Interaction between corruption and the GATT-WTO trade effect: a panel data analysis Suryadipta Roy * Phillips School of Business - 210, High Point Universy, High Point, NC - 27262, USA. Email: sroy@highpoint.edu Abstract Using cross-country time series data for 133 countries between 1984 and 2010, and implementing a fixed effects Poisson estimator using the gravy model of trade, we study the interaction between corruption and the treatment effect of a country joining the GATT-WTO on bilateral trade. Our results suggest an asymmetric effect of corruption on the GATT-WTO, in that domestic corruption has a negative impact on the GATT exporters while corruption in the trade partners has a posive impact on the GATT importers. The insight highlights the importance of accounting for instutional heterogeney among countries to explain mixed results on the effectiveness of the GATT-WTO for global trade. Keywords: corruption; GATT/WTO; interaction; gravy; panel data JEL codes: F13; C23; 73; O19. * Corresponding author. Phillips School of Business - 210, High Point Universy, High Point, NC - 27262, USA. Email: sroy@highpoint.edu Phone: 336-841-9163.
1. Introduction There exists a vast lerature on the impact of corruption on bilateral trade. Most of these studies (e.g. Kaufman and Wei (1999), Anderson and Marcouillier (2002), de Groot et. al. (2004), Jansen and Nordas (2004), Thede and Gustafson (2012)) find a detrimental effect of corruption on trade, while utt and Traca (2010) suggest a posive effect of corruption on trade in countries wh high rates of tariff protection (less than 15% of observations). In contrast, evidences on the effect of the General Agreement on Tariffs and Trade (GATT) and the World Trade Organization (WTO) on bilateral trade are more mixed. Rose (2004) and Eicher and Henn (2011) find that the effects of the WTO on trade are not statistically significant. On the other hand, Subramanian and Wei (2007) and Liu (2009) suggest posive GATT-WTO 1 trade effects. In this paper, we investigate the interaction between corruption and the GATT-WTO effect on bilateral trade using panel data for 133 countries. Our hypothesis is that the impact of a country s joining the GATT-WTO on bilateral trade, eher as an exporter or an importer, is contingent on domestic corruption as well as on the corruption levels of s trading partner. The paper contributes to the debate on the effectiveness of the GATT-WTO in fostering international trade. The interaction term between the GATT treatment effect and corruption in our empirical model enables us to uncover an asymmetry in the effect of corruption between the GATT and the non-gatt member countries. Moreover, we consider separate effects of the GATT-WTO on the importer and the exporter country. The paper also contributes to the lerature on estimations of the gravy model of bilateral trade. The empirical analysis employs a theoretically-consistent corruption-augmented gravy model by directly estimating bilateral imports using a country-pair and year fixed effects Poisson estimator in a panel setup. The 1 Throughout the paper the terms GATT, GATT-WTO, or GATT/WTO have been used interchangeably. The effect of the 1994 Uruguay Round of the GATT is denoted as WTO.
country-pair fixed effects control for multilateral resistance terms in the gravy equation (Anderson and van Wincoop (2003)) and the Poisson estimator enables us to account for the presence of zero trade flows between countries. Our results indicate negative effect of domestic corruption on the GATT exporters, and posive effect of trading partners corruption on the GATT importers. 2. ata and model The dependent variable in the regressions is the nominal value of bilateral imports by importer i from exporter j at time period t in current US dollars ( M i ) available from the IMF irection of Trade Statistics (IMF-OT). Corruption data are from the widely used International Country Risk Guide (ICRG) ratings between 1984 and 2010. The measure is an assessment of corruption whin the polical system reflected in the demand for special payments and bribes to obtain regular official services (financial corruption) as well as government patronage or special favors for select business interests (polical corruption). The index ranges from 0-6 wh higher values denoting lower corruption. For interpretation purposes, the measure has been standardized and rescaled wh higher values denoting greater corruption. The treatment effect of the GATT- WTO separately on the importer and the exporter country is captured by two dummy variables and, where the indicator turns one once country i (/j) joins the GATT as an importer (/exporter) in year t. For example, given that Albania joined the GATT on 2000, both the indicators equal zero between 1984 and 1999, and one between 2000 and 2010. On the other hand, since the Uned States (US) has been a member of the GATT since 1948, the dummy variables equal one for the US throughout the sample period.
