The J-curve revisited: an empirical analysis for Canada.
Understanding the relationship between the exchange rate and the trade account is essential to a successful monetary and trade policy, both of which are particularly important for a small open economy such as Canada. The exchange rate--trade account relationship (the exchange rate channel) is one link in the monetary transmission mechanism. The monetary transmission mechanism is a process that deals with actions taken by the central bank and the associated consequences on macroeconomic variables. In general, the transmission mechanism is the complex chain of cause and effect that connects the central bank's policy instrument (typically the setting of a short-term interest rate) with asset prices, aggregate demand, total output, the output gap, and, eventually, inflation. The goal of monetary policy is to contribute to raising living standards for all citizens through low and stable inflation. If the Bank sees inflation rising above its target level, it seeks to slow down the economy through a rise in the overnight rate and, consequently, longer term interest rates and a rise in the value of the currency, leading to a dampening of aggregate demand and inflation.
Successfully conducting an inflation targeting policy requires a thorough understanding of the monetary transmission mechanism, including the exchange rate channel. While the general patterns of the monetary transmission process are understood, there is a significant amount of uncertainty with respect to the net effect on each component of aggregate demand and the timing of such effects. Relating to the exchange rate channel, it must be determined how exports respond to a change in the exchange rate and the length of the time lag. How quickly and in what magnitude do imports respond to the same change in the exchange rate? The purpose of this paper is to further investigate and understand the exchange rate channel.
While the Canadian dollar experienced depreciations relative to the U.S. dollar over most of the 1990s and over the early part of this decade, the value of the Canadian dollar has increased dramatically over the past 3 years. The substantial movement of the value of the dollar has revived interest in Canada on the effects of such movements on the trade account.
The literature has long debated the question of the response of the trade balance following a domestic currency appreciation or depreciation. The textbook view is one of an improvement in the trade account following domestic currency depreciation: imports become relatively more expensive leading to a reduction in the purchase of imports, while foreigners will purchase more domestic exports as they are relatively cheaper. This view, however, overlooks two crucial points: the degree to which exporters pass through exchange rate movements into local currency prices of their exports, known as the degree of exchange rate pass through (ERPT), and the degree trade volumes respond to the exchange rate. If, for example, trade volumes are sluggish to respond in the short run following currency depreciation, short-run lags in the response of trade volumes combined with complete ERPT will initially worsen the trade account. This phenomenon is known in the literature as the J-curve theory, where the response of the trade balance traces out a J-curve tilted to the right.
Evidence on the existence of a J-curve is mixed. Kantano and Klein (1996) estimate trade elasticities between Canada and the U.S. using quarterly data over the period 1977:1-1992:1 and find a deterioration of the trade account over the period of the currency depreciation. Rose and Yellen (1989) investigate the response of U.S. trade using quarterly data over 1960:1-1985:4 through estimating a partial reduced form equation for the merchandise trade account and conclude there is no evidence of a stable J-curve. Rose (1990) finds similar results for the trade balances of a number of developing countries. Moffet (1989) estimates pass-through and quantity response coefficients for the U.S. using quarterly data over the period 1967:11987:4 and simulates a trade balance adjustment and discovers the response does not resemble a J-curve. For further work on the J-curve, see the literature review presented by Bahmani-Oskooee and Ratha (2004).
This paper presents updated work on this issue. Unique monthly trade volume data is used for the 1981:1-2005:12 period, where the monthly data better captures the dynamic features of the exchange rate mechanism. A useful technique in capturing the response of the trade balance, as suggested by Demirden and Pastine (1995), is to employ VARs and impulse response functions which explicitly account for feedback effects, a feature lacking in the methods of the above studies. Furthermore, cointegrating relationships are incorporated in the VARs, as suggested by Ericsson et al. (1998). The studies by Kantano and Klein (1996) and Moffet (1989) are susceptible to spurious relationships due to the possibility of nonstationary variables. Rose and Yellen (1989) and Rose (1990) avoid this potential problem using differenced data, but at the expense of potentially ignoring valuable long-run information.
The response of the trade balance depends on the degree of response of its components, export and import prices and export and import volumes. This paper attempts to measure the response of Canada's trade balance, along with its components, to currency depreciation. The results indicate the J-curve does not exist. As a robustness check, taking into account increases in intra-industry changes from changes in location and production processes of firms following the North American Free Trade Agreement (NAFTA) in 1994, we investigate possible changes in trade effects before and after 1994. Over the sample period 1981:1-1993:12, we find, while there is a slight dampening of the response of the trade account and its components over the NAFTA period relative to the pre-NAFTA period, a J-curve effect does not occur over both periods.
