Top

Empirical Economics

Published in:

01-08-2014

On the robustness of the trade-inducing effects of trade agreements and currency unions

Author: Jayjit Roy

Published in: Empirical Economics | Issue 1/2014

Activate our intelligent search to find suitable subject content or patents.

search-config

AI-assisted search

Off

Abstract

Regional trade agreements (RTAs) and currency unions (CUs) are potentially endogenous trade cost proxies in equations estimating their effects on bilateral trade. In case of both, this problem is magnified by the paucity of reliable instruments. Instead of resorting to the oft-employed alternative of panel data to address selection on just the time-invariant unobservables, this paper assesses the extent to which a positive association between CU or RTA membership and bilateral trade can be considered causal. In addition, it attends to recent concerns over the extensive margin of trade (at the country-level) and the issue of zero trade observations in log-linearized gravity models by relying more on a bivariate probit analysis. Despite not identifying point estimates, striking results are obtained. While most cross-sections exhibit a positive association between both RTAs and CUs and trade, the evidence in favor of a robust causal effect is strong mainly for CUs. However, the magnitude of the CU effect is still sensitive to the amount of selection on unobservables. Moreover, selection into RTAs (CUs) is mostly found to be positive (negative). Finally, the presence of spillovers across the policy regimes is also detected.

previous article TFP growth and its determinants: a model averaging approach

next article Evolving UK and US macroeconomic dynamics through the lens of a model of deterministic structural change

Dont have a licence yet? Then find out more about our products and how to get one now:

Springer Professional "Wirtschaft"

Online-Abonnement

Mit Springer Professional "Wirtschaft" erhalten Sie Zugriff auf:

über 67.000 Bücher
über 340 Zeitschriften

aus folgenden Fachgebieten:

Bauwesen + Immobilien
Business IT + Informatik
Finance + Banking
Management + Führung
Marketing + Vertrieb
Versicherung + Risiko

Jetzt Wissensvorsprung sichern!

inform now

Springer Professional "Wirtschaft+Technik"

Online-Abonnement

Mit Springer Professional "Wirtschaft+Technik" erhalten Sie Zugriff auf:

über 102.000 Bücher
über 537 Zeitschriften

aus folgenden Fachgebieten:

Automobil + Motoren
Bauwesen + Immobilien
Business IT + Informatik
Elektrotechnik + Elektronik
Energie + Nachhaltigkeit
Finance + Banking
Management + Führung
Marketing + Vertrieb
Maschinenbau + Werkstoffe
Versicherung + Risiko

Jetzt Wissensvorsprung sichern!

inform now

Available only for authorised users

The trade agreements database of the World Trade Organization, a source consulted extensively for this study, classifies RTAs according to their scale of integration. Since this paper does not attempt to distinguish between such regimes, the generic terminology of RTAs is used. In other words, as discussed below, RTAs include partial scope arrangements as well as relatively integrated regimes such as customs unions and free trade agreements.

See http://www.wto.org/english/tratop_e/region_e/region_e.htm.

While providing a review of the CU literature, Santos Silva and Tenreyro (2010) analyze the Euro’s effect by estimating gravity models in levels. Although this approach includes the zeros, see footnote 9.

Throughout the paper, extensive margin refers to the extensive country margin. Also, the intensive and extensive country margins are together referred to as overall trade.

While Altonji et al. (2005), Millimet et al. (2010), and Roy (2013) adopt a similar strategy in other contexts, to the author’s knowledge, Chang and Lee (2011) is the only study in the trade literature to pursue a related approach.

Persson (2001, p. 446) employs matching techniques to obtain a “very sizeable” CU effect. However, the statistical significance of the CU coefficient estimate is sensitive to the matching algorithm in Persson (2001).

According to Barro and Tenreyro (2007), two countries may share a common currency due to their independent decisions to maintain parity with a third anchor currency. As a result, the instrument for the CU dummy is obtained as the joint probability that a country-pair adopts the same anchor currency. However, unobservable historical and political ties between two countries may not only affect their bilateral trade, but may also lead to their choice of a common anchor thereby rendering the validity of the instrument doubtful. In addition, the lack of theoretically motivated multilateral resistance terms make the results unreliable.

