Estimating Tariff Equivalents of Service Trade Restrictions
Andrew Bridger
Contents
- Summary
- Key findings
- What are tariff equivalents or AVEs?
- Motivation
- Finding 1
- Finding 2
- Finding 3
- Conclusion
Summary
Summary
Aim
- Provide 2021 (new) ad-valorem tariff equivalents (AVEs) of trade restrictions in seven service sectors for 50 countries
How
International Trade and Production Database for Estimation (ITPD-E) Release 2
OECD’s Services Trade Restrictiveness Index (STRI)
Detailed sector level structural gravity model
Why
Tariff equivalents have never been calculated for 2021
Tariff equivalents have not been calculated using the ITPD-E dataset (all other estimates use different datasets)
Key findings
More restrictive services trade policies are associated with lower international trade in all seven service sectors analysed.
Estimated trade costs are considerably lower for trade between European Economic Area (EEA) members than between countries on a most-favoured nation (MFN) status.
There is significant sectoral heterogeneity in the estimated trade costs across service sectors.
What are tariff equivalents or AVEs?
Trade costs are typically expressed as ad-valorem tariff equivalents (AVEs).
Ad-valorem is latin for ‘to value’
An AVE quantifies the impact of the STRI on the relative attractiveness of imported services compared to domestic services
The AVE is essentially a tax (or tariff) equivalent for regulatory barriers to services trade
Presenting results in AVE terms is relatively straightforward and provides an easy to understand number that can be compared across countries.
Calculating tariff equivalents
To convert STRI scores to an AVE, I follow Benz (2017) as in the equation below:
\[AVE^k_{ij,t} = exp \left(\dfrac{-STRI^k_{ij,t} * \beta_1^k}{(\sigma^k-1)} \right)-1\]
What do I need to calculate an AVE?
\(STRI^k_{ij,t}\) score retrieved from the OECD STRI database
\(\beta_1^k\) is the trade elasticity calculated via a structural gravity model
\(\sigma^k\) is the elasticity of substitution between varieties of traded services sourced from four recent studies
Motivation
Services account for two-thirds of global GDP
Impact of trade costs in goods trade has been comprehensively studied...
...but little focus on the impact of trade restrictions in services
Key findings
Finding 1
More restrictive services trade policies are associated with lower international trade in all seven service sectors analysed.
Finding 2
Estimated trade costs are considerably lower for trade between European Economic Area (EEA) members than between countries on a most-favoured nation (MFN) status.
Transport |
Distribution |
Construction |
Insurance |
Finance |
Business |
Info. services |
||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Country | MFN | EEA | MFN | EEA | MFN | EEA | MFN | EEA | MFN | EEA | MFN | EEA | MFN | EEA |
| AUS | 31.9 | 37.8 | 22.0 | 38.3 | 137.8 | 23.4 | 40.5 | |||||||
| AUT | 37.2 | 11.3 | 51.6 | 15.1 | 20.2 | 4.1 | 49.8 | 3.0 | 137.8 | 27.3 | 49.7 | 11.1 | 44.1 | 6.7 |
