AVE of NTMs and overall protection
We run 5009 regressions based on specification (3), for each HS 6-digit product level, to estimate the tariff equivalent of core NTMs for 5009 imported products of 97 countries (28 EU countries are estimated separately), for each of the six points in time over the period 1997–2015. The average R2s of these regressions was 0.46, with a median of 0.43 and maximum of 0.99. Less than 1% of the adjusted R2s had a negative sign. Therefore, the fit of these regressions was generally satisfactory. The detailed product level estimates for all countries and years is available on the Links (data links) section of the GEP research centre website at: https://www.nottingham.ac.uk/gep/links/index.aspx. Here we seek to summarize the findings.
First, we estimate the AVEs of NTMs, using Eq. (4), across different dimensions. This enables us to compare the AVEs of NTMs with tariffs and overall protection, to assess the evolution of these measures over time.
Table 3 summarizes the average estimated AVEs of NTMs and provides a comparison with the corresponding average tariff and overall protection levels for products and countries over our sample period. A comparison of columns 4–5 identifies that the average AVE of NTMs is markedly higher than the average tariff throughout the period. Tariff rates are broadly decreasing over time, with the unweighted average tariff rate falling from 12% in 1997 to 5% in 2015. By contrast, the average AVE of NTM protection was 20% in 1997, and rose (with some fluctuation over time) to 57% in 2015. Therefore, NTMs were already a more important source of protection than tariffs at the start of our sample period, and have become even more important sources of trade protection over this period. When weighted by the import volume (columns 7–8), the relative magnitudes of the AVEs and tariff vary slightly, but the conclusion about the relative importance of NTMs and tariffs in overall protection is unaltered. We can conclude from Table 3 that on average the trade barrier effect due to NTMs was much greater than that induced by tariffs. This echoes the finding of Kee et al. (2009) on the dominance of NTMs relative to tariffs, but we further show that this dominance has increased over time.
A similar conclusion about the relative importance of the two trade policy tools can be drawn from an inspection of tariffs and the AVE of NTMs at the product level. Appendix Table 6 summarizes the percentage of product lines for each year and the full sample of countries where the tariff is greater, smaller or equal to the AVE of the core NTMs. At the start of the period, i.e. 1997, the tariff was higher than the AVE in just under 44% of product lines. By the end of our sample period (i.e., 2015), this was true for only about 27% of products, as compared to nearly two-thirds of products being subject to higher non-tariff than tariff protection.
Appendix Table 5 sets out the average AVE of NTMs for each country, presented in coefficient form, for the years for which information was available. Over the period from 1997 to 2015, the average AVE of NTMs for most countries was increasing in general, though there was variation across countries. Some high income countries such as Japan, Australia and New Zealand are identified as consistently ‘low protection’ countries. Countries with the highest AVEs of NTM are Morocco, Burkina Faso, Argentina, China, Mali, Niger and Nigeria. All of these are low-income countries. However, there was an increase in average AVEs towards the end of the sample period for a significant number of both low and high income countries. This appears to correspond with the post-financial crisis and the downturn in world trade.
NTMs across sectors
Table 4 reports the distribution of the AVEs of NTMs for different sectors. The AVEs are generally higher for agricultural products than for manufacturing products. There was an increase in the AVEs for most sectors over the period from 1997 to 2009, though the increase is most evident in manufacturing. Protection from NTMs is shown to be consistently high within the agricultural sector, but to be much more variable across industries in the manufacturing sector. By the end of the period, textiles, footwear, rubber and plastics, optical and medical instruments, machinery and electrical equipment are the most NTM-protected products in the manufacturing sector.
The comparability of the summary evidence in Table 4 with the evidence from other studies is constrained by a number of factors. Many other studies do not provide evidence over time or they use alternative classifications for identifying the incidence of NTMs or they adopt different metrics to measure the overall extent of NTM barriers or protection. One of the important sources of yearly, summary information over the last decade on policy interventions affecting international trade and other forms of international commercial exchange is the Global Dynamics (GD) database of the Global Trade Alert.Footnote 10 This data (for a larger number of countries than this study) includes count information on the total number of import-related interventions (harmful and liberalizing) implemented each year since 2008 to-date. The interventions include tariffs and the coverage of NTMs is not the same as that used in this study. For the years that are common with the present study, however, there is consistency in the pattern of change in trade protection over time between the evidence in Table 4 and the Global Trade Alert indicators. If one restricts the GD information to interventions reported within-year (i.e. up to the end of December in each year), both the overall average AVE in Table 4 (for both agriculture and manufacturing) and the count of new harmful interventions (reported in brackets for each year in what follows) fall between 2009 (274) and 2012 (220) and rise between 2012 (220) and 2015 (648).Footnote 11
NTMs across countries
The evolution of AVEs of NTMs, tariffs and overall protection can also be explored with the present results across countries, and in different regions and different income groups, as shown in Fig. 2.
