The service economy: U.S. trade coalitions in an era of deindustrialization

In one of the first studies examining the politics of trade in services, Chase (2008) examines labor groups’ lobbying in the motion picture industry. With a few notable exceptions (Manger 2009; Kim and Manger 2017; Gootiiz and Mattoo 2015), very little attention has been paid to firms’ lobbying over trade in services (Weymouth 2017). On the economics of services trade, there is a more extensive literature; see, e.g., Eschenbach and Hoekman (2006), Francois and Woerz (2008), Hoekman and Mattoo (2008), Francois and Hoekman (2010), and Jensen (2011).

Services represent around 75% of GDP in Organization for Economic Co-operation and Development (OECD) countries and 70% of the global economy (Francois and Hoekman 2010).

Feenstra and Sasahara (2017) find that growth in US exports generated 4.1 million additional jobs in services between 1995 and 2011.

In one of the first studies examining the politics of trade in services, Chase (2008) examines labor groups’ lobbying in the motion picture industry. He finds that low-skilled occupations were most likely to oppose the movement of motion picture production abroad. With a few notable exceptions (Manger 2009; Kim and Manger 2017; Gootiiz and Mattoo 2015), very little attention has been paid to firms’ objectives regarding services liberalization.

While a technical consultant may find it profitable to travel internationally to deliver a report to a foreign client, international travel for the purpose of cutting hair (almost) never occurs.

We also note that very large firms are still active as individuals in services, just less so than in manufacturing because there is less incentive to pay the costs of individually lobbying if the industry association is active and its liberalization objectives align with those of the firm. We therefore see part of our theoretical contribution as highlighting a key scope condition or intra-industry disagreement: that both trade partners must be reasonably competitive.

This portion of the analysis therefore necessarily excludes the US-Israel agreement, CUSFTA, and NAFTA.

On telecommunications, see Manger (2009); on financial services, see Cameron and Tomlin (2000); Roy et al. (2007); on insurance, see Cameron and Tomlin (2000).

US trade agreements share a similar structure, building on the text of the North America Free Trade Agreement (NAFTA), which entered into force in 1994 (Baccini et al. 2014). For details, see Appendix Table D.1 in Supplementary Material, which shows all the provisions related to service liberalization included in US trade agreements.

More services firms participated in the public campaigns for the trade agreements with Singapore, Chile, the CAFTA countries, Bahrain, Morocco, Oman, Peru, Colombia, Panama, and Korea than did goods-producing firms. Likewise, more services associations than goods associations supported the agreements with Singapore, Chile, Bahrain, Morocco, and Oman.

In Appendix C, we chart the growth of services lobbying over time, showing that lobbying on trade by services more than doubled from 1998 to 2016. We also examine the geographic dispersion of services firms’ support for trade, and find that services firms from a wide array of states have publicly supported trade in the US.

Note that productivity drives both ability to export and firm size. So while we often say that ‘larger firms are capable of exporting’ for ease of exposition, it is more precise to say that ‘highly productive firms are capable of exporting’. Size and productivity are likely to be closely linked in the long run, but some new entrants might be quite productive and still small, while some very large firms (especially in countries with state ownership) might be inefficient.

Osgood (2017).

On ordinary trade competition, see Milner (1987); Madeira (2016); Osgood (2016); Plouffe (2017). On the globalization of production, see Osgood (2017); Milner (1988).

On the empirics of firm heterogeneity, see Bernard et al. (2012) for a complete review. On models of firm heterogeneity and trade see, among others, Melitz (2003); Bernard et al. (2003). On models of firm heterogeneity and global production, see Antras and Helpman (2004); Helpman et al. (2004).

The US also has a markedly higher RCA in services than its trade agreement partners, so there is relatively little potential for foreign competition for market share inside the US. Data on relative FDI between the US and its trade partners show a similar asymmetry. Examining all sectors, the ratio of the stock of US foreign investment to the stock of its trade agreement partners in the US varies from 1.29 or 1.46 at the lowest, for South Korea and Canada respectively, to 2.08 (Mexico), 3.4 (Panama), and 7.42 (Singapore). All other ratios exceed 10. Data on FDI in services specifically is only available for a smaller set of countries, but is less than 1 for only one (Korea): 1.36 (Canada), 3.59 (Australia), 4.90 (Panama), 6.79 (Mexico), and 31.95 (Singapore). The ratio for the Middle East and Latin America regions is 1.75 and 17.62, respectively.

The indices are standardized to range between 0 and 1.

It is important to note that services are not equally open across all modes of delivery, however. The US’s complex system of occupational licensing, for example, restricts delivery of health care services by foreign nationals living in the United States. In the WTO’s parlance this is an instance of Mode 4: the movement of natural persons.

Labor costs as a share of revenue are only 10.0% in manufacturing in the US; in information, finance, professional services, education, health care, and the arts, they are 21.7, 14,4, 39.3, 43.8, 39.2, and 31.9%, respectively. Material inputs as a share of costs are 59.5% in manufacturing; comparable data for services are not provided by the US government. These figures are calculated using data from the Economic Census of the United States in 2012, available from https://​factfinder.​census.​gov.