Chintrakarn (2008) provides a concise review of this literature.

Note that Santos Silva and Tenreyro (2006) caution against the log model because heteroskedasticity in levels can induce correlation between the covariates and the error term in logs even if the covariates are exogenous in the levels model. Since the objective here is to assess sensitivity to correlation between covariates and the error term in the log model (regardless of the source of this underlying correlation), this issue is not particularly relevant. In addition, estimation in logs enables comparison to the literature.

In Baier and Bergstrand (2009b), the Taylor expansion is centred around a world of symmetric trade costs, i.e., \(t_{ij}=t\) for all country-pairs \(ij\). Here, \(t\) denotes trade costs.

Note that, Chang and Lee (2011) and Baier and Bergstrand (2009a) adopt the BVOLS approach while employing matching. Although Egger et al. (2011, p. 119) use country fixed effects, they allude to “convergence problems” as well.

Note that, the zero trade observations are discussed below in Sect. 4. Also, the US consumer price index is used to express GDP in 1995 dollars.

Here, area of a country is also considered to reflect trade costs. In fact, Melitz (2008, p. 676) considers area to be “a proxy for internal distance.”

Note that, the remainder of the paper focuses on \(k=1\). However, this is an ad hoc approach of including the zeros and the corresponding estimates are potentially biased due to this. Accordingly, estimation was also performed with \(k=0.1\) and \(k=10\). The results remained very similar and are available upon request. As discussed below, the fixed effects estimates were also obtained for each treatment after dropping the zero trade observations.

The country dummies are usually used to control for country-specific unobservables that do not vary across trading partners as well as the MR terms. In this case, they also capture the impact of GDP.

Before discussing the bivariate probit analysis in further detail, note that an alternative approach was also employed to examine selection into RTAs and CUs. First, the selection equation was estimated and the observation-specific predicted probabilities, \(\varPhi \left( Z_{ij}\widehat{ \lambda }\right) \), were obtained. Next, the outcome equation was estimated with \(D_{ij}-\varPhi \left( Z_{ij}\widehat{\lambda }\right) \) as an additional regressor and the corresponding coefficient estimate was used to examine selection. Interestingly, upon imposing the BV constraints, the coefficient estimate turned out to be statistically significant for more cross-sections in case of RTAs relative to CUs. The results are available upon request.

Here, \(\rho =0\) corresponds to the situation of exogeneity or no selection on unobservables. Hence, the estimates obtained from the bivariate probit with \(\rho \) constrained to 0 are identical to the corresponding estimates from the univariate probit in (1).

The method also involves an additional requirement that is weaker than independence between the observed (non-treatment) covariates and the remaining determinants of bilateral trade.

To be more precise, the following algorithm is adopted:

Estimate (5) constraining \(\rho \) to 0.

Use the estimates of \(\delta \) and \(\lambda \) to calculate \(\rho = \frac{\text {Cov}\left( Z\delta , Z\lambda \right) }{\text {Var}\left( Z\delta \right) }\).

Estimate the bivariate probit constraining \(\rho \) to the value from step 2.

Go back to step 2 and update value of \(\rho \), and then repeat step 3.

Continue until the convergence of \(\rho \) is attained.

Since \(Z\) represents the set of all regressors except \(D\), it includes \(X\). While it also includes the country fixed effects in (3), in case of (4) it denotes the GDPs and the MR terms from BVOLS, in addition to \(X\).

Note, \(\hbox {Cov}\left( \upsilon ,\varepsilon \right) = \hbox {Cov}\left( D-Z\lambda ,\varepsilon \right) = \hbox {Cov}\left( D,\varepsilon \right) \) since \(Z\) and \( \varepsilon \) are uncorrelated. Also, \(\hbox {Cov}\left( D,\varepsilon \right) = \hbox {Var} \left( D\right) \left\{ E\left[ \varepsilon |D=1\right] -E \left[ \varepsilon |D=0\right] \right\} .\)

Here, the BV constraints entail conceiving the trade cost variables as \( t_{ij}-\hbox {MR} t_{ij}\). Accordingly, if imposed, the constraints are relevant for (8) as well.