| BEL | 44.3 | 9.8 | 77.1 | 23.7 | 30.9 | 4.3 | 45.9 | 6.1 | 197.1 | 39.9 | 49.4 | 12.1 | 54.2 | 12.2 |
| BRA | 50.3 | 74.9 | 40.8 | 98.5 | 512.9 | 50.8 | 79.4 | |||||||
| CAN | 32.7 | 67.1 | 24.4 | 47.0 | 118.4 | 27.1 | 49.8 | |||||||
| CHE | 47.8 | 61.8 | 31.9 | 52.3 | 262.4 | 46.6 | 61.6 | |||||||
| CHL | 25.8 | 28.5 | 14.9 | 33.0 | 138.9 | 16.2 | 40.9 | |||||||
| CHN | 46.5 | 48.2 | 25.8 | 90.6 | 363.6 | 78.3 | 146.2 | |||||||
| COL | 31.6 | 31.7 | 24.6 | 47.0 | 240.8 | 20.7 | 52.9 | |||||||
| CRI | 39.6 | 43.8 | 21.4 | 38.1 | 113.3 | 32.6 | 41.7 | |||||||
| CZE | 30.8 | 10.3 | 31.4 | 12.8 | 14.4 | 2.5 | 21.0 | 1.3 | 54.6 | 9.9 | 24.5 | 10.0 | 28.3 | 7.4 |
| DEU | 29.9 | 9.1 | 39.9 | 13.6 | 16.8 | 1.6 | 27.5 | 2.4 | 111.3 | 42.6 | 29.4 | 9.6 | 27.8 | 5.1 |
| DNK | 35.3 | 7.2 | 45.3 | 14.5 | 22.4 | 2.5 | 36.8 | 3.6 | 126.8 | 20.8 | 33.9 | 5.9 | 38.1 | 5.0 |
| ESP | 34.5 | 8.7 | 41.3 | 12.8 | 22.6 | 2.5 | 37.3 | 4.9 | 59.0 | 4.8 | 37.0 | 7.7 | 32.0 | 4.2 |
| EST | 38.3 | 8.8 | 42.4 | 7.0 | 25.4 | 2.5 | 31.6 | 2.4 | 117.4 | 20.8 | 57.3 | 6.0 | 38.3 | 5.8 |
| FIN | 44.8 | 10.6 | 72.7 | 18.0 | 23.6 | 1.3 | 55.7 | 1.3 | 180.7 | 15.3 | 27.3 | 3.7 | 49.4 | 6.3 |
| FRA | 30.8 | 9.2 | 53.1 | 25.0 | 12.7 | 1.3 | 21.9 | 4.9 | 115.3 | 34.7 | 68.8 | 10.2 | 33.7 | 6.0 |
| GBR | 22.5 | 29.4 | 16.0 | 29.4 | 110.3 | 31.2 | 28.4 | |||||||
| GRC | 44.3 | 7.5 | 78.4 | 20.7 | 34.5 | 4.0 | 61.6 | 3.6 | 184.7 | 20.8 | 59.7 | 10.4 | 52.6 | 7.1 |
| HUN | 43.1 | 10.3 | 57.0 | 11.7 | 30.2 | 1.3 | 44.3 | 2.4 | 187.4 | 22.6 | 65.8 | 10.0 | 51.1 | 7.4 |
| IDN | 76.6 | 420.1 | 51.7 | 139.8 | 782.5 | 121.1 | 113.9 | |||||||
| IND | 78.0 | 177.3 | 37.7 | 161.7 | 856.4 | 147.8 | 70.9 | |||||||
| IRL | 33.7 | 8.7 | 48.2 | 15.9 | 17.1 | 1.3 | 25.4 | 1.3 | 94.9 | 15.3 | 28.5 | 8.2 | 35.5 | 6.9 |
| ISL | 62.5 | 9.8 | 130.3 | 33.7 | 65.7 | 10.2 | 88.1 | 4.9 | 361.4 | 46.7 | 59.8 | 11.0 | 101.4 | 13.5 |
| ISR | 64.6 | 44.2 | 38.3 | 57.5 | 198.5 | 49.4 | 69.9 | |||||||
| ITA | 41.7 | 10.9 | 50.8 | 19.2 | 37.2 | 5.4 | 61.3 | 4.4 | 172.8 | 33.5 | 68.9 | 7.6 | 52.4 | 10.9 |
| JPN | 30.1 | 26.9 | 11.4 | 30.1 | 134.5 | 40.4 | 33.4 | |||||||
| KAZ | 74.5 | 146.5 | 48.9 | 72.1 | 431.8 | 41.7 | 117.7 | |||||||
| KOR | 61.4 | 46.0 | 18.2 | 19.9 | 129.0 | 100.9 | 48.4 | |||||||
| LTU | 34.9 | 7.5 | 41.0 | 8.9 | 19.8 | 2.5 | 27.7 | 1.3 | 115.3 | 9.9 | 39.1 | 7.6 | 36.0 | 5.0 |
| LUX | 35.6 | 12.3 | 52.3 | 17.7 | 19.6 | 5.6 | 37.0 | 4.9 | 120.5 | 56.0 | 78.6 | 11.9 | 34.4 | 9.9 |
| LVA | 32.1 | 10.3 | 35.8 | 12.5 | 18.5 | 4.2 | 28.7 | 2.4 | 98.7 | 22.6 | 19.4 | 5.0 | 36.1 | 10.0 |
| MEX | 61.7 | 57.8 | 32.9 | 56.3 | 441.9 | 36.8 | 66.0 | |||||||
| MYS | 54.1 | 132.7 | 36.8 | 63.1 | 175.4 | 68.7 | 74.7 | |||||||
| NLD | 27.9 | 6.5 | 36.1 | 12.8 | 14.2 | 0.0 | 24.2 | 4.9 | 114.3 | 22.6 | 25.8 | 5.5 | 29.7 | 3.6 |
| NOR | 49.7 | 9.0 | 82.0 | 19.2 | 34.9 | 2.7 | 75.7 | 7.5 | 296.5 | 34.7 | 42.1 | 6.0 | 59.5 | 9.7 |
| NZL | 36.7 | 42.0 | 19.7 | 29.4 | 178.0 | 29.2 | 42.6 | |||||||