The organization of the paper is as follows. The "Trade Balance Components and the J-curve" section discusses the components of the trade balance, their determinants, and the J-curve. The "Data" section discusses the data. The "Empirical Framework" section presents the empirical framework. VARs and impulse response functions are employed to measure the response of the trade account and its components to exchange rate changes, where the VARs have taken into account long-run relationships among the determinants of the components. The Johansen and Juselius (1990) cointegration approach is used to measure the long-run cointegrating relations. The "Results" section discusses the results and robustness checks. The final section concludes and presents possibilities for further work.
Trade Balance Components and the J-Curve
Trade Balance Components and their Determinants
Nominal net exports for Canada (NNX) expressed in Canadian dollars are represented through the identity:
NNX = EP x x - IP x m
where EP is an index of home export prices denominated in Canadian dollars, x is the quantity of home exports, IP is an index of import prices denominated in Canadian dollars, and m is the quantity of home imports.
To measure the response of the trade balance, we will first estimate the long-run determinants of each variable within the trade balance. We will then incorporate these long-run relationships in the corresponding VARs, where the impulse responses of each of the components will be derived. We begin with a traditional functional relationship for the trade balance:
[NNX.sub.t] = f ([E.sub.t], [y.sub.t], [yus.sub.t]) (1)
where [E.sub.t] is the nominal Canada-U.S. exchange rate because approximately 90% of Canada's trade is with the U.S. (Statistics Canada 2005), [y.sub.t] is Canadian real GDP, and [yus.sub.t] is U.S. real GDP.
In listing the standard determinants of each component of the trade balance, we estimate cointegration relations for each of the following specifications:
[EP.sub.t] = f([E.sub.t], [IPP.sub.t], [yus.sub.t], [y.sub.t]), (2)
[x.sub.t] = f([E.sub.t], [yus.sub.t]), (3)
[IP.sub.t] = f([E.sub.t], [PPIMUS.sub.t], [yus.sub.t]), (4)
[m.sub.t] =f([E.sub.t], [y.sub.t]). (5)
The specification for export volume [x.sub.t] (Eq. 3) captures the traditional determinants of the exchange rate and foreign income, the latter of which is U.S. income. Imports are mainly determined by the exchange rate and domestic income, [E.sub.t] and [y.sub.t], respectfully. Our concern with import prices (IP) is their response to changes in the currency value. The percentage which import prices rise when the home currency depreciates 1% is called the degree of pass-through. If the response of import prices to exchange rate changes is one-to-one, there is complete exchange rate pass through. This implies the exporter does not change the price of the good, denominated in the exporter's currency. Alternatively, the literature takes an industrial organizational framework, where exchange-rate changes are viewed as marginal cost shocks for the exporter (Dombusch (1987), Goldberg and Knetter (1997), Yang (1997), Stennek and Verboven (2001)). In either case, measuring the degree of exchange-rate pass through involves estimating import price equations; a common empirical approach is to estimate the functional relationship:
[IP.sub.t] = f([E.sub.t], [C.sup.f.sub.t], [Z.sup.f.sub.t]) (6)
where [C.sup.f.sub.t] is a measure of the foreign (U.S.) exporter cost and [Z.sup.f.sub.t] includes import demand and supply shifters. The proxy for the U.S. export cost will be the producer price index for manufacturing (PPIMUS) because approximately 90% of Canada's imports from 1981-2005 are manufactured goods (Statistics Canada 2005). Since Canada is a relatively small country, a demand factor is not warranted. A proxy for supply will be U.S. real GDP (yus). The coefficient on [E.sub.t] captures the degree of ERPT. Given the variables are in logs, ERPT is complete if the coefficient is 1.
Export prices will be modeled in a similar fashion as import prices, where costs of Canadian exporters are proxied by the industrial price index of all goods (IPP), given Canada's exports are predominately commodities and manufacturing goods (Statistics Canada 2005). Looking at Canadian export prices denominated in Canadian dollars, complete ERPT implies export prices remain unchanged; U.S. import prices then change the same percentage as the exchange rate.
Empirical models of these expressions will be estimated to capture the long-run cointegrating relationships. Then each cointegrating relationship will be included in the corresponding VAR and impulse responses will be derived for each expression above.