The wholesale and retail sectors are important exceptions which rely on foreign-made products to stock warehouses and shelves. We do not see any evidence of intra-industry disagreement in this area, however. One explanation for this may be that smaller retailers are able to rely on globally connected wholesalers to stock their shelves, and so are not excluded from the gains from global sourcing. Trade politics in these sectors merits further detailed investigation.

Information services transform information into a commodity for distribution, and include publishing, motion pictures, broadcasting, and data processing, hosting and related services.

These differences reflect long-standing disparities in the treatment of goods and services in the GATT/WTO. For services liberalization under the General Agreement on Trade in Services (GATS), countries identify the specific service industries to which they will apply market access and national treatment obligations, along with any exceptions to those obligations. Services commitments thus “bind” the specified market access and national treatment for the particular industry, guaranteeing that conditions will not change in ways that would harm firms in other countries.

Responding to this feature of services liberalization, many of the advocates for services liberalization in the United States, like the Coalition of Services Industries, have pushed for a “negative list” approach to negotiations over services liberalization. Under this approach, all services are assumed to be open across all modes of delivery unless a specific reservation has been taken. Roy et al. (2007) find that services commitments in US PTAs tend to significantly reduce services trade barriers beyond countries’ GATS offers.

Table D.2 in the Supplementary Material shows that the language similarity among US annexes, which include reservations related to US service industries, is quite high for US PTAs. To obtain these indices, we rely on the Jaccard measures of language similarity implemented by the R package ‘textreuse.’ For a similar approach, see Wilkerson et al. (2015).

Our argument does not imply uniform support for liberalization among all US services firms or for all services industries. Relatively non-tradeable services, like construction, may be disinterested in trade liberalization though they may still benefit from liberalization of foreign investment. Likewise, smaller services firms are less likely to make significant gains from trade agreements, as exporting and horizontal FDI are heavily concentrated among the larger firms. Where America’s smaller services firms differ from its smaller goods-producing firms is that they are also less likely to face losses from trade agreements, and so have no obvious motive to oppose those agreements.

For an excellent review, see De Figueiredo and Richter (2014).

We focus on producers rather than labor. Many labor unions representing service workers have lobbied against US trade agreements. This may suggest concerns about the offshoring of services jobs, as in Chase (2008), Walter (2010, 2017), Owen (2016), and Owen and Johnston (2017), or a more general anti-trade orientation among the US labor movement. We sidestep these issues here, as we are operating under a standard assumption that firms’ engage in lobbying with the purpose of maximizing profits. We recognize that workers within the firm may have different trade policy preferences. However, a profit-maximizing firm may lobby for a trade agreement that provides new market opportunities even if some of its employees oppose trade deals due to their own employment insecurities. Intra-firm disagreements over trade are an interesting and important area for future research, but beyond the scope of our paper.

The results are similar if we use a variable that divides the number of associations that lobbied in industry i by the number of firms and associations that lobbied in the same industry. We label this variable Firm-centric Lobby and we show the results in Table D.3 in the Supplementary Material.

These data are introduced in Osgood (2018) and Osgood and Feng (2018).

Results are similar if we use a continuous measure of intra-industry divisions over trade which is equal to \( 1 - \frac {|\# \text {Opposing} - \# \text {Supporting} |}{\# \text {Opposing} + \# \text {Supporting} }\) where # Supporting is the count of all firms and associations that supported a trade agreement and # Opposing is the count of all firms and associations that opposed a trade agreement (see Table D.23 in the Supplementary Material).

Agricultural support activities (NAICS 115) and Support activities for mining (NAICS 213) are included among the services industries.

Our main findings are similar if we use a different set of controls including total factor productivity and capital–labor ratio (data from Orbis 2014), though we lose a large number of observations. These results are reported in Table D.6 in the Supplementary Material.

Clustering standard errors at the PTA level is problematic, given the small number of clusters. The results hold if we cluster standard errors at the industry level (see Table D.7 in the Supplementary Material) and if we use bootstrapped standard errors (see Table D.8 in the Supplementary Material).

The results are virtually the same if we use a probit or logit model for Divided and fractional regressions, which are particularly suitable when the outcome variable ranges between 0 and 1, for Lobby separate and Positions separate (see Table D.9 in the Supplementary Material).

We re-run our main models using 2-digit NAICS dummies for different services industries. The results, reported in Table D.10 in the Supplementary Material, indicate that lower fragmentation in services seems to be driven primarily by the retail, professional services, and finance. We note that the US has a clear comparative advatnage in these industries as showed in Fig. 2.

