Services trade policies and economic integration
Services trade policies and economic integration: New evidence for developing countries
Data weaknesses hamper analysis of how policies towards imports and exports of services, foreign direct investment and, more generally, regulation affects the operation of services sectors. Based on recently released regulatory policy data for 2016, this column uses machine learning methods to recreate to a high degree of accuracy the OECD’s Services Trade Restrictiveness Index to generate new estimates of services trade barriers for 23 developing countries. The analysis confirms that services policies are typically much more restrictive than tariffs on imports of goods, in particular in professional services and telecommunications. Developing countries tend to have higher services trade restrictions, but less so than has been found in research using data for the late 2000s.
Well-known data weaknesses hamper analysis of how policies towards imports and exports of services, foreign direct investment (FDI) and, more generally, regulation affects the operation of services sectors. Comparable time series data on relevant policy variables generally are not available on a cross-country basis for developing economies. The World Bank compiled information on services trade and investment policies applied in 103 countries in the late 2000s and associated services trade restrictiveness indicators (STRIs) (Borchert et al. 2014).1 Starting in 2014, the OECD has reported STRIs for its member countries as well as major emerging economies that span a broader range of policies and services sectors than the World Bank exercises. The OECD STRI is available on an annual basis starting in 2014, and covers 45 countries.2 The data have informed recent research on the role of services policies in influencing productivity, trade in value added, and FDI (e.g. De Backer et al. 2018).
A problem for applied policy research on developing country services trade policies is that the OECD panel dataset covers less than a dozen emerging economies. The World Bank has been collaborating with the WTO secretariat to update the information on developing countries. Policy data 23 developing countries has been posted on the Integrated Trade Intelligence Portal (I-TIP) website of the WTO. In a recent paper (Hoekman and Shepherd 2019), we utilise this information, which pertains to 2016, to generate new indicators of services policy restrictiveness in eight services sectors for these 23 countries, analyse their role as determinants of trade, and compare the potential impact of services trade policy reform with liberalisation of trade in goods.
Generating services policy indicators
A contribution of our analysis is applying a machine-learning algorithm to construct indicators that correlate well with the OECD STRIs for countries that are covered by the OECD dataset. Because the full detail of the methodology used by the OECD is not published, it is not possible to simply apply the OECD methodology to generate STRIs for the additional set of developing counties.
Our basic premise is that the OECD STRI, which relies on expert judgement to inform weighting and aggregation of policies (Grosso et al. 2015), is a good benchmark as it takes into account interaction effects as well as the raw weights attached to particular provisions by sector specialists. We use an elastic net as a prediction tool, identifying a subset of variables that have the best explanatory power in terms of the observed OECD STRI, and use the estimated values from the elastic net regression to predict values out of sample, where no OECD STRI exists (that is, for the 23 developing countries now included in I-TIP). We limit consideration to eight sectors that correspond well between the OECD and I-TIP databases: accounting, legal, commercial banking, insurance, air transport, road freight transport, distribution, and telecom. The resulting services policy indicators (SPIs) closely mimic the OECD STRI. SPIs and associated ad valorem equivalent (AVE) estimates derived from a structural gravity model of trade.
The AVEs are a measure of the equivalent tariff protection that would, if applied, generate the same degree of market insulation as the bundle of policies and regulations summarised by the SPI. Detailed results are reported in the paper on a sector-country basis. Figure 1 summarises the findings for eight sectors, aggregating countries into their respective regions. The most restrictive sectors are telecom, legal, and air transport.
Figure 1 Average AVEs by country group and services sector
In a qualitative sense, these estimates accord well with previous work based on the 2008 World Bank STRI, which also finds that professional services and telecom are sectors with the highest AVEs (Jafari and Tarr 2017). The main takeaway from this exercise is that AVEs in services sectors are high relative to applied rates of tariff protection in goods markets. An AVE in the 15-20%+ range represents a significant restriction to consumers and firms accessing services from foreign suppliers.
