Industries

The Constancy of Illusions or the Illusion of Constancies: Income and Price Elasticities for U.S. Imports, 1890-1992

The Dollar as an Official Reserve Currency under EMU

Inflation Targeting in the 1990s: The Experiences of New Zealand, Canada, and the United Kingdom

International Capital Mobility in the 1990s

The Effect of Changes in Reserve Requirements on Investment and GNP

International Economic Implications of the End of the Soviet Union

International Dimension of European Monetary Union:

Implications For The Dollar

European Monetary Arrangements: Implications for the Dollar, Exchange Rate Variability and Credibility

Fiscal Policy Coordination and Flexibility Under European Monetary Union: Implications for Macroeconomic Stabilization

The Federal Funds Rate and the Implementation of Monetary Policy: Estimating the Federal Reserve’s Reaction Function

Understanding the Empirical Literature on Purchasing Power Parity: The Post-Bretton Woods Era

Inflation, Inflation Risk, and Stock Returns

Author(s)

Jong-Wha Lee Phillip Swagel

Jaime Marquez

Michael P. Leahy John Ammer Richard T. Freeman Maurice Obstfeld

Prakash Loungani Mark Rush

William L. Helkie David H. Howard Jaime Marquez

Karen H. Johnson

Hali J. Edison Linda S. Kole

Jay H. Bryson

Allan D. Brunner

~ Hali J. Edison

Joseph E. Gagnon William R. Melick

John Ammer

Please address requests for copies to International Finance Discussion Papers, Division of International Finance, Stop 24, Board of Governors of the Federal Reserve System, Washington, D.C. 20551.

33

IFDP Number

462

461 460

459

458

457

456

455

454

453

452

451

450

International Finance Discussion Papers

Titles

1994

Are Apparent Productive Spillovers a Figment of Specification Error?

When do long-run identifying restrictions give reliable results?

199

(es)

Fluctuating Confidence and Stock-Market Returns

Dollarization in Argentina

Union Behavior, Industry Rents, and Optimal Policies

A Comparison of Some Basic Monetary Policy Regimes:

Implications of Different Degrees of Instrument Adjustment and Wage Persistence

Cointegration, Seasonality, Encompassing, and the Demand for Money in the United Kingdom Exchange Rates, Prices, and External Adjustment

in the United States and Japan

Political and Economic Consequences of Alternative Privatization Strategies

Is There a World Real Interest Rate?

Macroeconomic Stabilization Through Monetary and Fiscal Policy Coordination Implications for Monetary Union

Long-term Banking Relationships in General Equilibrium

The Role of Fiscal Policy in an Incomplete Markets Framework

Internal Funds and the Investment Function

34

Author(s)

Susunto Basu John S. Fernald

Jon Faust Eric M. Leeper

Alexander David

Steven B. Kamin Neil R. Ericsson

Phillip Swagel Dale W. Henderson Warwick J. McKibbin

Neil R. Ericsson David F. Hendry Hong-Anh Tran

Peter Hooper Jaime Marquez

Catherine L. Mann Stefanie Lenway Derek Utter

Joseoh E. Gagnon Mari D. Unferth

Jay H. Bryson

Michael S. Gibson

Charles P. Thomas

Guy V.G. Stevens