The condional mean function of our most comprehensive specification is: E( M i / X ) ( aij at 1 2 3Corr 4Corr 11 # Corr # Corr # Corr ) (1) 12 # Corr 21 22 Z Assuming that the dependent variable is distributed as Poisson, the cumulative distribution function takes the form (.) exp(.) where (.) denotes the expression whin the parentheses. Robustness and consistency properties of the Poisson estimator have been discussed in Wooldridge (2010). The seminal paper by Santos Silva and Tenreryo (2006) introduced the Poisson estimator in the gravy lerature. aij -s are fixed effects capturing time-invariant country pair-specific heterogeney, a t are year fixed effects capturing time-varying heterogeney, Corr Corr ) denotes corruption level for importer i (/exporter j) at time t, and Z denotes the (/ vector of time-varying control variables that explain bilateral trade in a gravy framework. These include (logarhms of) real GP, real GP per capa, and dummy variables indicating if the trading partners share a common currency, presence of a regional trading agreement (RTA), and if a developed country has accorded Generalized System of Preferences (GSP) treatment to a developing country trade partner. ata for real GP and per capa real GP are from the World Bank s World evelopment Indicators for different years. Indicators for common currency, existence of regional trading agreement, and GSP treatment have been obtained from Head et al. (2010). Since country-pair fixed effects 2 absorb all heterogeney that is constant over time, important variables that have been found to determine bilateral trade but remain constant over 2 According to Baier and Bergstrand (2007), the endogeney issue of the effect of trade agreements in gravy equations (i.e. GATT/WTO in our case) is effectively addressed using panel data wh bilateral fixed and country-and-time effects. Since we cannot incorporate the time-varying fixed effects in our model, we undertake robustness checks including multilateral resistance terms. As discussed in Section 4, the major results are found to remain unaffected.
time, e.g. distance or common language, get dropped from this specification. For this purpose, we implemented a specification by including only the country and year fixed effects to highlight the effect of these time-invariant variables on bilateral trade. The key variables of interest are the interaction terms between the treatment effect of a country joining the GATT-WTO and the corruption index. The terms # Corr and # Corr capture the interaction between domestic corruption and the GATT membership for the importer and the exporter country respectively. # Corr and Corr # capture the partner-country interaction effects. As noted in Ai and Norton (2003) and Greene (2010), given that equation (1) is non-linear, coefficients of the interaction terms do not measure the change in the partial effect of eher corruption or GATT-WTO on the condional mean function. For example, from equation (1), the partial effect of exporter-country corruption on bilateral trade is: E M / X ) Corr (.)( ) (2a) ( i 4 12 22 and the effect of a regime swch due to a country joining the GATT-WTO as an exporter (i.e. changing from 0 to 1) on the partial effect of exporter-country corruption is: ( E( M t ij / X ) Corr ) (.)( 4 22 12 ) ( 4 12 ) ( a ij a t Corr Corr Corr Corr ) (2b) 1 3 4 11 12 Z Greene (2010) suggests that graphical presentations are a very informative adjunct to numerical statistical results for interpreting interaction effects in non-linear models. Following his suggestion, we plot the predicted values of average bilateral imports from our comprehensive specification for the GATT and the non-gatt countries as functions of the corruption levels.
