This section presents some robustness checks results. The objective is essentially to test the consistency of the results obtained for our measures of market access by stretching the set of explanatory variables.
Survival and Export Value
The set of dummy variables included in all of our estimations prevent us from test-ing the impact of variables such as the standard gravity ones. We thus re-run all specifications without destination dummies and add a set of bilateral indicators.
We control for the existence of a common language (i.e. Spanish) and the bilat-eral distance. We also control for the existence of an RTA distinguishing between Andean, MERCOSUR and the Rest of the World. Finally, we include a dummy indicating whether the destination country is part of the GSTP scheme and another indicating whether it is a member of the WTO. Generally speaking results on other control variables are not dramatically affected. Not surprisingly we lose some effi-ciency in our estimation. Coefficients of the bilateral controls are significant only for the GSTP dummy and to some extent the common language dummy. Moreover the significance level is never lower than 5 percent. In order to avoid any possible overlap in the dummy definition we included the trade policy dummies one by one.
Changes are only minor.
As a robustness check we also consider a simpler version of our market access measures by removing trade weights but still correcting for demand responsiveness, as in equations (5) and (6) withTd,hsw defined as in equation (2).
T T RIjd =
Results in all specifications are qualitatively similar to those obtained with the trade weighted versions. Coefficients of all explanatory variables are not found to be statistically different between comparable specifications except for market access variables in two cases.
We impose a linear form for our RPM measure essentially because it allows an easy quantitative interpretation of its coefficients. However, we re-estimate all relevant specifications by including a natural log version of our RPM measure as presented in equation (7).17
Results are similar to those obtained with the original RPM measure both in terms of sign and size of impact. This essentially reflects the fact that both measures are closely related. This is not surprising as in general, for small values of τ, we have that ln(1 +τ) ≈τ. In the same wake we also included our TTRI measure in levels. Once again results are only marginally modified and remain identical from a qualitative point of view. In addition, the impact of other explanatory variables remains essentially the same whatever version of our market access measures is taken into consideration.
Preferential access is often subject to stringent rules and regulations, such as rules of origin, which add to overall trade costs. The costs of using preferential access could outweigh the benefits if margins are small. As a consequence traders may find it more profitable to pay MFN rates. As a test, we check whether our results are robust to this issue by applying the simple rule that preferences are used only when the preferential margin is larger than 2.5 percent (Estevadeordal et al. (2008)). We recalculate the indices and then re-estimate all core specifications. Results show
17The RPM, which is equal to
no substantial variation compared to the reference ones. Finally, regressions were also run in which the top trading partner, namely the United States of America, is excluded from the sample. Once again, the results are very close to the ones in the base specification.
Export Value
As in the case of survival analysis, periods are treated independently conditional on explanatory variables in estimating equation (4). At the firm level this could be seen as not fully realistic as the history of a trade relationship is likely to matter in explaining its continuation and the evolution of its intensity across time. For instance, using Colombian firms’ data Eaton et al. (2008) observe that many do-mestic firms often start selling small quantities to a single neighbor country. Those who survive, slightly more than half of them, do tend to expand their presence as time goes by in their current destinations. This again could be explained by profit uncertainty revealed through exporting and correlated across time. If a firm’s ex-port profit in a market is uncertain but correlated over time, entry allows the firm to learn its profit potential there in the current period and in the forthcoming ones.
This feature which we believe may be an important one could be reflected in a dynamic specification where the present level of exports is a function of the first lag of it. However, the first lag of exports would capture much more than trade relationship history. Moreover, estimation of the resulting dynamic specification may end up in arguable choices especially in terms of instruments. We thus opt for the inclusion of duration dummies in the wake of our survival analysis. Results are in line with those obtained previously. The noteworthy characteristic is the gain in significance of coefficients of the TTRI variables. As to coefficients of duration dummies, they are all positive (the reference is duration equal to one), significant at least at the 5 percent level and slightly increasing with duration. This is consistent with the existence of profit uncertainty which is correlated across time as mentioned previously. However, this result can also be interpreted as evidence of the positive influence of tenure in a specific trade relationship on exports in that specific trade relationship. A possible explanation is that trust amongst trade partners increases with the aging of the relationship itself. These two explanations are not mutually exclusive.
