Economic Implications of Data Regulation

Annex B. Econometric analysis

The gravity model of trade expresses trade flows as a function of the (economic) size of the trading countries and trade costs. A generic sector-specific structural gravity equation can be expressed as:

${\left(X_{ij}\right)}^k=\frac{Y_i^k{E}_j^k}{Y^k}{\left(\frac{T_{ij}^k}{{{\mathrm{\Pi }}_i^{k}P}_j^k}\right)}^{-{Ѳ}^k}$ (1)

where trade flows from economy i to economy j in sector k, $X_{ij}^k$, are a function of the supply of sector k-goods from economy i, $Y_i^k$, and expenditure for sector k-goods in economy j, $E_j^k$. $T_{ij}^k>1$are trade costs when sector k-goods are shipped from exporter-economy i to importer-economy j. ${k }^{}$is the sector-specific trade elasticity, and ${\mathrm{\Pi }}_i^k$and $P_j^k$are the price indices representing outward and inward multilateral resistance terms, respectively. The size term is captured by $\frac{Y_i^k{E}_j^k}{Y^k}$ and shows the hypothetical level of frictionless trade between two countries, which is proportional to their overall share of global economic activity. The trade cost term, $\frac{T_{ij}^k}{{{\mathrm{\Pi }}_i^{k}P}_j^k}$, is a scaling factor that takes into account trade frictions.

Trade flows ($X_{ijt}^k$), taken from the OECD Trade in Value Added database, are regressed against standard gravity variables including:

• The log of bilateral distance (${lndist}_{ij}$); contiguity (${contig}_{ij})$; common official language (${comlang}_{ij})$; and colonial history (${colony}_{ij})$ – from the CEPII database.

• To capture digital connectivity, the minimum value of the log of the percentage of the population with access to the Internet between economy pairs is used (${minlndigi}_{ijt})$ – from the International Telecommunications Union database.

• A dummy variable capturing the presence of Regional Trade Agreements (RTAs) is also used – from the CEPII database.

• To identify the domestic policy environment for digital trade, the Digital STRI (${DSTRI}_{it}$) is included.1

The specifications include exporter-sector-year and importer-sector-year fixed effects (${\mathrm{\eta }}_{it}^k$ and ${\mathrm{\mu }}_{jt}^k)$ controlling for all time-varying economy-specific unobservable variables, including multilateral resistance terms. They are estimated using PPML with high dimensional fixed effects. PPML (Poisson Pseudo Maximum Likelihood) allows to account for hetereoscedasticity and for zero trade flows (Figure X).

$X_{ijt}^k=\exp \left({lndist}_{ij}+{border}_{ij}+{minlndigi}_{ijt}+{border}_{ij}*{DSTRI}_{it}+{RTA}_{ijt}+{RTA\_ecomm\_\_chapter}_{ijt}+ {contig}_{ij}+{comlang}_{ij}+{colony}_{ij}+ {\mathrm{\eta }}_{it}^k+{\mathrm{\mu }}_{jt}^k\right)*{\mathrm{\epsilon }}_{ijt}^k$ (2)

Table A B.1. Impact of digital connectivity and digital trade policies on trade, 2014-18

	All	Agriculture	Manufacturing	ICT goods	ICT services	DD services	Other services
Log of distance	-1.437***	-1.452***	-1.140***	-0.971***	-1.416***	-1.246***	-1.551***
	(-128.08)	(-102.52)	(-162.23)	(-48.36)	(-72.65)	(-90.22)	(-126.40)
Minimum digital connectivity	0.218***	0.432***	0.360***	0.749***	0.337	0.530***	0.275***
	(4.98)	(3.57)	(8.08)	(7.61)	(1.39)	(4.39)	(2.95)
Border*digital STRI of reporter	-8.964***	-10.16***	-7.784***	-3.067***	-9.482***	-13.28***	-11.21***
	(-40.48)	(-40.70)	(-60.30)	(-11.59)	(-19.47)	(-50.84)	(-39.97)
RTA	-1.012***	-0.779***	-0.356***	-0.186***	-1.024***	-1.183***	-1.445***
	(-65.68)	(-22.67)	(-21.17)	(-4.49)	(-21.10)	(-38.28)	(-51.48)
Contiguity	-0.816***	-0.525***	-0.138***	-0.171**	-1.797***	-1.815***	-1.003***
	(-19.09)	(-7.47)	(-4.20)	(-2.28)	(-10.11)	(-14.53)	(-10.60)
Common language	-0.736***	-0.914***	-0.441***	0.0135	-0.001	0.242***	-0.897***
	(-15.34)	(-7.04)	(-8.37)	-0.21	(-0.01)	-3.76	(-9.50)
Colony	-0.294***	-0.0779	-0.183***	-0.403***	-0.551***	-0.493***	-0.552***
	(-9.61)	(-0.84)	(-5.22)	(-4.86)	(-5.83)	(-7.59)	(-10.16)
Constant	20.60***	20.10***	17.67***	14.94***	19.12***	18.17***	21.19***
	-102.65	-40.23	-92.58	-31.88	-18.38	-33.32	-53.23
N	822 984	55 250	257 984	37 050	37 050	74 100	290 375

Note: Estimated using ppml and reporter-sector-year and partner-sector-year fixed effects. Digital connectivity defined as the minimum, across a dyad of the share of people connected to the Internet.

Source: Own calculations using TiVA 2021 database.

Note

← 1.  Introducing the DSTRI leads to a shorter period covered in the regressions since availability of data for this indicator begins in 2014.