| PER | 43.9 | 47.5 | 23.2 | 47.3 | 156.5 | 26.8 | 59.7 | |||||||
| POL | 40.0 | 9.7 | 59.0 | 18.6 | 30.3 | 2.5 | 42.5 | 6.1 | 166.4 | 32.8 | 105.8 | 9.3 | 48.4 | 4.4 |
| PRT | 30.2 | 9.0 | 42.4 | 17.7 | 21.8 | 2.5 | 42.7 | 2.4 | 126.8 | 34.7 | 61.7 | 9.5 | 29.3 | 4.4 |
| RUS | 103.6 | 88.0 | 44.5 | 97.1 | 536.6 | 47.5 | 105.7 | |||||||
| SGP | 42.5 | 62.6 | 25.8 | 47.9 | 240.8 | 41.2 | 60.9 | |||||||
| SVK | 36.4 | 9.6 | 35.4 | 9.7 | 22.3 | 2.5 | 26.3 | 1.3 | 92.2 | 15.3 | 52.4 | 11.5 | 29.9 | 5.2 |
| SVN | 44.1 | 11.6 | 48.6 | 9.7 | 30.5 | 4.2 | 36.5 | 8.9 | 154.1 | 28.5 | 72.1 | 8.8 | 48.6 | 6.0 |
| SWE | 43.5 | 9.7 | 64.2 | 19.8 | 25.6 | 4.3 | 50.1 | 3.6 | 170.2 | 20.8 | 32.0 | 4.6 | 41.7 | 4.9 |
| THA | 97.8 | 129.2 | 53.0 | 180.7 | 640.7 | 142.4 | 91.9 | |||||||
| TUR | 58.4 | 57.0 | 34.0 | 45.7 | 272.8 | 86.9 | 73.0 | |||||||
| USA | 49.2 | 39.9 | 26.6 | 70.2 | 152.9 | 26.2 | 33.9 | |||||||
| VNM | 57.5 | 110.5 | 34.7 | 123.9 | 495.8 | 58.1 | 111.2 | |||||||
| ZAF | 48.6 | 61.0 | 25.4 | 39.1 | 317.7 | 39.5 | 60.5 | |||||||
Finding 2 continued
Additional data on Scotland and England in 2019. Note: not in my final dissertation.
Distribution |
Construction |
Business |
||||
|---|---|---|---|---|---|---|
| Country | MFN | EEA | MFN | EEA | MFN | EEA |
| SCO (2019) | 33.8 | 17.4 | 18.4 | 5.8 | 42.8 | 16.8 |
| ENG (2019) | 37.1 | 20.2 | 18.4 | 5.8 | 37.2 | 14.0 |
| GBR | 29.5 | 16.0 | 31.2 | |||
| IRE | 48.2 | 15.9 | 17.1 | 1.3 | 28.5 | 8.2 |
| NOR | 82.0 | 19.2 | 34.9 | 2.7 | 42.1 | 6.0 |
Finding 3
There is significant sectoral heterogeneity in the estimated trade costs across service sectors.
Conclusion
Future research
As more data becomes available, estimations should become more robust and allow for the use of country-pair fixed effects.
Bilateral STRIs could be created by incorporating services PTAs.
A ‘water’ variable could be used to quantify the effects of binding commitments in services trade.
Further research could investigate the impact of mode of services restrictions and their impact on services trade.
Counterfactual STRIs as in Shepherd (2019) could also be created to investigate the effects of potential policy changes.
Conclusion
The findings are broadly similar to the recent literature.
Trade costs in services are high but large differences in trade costs across service sectors.
Policymakers should be aware of different business models, competition challenges, and regulatory frameworks within sectors.
In contrast with previous studies, I find mostly positive and significant results for the EEA dummy, suggesting not all variation is captured using the intra-EEA data.
Acknowledgements
I would like to give a special thanks to Julija Harrasova and Ludovic Maguire (Ludo). Their intelligence, guidance and calming influence was invaluable.
Thanks to my mate, Joe Paul, for helping me render this presentation in Quarto.
Code will be available to reproduce my research soon. Available on my GitHub or my Website.