Since the data includes zero trade observations and more countries than used in most previous analyses, Liu (2009) considers it to be relatively more complete.

See http://www.wto.org/english/news_e/pres09_e/pr548_e.htm.

See http://www.efta.int/about-efta/history.aspx.

A few minor modifications were also made to the WTO membership variables. As a result of the modifications, Czechoslovakia (for 1950–1990), Lebanon (for 1950), Liberia (for 1950), and Syria (for 1950) are treated as members.

Note that in Rose (2000, p. 15), “many currency union pairings have no trade.”

A relevant concern in the context of extensive margin is the issue of birth of nations (e.g., due to decolonization). However, Liu (2009) does not consider the pattern of extensive margin in this data set to be driven by the emergence of new countries. Moreover, Felbermayr and Kohler (2010) consider the issue to be less significant with cross-section data.

Given the iterative approach outlined above, in the end, \(\rho \) is constrained and not estimated. Hence, the corresponding standard errors are not available.

Note that, the primary interest lies in the treatment variables. Hence, in keeping with the bivariate probit results in Altonji et al. (2005) and Millimet et al. (2010), coefficient estimates of the other regressors are not displayed. However, they are available upon request.

The author reports non-robust standard errors to be more conservative (i.e., the author does not want to find the statistically significant effect obtained under exogeneity to disappear quickly due to large standard errors). Nonetheless, the results are virtually unchanged if one uses robust standard errors.

Using the notation in (5), the marginal effects are computed as \(\varPhi \left( \widehat{\tau }+\overline{Z}\widehat{\delta }\right) -\varPhi \left( \overline{Z}\widehat{\delta }\right) \), where \(\overline{Z}\) depicts the sample mean of all covariates except \(D\), and \(\widehat{\tau }\) and \(\widehat{\delta }\) are the coefficient estimates from the bivariate probit. In this context, it should be noted that the marginal effects only correspond to a partial equilibrium change.

Note that according to Egger et al. (2011, p. 134), RTA membership “does not significantly affect the extensive margin of trade.”

Note that the estimates obtained without the BV constraints indicate a robust causal effect as well. However, if the estimates of \(\rho \) derived solely from the parametric assumption are relied upon, imposition of the BV constraints suggests otherwise. Also noteworthy, if the BV constraints are absent, the estimated treatment effects are accompanied by higher marginal effects.

Note that for 1970–1990, and for 2000 without the Euro, the marginal effects are fairly stable given \(\rho \), once the BV constraints are imposed. The smaller marginal effects in case of 2000 with Euro are not unexpected given the differences between the Euro countries and members of other CUs.

Incidentally, countries which adopted the Euro are also RTA members.

Note that, the estimates of \(\rho \) are unavailable for some of the 1960 and 1980 specifications due to difficulty in convergence. However, while 1960 is anyway characterized by a lack of positive association, the estimates obtained upon invoking equal selection are available for 1980.

Note that the RTA and CU coefficient estimates in Table 4 are higher than those typically obtained in the literature. However, this is largely due to the omission of zero trade observations in most existing studies. Nonetheless, the regressions were also performed at the intensive margin, i.e., after dropping the zeros. While the results are available upon request, the coefficient estimates at the intensive margin are noteworthy. The RTA estimates (except 1950) vary between 0.3 and 0.9 and are always statistically significant. Similarly, the CU coefficient estimates are positive and statistically significant for all cross-sections except 2000 (with Euro); they vary between 0.3 and 1.4.

Here, in the context of the log specification, note that Frankel’s (2010) findings do not support either hypothesis as an explanation for the typically observed discrepancy in trade promoting effects of the Euro and other CUs. Although Frankel (2010) suspects the typically small sample sizes in most Euro studies as a possible (but unlikely) explanation, the suspicion is driven by gravity equations which do not control for the MR terms.