An important additional step in validating the SPIs is demonstrating their ability to act as statistically significant predictors of aggregate trade flows – i.e. combining goods and services on the basis of the input-output relationships that exist between services and other sectors (e.g. Hoekman and Shepherd 2017). To do so, we estimate a structural gravity model of total trade (goods and services combined), given that services policies affect both trade in services and trade in goods that use services as inputs. The results indicate that the SPIs generally exhibit very similar performance to the OECD STRI, illustrating the value of our machine learning-based reproduction of the OECD approach. Like the STRI, the SPI has a negative association with trade that is statistically significant at the 1% level. We also find a positively signed and statistically significant interaction term when controlling for joint membership of an economic integration agreement.
Finally, we use the structural gravity model to simulate the potential effects on trade of reducing the restrictiveness of services policies by 10% as opposed to a similar proportional reduction in applied tariffs on goods imports, controlling for preferential trade arrangements. For the 27 non-high-income countries in the sample, reducing the restrictiveness of services policies by 10% would boost real income by 0.5%, compared with 0.4% for a 10% cut in applied tariffs. Both figures are modest, but given that the policy changes are relatively small, that should not be surprising. Given higher barriers to trade in services than for goods, the results suggest developing countries stand to benefit from reforming services policies.
Our SPIs provide a first quantitative snapshot of applied services policies in a sample of developing countries in 2016. While OECD member countries are typically more liberal than developing economies, the differences are not always large in terms of the index scores and AVEs. This finding requires cautious interpretation, as the number of countries is relatively small, but the SPIs line up well with those of Borchert et al. (2014) using the World Bank STRI for 2008.
A contribution of the analysis is to provide a ‘proof of concept’ for the use of statistical tools, such as machine learning, to capture the complexities, nonlinearities, and interdependencies of different services policy measures. The use of such techniques allows extension of the OECD STRI to developing countries for which the underlying policy data are available. Such tools can also help identify subsets of measures that do most of the explanatory work in terms of bilateral trade flows. In our view, the primary value of generating these kinds of indices is to explain important economic outcomes, not merely summarise data. Given the success of the OECD approach, which we have mimicked, the bar is now set very high for designing a ‘better’ index.
Although the release of services trade-related policy data in I-TIP is laudatory, it spans only a subset of developing countries. Most low-income economies are not covered. Doing more to generate services policy data on a regular basis for a broad range of developing countries to complement the OECD STRI is a necessary condition for causal analysis of the effects of policies and to inform liberalization strategies and regional integration processes.
Borchert, I, B Gootiiz, and A Mattoo (2014), “Policy Barriers to International Trade in Services: Evidence from a New Dataset”, World Bank Economic Review 28(1): 162-188 (see also the related Vox column here).
De Backer, K, S Miroudot and D Rigo (2018), “Barriers to trade in services have an impact on multinational production in the manufacturing sector”, VoxEU,org, 19 April.
Grosso, M F Gonzales, S Miroudot, H Nordås, D Rouzet and A Ueno (2015), “Services Trade Restrictiveness Index (STRI): Scoring and Weighting Methodology”, OECD Trade Policy Papers 177.
Hoekman, B and B Shepherd (2017), “Services productivity, trade policy and manufacturing exports”, The World Economy 40(3): 499–516.
Hoekman, B and B Shepherd (2019), “Services Trade Policies and Economic Integration: New Evidence for Developing Countries”, CEPR Discussion Paper 14181.
Jafari, Y and D Tarr (2017), “Estimates of Ad Valorem Equivalents of Barriers against Foreign Suppliers of Services in Eleven Services Sectors and 103 Countries”, The World Economy 40(3): 544-73.
 The World Bank data are at https://datacatalog.worldbank.org/dataset/services-trade-restrictions-database but were not accessible at the time of writing this column (last accessed 8 December 2019).
 The OECD produces STRIs for OECD member countries and nine emerging economies: Brazil, China, Colombia, Costa Rica, India, Indonesia, Malaysia, the Russian Federation, and South Africa (see https://qdd.oecd.org/subject.aspx?Subject=063bee63-475f-427c-8b50-c19bffa7392d).