5. Assessing Effects of the CCFTA

Through the previous analyses, we see that Canada's exports to Chile and its imports from Chile both grew impressively upon trade liberalization, and we also see that a large amount of these increases consisted of new trade. However, as noticed in the previous analyses, there are always some "anomalies" that obscure the data and more advanced econometric techniques are required to obtain a more comprehensive and precise assessment of the trade-enhancing effect of the CCFTA.

Similar to other studies on assessing the effect of FTAs, the average treatment effect approach is adopted to establish a causal link between the CCFTA and the expansion of bilateral trade between Canada and Chile Footnote 12. In this case, both Canada and Chile are considered as having received a treatment (i.e., CCFTA), the average treatment effect is estimated by comparing the trade performance between the "treated" and the "untreated".

Since the level of trade for the "treated" in the absence of the CCFTA is not observable, we must compare its trade performance to countries that are not part of the CCFTA. The effect found will be the expected (average) difference in trade performance between the treated and untreated controlling for all common characteristics of each country-pair.

The estimating equation can be set up as follows,

ln yijkt = α + β1CCFTAijt + β2Tikt + β3 ln GDPit + β4 ln GDPjt + εijkt

where ln yijkt is the natural logarithm of imports in product k from country i to country j at time t, CCFTAijt is the treatment variable that equals to one when country i and country j are Canada and Chile (or vice versa) at time tt ≥ 1997, Tijkt is one plus the tariff of product k that country i imposed on country j at time t , lnGDPit is the GDP of country i at time t, lnGDPjt is the GDP of country j at time t , and εijkt is random error. Applying a panel set-up (cross-sectional and time series data) with fixed effect will control for the effect between country-pair differences; for instance, different responses of different country-pairs to common macroeconomic shocks. The treatment effect found is interpreted as the average treatment effect on the treated (ATT), in this case, the country-pair Canada-Chile after 1997.

We augmented the basic specification listed above with the inclusion of control variables such as Inimrikt, the import penetration ratio of product k in country i at time t, representing import demand at the product level; ln pcgjt, productivity of country j at time t, representing country's production capacity; ln popit, the population of country i to control for the size of import markets; and ftait to control for preference erosion stemming from FTAs with third parties.

We run the regression above on the trade of Canada's top 50 trading partners over the period between 1995 and 2011 Footnote 13. In this average treatment evaluation set-up, all variables except the CCFTA are considered control variables and by doing this, the effect of the treatment variable, CCFTA is expected to be enhanced. It is expected that the treatment effect of tariff on imports would be negative because tariff is assumed to restrict or lower imports. Also expected is a negative effect of FTAs on imports because the existence of other FTAs may divert trade of CCFTA members to their other FTA partners and erodes preference advantages over time. The following table reports the estimation results:

Table 13: Regression Results from the Average Treatment Effect Estimation

Note: Superscripts ***, ** and * represent significance levels of 0.01, 0.05 and 0.1 respectively. Standard errors are presented in brackets.

VariableEstimated coefficient
CCFTAijt0.115***0.135***
(0.005)(0.005)
Tijkt0.018***-0.010**
(0.004)(0.004)
ln GDPit0.507***0.523***
(0.006)(0.006)
ln GDPjt0.653***0.542***
(0.018)(0.019)
ln pcgjt-0.692***-0.584***
(0.019)(0.020)
ftait-0.050***-0.046***
(0.009)(0.009)
ln imrikt0.200***0.201***
(0.001)(0.001)
ln popit1.214***1.031***
(0.041)(0.041)
Tariff StandardizationYesNo
Parametric weighingYesYes
Non-parametric matchingYesYes
Panel set-upYesYes

Source: Author's own calculations.

