The Bilateral J-Curve: Sweden versus her 17 Major Trading Partners

Transcription

1 Bahmani-Oskooee and Ratha, International Journal of Applied Economics, 4(1), March 2007, The Bilateral J-Curve: Sweden versus her 17 Major Trading Partners Mohsen Bahmani-Oskooee and Artatrana Ratha The University of Wisconsin-Milwaukee and St Cloud State University Abstract The main purpose of this paper is to use disaggregate data at bilateral level between Sweden and her 17 trading partners and investigate the short-run and the long-run effects of real depreciation of Swedish krona on her bilateral trade balances. The methodology that deemed to be appropriate and is said to achieve our goal of investigating both effects at the same time is the bounds testing approach. The empirical results reveal that depreciation of krona has short-run effects on the trade balance in 14 out of 17 cases. However, the J-curve effect is present only in five cases, i.e., in the trade balance between Sweden and Austria, Denmark, Italy, Netherlands, and the U.K. In majority of the cases, the short-run effects do not last into the long run. Keywords: Bilateral J-Curve, Bounds Testing, Cointegration, Sweden JEL Classification: F31 1. Introduction A traditional method of assessing the impact of currency devaluation on the trade balance prior to 1973 and of currency depreciation since 1973 has been one of estimating the well-known Marshall-Lerner (ML) Condition. The ML condition asserts that for a small country if the sum of import and export demand elasticities add up to more than unity, devaluation or depreciation could improve the trade balance in the long run. Since introduction of the J-Curve concept by Magee (1973) researchers have tried to distinguish the short-run response of the trade balance from its long-run response. Magee postulated that in the short run due to adjustment lags, it is possible for the trade balance to deteriorate first and improve later, resulting in a pattern of movement that resembles the letter J, hence the J-Curve phenomenon. Since its introduction researchers have tried to test the J-Curve for almost every country. Bahmani-Oskooee and Ratha (2004) provide a detailed review of the literature. From the literature we gather that some countries have received relatively more attention than the others. Since this paper concentrates on the experience of Sweden with the J-Curve, a short review of the related studies could set the stage for this paper so that we can easily distinguish the contribution of this paper from the others. Sweden was one of many countries for which Bahmani-Oskooee and Niroomand (1998) estimated the Marshall-Lerner condition using Johansen s cointegration and maximum likelihood approach. They showed that there exist two

2 Bahmani-Oskooee and Ratha, International Journal of Applied Economics, 4(1), March 2007, cointegrating vectors. While in the first vector the ML condition was not met, in the second one it was. However, Bahmani-Oskooee and Niroomand failed to adjust Johansen s λ-max and trace statistics for number of observations, number of variables and number of lags imposed on the VAR. Following Cheung and Lai (1993, p. 317) once the adjustment is made, the results show that there is only one vector and estimate of this vector (i.e., the first vector) fails to support the ML condition, implying that in the long run depreciation of Swedish krona cannot improve her trade balance with the rest of the world. The ML condition is an indirect method of assessing the impact of currency depreciation on the trade balance. Existing time-series techniques allow us to test directly the long-run relationship between the trade balance and the exchange rate. This is exactly what Hacker and Hatemi-J. (2003) do for five small north European countries including Sweden. They directly relate a measure of the trade balance to real effective exchange rate in addition to domestic and world income. Although they apply Johansen s cointegration technique to justify the long-run relationship, they do not report cointegration test results, nor do they report estimates of cointegrating vectors. They then use an error-correction model and engage in impulse response analysis and provide support for the J-Curve for most cases including Sweden. The two studies reviewed above use aggregate data between Sweden and the rest of the world and could suffer from aggregation bias. To overcome the bias, Hatemi-J. and Irandoust (2005) disaggregate the data and estimate the ML condition between Sweden and her six major trading partners using annual data over period. Their bilateral trade elasticities reveal that except for the case of Germany, the ML condition is not satisfied at bilateral level, implying that real depreciation of krona has no long-run favorable impact on the trade balance between Sweden and each of the remaining five trading partners (i.e., Denmark, France, Norway, the UK and the US). Australia could also be added to the list of five partners due to evidence from Bahmani-Oskooee et al. (2005) who investigated the short run and the long run effects of real depreciation of the Australian dollar on her bilateral trade balance with her 23 partners that included Sweden. They showed that while real depreciation of the Australian dollar against Swedish Krona had short-run effects on Australia-Sweden bilateral trade balance, these effects did not last into the long run. Finally, bilateral trade flow models between Sweden and her eight trading partners were subject to panel cointegration analysis by Irandoust et al. (2006). Due to annual and limited number of observations, they establish cointegration using panel approach and use the estimates for bilateral analysis. From the estimates they infer that the ML condition is met between Sweden and only two partners (France and Netherlands) As the above review indicates the mixed evidence on the effectiveness of depreciation of Swedish Krona on her bilateral trade balance has been limited to no more than eight major trading partners from Europe. In this paper we try to assess the short-run as well as the long-run effects of depreciation of Swedish krona on her bilateral trade balance with her trading partners. To do so we employ bounds testing approach to error-correction modeling and cointegration which is designed to capture the short-run and long-run effects of changes in one variable on another. Furthermore, to make the study as comprehensive as possible, we extend the list of trading partners to include 17 partners. To learn about the relative importance of each partner, we report the trade shares as well as the list of the partners in Table 1. The rest of the paper is organized in the following manner. Section 1 introduces the bilateral trade balance model and the