43.9 43.9 44 44 44.1 44.2 44.3 44.1 44.2 44.3 43.8 43.9 43.6 43.8 44 44 44.1 44.2 44.3 44.2 44.4 3. Results and iscussion Results for the fixed effects Poisson regression model for our sample of 133 countries are reported in Table 1. For hypothesis testing, we implement heteroskedasticy-robust standard errors that allow for clustering whin country pairs- this addresses issues of overdispersion associated wh Poisson distributions as well as that of serial correlation (Cameron and Trivedi (2009)). First, we examine the interaction effects separately for the GATT importers and exporters wh respect to domestic and partner-country corruption levels; the results are reported in cols. (1) through (4). Col. (5) reports results for the comprehensive specification in equation (1). -2.19 0 2.25 Corruption-exporter j=0 j=1-2.19 0 2.25 Corruption-exporter i=0 i=1-2.19 0 2.25 Corruption-importer i=0 i=1-2.19 0 2.25 Corruption-importer j=0 j=1 Fig.1: Interaction between domestic and partner-country corruption and GATT-WTO; i=1 (/j=1) denotes importer i (/exporter j) to be in the GATT-WTO
In order to interpret the interaction effects, Fig.1 plots the predicted mean of bilateral imports from the comprehensive specification as a function of corruption (on the x-axis) wh separate lines to represent the GATT- and the non-gatt countries. The comprehensive specification is preferred to other models based on the Akaike Information Crerion (AIC) and the Bayesian Information Crerion (BIC) model selection creria. The significant interaction effects can be viewed as the difference in slopes for the two groups. The upper-left corner of Fig.1 suggests a highly negative impact of domestic corruption on the GATT exporters- the interaction term # Corr (in col. (2)) is negative and highly significant (at 0.1% level). The diagram shows that at very high corruption levels, bilateral exports of the GATT members and the non-gatt countries are almost equal. The upper-right corner of Fig.1 indicates a posive effect of partner-country corruption on the GATT importers, since the trade gap between the GATT importers and the non-gatt importers progressively widens wh increase in exporters corruption levels. This is based on the posive and significant interaction term # Corr in col. (4). As shown in the bottom-left corner of Fig.1, domestic corruption has had a slightly greater effect on the GATT importers ( # Corr term negative but not significant). The GATT membership dummies are posive and significant- this is reflected in the higher trade levels of the GATT members compared to the non-gatt countries in Fig. 1. Col. (6) reports the specification that controls for country and year fixed effects only- this allows to us to incorporate important time-invariant variables in the gravy model. We observe that distance reduces trade, while common language, contiguy, and colonial relationship lead to greater bilateral trade. Overall, results for the control variables in all the specifications mostly conform to the existing lerature on trade gravy models (see e.g. the survey in Bergstrand and Egger (2011)). Real GP for both the importer and the exporter countries are posively related
to bilateral trade- the coefficients are less than one supporting the stylized fact that large economies are less open to trade. Having a common currency and sharing an RTA are posively related to trade while the GSP dummy is found to be negatively associated wh trade. Table 2 reports the partial effects of domestic and partner-country corruption for the different GATT-WTO membership status based on equation (2b). One standard deviation increase in domestic corruption among the GATT exporters is found to be associated wh 9% reduction in bilateral exports. On the other hand, among the GATT importers, a similar increase in partners corruption is associated wh 11% increase in bilateral imports. The results support our hypothesis that the impact of the GATT-WTO on bilateral trade, eher as an exporter or an importer, is contingent on domestic corruption as well as on the corruption levels of s trading partner. While trade liberalization pursued under the GATT and the WTO has been successful in general, members wh high corruption levels have failed to generate higher exports. It is to be noted that epken and Sonora (2005) and Francois and Manchin (2013) have suggested that the qualy of instutions matter strongly for the exporter countries to generate higher exports. This is because countries wh high corruption levels face significant fixed costs of exporting (besides high transactions costs) that prevent them for participating in international trade. By including observations on country-pairs wh zero bilateral trade, the Poisson regression model implemented in our study captures the effect of these fixed costs due to corruption on trade participation and thereby on bilateral trade. While the significant posive effect of exporters corruption on the GATT importers is counterintuive, is likely that membership in the multilateral trade liberalization program has migated the negative effects of partners corruption on the GATT importers.