A typical J-curve scenario follows. A depreciation of the Canadian dollar raises the Canadian price of imports, IP. There is only a small immediate impact on the volume of trade flows, i.e. x and m change little. This is mainly due to contracts because changes in supply and sourcing require long-term investment decisions, a non-trivial cost. Hence most import and export orders are placed several months in advance. The demand side response is sluggish as consumers are slow change their habits. Hence, export and import volumes reflect buying decisions made on the basis of the old exchange rate. Given the domestic currency price of exports does not change (from the foreign country's view, there is complete ERPT for their imports), the value of imports (IP x m) rises substantially, and the value of exports (EP x x) only rises slightly, and the balance of trade deteriorates in the short run. As time passes and new contracts are negotiated, the increased price of imports eventually reduces the quantity of imports, while the volume and value of exports also increase, leading to an improvement in the trade balance. The response of the trade balance resembles a "J" tilted to the right. More formally, this short-run response implies the Marshall-Lerner condition is not satisfied. The condition states, under the assumption of infinitely elastic supply curves of imports and exports of a country and a given depreciation of the domestic currency, the domestic trade balance will improve as long as the sum of the price elasticities of import demand and export demand, in absolute terms, is greater than 1. The above reasons for a lagged response in import and export volumes translate into low demand elasticities that sum to less than 1.
Concern arises over the assumption of infinitely elastic supply curves. Specifically, would export prices remain constant in spite of the devaluation? It may be the case the foreign country (the U.S.) may not want to pass through exchange rate movements to the local (Canadian) currency prices of their exports, for fear of losing market share in the importing country (Canada); thus, the domestic price of the home country's imports does not increase by the same percentage as the depreciation of its currency. In this case, foreign firms lower their markup, resulting in revenues measured in its own currency from sales in the home country declining.
From the home country's (Canada's) point of view, in light of an inelastic demand for its exports, it is plausible producers will raise export prices, preventing the importing (foreign) country from facing a lower price of imports (denominated in the foreign country's currency) as a result of its currency appreciation. In this case, the domestic exporter increases its markup, raising total domestically-denominated revenue. In either case, import prices do not respond one-to-one to exchange rate changes. If either of these cases occur, it is possible for the trade balance to improve following a depreciation or, at a minimum, the J-curve should be weakened or dampened. Measuring the response of each component of the trade balance will reveal whether the trade balance resembles a J-curve response.
The data consists of monthly data for 1981:1-2005:12. The data used for nominal net exports (NNX) is merchandise exports less imports on a balance of payments basis. The data used for export prices (EP) is the Paasche current weighted export price index and for import prices (IP) is the Paasche current weighted import price index. For the volume of exports (x), constant dollar exports were calculated using the Laspeyres fixed weighted export volume index and the corresponding current value export series. Constant dollar imports (m) were calculated in a similar fashion. To our knowledge, these import and export volumes have not been used in trade account studies for Canada. The Canadian industrial price index was used as a proxy for the cost of production in the export sector (IPP) and the U.S. producer price index for manufacturing was used for the cost of production in the U.S. export sector (PPIMUS). Canadian output (y) is proxied by real (1992) GDP at factor cost and U.S. output (yus) is proxied by real (1992) industrial production. The nominal exchange rate (E) is the amount of Canadian dollars per U.S. dollar. All the variables were converted to logarithmic form. The Canadian non-GDP data was sourced from CANSIM and Statistics Canada. The U.S. producer price data was obtained from CITIBASE. GDP data was found in the International Monetary Fund's (IMF's) International Finance Statistics database.
Vector auto-regressions (VAR) are employed to investigate the response of the trade balance to nominal exchange rate innovations. The approach begins with a kdimensional unrestricted VAR model of order p:
[Y.sub.t] = [A.sub.1][Y.sub.t-1] + ... + [A.sub.k][Y.sub.t-k] + [[epsilon].sub.t], t = l, ..., T (7)
where [Y.sub.t] is a vector of p variables, [[epsilon].sub.t] is a p-dimensional vector of disturbances, E([[epsilon].sub.t][[epsilon]'.sub.t]) = V, and [[epsilon].sub.t] is uncorrelated with all variables dated t - 1 and earlier.
If some of the variables in the VAR are nonstationary then it will be difficult to describe their dynamics. Furthermore this may lead to improper statistical inference (estimation and hypothesis testing) as the VAR may not be stable or stationary. In the terms of the trade account and changes in the exchange rate, there is no theoretical explanation for a change in the nominal exchange rate to have a permanent effect on real exports and real imports in the long run. Instead nominal exchange rate changes only have nominal effects in the long run. In this context, real imports and real exports are mean reverting. Also, the results from using an unrestricted VAR to estimate impulse responses over long horizons without taking into account of cointegrating relations may be inconsistent (Phillips 1998). The Augmented Dickey and Fuller (1981) test for integrated variables was employed. Although not reported here, the results show the variables NNX, E, EP, IP, IPP, m, PPIMUS, yus and y behave as a random walk with a drift, while x behaves as a random walk. One solution is to ensure stationarity by differencing the data. A concern with this approach is the differenced data potentially ignores valuable information about a system's long-run dynamics if the level of some of the variables have long-run equilibrium relationships, if they are cointegrated. One solution adopted by Ericsson et al. (1998) is to incorporate the long-run cointegrating relations with the differenced data in the VAR. Measuring and accounting for both the short-run and long-run properties of the data is a necessary feature in correctly estimating impulse responses. The cointegrating relations for the components of the trade balance are estimated in the next section. Finally, the structural shocks are identified through the use of the recursive or Choleski factorization approach.
