3.2 Empirical specification for the impact of PTA trade facilitation provisions on trade

Box 3.2 Empirical specification for the impact of PTA trade facilitation provisions on trade

The following equation estimates the heterogeneous impact of preferential trade agreements (PTAs) on export performance of different types of firms (Lee, Rocha, and Ruta 2021):

yijt = β1 PTAjt + β2 (PTAjt × GVCfirmi( j)) + β3 log(GDPjt) + αit + αj + εijt. (B3.2.1)

The baseline outcome variable yijt is the export participation and log export value of firm i ’s exports to country j in year t. PTAjt is a dummy variable, equal to 1 if Peru has a PTA in force with country j in year t. This variable is then interacted with GVCfirmi( j), which is a time-invariant dummy variable, equal to 1 if exporter i imports intermediate inputs (for a GVC firm) or if exporter i imports inputs from country j (for a bilateral GVC firm). With this specification, β1 would capture the average effect of the PTA with country j across exporters that do not import (traditional exporters), and β2 would capture the differential impact on GVC firms, if any.

The estimation also includes a rich set of fixed effects to control for a wide range of other factors that can affect firms’ exports to a particular country. Firm-year fixed effects, αit, control for time-varying firm characteristics, such as productivity shocks, that might affect a firm’s export performance to all countries. Destination fixed effects, αj, capture destination country characteristics, including standard gravity variables—such as distance from Peru, shared language, or cultural or regulatory similarities—that could also affect Peru’s exports to the country j. In addition, all regressions include the gross domestic product (GDP) of country j in year t to control for demand shocks.

Equation (B3.2.2) includes the key variable of interest, TFjt, which captures the extent of TF commitments in the PTA by counting the number of provisions included. As was shown in figure 3.6, the number of TF provisions in Peru’s PTAs ranges between 15 and 34. For ease of interpretation, this is normalized to between 0 and 1 in the regressions. The interaction terms allow the effect of TF provisions to vary by firm type.

yijt = β1 RTAjt + β2 TFjt + β3 (TFjt × GVCfirmi( j)) + β4 Depthjt + β5 log(GDPjt) + αit + αj + εijt. (B3.2.2)

A key challenge in identifying the effect of specific provisions in DTAs is that modern PTAs increasingly cover many policy areas, and agreements that are deep overall typically include many commitments across all chapters. The large number of possible provisions and the multicollinearity pose econometric challenges to perfectly control for the commitments in various chapters. The econometric specification takes the simplest intuitive approach to control for the overall depth of the agreement by counting the number of policy areas it covers (like the construction of the TF variable). The variable Depthjt in equation (B3.2.2), therefore, captures the “horizontal” depth of PTAs, which represents the inclusion of 52 policy areas in DTAs (Hofmann, Osnago, and Ruta 2017). The depth variable is constructed by counting the number of strictly enforceable provisions.a

a. For detailed information on the legal enforceability of provisions, see Hofmann, Osnago, and Ruta (2017).