4. Robustness checks Other robustness checks have also been conducted. First, corruption might be endogenous in equation (1) due to feedback effect from bilateral trade on corruption levels. We address this problem by using one-period lagged value of corruption as explanatory variable. Second, our definion of the treatment effects of the GATT/WTO precludes us from including time-varying importer-year and exporter-year fixed effects to control for multilateral resistance in the gravy model, since / will be collinear wh the fixed effects and hence get dropped from the regression. For this purpose, following Head (2003), we control for multilateral trade resistance by including a proxy measure for remoteness for any country i using the following formula: is tan ceij Re moteness, where GP wt denotes world GP in year t. Third, GP / GP j wt we redefine the treatment effects to capture the interaction between corruption and the effect of the 1994 Uruguay Round and the establishment of the WTO on bilateral trade. This is based on the motivation that the conclusion of the Uruguay Round in 1994 and the founding of the WTO beginning 1995 have been associated wh greater obligations on the new member countries, viz. the developing nations, to undergo substantial trade liberalization (Subramanian and Wei (2007)). Thus the post-1995 effect of the WTO and s interaction wh corruption on bilateral trade might be different from the pre-1995 effect of the GATT. Thus, we redefine (/ ) such that the indicators equal zero for all countries between 1984 and 1994, and one between 1995 and 2010. Finally, instead of the ICRG corruption data, we use the Control of Corruption measure from the World Governance Indicators (WGI) compiled by Kaufman, Kraay, and Mastruzzi (2010). The WGI corruption data are available every alternate years from 1996 till 2004, and yearly onwards.
Results for the robustness checks based on the comprehensive specification in eq. (1) as reported in Table 3 mostly reinforce our main conclusions. While the GATT-WTO treatment effects are found to be posive and highly significant, there is evidence of the effect of corruption on the GATT members. This is irrespective of (i) addressing the endogenous effect of corruption (in col. (7)); (ii) endogeney of trade agreements (in col. (8)); (iii) effect of the Uruguay Round in col. (9); and (iv) using an alternate measure of corruption (col. (10)). Based on the sign of the interaction terms, the posive effect of partner country corruption on the GATT importers remains significant. The negative effect of domestic corruption among the GATT exporters also remains significant, except for the final specification (col. (10)) which uses the WGI corruption measure as the explanatory variable. This specification, on the other hand, reports a negative impact of domestic corruption on the GATT importers. However, is to be noted that this regression uses a much smaller number of observations due to the shorter time span of the WGI dataset as compared to the ones based on the ICRG corruption measure. Overall, the results suggest the importance of accounting for the heterogeneous effect of domestic and partners corruption levels on the effect of the GATT/WTO on bilateral trade. 5. Conclusion We implement a non-linear fixed effects Poisson estimator in the gravy model of trade to empirically assess the impact of corruption, the trade effect of the GATT-WTO, and their interaction on bilateral trade in 133 countries over the period 1984-2010. We introduce a new result in the trade lerature on the effectiveness of the GATT-WTO in promoting trade between countries. Greater domestic corruption is found to reduce bilateral exports for the GATT members. We also find that imports for the GATT members are posively associated wh
corruption levels in their trading partners. The results suggest important sources of asymmetry in the effect of corruption on bilateral trade. We believe that this instutional asymmetry needs to be given serious consideration to evaluate the effectiveness of the GATT-WTO for global trade liberalization.