5 Concluding remarks
The paper provides an empirical assessment of the impact of changes in tariff-wise market access conditions for Peru using firm level customs data over the period 2002-2008. We use two indicators of tariff-wise market access. One is the average tariff level corrected for the responsiveness of foreign demand Peruvian firms face in their destination market in their sector of activity. This indicator reflects absolute conditions of market access. The other reflects relative market access conditions.
It is the average, again corrected for the responsiveness of foreign demand, between the average tariff of foreign competitors and the tariff faced by Peruvian firms in a given destination country and sector. The latter indicator closely measures the
effective preference margin, which could be either positive or negative, enjoyed by Peruvian firms.
We consider both survival and export performance as dependent variables. Based on probit estimations with random effects we obtain that better market access con-ditions increase the probability of survival of a trade relationship. However, the impact of our measure of preference margin appears to dominate that of our ab-solute measure of market access. Although intuitively consistent, precise insights from theoretical analysis are missing and existing ones may even be misleading.
Other interesting features have emerged. For instance, diversification both in terms of destination country and products exported positively affects survival rates. Our results also show that firms that are part of some international production network, namely those which are both importing and exporting, enjoy lower hazard rates.
Similar results are found in our export performance analysis. Although less marked, the predominance in influencing export performance of relative market access condi-tions is also observed. About 20 percent of the increase in exports to MERCOSUR markets observed during the 2002-2008 period is due to improvements in the effec-tive preference margin perceived by Peruvian firms. Diversification and integration into international production chains also improve export performance.
Complexity in exporting to some preferential trade partner market is often pre-sented as the result of overlapping complicated rules of origin and administrative procedures to fulfill the latter. But complexity is also synonymous of opacity in effective tariff treatment. Taking the case of Peruvian firms facing a phase of trade integration into MERCOSUR markets, our results suggest that a fundamental com-ponent of trade creation in a trade agreement is the effective preference margin firms in the various signatory countries would eventually enjoy. From the point of view of policy makers this could easily be computed and should be at the core of their approach to negotiations. Other benefits could obviously be retrieved from a trade agreement, especially if it goes beyond strictly speaking market access matters.
However, if trade agreements are used as a trade promotion instrument this could only be effective if the advantage provided by the preferential market access is truly effective. This is to say that with the ongoing multiplication of trade agreements of all kinds and the inherent risk of preferences erosion, the trade promotion aspect of them could never stay the core element.
Appendix
Table 1: Survival of exports.
Probit MCW Cloglog Probit MCW Cloglog
ln(entry exp value) -0.087*** -0.057*** -0.087*** -0.086*** -0.057*** -0.087***
(0.003) (0.002) (0.003) (0.003) (0.002) (0.003) multiproduct -0.482*** -0.314*** -0.488*** -0.482*** -0.314*** -0.487***
(0.017) (0.013) (0.017) (0.017) (0.013) (0.017) multi destination -0.250*** -0.106*** -0.253*** -0.251*** -0.107*** -0.254***
(0.014) (0.011) (0.014) (0.014) (0.011) (0.014) importer 0.030*** 0.043*** 0.035*** 0.030** 0.042*** 0.035***
(0.012) (0.010) (0.012) (0.012) (0.010) (0.012) multi spell -0.258*** -0.920*** -0.239*** -0.257*** -0.920*** -0.238***
(0.021) (0.018) (0.021) (0.021) (0.018) (0.021) ln(gni per cap) -0.119** 5.612*** -0.128** -0.117** 5.610*** -0.125**
(0.054) (0.065) (0.057) (0.054) (0.065) (0.057) ln(imports) -0.028*** 0.143*** -0.027*** -0.028*** 0.145*** -0.028***
(0.004) (0.015) (0.004) (0.004) (0.0.015) (0.004)
ln(1+TTRI) 0.612*** 1.098** 0.617***
ln(σ2u) -1.550*** -13.802*** -2.001*** -1.556*** -13.086*** -2.016***
(0.126) (4.049) (0.196) (0.127) (3.166) (0.198)
duration dummies Yes Yes Yes Yes Yes Yes
sector dummies Yes Yes Yes Yes Yes Yes
year dummies Yes Yes Yes Yes Yes Yes
destination dummies Yes Yes Yes Yes Yes Yes
N 94314 94314 94314 94314 94314 94314
pseudoR2 0.050 0.081 0.051 0.050 0.081 0.051
Standard errors in parentheses
*p <0.10, **p <0.05, *** p <0.01
Table 2: Survival of exports (weighted and unweighted TTRI and RPM).