Long-run Relations and Cointegration
The Johnansen and Juselius (1990) technique provides a procedure to examine the cointegration in a multivariate setting, allowing for the testing and estimation of more than one cointegrating vector in the data. The approach makes use of all the information available in the long-run and short-run fluctuations of each variable. This approach beings with the unrestricted vector error correction (VEC) form:
[DELTA][Y.sub.t] = [[GAMMA].sub.1][DELTA][Y.sub.t-1] + ... + [[GAMMA].sub.k-1][DELTA][Y.sub.t-k+1] + [PI][Y.sub.t-1] + [PSI][D.sub.t] + [[epsilon].sub.t], t = 1, ..., T (8)
where [[epsilon].sub.1], ..., [[epsilon].sub.T] are niid (0,[summation]) and [D.sub.t] is a vector of nonstochastic variables, such as deterministic variables. The [PI] matrix conveys the long-run information. This approach makes use of all the information available in the long- and short-run fluctuations of each variable. When 0 < rank ([PI]) = r < p, [PI] can be written [PI] = [alpha][beta]', where [beta] may be interpreted as a p x r matrix of cointegrating vectors and [alpha] as a p x r matrix of error correction parameters. The remaining (p - r) unit root combinations are common stochastic trends.
The Schwarz Information Criteria was chosen in determining the optimal lag for the Vector Error Correction models. The left panel of Table 1 reports the cointegration trace test statistics and the right panel shows the resulting cointegrating parameters for each normalized variable in the trade balance. In all cases, the trace tests support one cointegrating vector. All of the estimated parameters have theoretically consistent signs. For the normalized nominal net exports variable (NNX), a rise in the nominal exchange rate (E) leads to an increase in nominal net exports in the long run, while a rise in U.S. real GDP (yus) increases net exports and a rise in Canadian real GDP (y) worsens the trade account.
In the export price equation (EP), export prices are sensitive to domestic producer price movements (IPP) as cost increases are passed through to export prices. A rise in the exchange rate or a currency depreciation results in export prices falling, although only by a relatively small amount. The signs on U.S. output and Canadian output can be interpreted as output demand and supply responses of export prices. The volume of exports (x) increases following currency depreciation, while a rise in output in the U.S. stimulates exports; both are theoretically correct responses.
The results of the import price equation indicate the degree of exchange rate pass through on import prices is 0.524 (i.e. a 1% depreciation of the home (Canadian) currency leads to a 0.524% increase in import prices). The less than complete ERPT reflects the unwillingness of U.S. exporters to pass through exchange rate movements to their Canadian currency export prices because of a fear of losing market share. Thus, U.S. firms lower their markups, where revenue from sales in Canada measured in U.S. currency, declines. This value is consistent with the studies of Goldberg and Knetter (1997) and Moffet (1989), which investigate the degree of ERPT for U.S. imports. Import prices increase as the cost of production in U.S. manufacturing increases. The negative sign on yus captures an output supply response of import prices. Canadian output was not included because Canada's small economy would have a negligible effect on U.S. import prices. The volume of imports (m) falls in response to currency depreciation. Finally, imports rise when Canadian output rises.
In analyzing the response of net exports and its components to exchange rate shocks, the long-run influences should be taken into account. Specifically, the estimated cointegrating vectors will be used to construct the equilibrium correction terms. These estimated dominant long-run feedback effects will be included in the corresponding VAR from which the impulse response functions will be derived. The constructed error correction terms are:
[ec1.sub.t] = [NNX.sub.t] - 0.568 - 0.097 [E.sub.t] - 0.577 [yus.sub.t] + 0.680 [y.sub.t],
[ec2.sub.t] = [EP.sub.t] - 1.51 + 0.006 [E.sub.t] - 2.84 [IPP.sub.t] + 0.311 [yus.sub.t] + 1.20 [y.sub.t],
[ec3.sub.t] = [x.sub.t] - 0.798 - 0.078 [E.sub.t] - 2.04 [yus.sub.t],
[ec4.sub.t] = [IP.sub.t] - 4.88 - 0.524 [E.sub.t] - 0.002 [PPIMUS.sub.t] + 0.092 [yus.sub.t],
[ec5.sub.t] = [m.sub.t] + 2.93 + 0.295 [E.sub.t] - 2.90 [y.sub.t].