The (fixed effects) coefficient estimates of the other regressors are similar to those found in the literature, but are not the focus of the paper. As a result, they do not find mention, but are available upon request.

At the intensive country margin, note that the evidence in favor of a robust causal effect is largely similar for all cross-sections except 1970. The results are available upon request.

Note that Baier and Bergstrand (2007) fail to reject the null of strict exogeneity in case of free trade agreements.

Note that the 1970 results across the log and probit specifications should not be viewed as inconsistent. It does not seem implausible for RTAs to help country-pairs move from zero to positive trade but turn less effective thereafter.

Again, the evidence in favor of causality is similar at the intensive margin but strong during 1960 as well. The results are available upon request.

Before proceeding, note that some gravity studies include both RTA and CU as regressors in the same equation. However, the estimated RTA (CU) effects were hardly found to change upon the inclusion of CU (RTA) as a regressor. In fact, this was true regardless of the BVOLS or fixed effects approach. For the fixed effects approach, the (cross-section) correlations between CU and RTA, after conditioning on all regressors, were found to be less than 0.085 in the majority of cases, and reached a maximum of 0.131 in 1970. Nonetheless, future work should address the selection issue by treating both RTA and CU as potentially endogenous treatments in the same equation.

Note that, this is not inconsistent since unobservables that move two countries from zero to positive trade need not be the same as unobservables that encourage a greater amount of bilateral trade. For example, two countries that do not trade and have unobservables that are negatively associated with trade may form a CU to encourage the start of trade. This would correspond to negative selection at the extensive margin. Similarly, consider two country-pairs with unobservables such that one of the pairs (A) trades very little while the other pair (B) engages in substantial trade. Now, under positive trade, suppose that the benefits from trade accrue only if trade exceeds a certain threshold which is slightly (much) greater than B’s (A’s) trade. Here, due to uncertainty over attaining the threshold, pair A may not incur the costs associated with negotiating a CU while pair B may do so. This would lead to positive selection with respect to the amount of bilateral trade.

The evidence in favor of a robust causal effect is somewhat similar at the intensive margin except for 1960 and 1970. Not only are the estimated treatment effects positive and significant for all cross-sections except 2000 (with Euro), they are also evidenced to be robust to selection on unobservables. Moreover, the positive and significant effects obtained under exogeneity vary between 0.6 and 2.6. The results are available upon request.

Note that although the dataset contains a large number (about fifty percent) of zero trade observations, estimation was also performed after dropping observations for a country prior to its independence if all pairwise observations containing that country had zero trade, prior to its independence. The results remained very similar and are available upon request. Next, as discussed above, the log models were also estimated after omitting the zero trade observations. However, the corresponding results are susceptible to a potential sample selection bias.

Altonji JG, Elder TE, Taber CR (2002) Selection on observed and unobserved variables: assessing the effectiveness of catholic schools. Unpublished manuscript, Department of Economics, Northwestern University

Altonji JG, Elder TE, Taber CR (2005) Selection on observed and unobserved variables: assessing the effectiveness of catholic schools. J Polit Econ 113:151–184CrossRef

Anderson JE, van Wincoop E (2003) Gravity with gravitas: a solution to the border puzzle. Am Econ Rev 93:170–192CrossRef

Anderson JE, van Wincoop E (2004) Trade costs. J Econ Lit 42:691–751CrossRef

Anderson JE, Yotov YV (2011) Terms of trade and global efficiency effects of free trade agreements, 1990–2002. NBER working paper no. 17003

Baier SL, Bergstrand JH (2007) Do free trade agreements actually increase members’ international trade? J Int Econ 71:72–95CrossRef

Baier SL, Bergstrand JH (2009a) Estimating the effects of free trade agreements on international trade flows using matching econometrics. J Int Econ 77:63–76CrossRef

Baier SL, Bergstrand JH (2009b) Bonus vetus OLS: a simple method for approximating international trade-cost effects using the gravity equation. J Int Econ 77:77–85CrossRef