Table 1: Fixed effects Poisson regressions ependent variable - Bilateral Imports in current US dollars M ) (1) (2) (3) (4) (5) (6) Condional fixed effects Uncondional fixed effects 0.28*** 0.33*** 0.35*** 0.35*** [0.06] [0.05] [0.06] 0.25** 0.18* 0.26** 0.22** [0.08] Corr -0.02 0.01 0.004 0.03 [0.05] Corr 0.10** -0.07* 0.02 0.01 [0.05] [0.05] # Corr 0.002-0.02-0.02 # Corr -0.11*** -0.11*** -0.07* # Corr -0.03 0.01-0.03 [0.05] # Corr 0.08* 0.08* 0.06 [0.05] log(real GP ) 0.82*** [0.13] 0.97*** [0.13] 0.96*** [0.13] 0.84*** 0.83*** 0.60*** [0.17] log(percapa Real GP ) 0.15 0.12 0.13 0.13 0.13 0.31+ log(real 1.06*** 0.92*** 0.93*** 1.05*** 0.91*** 0.73*** GP ) log(percapa 0.08 Real GP ) Common 0.20*** currency RTA 0.27*** GSP -0.20* Log( ist ) ij Common Language Common Border Former Colony Landlocked 0.09 0.20*** 0.28*** -0.16+ 0.10 0.20*** 0.27** -0.18* 0.09 0.19*** 0.27*** -0.20* 0.09 0.20*** 0.28*** -0.16+ ( i 0.25 0.11* [0.06] 0.69*** 0.05 [0.10] -0.55*** 0.18* 0.45*** [0.08] 0.28** -0.29 [0.48]
N 399648 399648 399648 399648 399648 413185 Pseudo-R 2 0.93 Country-pair Yes Yes Yes Yes Yes fixed effects Year fixed Yes Yes Yes Yes Yes Yes effects Importer Yes fixed effects Exporter Yes fixed effects AIC 1.80e+13 1.80e+13 1.80e+13 1.80e+13 1.78e+13 7.23e+13 BIC 1.80e+13 1.80e+13 1.80e+13 1.80e+13 1.78e+13 7.23e+13 Note: Robust standard errors adjusted for clustering whin country-pair in brackets; + p<0.1, * p<0.05, **p<0.01, *** p<0.001; estimates for year and country-pair/importer/exporter fixed effects not reported; / denotes importer i (exporter j) joining GATT-WTO in year t Table 2: Marginal effects of corruption dm i dicrg dm i dicrg 0, 0 0.003+ 0.02 1, 0 0.01-0.09** 1, 0-0.02 0.11** Note: Based on model in col. (5) of Table 1; + p<0.1, * p<0.05, **p<0.01, *** p<0.001
Table 3: Condional Fixed effects Poisson regressions ependent variable - Bilateral Imports in current US dollars M ) (7) (8) (9) (10) 0.33*** 0.35*** 0.37*** 0.47*** [0.06] [0.06] 0.25** 0.25** 0.19** 0.39*** [0.08] [0.06] Corr 0.01-0.01-0.06** -0.01 [0.02] Corr 0.03 0.01-0.05* -0.08 [0.05] [0.05] [0.02] [0.11] # Corr -0.03-0.02 0.02-0.19* # Corr -0.10** -0.09** -0.05+ -0.09 [0.02] [0.11] # Corr 0.01 0.02 0.04 0.09 [0.06] # Corr 0.08* 0.08* 0.11*** 0.15*** log(real GP ) 0.91*** 0.84*** 0.98*** 1.39*** [0.13] log(percapa Real GP ) 0.07 0.13 0.001 [0.15] -0.25 log(real GP ) 1.10*** 0.83*** 1.07*** 1.16*** [0.15] [0.15] [0.15] log(percapa Real GP ) -0.01 0.17-0.01 [0.17] -0.31+ [0.17] Common currency 0.20*** 0.22*** 0.23*** 0.26*** [0.05] RTA 0.26*** 0.27*** 0.25** -0.11 GSP -0.17+ -0.16* [0.08] -0.18+ -0.32** Re moteness -2.80e-10 [4.85e-10] Re moteness 2.58e-09*** [8.08e-10] N 382254 403693 317088 136284 Country-pair Yes Yes Yes Yes fixed effects Year fixed effects Yes Yes Yes Yes Note: Robust standard errors adjusted for clustering whin country-pair in brackets; + p<0.1, * p<0.05, **p<0.01, *** p<0.001; estimates for year and country-pair fixed effects not reported; Corr / Corr in column (7) denotes lagged corruption; / in column (9) denotes the beginning of the Uruguay Round and establishment of WTO; ( i Corr / Corr in column (10)