Impulse Response Analysis
The immediate or short-run response of the trade account to an exchange rate change is best captured through impulse response functions. The VARs contain the differenced variables representing the components of the trade balance, along with the error correction terms calculated above. Figures 1, 2, 3, 4, and 5 present the impulse responses of the differenced variables in each equation in Table 1, along with their corresponding error-correction terms, from a shock to the nominal exchange rate. The letter d denotes a difference variable. For example, Fig. 1 presents the response of dNNX from a VAR composed of the variables ordered dyus, dE, dNNX, dy, and ec1 to a positive, one standard-deviation orthogonal shock to dE, along with the 95% confidence intervals (CI). Figures 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, and 15 show the response of a particular variable to a shock in the exchange rate (dE).
[FIGURE 1 OMITTED]
[FIGURE 2 OMITTED]
Figure 1 shows the response of the trade balance variable, dNNX, from a shock to the exchange rate, dE. Within the same period of the depreciation (time 0 on the horizontal axis), there is a small positive change in the trade balance, i.e. an improvement in the level of the trade balance. However, this improvement is not statistically significant as the confidence band spans zero, although marginally as the lower bound is just below the zero axis. Nevertheless, the change is not negative, suggesting there is not J-curve effect during the period of the exchange rate change. For the remaining periods, the responses are statistically insignificant. Overall, the results do not show a J-curve response to Canada's trade balance as the trade account remains relatively unresponsive to a currency depreciation.
Figures 2, 3, 4, and 5 present the trade account components' responses to shocks to the exchange rate, where the VARs that produce Figs. 2, 3, 4, and 5 relate to Eqs. 2, 3, 4, and 5, with each containing the corresponding error correction term. The results of each are, in general, consistent with the unresponsive result of the trade account, in particular, the result of not showing a J-curve effect. Beginning with the trade volumes, the volume of exports (dx) in Fig. 3 shows a marginal negative change and this result is statistically insignificant. The first period after the shock shows a marginal positive change in the trade account, but again it is statistically insignificant. For the import volumes (dm) in Fig. 5, there is very little, if any, change during the period of the exchange rate change. The remaining periods show very marginal changes and, again, are all statistically insignificant.
[FIGURE 3 OMITTED]
[FIGURE 4 OMITTED]
The responses of export prices (dEP) and import prices (diP) in Figs. 2 and 4 are relatively similar and are thus consistent with the response of the trade account. After the currency depreciation, they initially show a positive change, by approximately the same amount. Furthermore, these initial responses are statistically significant. For both prices, price changes stop after approximately two periods after the exchange rate shock. For export prices, the positive change suggests Canadian exporters raise the Canadian price of goods faced by U.S. importers. Hence, Canadian firms do not pass on the Canadian dollar depreciation to their U.S. customers. The result is consistent with the careful study conducted by Schembri (1989). Using data from the Census of Manufacturers on a Canadian export industry, when the Canadian dollar depreciated, U.S. dollar prices decreased only 15% of the total expected fall. This showed price stickiness or incomplete pass through to export prices, implying Canadian exporters are able to keep their prices and maintain relatively stable profit margins, reflecting some degree of monopoly power. This effect is also present in this study. Specifically, a one standard-deviation orthogonal shock to dE translates the response function of the exchange rate log to 0.01 on the vertical axis (not reported here). As Fig. 2 shows, Canadian dollar export prices change positively (incomplete pass through to U.S. importers) and by a less amount, approximately 0.005. These two effects lead to a drop in U.S. dollar prices, but by proportionately less than the rate of depreciation of the Canadian dollar, resulting in an incomplete ERPT.
[FIGURE 5 OMITTED]
[FIGURE 6 OMITTED]
Figure 4 shows the response of import prices. The response shows a positive change and this effect is statistically significant. Furthermore, there is incomplete pass through to import prices. Figure 4 shows the change in import prices is approximately 0.006, while the exchange rate changes by 0.01. Given that Canadian import prices do not fully adjust to changes in the exchange rate, this suggests that foreign firms are partly absorbing exchange rate changes in their profits. This response of Canadian import prices is another result consistent with the import price results in Schembri (1989).
Overall, both export and import volumes are unresponsive to exchange rate changes, perhaps due to quantities determined by contracts set in previous periods. Another potential explanation is the time take for Canadians importers to shift to cheaper products and for foreign consumers to realize prices have changed. With both import and export prices rising by approximately the same proportion and unresponsive export and import volumes, together, these responses are consistent with the unresponsiveness of the trade account in Fig. 1. Thus, there is not a J-curve effect in Canada following a currency depreciation.