Baldwin R (2006) The Euro’s trade effects. European Central Bank working paper no. 594

Barr D, Breedon F, Miles D (2003) Life on the outside: economic conditions and prospects outside euroland. Econ Policy 18:573–613CrossRef

Barro RJ, Tenreyro S (2007) Economic effects of currency unions. Econ Inq 45:1–23CrossRef

Baur P, Winschel V (2009) Bayes network analysis of trade effects of currency unions and free trade agreements. Unpublished manuscript, Department of Economics, University of Mannheim

Bogetić Z (2000) Official dollarization: current experiences and issues. Cato J 20:179–213

Boughton JM (1991) The CFA franc zone: currency union and monetary standard. IMF working paper no. 91/133

Carrère C (2006) Revisiting the effects of regional trade agreements on trade flows with proper specification of the gravity model. Eur Econ Rev 50:223–247CrossRef

Chang P, Lee M (2011) The WTO trade effect. J Int Econ 85:53–71CrossRef

Chintrakarn P (2008) Estimating the euro effects on trade with propensity score matching. Rev Int Econ 16:186–198CrossRef

Ederington J, Minier J (2003) Is environmental policy a secondary trade barrier? an empirical analysis. Can J Econ 36:137–154CrossRef

Egger P, Larch M (2008) Interdependent preferential trade agreement memberships: an empirical analysis. J Int Econ 76:384–399CrossRef

Egger H, Egger P, Greenaway D (2008) The trade structure effects of endogenous regional trade agreements. J Int Econ 74:278–298CrossRef

Egger P, Larch M, Staub KE, Winkelmann R (2011) The trade effects of endogenous preferential trade agreements. Am Econ J Econ Policy 3:113–143CrossRef

Feenstra RC (2004) Advanced international trade: theory and evidence. Princeton University Press, Princeton

Felbermayr GJ, Kohler W (2010) Modelling the extensive margin of world trade: new evidence on GATT and WTO membership. World Econ 33:1430–1469CrossRef

Frankel JA (1997) Regional trading blocs. Institute for International Economics, Washington

Frankel JA (2010) The estimated effects of the euro on trade: why are they below historical effects of monetary unions among smaller countries? In: Alesina A, Giavazzi F (eds) Europe and the euro. University of Chicago Press, Chicago, pp 169–218CrossRef

Frankel JA, Rose AK (2002) An estimate of the effect of currency unions on trade and output. Q J Econ CXVII:437–466CrossRef

Freund C, Ornelas E (2010) Regional trade agreements. Ann Rev Econ 2:139–166CrossRef

Friberg R, Matha TY (2004) Does a common currency lead to (more) price equalization? the role of psychological pricing points. Econ Lett 84:281–287CrossRef

Ghosh S, Yamarik S (2004) Are regional trading arrangements trade creating? an application of extreme bounds analysis. J Int Econ 63:369–395CrossRef

Ghosn F, Palmer G (2003) Codebook for the militarized interstate dispute data, version 3.0. http://cow2.la.psu.edu

Gibler DM, Sarkees MR (2004) Measuring alliances: the correlates of war formal interstate alliance data set, 1816–2000. J Peace Res 41:211–222CrossRef

Glick R, Rose AK (2002) Does a currency union affect trade? the time series evidence. Eur Econ Rev 46:1125–1151CrossRef

Gómez-Oliver A, Itam S, Piñerúa C, Shah S (1999) The Bahamas: selected issues and statistical appendix. IMF staff country report no. 99/106

Havránek T (2010) Rose effect and the euro: is the magic gone? Rev World Econ 146:241–261CrossRef

Helpman E, Melitz M, Rubinstein Y (2008) Estimating trade flows: trading partners and trading volumes. Q J Econ CXXIII:441–487CrossRef

Henderson DJ, Millimet DL (2008) Is gravity linear? J Appl Econ 23:137–172CrossRef

Jovanovic MN (1998) International economic integration: critical perspectives on the world economy. Taylor & Francis, London