3 Bahmani-Oskooee and Ratha, International Journal of Applied Economics, 4(1), March 2007, methodology. Section 2 reports the results. Our summary and conclusion are provided in Section 4. Finally, the sources of the data and definition of variables are discussed in an Appendix. 2. The Model and the Methodology 2.1 The Model Almost every study that has tried to assess the impact of currency depreciation on the trade balance, has directly related a measure of the trade balance to home income, foreign income, and a measure of the real exchange rate. Therefore, we follow that tradition in this paper and adopt the model from Bahmani-Oskooee and Brooks (1998) as outlined by equation (1): Log TB = α + β Log Y + γ LogY + λ Log REX ε (1) i, t S,t i,t i,t + In (1) TB i is a measure of the trade balance between Sweden and trading partner i. Since the model is specified in log-linear term, it is defined as the ratio of Swedish exports to partner i over her imports from partner i. 1 The two measures of income are denoted by Y S and Y i. While Y S denotes the Swedish income, Y i denotes the income of trading partner i. The real exchange rate between krona and the currency of trading partner i is denoted by REX i. Finally, ε i is an error term. We expect an estimate of β to be negative or positive. If increase in Swedish income results in an increase in her imports from partner i, then an estimate of β is expected to be negative. However, if the increase in Swedish income is due to an increase in production of import-substitute goods, actually her imports could decrease (Bahmani-Oskooee, 1986). By the same token, an estimate of γ could also be positive or negative. Finally, as the Appendix indicates, the real bilateral exchange rate between Swedish krona and trading partner i s currency is defined in a way that an increase reflects real depreciation of krona. If real depreciation of krona against i s currency is to improve her trade balance with partner i, an estimate of λ is expected to be positive. 2.2 Methodology If we estimate equation (1) by any technique, we will only be able to assess the long-run effects of the right-hand side variables on the dependent variable. Since our purpose is to also assess the short-run effects of currency depreciation on the bilateral trade balance between Sweden and each of her 17 trading partners, following recent advances in time-series econometrics we must specify (1) in an error-correction modeling format. On this regard we follow Pesaran et al. (2001) as in equation (2): i, t Δ LogTB i,t = α + + δ LogTB 1 n k= 1 ω Δ LogTB k i,t-1 + δ LogY 2 i,t k + n k= 0 S,t 1 β k Δ LogY + δ LogY 3 S,t k i,t 1 + γ Δ LogY + δ LogREX 4 n k= 0 k i,t 1 i,t k + + u i,t n k= 0 λ k Δ LogREX i,t k (2)