A consistent picture is evident across all the regions; namely one of stable levels or modest declines in average tariff levels, combined with much higher levels of overall protection resulting from much higher levels of NTM than tariff protection. Indeed, the evolution of overall protection in all regions is predominantly driven by changes in NTM protection. Except for Sub-Saharan Africa, overall protection is higher in all regions by the end of the period than at the beginning, and substantially so in the case of some regions (e.g. North America and South Asia). Indeed, in the case of North America, the AVEs of NTMs and overall trade protection rose consistently after 2003. In most regions, other than North America (for which the data starts in 2003), the AVEs of NTMs tended to increase before 2003. The clear exception to this is Europe and Central Asia for which a sharp fall in NTM protection is identified between 2000 and 2003. This may be due to the ending of the Multi-Fiber Agreement (MFA), and the elimination of the quantity restrictions on textiles imports from developing countries by the developed countries. However, after 2006, NTM protection and overall trade protection rose again sharply across all regions. The estimates seem to be capturing the effects of the more protectionist trade policies adopted globally following the 2008 financial crisis. By 2012, we identify some reversal in this more protectionist stance, though NTM and overall protection generally increased again after 2012.
Figure 3 depicts the evolution of tariffs, AVEs of NTMs and overall protection using a classification of countries based on income groupings. The average tariff for high income countries is significantly lower than in the case of middle and low income countries, but the difference in overall protection between higher and lower income countries declined markedly over the period as protection from NTMs rose more sharply in high income countries (especially the OECD countries and after 2006). Average levels of overall protection in 2015 are identified by this study to be at a tariff-equivalent of about 60% in both OECD and low income countries. Having changed relatively little over the period in the low income countries but risen sharply, from a little over 20% at the start of the period, in the case of the OECD countries. Clearly the evolution of tariffs fails completely to reflect the changing stance of trade policy in this period.
Comparison with Kee et al. (2009) results
Appendix Table 7 provides the average AVEs estimates for a comparable set of countries covered by Kee et al. (2009) in their study (i.e., re-estimated here) and this present study, for estimation surrounding 2002 in the former and 2003 in the latter. There are some similarities between the two sets of results. The relative importance of NTMs and tariffs as sources of protection is a feature of both studies; non-tariff being more dominant than tariff protection. This is evident from the average AVEs and tariff levels in both studies. More than half of the product lines subject to core NTMs are identified as being more restricted by NTMs than tariffs in both studies. In addition, the most protected industries (or imports competing with products produced by these industries subject to most restriction) are identified to be similar in both studies. It is also the case that the individual countries with the highest level of NTM protection are identified by both studies to be generally low-income countries.
However, there are also some differences in the average levels of NTM protection across countries in the two studies, despite the common estimation method. It is evident from Table 7 that average AVEs are generally higher for the comparable sample than the present study; only for 24 countries is the average AVE higher in the present study, while it is lower in the case of 54 countries. The simple average AVE across the common set of 82 countries is 29.5% in the current study and 42.7% for Kee et al. (2009). These differences are likely to stem from the different datasets on NTM incidence adopted, and the comparison is based on simple averages. Notwithstanding this, both studies reveal the dominance of NTMs relative to tariffs and the importance of non-tariff barriers in determining overall protection levels.
Our base modelling recognizes the possible endogeneity of NTMs. Nonetheless, as a further check, we re-estimated the regressions using the 3-year lags of NTMs and tariffs. The NTM incidence variable continues to be instrumented (now with 3-year lagged instruments). Appendix Tables 8 and 9, and R-squares plot depicted by Appendix Fig. 4, report these additional findings. While the magnitude of the average effect differs from the original results (expected given differences across observations), the key point is the non-negligible importance of AVE of NTMs still holds. Looking at the correlation between original and new estimates (see column 3 in Table 8), of the more than 5000+ coefficients estimated, we find a correlation ranging from 0.36 and 0.75. Furthermore, the R-squares for new estimates mirror those of the original estimates. Table 9 shows the correlations between the incidence of NTMs over time. The high correlation over time indicates persistence in the incidence and non-incidence of NTMs, with the correlation in incidence between any two ‘adjacent’ points in time being at least over 0.7 and generally over 0.8. This indicates a ‘slow changing NTM variable’, where cross-sectional, rather than time, variation tends to drive our results and in turn implying that our instrumented contemporaneous variable is robust.
Next, we re-run the analysis for a balanced sample. Appendix Table 10 and the R-squares in Appendix Fig. 5 report the results in summary for this sample. Although the R-squared graph suggests a slightly lower fit for some regressions, the average effect doesn’t differ as much and the correlation between the matched coefficients for the balanced and unbalanced samples is generally high.
Finally, we obtain the AVE of NTMs from estimating the linear specification (2), rather than the non-linear specification (3). Given the difference in specifications and the susceptibility of the means to be affected by extreme values, the R-squares and average AVE of NTMs for the linear and non-linear estimation are not strictly comparable. Therefore, we follow Kee et al. (2009) to find the proportion of estimates AVE of NTMs from the linear specification that are negative (i.e. have a trade promoting effect). We find around 12–18% of the sample to be so. This is similar to Kee et al. who find 13% of AVE of NTMs to be negative. Even though specific NTMs, such as sanitary and phytosanitary measures or technical measures, could have positive trade effects under some circumstances, we do not expect the incidence of all core NTMs at the tariff line level to be net trade-promoting for other than a very small proportion of tariff lines. Indeed, even in the case of the unrestricted estimation, the overwhelming majority of NTMs are trade-restricting according to our estimates. In line with Kee et al. (2009), our preferred estimates for comprehensive measurement of the trade effects of non-tariff barriers are those based on a non-linear estimation method.Footnote 12