The estimation results of two separate specifications are presented in Table 13 and both results indicate that the treatment effect of the CCFTA is significant and confirm our expectation that the CCFTA did contribute positively to the growth of trade. In the first specification, Tijkt is standardized to represent the tariffs of other countries relative to CCFTA countries Footnote 14. This approach allows the estimated coefficient of CCFTAijt to capture the "all-in" effect of the CCFTA including both the tariff and non-tariff effects of the CCFTA. Because of this adjustment, the coefficient of Tijkt should be interpreted as a net effect experienced only by the non-CCFTA partners in the presence of CCFTA. Normally, the effect of tariffs on imports is expected to be negative; however in this case, because CCFTA countries had more extensive tariff reductions relative to those of non-CCFTA trading partners, the standardized tariffs for the same products imported from non-CCFTA countries actually increased relative to CCFTA tariffs. Since Canada's and Chile's trade with non-CCFTA countries continued irrespective of CCFTA tariff reductions, the relation between the changes in relative tariffs of non-CCFTA countries and the growth in their trade flows can be positive. In addition to this adjustment, we apply a pre-estimation propensity score matching to ensure a balanced set of observations between the treated and the non-treated. Only the matched pairs will then be included in the regression where we apply a parametric propensity score weighing in the estimation. The treatment effect of CCFTA is estimated as 0.115. In other words, if the country-pair is composed of both CCFTA members, the import of product k by country i from country j will grow 12.2 percent (= [exp(0.115) -1]*100) faster than if the country-pair is not composed of both CCFTA members. With the presence of CCFTA, the estimated effect of relative tariffs for a non-CCFTA country pair to that of Canada-Chile on trade growth is positive. The effects of economic sizes measured in terms of GDP on trade are, as expected, strongly positive and significant. The treatment effect for the existence of other FTAs confirms the diversion effect of other FTAs on trade between CCFTA countries. Lastly, a high import penetration ratio or weak productivity performance is expected to increase more imports.

The second specification reported in Table 13 is estimated without standardization of tariffs. Thus, the effect of CCFTAijt is interpreted as a non-tariff effect, while the estimated coefficient for Tijkt represents the tariff effect including the effects for both CCFTA and non-CCFTA countries. Again, the estimation is carried out by applying both the matching and parametric weighting. The estimated effect of CCFTA is 0.135. This means that, conditional on all control variables, non-tariff effect of the CCFTA generate additional growth of trade by 14.5 percent (= [exp(0.135) -1]*100) between CCFTA countries relative to non-CCFTA countries. With the presence of CCFTA, the tariff rate has on average a negative effect on imports. The existence of other FTAs is estimated to decrease imports while a high import penetration ratio or weak productivity performance is expected to increase more imports.

The econometric results show that the effect of a free trade agreement to increase trade growth between the two countries is strong, leading to 12.2 percent growth of trade per year. Results of other specifications for robust check are reported in Annex 1.

5.1 Assessing Effect of Extensive Margin

While counting the number of, or examining the value contributed by, existing and new products as presented in Section 4.3 is a simple and useful way to gauge the gains in new trade, a more sophisticated approach is to decompose the increase in bilateral trade between Canada and Chile into contributions from greater volumes of traded product (intensive margin) and from a larger set of products (extensive margin). To this end, we follow the seminal work of Feenstra (1994) and Hummels and Klenow (2005) Footnote 15 to decompose bilateral merchandise trade data. The trade data at the HS06 product level from 1995 to 2010 are used in the estimation of extensive and intensive margins Footnote 16.

According to Hummels and Klenow, the intensive margin of trade is defined on the basis of shares of bilateral trade in total trade within the categories of products already traded while extensive margins are defined on the basis of shares of bilateral trade in total trade across categories of products that were not traded in the initial period. Therefore, both extensive and intensive margin should always be less than one. The product of the extensive and intensive margins should be equal to a country's market share in the partner country's market (See Annex 2). The extensive margin gauges the importance of categories of products compared to that of volume of trade within categories. With all other things being equal, if a country concentrates all of its exports in a small number of product categories, it will have a higher intensive export margin and a lower extensive margin. If that country divides its exports thinly over many product categories, it will have a lower intensive export margin and a higher extensive margin.

Table 14 reports the extensive and intensive margins for Canada's exports to Chile. The imputed average index of extensive margins for Canada's exports to Chile was 0.606, much higher than the index for intensive margins of 0.036, which means that Canadian exports to Chile were mainly driven by the extensive margins of trade. Canada exported a broader spectrum of products rather than more existing products. An interesting observation from this table is that the extensive margins during the studied period increased from an average 0.57 to 0.60. This implies that the CCFTA opened opportunities for new exporters and new products by reducing the entry threshold to the Chilean market. Opposite to the upward trend of extensive margins is a gradual decline of intensive margins over the studied period, which indicates the worsening sale volume of each existing product. This observation is consistent with the data analysis presented in Section 4.3 which suggested that the erosion of Canadian CCFTA preferences in Chile curtailed the expansion of exports of existing products to Chile, but not the expansion of new products. The last column of Table 14 shows a decline in intensive margins contributed importantly to the drop in Canada's import share in the Chilean market. In other words, without the offsetting effect of the expansion of extensive margins or the continuous introduction of new products, the decline in Canada's import share in Chile would have been even more significant.