denotes WGI corruption measure; for importer i/exporter j. Re moteness / Re moteness indicates remoteness measure
References: Ai, C. and Norton, E. (2003) Interaction terms in log and prob models Economics Letters 80, 123-129. Anderson, J. and Marcouiller,. (2002) Insecury and the pattern of trade: an empirical investigation The Review of Economics and Statistics 84, 342-352. Anderson, J.E. and van Wincoop, E. (2003) Gravy wh Gravas: A Solution to the Border Puzzle American Economic Review 93, 170-192. Baier, S.L. and J. Bergstrand (2007) o free trade agreements actually increase members international trade? Journal of International Economics 71, 72-95. Bergstrand, J.H., and Egger, P. (2011) Gravy equations and economic frictions in the world economy in: Bernhofen,., Falvey, R., Greenaway,., Kreickemeier, U. (Eds.), Palgrave handbook of international trade, 532-570. Cameron, A.C. and P. Trivedi (2009) Microeconometrics using Stata Revised Edion, Stata Press. de Groot, H., Linders, G., Rietveld,P. and Subramanian, U. (2004) The instutional determinants of bilateral trade patterns Kyklos 57, 103-123. epken, C.A. and R.J. Sonora (2005) Asymmetric effects of economic freedom on international trade flows International Journal of Business and Economics, 4(2), 141-155. utt, P., and Traca,. (2010) Corruption and Bilateral Trade Flows: Extortion or Evasion? The Review of Economics and Statistics 92, 843-860. Eicher, T., and Henn, C. (2011) In search of WTO trade effects: preferential trade agreements promote trade strongly, but unevenly Journal of International Economics 83, 137-153.
Francois, J. and Manchin, M. (2013) Instutions, Infrastructure, and Trade World evelopment 46, 165-175. Greene, W. (2010) Testing hypotheses about interaction terms in nonlinear models Economics Letters 107, 291-296. Head, K. (2003) Gravy for beginners, mimeo, Universy of Brish Columbia. Head, K., Mayer, T. and Ries, J. (2010) The erosion of colonial trade linkages after independence Journal of International Economics 81, 1-14. Jansen, M. and Nordas, H.K. (2004) Instutions, trade policy and trade flows WTO Staff Working Paper ERS-2004-02. Kaufmann,. and Wei, S. (1999) oes grease money speed up the wheels of commerce? International Monetary Fund Working papers 00/64. Kaufmann,. & Kraay, A. and Mastruzzi, M. (2010) The worldwide governance indicators : methodology and analytical issues, Policy Research Working Paper Series 5430, World Bank. Liu, X. (2009) GATT/WTO promotes trade strongly: sample selection and model specification Review of International Economics 17, 428-446. Rose, A. (2004) o we really know that the WTO increases trade American Economic Review 94, 98-114. Santos Silva, J. and S. Tenreyro (2006) The Log of Gravy Review of Economics and Statistics 88(4), 641-658. Subramanian, A., and Wei, S. (2007) The WTO promotes trade, strongly but unevenly Journal of International Economics 72, 151-175. Thede, S., and Gustafson, N. (2012) The multifaceted impact of corruption on international trade The World Economy 35, 651-666.
Wooldridge, J. M. (2010) Econometric analysis of cross section and panel data. Second ed., MIT Press, Cambridge, Mass.