Probit MCW Cloglog
ln(entry exp value) -0.086*** -0.057*** -0.087***
(0.003) (0.002) (0.003)
ln(gni per cap) -0.115** 5.608** -0.124**
(0.054) (0.066) (0.057)
pseudo R2 0.050 0.081 0.051
Robust Standard errors in parentheses
*p <0.10, **p <0.05, ***p <0.01
Table 3: Regression on the value of exports.
RE MCW RE MCW RE MCW
multiproduct -1.434*** -1.428*** -1.434*** -1.428*** -1.434*** -1.428***
(0.028) (0.028) (0.028) (0.028) (0.028) (0.028) multi destination 0.534*** 0.539*** 0.538*** 0.546*** 0.537*** 0.545***
(0.024) (0.024) (0.024) (0.024) (0.024) (0.024) importer 0.466*** 0.456*** 0.466*** 0.457*** 0.467*** 0.457***
(0.023) (0.023) (0.023) (0.023) (0.023) (0.023)
ln(gni per cap) 0.132 0.366* 0.125 0.353* 0.121 0.353*
(0.081) (0.081) (0.081) (0.090) (0.081) (0.090) ln(imports) 0.156*** 0.091*** 0.156*** 0.091*** 0.155*** 0.091***
(0.008) (0.016) (0.008) (0.016) (0.008) (0.016)
ln(1+TTRI) -1.747*** 0.006 -0.508 -0.136
(0.300) (0.486) (0.409) (0.589)
RPM 0.586*** -0.054 0.519*** -0.072
(0.119) (0.109) (0.141) (0.131)
AVG(ln(gni per cap)) -0.550*** -0.517*** -0.530***
(0.082) (0.082) (0.082)
AVG(ln(imports) ) 0.083*** 0.084*** 0.084***
(0.016) (0.018) (0.018)
AVG(ln(1+TTRI)) -2.89*** -0.667
(0.608) (0.906)
AVG(RPM) 0.882***
(0.311) CONSTANT 5.466*** 9.676*** 5.537*** 9.139*** 5.581*** 9.368***
(0.815) (0.791) (0.816) (0.784) (0.815) (0.793)
sector dummies Yes Yes Yes Yes Yes Yes
year dummies Yes Yes Yes Yes Yes Yes
destination dummies Yes Yes Yes Yes Yes Yes
N 94314 94314 94314 94314 94314 94314
R2 overall 0.252 0.253 0.253 0.253 0.253 0.252
Standard errors in parentheses
*p <0.10, **p <0.05, *** p <0.01
Table 4: Survival of exports with marginal effects.
Probit MCW Cloglog Probit MCW Cloglog
ln(entry exp value) -0.030*** -0.019*** -0.028*** -0.030*** -0.019*** -0.028***
(0.001) (0.001) (0.001) (0.001) (0.001) (0.001) multiproduct -0.162*** -0.104*** -0.157*** -0.162*** -0.104*** -0.157***
(0.006) (0.004) (0.006) (0.006) (0.004) (0.006) multi destination -0.087*** -0.035*** -0.082*** -0.088*** -0.035*** -0.083***
(0.005) (0.004) (0.004) (0.005) (0.004) (0.004) importer 0.011*** 0.014*** 0.011*** 0.010** 0.014*** 0.011***
(0.004) (0.003) (0.004) (0.004) (0.003) (0.004) multi spell -0.092*** -0.293*** -0.076*** -0.092*** -0.293*** -0.076***
(0.007) (0.005) (0.007) (0.007) (0.005) (0.007) ln(gni per cap) -0.042** 0.446*** -0.041** -0.041** 0.446*** -0.040**
(0.019) (0.001) (0.018) (0.019) (0.001) (0.018) ln(imports) -0.010*** 0.046*** -0.009*** -0.010*** 0.047*** -0.009***
(0.001) (0.005) (0.001) (0.001) (0.005) (0.001) ln(1+TTRI) 0.214*** 0.292*** 0.199***
(0.057) (0.089) (0.054)
RPM -0.064*** -0.153*** -0.060***
(0.015) (0.039) (0.013)
Avg(ln(gni per cap)) -0.554*** -0.554***
(0.001) (0.001)
Avg(ln(imports)) -0.068*** -0.069***
(0.005) (0.005)
Avg(ln(1+TTRI)) 0.019
(0.163)
Avg(RPM) 0.058
(0.040)
duration dummies Yes Yes Yes Yes Yes Yes
sector dummies Yes Yes Yes Yes Yes Yes
year dummies Yes Yes Yes Yes Yes Yes
destination dummies Yes Yes Yes Yes Yes Yes
N 94314 94314 94314 94314 94314 94314
Marginal effects; Standard errors in parentheses (d) for discrete change of dummy variable from 0 to 1
*p <0.10, **p <0.05, *** p <0.01
Table 5: Survival of exports with marginal effects.