The existing literature in trade theory does not offer any compelling explanation as to what may lead to a coincidental decline of intensive margins and increase in extensive margins in the context of multiple trade agreements. Preference erosion should logically affect both extensive and intensive margins. It is highly plausible that what has been observed here is anomalous and the effect of the CCFTA on the margins is non-representative. However, given the parallel between our trade data and the international trade theory, the following explanation can be offered.

There are two opposite and competing forces affecting Canada's exports to Chile. First, consumers in both countries value variety and are willing to pay a premium for a broader range of products. The CCFTA created opportunities for new products and new firms in the partner country's market by reducing the entry thresholds. Second, Chile was a relatively small country with a population of 17 million and had a relatively small domestic market. When the subsequent FTAs between Chile and other third countries came into effect, the erosion of preferences caused the sales of existing Canadian products in Chile to be affected first. Canadian exporters continued to introduce new varieties; however, further expansion of each new Canadian variety would be constrained by the small size of the Chilean market and the competitive pressures from third countries.

Table 14: Extensive and Intensive Margins of Canadian Exports to Chile, 1995-2010

YearExtensive MarginIntensive MarginImport Share
19950.5650.0370.021
19960.5790.0420.024
19970.6130.0380.023
19980.6600.0440.029
19990.6250.0460.029
20000.5690.0530.030
20010.5730.0460.026
20020.5780.0350.020
20030.6560.0300.020
20040.5240.0300.016
20050.5580.0240.013
20060.5470.0250.014
20070.6100.0370.023
20080.6120.0280.017
20090.6100.0310.019
20100.5990.0220.013

Source: Author's own calculation

The market situation for Chilean exports to Canada was completely different compared to Canadian exports to Chile (see Table 15). Chile was the only major South American country with which Canada had an FTA. Chile's products enjoyed exceptional preferences in the Canadian market that most other South American products did not have. Most importantly, the Canadian market was much bigger and broader than Chile's, and therefore more capable to absorb any new varieties of Chilean industrial and agricultural products. As a result, both extensive and intensive margins for Chilean exports to Canada increased significantly following the implementation of the CCFTA. Chile's share in total Canada's imports from the world increased from 0.1 percent before the CCFTA to 0.4 percent in 2011 with extensive and intensive margins increased by 0.25 points and 0.002 points, respectively.

Table 15: Extensive and Intensive Margins of Canadian Imports from Chile, 1995-2011

YearExtensive MarginIntensive MarginImport Share
19950.1990.0060.001
19960.1940.0080.001
19970.2260.0050.001
19980.2650.0050.001
19990.2800.0050.001
20000.3400.0050.002
20010.3440.0050.002
20020.3150.0060.002
20030.3720.0070.003
20040.3620.0100.004
20050.3800.0110.004
20060.4200.0110.005
20070.3730.0110.004
20080.3900.0110.004
20090.4440.0110.005
20100.4400.0110.005
20110.4440.0100.004

Source: Author's own calculation

5.2 Assessing the CCFTA Effect on Economic Welfare

The analysis of the welfare effect of the CCFTA follows the approach of Arkolakis et al. (2012) by estimating Canada's overall welfare gains from the CCFTA. Estimating the contribution of welfare gains from the trade agreement could be challenging, but the recent paper by Arkolakis et al. provides an effective solution. The paper suggests that welfare prediction from all trade models could be pinned down to two key statistics:

  • (i) the share of expenditure on domestic goods; and
  • (ii) the elasticity of imports with respect to variable trade costs or "trade elasticity".

This approach significantly simplifies welfare calculations for trade agreements (See Annex 3).

Essentially, the welfare gain is calculated as,

Upper case W circumflex equals to lambda to the power of one over epsilon circumflex, also equals to one minus lambda to the power of negative one over epsilon, also equals to one minus open bracket one minus R close bracket to the power of negative one over epsilon. 

where

Upper case w circumflex

is the change in income, λ is the share of Canada's expenditure on domestic goods, R is the import penetration ratio of imports from Chile and ε is the trade elasticity. The share of Canada's expenditure on domestic goods could be replaced by (1 - R) where R is the import penetration ratio of imports from Chile and can be calculated by the following expression,

Upper case R equals to the summation on lower case k of upper case M subscript lower case k divided by upper case GDP plus summation on lower case k of upper case M subscript lower case k minus the summation on lower case r of upper case X subscript lower case r.

where GDP is the gross domestic product of Canada,

Summation on lower case k of upper case M subscript lower case k.

is the sum of all imports from Chile and

Summation on lower case r of upper case X subscript lower case r.

is the sum of all exports to Chile.

Table 16 reported the estimated welfare gains from trade with Chile for the period from 1990 to 2010 Footnote 17.