Table D.24 in Supplementary Material shows heterogeneous effects across partners. In particular, we interact Services with a dummy capturing large trading partners (Australia, Korea and NAFTA [for position-taking only]) and a dummy capturing developed trading partners (Australia, Bahrain, Morocco, NAFTA [for position-taking only], Oman, Singapore, and South Korea). The expectation is that we should observe more industrial disagreement in the case of large developed economies, which are particularly competitive in services, compared to smaller developing countries. This expectation is met for data on position-taking, whereas we find no heterogeneous effects for the data on lobbying. The latter result may be explained by the lack of variation across PTAs due to a relatively small sample and by the fact that firms and associations lobby typically for several PTAs at the same time. For instance, firms and associations lobbying for the PTA with South Korea are often lobbying also for the PTAs with Colombia, Panama, and Peru. This interdependence across PTAs makes it difficult to pin down the effect of Services on lobbying activities related to specific PTAs.

Industry tradability is characterized according to the geographic concentration of the 6-digit NAICS industry in the United States. When production exceeds local demand, the excess supply must be either consumed or exported to another region. Thus, low concentration implies low tradability. An advantage of this approach is that it can be applied to services as well as goods.

The goods trade data are from Comtrade.

The services data are from the World Bank’s Trade in Services Database, available at https://​data.​worldbank.​org/​data-catalog/​trade-in-services. We use EBOPS classifications that roughly equate to 2-digit NAICS services industries. In analyzing goods trade, researchers have access to monthly data on US goods exports and imports for over 8,000 product categories. In services, the US trade statistics cover only around 40 categories annually since 2006, and fewer categories prior to that. For the vast majority of US PTA partner countries, there is no disaggregated bilateral services trade data prior to 2006. The mismatch in the level of aggregration between the services and goods data, as well as the relative coarseness of the services data, force us to adopt a summary proxy for comparative advantage.

Our results are similar if we use an ordinal measure of comparative advantage (see Table D.12 in Supplementary Material). While an RCA measure – ideally one capturing US RCA relative to the PTA partner – would be preferable, we are unable to build such a measure at the 2-digit level due to the aforementioned data limitations.

The results hold if we cluster standard errors at the industry level (see Table D.13 in the Supplementary Material) and if we use bootstrapped standard errors (see Table D.14 in the Supplementary Material).

The data correspond to the year 2014 and are derived from publicly available BEA statistics.

The goods trade data are from Comtrade.

The services data are from the World Bank’s Trade in Services Database. We use the EBOPS classifications that are roughly equivalent to 2-digit NAICS services industries.

Our identification strategy is similar to that of Autor et al. (2013), who use Chinese exports to other developed countries to instrument for Chinese exports to the US

The U.K. was the second-largest exporter of services in the 1990s.

We considered netting out U.K. trade with the US, but bilateral country-industry trade data (as opposed to country-industry global imports and exports data) are incomplete for services industries in the 1990s.

The correlation between France and US comparative advantage is close to zero for services, whereas the correlation between Germany and US comparative advantage is 0.3.

Autor et al. (2013) confirm the robustness of their instruments using a measure of comparative advantage in a gravity model.

China’s entry into the WTO in 2001 is not a threat to our identification strategy using data from this time.

As suggested by Conley et al. (2012).

Our results are similar if we use an ordinal measure of comparative advantage (see Table D.12 in Supplementary Material).

The (unreported) Anderson-Rubin Wald test shows that orthogonality conditions are valid, i.e. the coefficients of the endogenous regressor in the structural equation are not equal to zero. Indeed, when we estimate the reduced form of the equation with the instrument as the regressor, its coefficient is always negative and significant (results available upon request).

Reports are available at https://​ustr.​gov/​trade-agreements/​free-trade-agreements.

https://​ustr.​gov/​archive/​assets/​Trade_​Agreements/​Bilateral/​Australia_​FTA/​Reports/​asset_​upload_​file553_​3390.​pdf.

Not every industry prepared a report for all trade agreements, which explains the missing information. The breakdown of the industries by sector is reported in the appendix (Tables D.20, D.21, and D.22 in Supplementary Material.

https://​ustr.​gov/​archive/​assets/​Trade_​Agreements/​Bilateral/​Australia_​FTA/​Reports/​asset_​upload_​file118_​3412.​pdf.

Ibid.

https://​ustr.​gov/​sites/​default/​files/​ITAC-10-Services-and-Finance-Industries.​pdf.

For instance, see https://​ustr.​gov/​archive/​assets/​Trade_​Agreements/​Bilateral/​Chile_​FTA/​Reports/​asset_​upload_​file480_​4953.​pdf.

We do not theorize about the agricultural sector and thus do not hold priors about lobbying in this sector. Rather than exclude the agricultural trade advisory committee reports, we include them for comparative purposes and in hopes of spurring future research on the sector’s trade policy stances.

For example, “the Agreement creates the framework for improved markets” or “the Agreement creates significant new opportunities for market access.” To make the text analysis comparable across all three sectors, we divided the word frequency by the total number of words in each sector, e.g. in all the reports issued by services industries.

For instance, we capture statements like “the Committee is disappointed by the absence of provisions that will facilitate business travel” or “The Committee remains disappointed by provisions that could allow governmental restrictions.”

Future work should also seek to uncover why previous efforts at multilateral services liberalizations (e.g. the Trade in Service Agreement) have not been successful.