Probit MCW Cloglog
ln(entry exp value) -0.030*** -0.019*** -0.028***
(0.001) (0.001) (0.001) multiproduct -0.162*** -0.104*** -0.157***
(0.006) (0.004) (0.006) multi destination -0.088*** -0.035*** -0.083***
(0.005) (0.004) (0.004)
importer 0.010** 0.014*** 0.011***
(0.004) (0.003) (0.004) multi spell -0.092*** -0.293*** -0.076***
(0.007) (0.005) (0.007) ln(gni per cap) -0.040** 0.446*** -0.040**
(0.019) (0.001) (0.018) ln(imports) -0.010*** 0.047*** -0.009***
(0.001) (0.005) (0.001)
ln(1+TTRI) 0.090 0.078 0.071
(0.073) (0.171) (0.069)
RPM -0.050*** -0.142*** -0.049***
(0.018) (0.047) (0.017) Avg(ln(gni per cap)) -0.554***
(0.001)
Avg(ln(imports)) -0.068***
(0.005)
Avg(ln(1+TTRI)) 0.163
(0.161)
Avg(RPM) 0.087*
(0.046)
duration dummies Yes Yes Yes
sector dummies Yes Yes Yes
year dummies Yes Yes Yes
destination dummies Yes Yes Yes
N 94314 94314 94314
Marginal effects; Standard errors in parentheses (d) for discrete change of dummy variable from 0 to 1
*p <0.10, **p <0.05, *** p <0.01
Figure 1: Market access conditions in MERCOSUR countries.
Figure 2: Market access conditions in non MERCOSUR countries.
Figure 3: Positive versus Negative Preference Margins (# of trade relationships)
Figure 4: Trade Relationship Status (frequency)
Figure 5: Entry versus Continuing Trade Relationships: share in current year ex-ports
The discriminatory effect of a non-discriminatory relative preference margin.
1 Introduction
The effect of tariffs on trade is at the very center of trade negotiations between countries. They are known to affect trade patterns, it is therefore not surprising that many studies in international trade have looked at the effects of tariffs on trade. The general consensus is that higher tariffs make entry into and survival in a destination country more difficult, by increasing the gap between the price paid by the importer and the price received by the producer.
However, research has recently been looking beyond the simple tariff that an exporter faces. It has been shown that the relative tariff he faces compared to the one that competitors from other countries face, usually called the preference margin, may play a determinant role in determining the extent to which he exports.1
Another topic that has led to a huge amount of studies is the effect of the various costs associated with distance on trade. This paper shows that the magnitude of the effect of the Relative Preferential Margin (RPM) on exports may be dependent on the magnitude of other destination specific costs related to distance.
A large trend of literature has focused on the fact that distance remains a very important, if not growing, determinant of trade. This is for example the case of Disdier & Head (2008), who look at over 100 papers that estimate the impact of distance and find that it is on average not diminishing over time. A reason they put forward is that transport costs only represent a small fraction of costs related to distance. This is confirmed by the study of Lendle et al. (2012), who compare usual trade to trade done online on ebay, and show that the distance effect is much lower in the latter case, which they conclude as being due to lower search costs.
The other components of distance found in the literature that they mention are shipping costs, information frictions or trust frictions and taste differences. Two things can be retained from these findings, one is that distance must not exclusively
1See for instance Low et al. (2009), Carrere et al. (2010), Hoekman & Nicita (2011), Fugazza
& Nicita (2013), Fugazza & McLaren (2014).
61
be seen as shipping costs and the other is that many costs are destination specific but independent of the quantity exported.
Many studies have pointed out the fact that fixed costs can also be destination specific, such as Arkolakis & Muendler (2011), Johnson (2012) or Helpman et al.