[FIGURE 7 OMITTED]
[FIGURE 8 OMITTED]
Robustness Check: Pre-NAFTA and NAFTA Periods
We investigate whether the results from above are significantly affected with the implementation of NAFTA. It is well documented that this policy resulted in a rationalization and reallocation of production activity in North American, leading to greater intra-industry trade across the Canada-U.S.-Mexico borders (Head and Ries (1999), Statistics Canada (2002), Dion (2000), Cameron and Cross (1999). Increased intra-industry trade would imply trade flows being determined more by intra-industry or upstream-downstream production factors and timing and less on the exchange rate.
Cointegrating relations are estimated for Eqs. 1, 2, 3, 4, and 5 over the two sample periods 1981:1-1993:12 (pre-NAFTA) and 1994:1-2005:12 (NAFTA). The results are presented in Table 2. Most of the pre-NAFTA estimated coefficients are similar to the full period sample, including the parameters on the exchange rate. However, the signs on the exchange rate for the Wade balance (Eq. 1) and volume of exports (Eq. 3) changed during the NAFTA period. Error correction terms are constructed with these parameters and included in the corresponding VARs, to follow the procedure set by the full sample analysis. Figures 6, 7, 8, 9, and 10 present the impulse response functions over the pre-NAFTA period (1981:1-1993:12) and Figs. 10, 11, 12, 13, 14, and 15 show the impulse response functions over the NAFTA period 1994:1-2005:12.
[FIGURE 9 OMITTED]
[FIGURE 10 OMITTED]
Figure 6 shows an improvement in the trade account in the period of currency depreciation. However, this effect is statistically insignificant, but only marginally so because the lower bound of the confidence interval is just below the zero axis. In the remaining periods, the changes are also insignificant. Overall, as in the full sample case, in the pre-NAFTA period, there is no J-curve effect.
The response in the components of the trade account is consistent with no J-curve effect. Both impulse responses for export volume changes and import volume changes (Figs. 8 and 10) show a slightly greater response than over the whole sample. However, the responses are again insignificant, but only marginally so for export volumes. The positive responses of both export and import prices in Figs. 7 and 9 are similar to the full sample cases, both showing incomplete ERPT and both are statistically significant. Overall, relatively unresponsive trade quantities and export and import prices rise in similar proportions confirm a relatively unresponsive trade account, implying no J-curve effect over the pre-NAFTA period.
Over the NAFTA period, the trade account responds slightly less than in the pre-NAFTA period. Export and import volumes responses are small and more dampened than in the pre-NAFTA case; where the responses are again statistically insignificant. The response of export and import prices are similar as in the pre-NAFTA case; both are positive and statistically significant. Again, incomplete exchange rate pass through is present in both prices.
[FIGURE 11 OMITTED]
[FIGURE 12 OMITTED]
Overall, while there are some dampening effects on the trade account over the NAFTA period, the responses are, for the most part, similar for this variable and the components of the trade account over the pre-NAFTA and NAFTA periods. More importantly for this paper, the result of no J-curve effect for Canada is robust, even when taking into account NAFTA.
Robustness Checks on Levels of Variables
To explore the sensitivity of the impulse responses, we conducted the impulse responses derived from the VEC models of Eqs. 1, 2, 3, 4, 5, and 6. Hamilton (1994, pp. 652-653) proposes such a sensitivity test. Specifically, we investigate whether the first period impulse response of the nonstationary variable is similar to the response of the differenced variable. For all the variables, the initial (first period) responses were all consistent with the difference variable responses in Figs. 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, and 15 (full and sub samples). Most notably, the initial response of NNX was positive, not negative, not showing a J-curve effect after currency depreciation. Thus, the impulse response results in Figs. 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, and 15 are robust with respect to the corresponding VECM impulse responses. The results are available upon request.
[FIGURE 13 OMITTED]
[FIGURE 14 OMITTED]
Using an impulse response analysis that incorporates cointegrating relations and a unique data set of export and import volumes, evidence show the J-curve effect does not exist for Canada. Specifically, the impulse response showed a positive change in the period of the currency depreciation and not a negative response. The response of the trade account was statistically insignificant. The responses of the components of the trade account were consistent with the trade account effect, where export and import volumes were essentially not affected and export and import prices showed a significant positive change and incomplete pass through. The pricing responses showed the effect of pricing to market, reflecting a degree of monopoly power of Canadian exporters and firms importing into Canada. Robustness checks showed that while the response of the trade account and trade quantities here marginally dampened over the NAFTA period relative to pre-NAFTA, the result of no J-curve effect in Canada held in both periods.