Table 16: Estimated Welfare Gains from Trade with Chile, 1990-2010

YearEstimated Welfare Gain (%)Expected Welfare Gain without the CCFTA Effect (%)
19900.008N/A
19910.008N/A
19920.009N/A
19930.009N/A
19940.009N/A
19950.011N/A
19960.013N/A
19970.0110.012
19980.0120.013
19990.0130.014
20000.0160.014
20010.0180.015
20020.0180.016
20030.0230.016
20040.0310.017
20050.0370.018
20060.0390.018
20070.0340.019
20080.0350.020
20090.0350.020
20100.0360.021

Source: Author's own calculation

The import penetration ratios for imports from Chile had been rather low, which implies the calculated λ are on average about 0.999. Thus the calculated welfare gain on average equals 0.020 percent of Canada's GDP given the trade elasticity is -3.24, which was estimated based on recent tariff and trade data. It can be seen that on average Canada's welfare gains from trade with Chile before the CCFTA is only 0.009 percent of Canada's GDP while after the implementation of CCFTA, the welfare gains rose to an average of 0.025 percent.

Next, we simulate the welfare gains from any expanded trade with Chile as if there were no agreement in place to estimate the amount of net increases to Canada's welfare gains induced by the CCFTA. The first step is to perform a trend regression on λ from 1990 to 1996 to predict the expected 's if the economic conditions in 1990-1996 were to continue. In which case, the constant is found to be 0.99978 and the coefficient for the trends is -0.00002. The expected welfare gains for trade with Chile without the CCFTA effect are listed side by side with the estimated total welfare gains for trade with Chile in the above table. As can be seen, the differences between the estimated welfare gains with and without the CCFTA effect were rather large. The difference expands quite rapidly after the implementation of the CCFTA, and then stabilizes at the 0.015 percent of GDP during the period from 2007 to 2010. Based on Canada's GDP in 2010 of $1.66 trillion, Canada's overall net welfare gains from the CCFTA would be around $250 million annually.

Figure 11: Difference between Estimated and Expected Welfare Gains from Trade with Chile, 1990-2010

See the following figure 11 text alternative for description.

Figure 11 Text Alternative
YearEstimated welfare gainsExpected welfare gains without the CCFTA effect
19900.008n/a
19910.008n/a
19920.009n/a
19930.009n/a
19940.009n/a
19950.011n/a
19960.013n/a
19970.0110.012
19980.0120.013
19990.0130.014
20000.0160.014
20010.0180.015
20020.0180.016
20030.0230.016
20040.0310.017
20050.0370.018
20060.0390.018
20070.0340.019
20080.0350.02
20090.0350.02
20100.0360.021

Source: Author's own calculation

Footnotes

Footnote 12

Wooldridge, Jeffrey M. (2002) "Econometric Analysis of Cross Section and Panel Data", pp. 603-642, The MIT Press, Cambridge, Massachusetts, U.S.A.

Return to footnote 12 referrer

Footnote 13

Data from the Global Trade Atlas.

Return to footnote 13 referrer

Footnote 14

The standardized tariff is expressed as a ratio of Canada/Chile's tariff for the product imported from a third country relative to Canada/Chile's tariff for the same product traded between Canada and Chile. As Canada/Chile tariffs asymptotically approach zero during the implementation period of the CCFTA, Canada/Chile tariffs for imports from third countries actually increase relative to Canada/Chile tariffs. For instance, assume a Canadian MFN tariff is 5% or (1+0.05) before the CCFTA, the standardized tariff or the tariff ratio for the same product imported from both Brazil and Chile is 1. After the CCFTA, however, the Canadian tariff for the product imported from Chile goes to zero, the standardized tariff for the same product imported from Brazil increases from 1 to 1.05.

Return to footnote 14 referrer

Footnote 15

Details of the derivation of margins can be found in the Annex 2.

Return to footnote 15 referrer

Footnote 16

Data are drawn from Global Trade Atlas. Any missing data are augmented by UNCTAD data from the World Integrated Trade Solutions.

Return to footnote 16 referrer

Footnote 17

All imports and exports data are taken from the Global Trade Atlas for 1995 to 2010 and augmented with additional data from World Integrated Trade Solutions for 1990 to 1994. The U.S.- Canada exchange rates are taken from the Bank of Canada website while the historical ones are taken from the Pacific Exchange Rate Service operated by the Sauder School of Business at University of British Columbia. The gross domestic product data at current prices are taken from Statistics Canada CANSIM 038-00017 for all years.

Return to footnote 17 referrer