(2008). The latter have a model in which firms’s decisions to enter or exit the export market depend on destination specific fixed costs. Arkolakis et al. (2008) also consider a country specific fixed cost in their model and Das et al. (2007) give some examples of fixed costs that are destination specific, and potentially incurred each year, such as minimum freight and insurance charges, and the costs of monitoring foreign customs procedures and product standards. The empirical results of this paper tend to suggest that fixed costs may be an important barrier to exporters, such that the extensive margin is particularly affected by changes in such costs.
The intuition for the preference margin to play a stronger role in closer countries is that the share related to relative tariffs in the exporter’s overall costs is larger than in further away countries. These destination specific costs will include other variable costs such as transport costs and also fixed costs. Therefore, changes in the preference margin will have a lesser impact if the former only represents a small portion of the exporter’s costs. Whether the effect takes place along the intensive or the extensive margin will then depend on the importance of destination specific fixed costs relative to destination specific variable costs.
The fact that the extensive margin seems to play a strong role in this paper is consistent with studies such as Chaney (2008) who shows that for US firms, the impact of trade costs on the extensive margin is very strong, in fact stronger than on the intensive margin. Ruhl (2008) also theoretically shows that a permanent change in tariffs will affect the extensive margin. For example, if it is a tariff decrease that is considered, it increases the profit from exporting such that new firms enter the market.
The result found in this paper, namely that firms exporting to closer countries are affected more by a same change of the RPM, may have important policy impli-cations. A country may be able to affect a particular country, or group of countries, more than others even by applying the same change in tariff to all countries. Like-wise, a new Regional Trade Agreement (RTA) with countries that are far away will have a smaller effect than an RTA with close partners. This discriminatory effect may then take place hidden behind what appear to be non-discriminatory tariffs.
This may mean that trade regulators should consider the fact that the effect of tariffs, or relative tariffs, may depend on the distance to the destination country.
The remainder of the paper is organized as follows: section 2 presents the the-oretical reasoning that explains why firms may be less affected by a same change of the RPM in further away countries. The data used to test this idea empirically is presented in section 3. The empirics that test the effect of the RPM on firms depending on distance are in section 4 and the results are presented in section 5.
Section 6 shows an alternative specification where one can see to what extent the RPM plays a role in explaining the margins of exports relative to other variables.
Robustness checks are run in section 7 and section 8 concludes.
2 Relative tariffs and overall costs
The analysis below gives the intuition of why relative tariffs may have an impact on exporting firms that depends on distance. It shows how the different margins of trade are expected to be affected by the RPM, and how this effect may vary with distance.
First of all, the choice of the RPM as the measure of trade barriers is chosen rather than the direct tariff as it better reflects the trade costs that the exporter faces compared to the costs faced by competitors from other countries. As mentioned above, recent studies have shown that this relative measure plays a larger role than the simple tariff that the firms’ face. However, other costs will play a role in the exporting firm’s level of competitiveness. These costs will be a mix of variable and fixed costs, in which both have a destination specific component.
Further away countries will have higher variable and fixed costs, such that the share of the tariff in overall costs is lower. As mentioned above, the exporter is also affected by tariffs that competitors of other countries face. Once again, the advantage gained on competitors due to a lower relative tariff will also only represent a smaller fraction of the overall barriers that an exporter faces when exporting to a further away country.
As in Fugazza & Nicita (2013) and Fugazza & McLaren (2014), the RPM mea-sures the advantage that Peru has in exporting its goods in sector j to country d. Equation (1) presents this measure, where v denotes countries competing with Peru in sectorj in exporting to countryd. Td,hsw is the trade weighted average of the tariffs applied by country d to imports for each hs product (HS 6-digit products) originating from each country v. x indicates exports from Peruvian firms in sector j to countryd at the product level, is the bilateral import demand elasticity, T is the applied tariff, and hs are HS 6-digit categories.
RP Mjd =
A positive RPM means that the exporter has an advantage on exporters from other countries when exporting to destination d whereas a negative RPM would mean that he has a disadvantage. It can also be noted that it reflects the discrimi-natory effects of the overall system of preferences.
A firm exporting to a far away country will have higher variable (excluding the RPM) and fixed costs, as explained above. These high costs will remain high and unchanged whatever the advantage or disadvantage the firm has through the RPM. The RPM will therefore be of smaller importance in his choice of entry and