From a policy perspective, identifying the trade account's relative unresponsiveness to exogenous changes in the exchange rate is relevant information in the measurement of the effect of the exchange rate channel on aggregate demand and consequently inflation, i.e. the exchange rate channel. After currency depreciation, possibly caused by monetary policy through an increase in the rate of growth of money, there will only be a marginal effect on aggregate demand through the trade account. Thus, there would be a weak influence on inflation, given aggregate demand relative to aggregate supply has not changed significantly, the results show import prices experiencing incomplete exchange rate pass through, implying a diminished inflationary effect of currency depreciation as import prices will not rise proportionately. In light of these results, in conducting monetary policy, to reduce output and consequently inflation, given the relative unresponsiveness of the exchange rate on the trade account, a greater change in the interest rate is necessary to achieve a given change in output.
[FIGURE 15 OMITTED]
Published online: 27 July 2008
Acknowledgement I am grateful to Gregor Smith, Nadia Soboleva, and Alex Maynard for their comments. Of course, the usual disclaimer applies.
Bahmani-Oskooee, M., & Ratha, A. (2004). The J-curve: A literature review. Applied Economics, 36, 1377-1398.
Cameron, G., & Cross, P. (1999). The importance of exports to GDP and jobs. Canadian Economic Observer, 12(11), 35-42.
Demirden, T., & Pastine, I. (1995). Flexible exchange rates and the J-curve: An alternative approach. Economic Letters, 48, 373-377.
Dickey, D. A., & Fuller, W. A. (1981). Likelihood ratio statistics for autoregressive time series with a unit root. Econometrica, 49, 1057-1072.
Dion, R. (2000). Trends in Canada's merchandise trade. Bank of Canada Review, 2000(Winter), 55-65.
Dornbusch, R. (1987). Exchange rates and prices. American Economic Review, 77, 93-106, March.
Ericsson, N. R., Hendry, D. F., & Mizon, G. E. (1998). Exogeneity, cointegration, and economic policy analysis. Journal of Business and Economic Statistics, 16(4), 370-387.
Goldberg, P., & Knetter, M. (1997). Goods prices and exchange rates: What have we learned? Journal of Economic Literature, XXXV 1243-1272, September.
Hamilton, J. D. (1994). Time series analysis. Princeton, NJ: Princeton University Press.
Head, K., & Ries, J. (1999). Rationalization effects of tariff reductions. Journal of International Economics, 47(2), 295-320.
Johansen, S., & Juselius, K. (1990). Maximum likelihood estimation and inference on cointegration--With applications to the demand for money. Oxford Bulletin of Economics and Statistics, 52(2), 169-210.
Kantano, M., & Klein, L. (1996). Estimation of J-curve: United States and Canada. Canadian Journal of Economics, 29(3), 523-539.
Moffet, M. H. (1989). The J-curve revisited: An empirical examination for the United States. Journal of International Money and Finance, 8, 425-444.
Phillips, P. C. B. (1998). Impulse response and forecast error variance asymptotics in nonstationary VARs. Journal of Econometrics, 83, 21-56.
Rose, A. K. (1990). Exchange rates and the trade balance: Some evidence from developing countries. Economies Letters, 34, 271-275.
Rose, A. K., & Yellen, J. L. (1989). Is there a J-curve? Journal of Monetary Economics, 24, 53-68.
Schembri, L. (1989). Export prices and exchange rates: An industry approach. In R. C. Feenstra (Ed.), Trade policies for international competitiveness, chapter 6. Chicago: University of Chicago Press and NBER.
Statistics Canada (2002). Canadian trade review (pp. 46-55). Canada: Department of Foreign Affairs and International Trade.
Statistics Canada (2005). Trade update 2005. Second annual report on Canada's state of trade (pp. 24-33). Canada: The Economic and Trade Analysis Division of the Department of Foreign Affairs and International Trade.
Stennek, J., & Verboven, F. (2001). Merger control and enterprise competitiveness: An empirical analysis and policy recommendations. Sweden: Centre for Economic Policy Research.
Yang, J. (1997). Exchange rate pass-through in U.S. manufacturing industries. Review of Economics and Statistics, 79(1), 95-104.