Theory
The gravity model
Newton’s Law of Gravity
\[F_{ij} = G \frac{M_{i}M_{j}}{D_{ij}^2}\] Where:
\(F_{ij}\) is the magnitude of the
gravitational force between objects \(i\) and \(j\)\(G\) is the gravitational constant
\(M_{i}\) is object \(i\)’s mass
\(M_{j}\) is object \(j\)’s mass
\(D_{ij}\) is the distance between objects \(i\) and \(j\)
Gravity Trade Model
\[X_{ij} = G \frac{Y_{i}E_{j}}{T^{\theta}_{ij}}\] Where:
\(X_{ij}\) is trade from country \(i\) to \(j\)
\(G\) is the inverse of world production \(G=\frac{1}{Y}\)
\(Y_{i}\) is country \(i\)’s domestic production
\(E_{j}\) is country \(j\)’s aggregate
expenditure\(T_{ij}\) is total trade frictions between countries \(i\) and \(j\)
\(\theta\) is the trade elasticity
Best practice structural gravity modelling
Table adapted from Yotov (2022).
Review of OECD STRI gravity literature
*Includes service FTAs in STRI framework. Table adapted from Fraser (2021).
Estimation strategy
\[X^k_{ij,t} = exp[\beta_1 STRI^k_{j,t}*INTER_{ij} + \beta_2 INTER_{ij} + Z_{ij,t} + \pi^k_{i,t} + \chi^k_{j,t}] * \epsilon^k_{ij,t}\]
Where:
\(X^k_{ij,t}\) is exports from country \(i\) to country \(j\) in sector \(k\) in year \(t\)
\(STRI^k_{jt}*INTER_{ij}\) represents the interaction between international trade and sector-specific STRI score \(k\) of the importing country \(j\) in year \(t\) (that is, the trade elasticity)
\(INTER_{ij}\) is a dummy for international trade observations
\(Z_{ij,t}\) is a vector of bilateral gravity covariates
\(\pi^k_{i,t}\) is an exporter-year fixed effect
\(\chi^k_{j,t}\) is an importer-year fixed effect
\(\epsilon^k_{ij,t}\) is the error term
Estimation strategy covariates
\(Z_{ij,t}\) is a vector of bilateral gravity covariates that includes:
\(ln(DIST_{ij})\) is the log of population weighted distance between country pairs
\(CNTG_{ij}\) is a dummy for contiguity (country pairs share a common border)
\(CLNY_{ij}\) is a dummy for ever being in a colonial relationship
\(LANG_{ij}\) is a dummy for common language
\(SPTA_{ij,t}\) is a dummy for services preferential trade agreement (PTA)
\(EEA_{ij}\) is a dummy if both countries are in the EEA
Trade elasticity results
Regression results
| Transport | Distribution | Construction | Insurance | Finance | Business services | Info. services | |
|---|---|---|---|---|---|---|---|
| Trade elasticity | -3.473*** | -7.167*** | -3.240** | -3.818*** | -8.426*** | -3.351*** | -4.722*** |
| (0.483) | (1.527) | (1.429) | (0.846) | (1.033) | (0.592) | (0.589) | |
| Int. Border | -4.379*** | -8.138*** | -7.048*** | -5.258*** | -3.739*** | -4.555*** | -4.057*** |
| (0.218) | (0.382) | (0.544) | (0.458) | (0.464) | (0.340) | (0.282) | |
| Log distance | -0.007 | 0.256** | -0.434** | -0.326** | 0.153 | -0.040 | -0.289*** |
| (0.085) | (0.122) | (0.210) | (0.144) | (0.140) | (0.096) | (0.098) | |
| Contiguity | 0.875*** | 1.497*** | 0.410 | 0.278 | 0.820** | 0.730*** | 0.285 |
| (0.155) | (0.247) | (0.342) | (0.366) | (0.374) | (0.208) | (0.202) | |
| Colony ever | 0.646*** | 0.163 | 0.920*** | 1.089*** | 0.879** | 0.579** | 0.510** |
| (0.203) | (0.328) | (0.328) | (0.329) | (0.355) | (0.268) | (0.208) | |
| Common language | 0.294** | 0.244 | -0.032 | 0.627*** | 0.680*** | 0.304 | 0.506*** |
| (0.119) | (0.189) | (0.327) | (0.234) | (0.262) | (0.202) | (0.145) | |
| SPTA | 0.167 | 0.284 | 0.663** | -0.187 | -0.481* | -0.130 | -0.123 |
| (0.123) | (0.254) | (0.296) | (0.203) | (0.272) | (0.177) | (0.187) | |
| EEA | 0.773** | 1.195* | 1.523** | 0.252 | 0.266 | 0.045 | 0.429 |
| (0.331) | (0.617) | (0.716) | (0.537) | (0.514) | (0.492) | (0.357) | |
| Num.Obs. | 10153 | 7729 | 7865 | 8479 | 9218 | 10393 | 10079 |
| R2 Adj. | 0.990 | 0.999 | 0.999 | 0.997 | 0.982 | 0.991 | 0.994 |
Note: ^^ Robust standard errors are clustered by exporter, importer and year in parentheses below the parameter estimates. Statistical significance is indicated as follows: * (10%), ** (5%), and *** (1%).