Krishna P (2012) Preferential trade agreements and the world trade system: a multilateralist view. NBER working paper no. 17840

Liu X (2009) GATT/WTO promotes trade strongly: sample selection and model specification. Rev Int Econ 17:428–446CrossRef

Magee CSP (2003) Endogenous preferential trade agreements: an empirical analysis. Contrib Econ Anal, Policy 2

Magee CSP (2008) New measures of trade creation and trade diversion. J Int Econ 75:349–362CrossRef

Melitz J (2008) Language and foreign trade. Eur Econ Rev 52:667–699CrossRef

Millimet DL, Tchernis R, Hussain M (2010) School nutrition programs and the incidence of childhood obesity. J Human Resour 45:640–654CrossRef

Millimet DL, Tchernis R (2009) On the specification of propensity scores: with applications to the analysis of trade policies. J Bus Econ Stat 27:397–415CrossRef

Moheeput A (2008) Issues on the choice of exchange rate regimes and currency boards: an analytical survey. Working paper no. 855, Department of Economics, The University of Warwick

Orefice G (2013) International migration and trade agreements: the new role of PTAs. FIW-Working paper

Persson T (2001) Currency unions and trade: how large is the treatment effect? Econ Policy 33:435–448

Pomfret R (2007) Is regionalism an increasing feature of the world economy? World Econ 30:923–947CrossRef

Rose AK (2000) One money, one market: estimating the effect of common currencies on trade. Econ Policy 30:9–45

Rose AK (2001) Currency unions and trade: the effect is large. Econ Policy 33:449–461CrossRef

Rose AK (2006) Checking out: exits from currency unions. http://faculty.haas.berkeley.edu/arose

Rose AK, Engel C (2002) Currency unions and international integration. J Money Credit Bank 34:1067–1089CrossRef

Rose AK, van Wincoop E (2001) National money as a barrier to trade: the real case for monetary union. Am Econ Rev 91:386–390CrossRef

Rose AK, Stanley TD (2005) A meta-analysis of the effect of common currencies on international trade. J Econ Surv 19:347–365CrossRef

Roy J (2010) Do customs union members engage in more bilateral trade than free trade agreement members? Rev Int Econ 18:663–681CrossRef

Roy M (2013) How well does the U.S. government provide health insurance for infants? Empir Econ forthcoming

Santos Silva JMC, Tenreyro S (2006) The log of gravity. Rev Econ Stat 88:641–658CrossRef

Santos Silva JMC, Tenreyro S (2010) Currency unions in prospect and retrospect. Ann Rev Econ 2:51–74CrossRef

Schuler K (2005) Some theory and history of dollarization. Cato J 25:115–125

Vicard V (2009) On trade creation and regional trade agreements: does depth matter? Rev World Econ 145:167–187CrossRef

Whalley J (2008) Recent regional agreements: why so many, why so much variance in form, why coming so fast, and where are they headed? World Econ 31:517–532CrossRef

World Bank Group (2008) World development indicators. World Bank Publications, Washington

Title: On the robustness of the trade-inducing effects of trade agreements and currency unions
Author: Jayjit Roy
Publication date: 01-08-2014
Publisher: Springer Berlin Heidelberg
Published in: Empirical Economics / Issue 1/2014
Print ISSN: 0377-7332
Electronic ISSN: 1435-8921
DOI: https://doi.org/10.1007/s00181-013-0739-9

Springer Professional

Abstract

Please log in to get access to your license.

Dont have a licence yet? Then find out more about our products and how to get one now:

Springer Professional "Wirtschaft"

Springer Professional "Wirtschaft+Technik"

Other articles of this Issue 1/2014

The European unemployment problem: its cause and cure

Real-time forecasting US GDP from small-scale factor models

Evaluating FOMC forecast ranges: an interval data approach

Testing exclusion restrictions and additive separability in sample selection models

Is eco-efficiency in greenhouse gas emissions converging among European Union countries?

Income inequality between Chinese regions: newfound harmony or continued discord?