G. J. Georgopoulos ([mail]) Department of Economics, School of Analytic Studies and Information Technology, York University, 4700 Keele Street, Toronto, Canada, M3J 1P3 e-mail: firstname.lastname@example.org
Table 1 Johansen trace tests and estimation results, 1981:1-2005:12 Johansen trace tests and estimation results, 1981:1-2005:12 Equation and Trend in Data Levels and Constant in Normalized Variable Cointegrating Relations Trace Test Rank Statistic 95% Critical Value (1) NNX (lags = 2) [less than 54.15 * 47.21 or equal to] 0 [less than 18.13 29.68 or equal to] 1 (2) EP (lags = l) [less than 89.92 * 68.52 or equal to] 0 [less than 43.70 47.21 or equal to] 1 (3) x (lags = 2) [less than 166.60 * 29.68 or equal to] 0 [less than 3.91 15.41 or equal to] 1 (4) IP (lags = 3) [less than 48.02 * 47.21 or equal to] 0 [less than 12.24 29.68 or equal to] 1 (5) m (lags = 2) [less than 38.74 * 29.68 or equal to] 0 [less than 6.04 15.41 or equal to] 1 Equation and Trend in Data Levels and Constant in Normalized Variable Cointegrating Relations Cointegrating Parameters Constant E IPP (1) NNX (lags = 2) 0.568 0.097 -- (2) EP (lags = l) 1.51 -0.006 2.84 (3) x (lags = 2) 0.798 0.078 -- (4) IP (lags = 3) 4.88 0.524 -- (5) m (lags = 2) -2.93 -0.295 -- Equation and Trend in Data Levels and Constant in Normalized Variable Cointegrating Relations Cointegrating Parameters PPIMUS yus Y (1) NNX (lags = 2) -- 0.577 -0.680 (2) EP (lags = l) -- 0.331 -1.2 (3) x (lags = 2) -- 2.04 -- (4) IP (lags = 3) 0.002 -0.092 -- (5) m (lags = 2) -- -- 2.90 Table 2 Johansen trace tests and estimation results, pre NAFTA and NAFTA periods Johansen trace tests and estimation results, pre NAFTA and NAFTA periods Equation and Trend in Data Levels and Constant in Normalized Cointegrating Relations Variable Trace Test Rank Statistic 95% Critical Value Pre-NAFTA, 1981:1- 1993:12 (1) NNX [less than 58.82 (a) 47.21 or equal to] 0 [less than 15.30 29.68 or equal to] 1 (2) EP [less than 89.84 (a) 68.52 or equal to] 0 [less than 41.64 47.21 or equal to] 1 (3) x [less than 63.72 (a) 59.68 or equal to] 0 [less than 4.88 4.27 or equal to] 1 (4) IP [less than 62.14 (a) 47.21 or equal to] 0 [less than 27.39 29.68 or equal to] 1 (5) m [less than 29.81 (a) 29.68 or equal to] 0 [less than 5.66 15.41 or equal to] 1 NAFTA, 1994:1- 2005:12 (1) NNX [less than 60.10 (a) 47.21 or equal to] 0 [less than 27.84 29.68 or equal to] 1 (2) EP [less than 91.71 (a) 68.52 or equal to] 0 [less than 46.00 47.21 or equal to] 1 (3) x [less than 69.86 (a) 29.68 or equal to] 0 [less than 12.90 15.41 or equal to] 1 (4) IP [less than 59.23 (a) 47.21 or equal to] 0 [less than 19.89 29.68 or equal to] 1 (5) m [less than 73.26 (a) 29.68 or equal to] 0 [less than 8.54 15.41 or equal to] 1 Equation and Trend in Data Levels and Constant in Normalized Cointegrating Relations Variable Cointegrating Parameters Constant E IPP Pre-NAFTA, 1981:1- 1993:12 (1) NNX 2.21 0.258 -- (2) EP 1.55 -0.015 2.26 (3) x -0.865 0.269 -- (4) IP 5.36 0.552 -- (5) m -2.54 -1.15 -- NAFTA, 1994:1- 2005:12 (1) NNX 3.15 -0.486 -- (2) EP 4.34 -0.111 -0.064 (3) x 3.38 -0.266 -- (4) IP 5.42 0.151 -- (5) m 1.20 -0.123 -- Equation and Trend in Data Levels and Constant in Normalized Cointegrating Relations Variable Cointegrating Parameters PPIMU yus y Pre-NAFTA, 1981:1- 1993:12 (1) NNX -- 0.161 -0.675 (2) EP -- -2.05 0.389 (3) x -- 2.43 -- (4) IP -0.058 -0.145 -- (5) m -- -- 2.85 NAFTA, 1994:1- 2005:12 (1) NNX -- 0.227 -0.418 (2) EP -- 0.235 -0.064 (3) x -- 1.50 -- (4) IP -0.169 -0.005 -- (5) m -- -- 1.97 (a) Rejected at the 5% level of significance
|Printer friendly Cite/link Email Feedback|
|Author:||Georgopoulos, George J.|
|Publication:||Atlantic Economic Journal|
|Date:||Sep 1, 2008|
|Previous Article:||An inquiry into the contemporary differential between female and male voter turnouts.|
|Next Article:||Running the numbers on lotteries and the poor: an empirical analysis of transfer payment distribution and subsequent lottery sales.|