AVEs
AVE summary statistics
MFN |
Intra-EEA |
|||||||
|---|---|---|---|---|---|---|---|---|
| Industry | Median | Mean | Min | Max | Median | Mean | Min | Max |
| Transport | 42.8 | 45.9 | 22.5 | 103.6 | 9.7 | 9.5 | 6.5 | 12.3 |
| Distribution | 51.2 | 68.6 | 26.9 | 420.1 | 15.5 | 16.3 | 7.0 | 33.7 |
| Construction | 25.4 | 27.9 | 11.4 | 65.7 | 2.5 | 3.2 | 0.0 | 10.2 |
| Insurance | 45.8 | 55.1 | 19.9 | 180.7 | 3.6 | 3.8 | 1.3 | 8.9 |
| Finance | 168.3 | 233.9 | 54.6 | 856.4 | 22.6 | 26.4 | 4.8 | 56.0 |
| Business services | 44.3 | 52.4 | 16.2 | 147.8 | 9.0 | 8.5 | 3.7 | 12.1 |
| Info. services | 49.0 | 56.2 | 27.8 | 146.2 | 6.1 | 7.0 | 3.6 | 13.5 |
| a Note: Intra-EEA refers to services trade between two EEA members. MFN refers to all other trading relationships. |
||||||||
AVEs on most-favored nation status
Further information
Limitations (Trade elasticity)
Relatively short time span of the panel
Interaction term and controlling for service PTAs
Omitted variable bias
Granularity of services sector trade data
Service sector performance
Mode of supply in service sector
Limitations (AVEs)
AVEs are based on observed values not prices
Assumes trade costs are symmetrical
Might not be actionable from a policy perspective
From traditional gravity to structural gravity
“The golden age of structural gravity” - the inclusion of multilateral resistance terms.
The equation below provides the theoretical demand side sectoral structural gravity model from Anderson (2004): \[\label{eq3}
X^k_{ij} = \frac{Y^k_i E^k_{j}}{Y^k} \left( \frac{t^k_{ij}}{\Pi^k_i P^k_j} \right)^{1-\sigma^k}\]
The important multilateral resistance terms:
\(\Pi^k_{i,t}\) is the outward multilateral resistance term
\(P^k_{i,t}\) is the inward multilateral resistance term
From traditional gravity to structural gravity
Issues with OLS because:
Heteroscedasticity often affects trade data
OLS estimates of log-linearised forms are biased and inconsistent in the presence of heteroscedasticity
Further, OLS estimation requires the removal of observations with zero trade value, which are common in trade data
Solution is to use Poisson Pseudo Maximum Likelihood (PPML) estimator.
Data utilised
Trade data is the ITPD-E Release 2 compiled for the United States International Trade Commission (USITC).
I supplement the ITPD-E data by appending missing domestic trade data for a small group of countries with data sourced and transformed from the OECD, IMF and Asian Development Bank. Details are in Appendix C of the paper.
Policy variable are the OECD’s Services Trade Restrictiveness Indexes (STRI):
MFN STRI
intra-EEA STRI
Gravity covariates are sourced from the the USITC’s Dynamic Gravity Dataset (DGD).
ITPD-E Release 2
The ITPD-E is well suited for estimation with structural gravity models of trade.
Published in July 2022
Superior country coverage - covers 265 countries and includes data on domestic trade for most countries, which is required for theory consistent estimates
Not restricted to just manufacturing - 170 industries including 17 service industries
ITPD-E provides consistent panel data over many years and is developed using administrative data, which does not include information estimated by statistical techniques.
Statistically constructed data should be avoided in estimation, as it is likely to bias the estimates through overfitting.
Data compilation
An overview of how the data was compiled is provided below. The detailed methodology for data compilation can be found in the reproducible R code.
Filter ITPD-E to obtain trade data from 2014 to 2019 and sectors included in the analysis.
Extend the ITPD-E data by filling in missing domestic trade data for several countries.
Map the MFN STRI and the intra-EEA STRI to the ITPD-E data. Only four of the seven services sectors analysed in the ITPD-E correspond directly to the STRI sectors.
If both countries are in the EEA, use the intra-EEA STRI and use the MFN STRI if otherwise.