4 Bahmani-Oskooee and Ratha, International Journal of Applied Economics, 4(1), March 2007, Equation (2) without lagged level variables is a standard VAR model. The linear combination of lagged level variables is added as a proxy for lagged error-correction term. Are we justified to add the linear combination of lagged level variables? Pesaran et al. (2001) propose applying the familiar F test for their joint significance. However, they show that the F test in this context is non-standard and has its own new critical values that depend upon whether variables in (2) are integrated of order zero, I(0) or order one, I(1). By assuming all variables are I(1), Pesaran et al. (2001) provide an upper bound critical value and by assuming all variables are I(0), they provide a lower bound critical value. For joint significance of all lagged level variables, the calculated F statistic should be greater than the upper bound critical value which is also an indication of cointegration among the variables. Since integrating properties of the variables are incorporated in calculation of the critical values, Pesaran et al. (2001) argue that there is no need for pre-unitroot testing and variables could be I(1) or I(0) or combination of the two. While this is the main advantage of this approach, there is another advantage, that is, the short-run and the long-run effects of the right hand side variables on the dependent variable are assessed simultaneously. For example, the short-run effects of currency depreciation on the bilateral trade balance are inferred by the estimates of λ K s. The J-curve will be supported if λ takes negative values at lower lags and positive values at higher lags. The long-run effects of real depreciation are inferred by the estimate of δ 4 that is normalized by δ The Results 3.1 The F-Test The error-correction model (2) is estimated between Sweden and each of her 17 trading partners identified in Table 1 using quarterly data over the 1980I-2005IV period. One problem in applying the F test for cointegration is the number of lags that need to be imposed on each first differenced variable. The evidence from the literature suggests that the results of the F test will be sensitive to the lag length. To avoid this problem we follow Bahmani-Oskooee and Gelan (2006) and impose a maximum of 12 lags on each first differenced variable. We then use AIC criterion and select the optimum number of lags. All results reported, therefore, belong to optimum models. First, we carry out the F test for the joint significance of all lagged level variables at optimum lags and report the results in Table 2. The results in Table 2 reveal that the calculated F statistic is greater than its critical value of 3.52 at the 10% level of significance in all countries except in the results for Australia, Canada, Italy, Portugal, Switzerland and the U.S., implying that in 11 cases there is evidence of cointegration. In six countries where the F statistic rejects cointegration, we will proceed with our analysis and provide an alternative way of supporting cointegration later. 3.2 The Short-Run Results Next, we consider the short-run coefficient estimates. Since the central theme of the paper is to infer the J-curve effect, for brevity we only report the short-run coefficient estimates of the bilateral exchange rate in Table 3.

5 Bahmani-Oskooee and Ratha, International Journal of Applied Economics, 4(1), March 2007, From Table 3 we gather that there are 14 cases in which there is at least one coefficient estimate that is significant at the 10% level, implying that in these cases the bilateral exchange rate has short-run effects on the trade balance between Sweden and each of these 14 trading partners. The exception cases are Australia, Canada, and France. Furthermore, negative coefficients are followed by positive ones, supporting the J-curve in the cases of Austria, Denmark, Italy, Netherlands, and the U.K. There is evidence of an inverse J-curve in the results for Japan, Norway and the U.S. 3.3 The Long-Run Results The question remains to be answered is: do these short-run effects last into the long-run? To this end, we report the long-run coefficient estimates in Table 4. Considering 10% level of significance, Table 4 reveals that the bilateral real exchange rate has favorable effect on the trade balance in the results for Norway and unfavorable effect in the cases of Germany, Italy, Switzerland, U.K. and the U.S. Thus, in these five cases, the negative shortrun effects do last into adverse long-run effects. Lack of any significant or adverse long-run effects of depreciation of Swedish krona on her trade balance with Denmark, France, U.K. and U.S. is consistent with Hatemi-J. and Irandoust (2005) who used an alternative methodology to show that the Marshall-Lerner condition is not satisfied in the trade between Sweden and each of these four countries. However, our finding in the case of Germany contradicts them. Our finding of, again, no significant impact of the bilateral exchange rate on the trade balance with Australia is consistent with Bahmani-Oskooee et al. (2005) who investigated the impact of depreciation of the Australian dollar on her trade balance with her 23 trading partners one of which was Sweden. Turning to the income effects we gather that Swedish income carries a significant coefficient at the 10% level in 10 cases. In the cases of Australia, Austria, Denmark, Germany, and Spain the estimate is negative, implying that economic growth in Sweden that results in more Swedish imports, has adverse effect on her trade balance with each of these five countries. On the other hand, in the results for Italy, Netherlands, Portugal, U.K. and the U.S., the estimated income elasticity is positive. In these cases, increase in Swedish income is due to increase in production of import-substitute commodities which leads to actually a decrease in Swedish imports from each of these partners. The decreased imports, in turn, lead to an improvement in the trade balance of Sweden with each of these partners. As for the effect of trading partner s income, the results reveal that Log Y i carries a significant coefficient at the 10% level in 12 cases. While the estimated elasticity is positive in nine cases, it is only negative in the cases. The positive elasticity in most cases indicates that as Sweden s trading partners grow, they import more from Sweden, leading to an improvement in Sweden s trade balance with each of these nine partners. The list includes Australia, Austria, Denmark, Finland, Germany, Japan, New Zealand, Spain and Switzerland. 3.4 Alternative Evidence of Cointegration The last question we would like to address using the results reported so far is whether the shortrun adjustment of all variables is toward long-run equilibrium. To answer this question, following Pesaran et al. (2001) we use the long-run coefficient estimates reported in Table 4 and calculate the lagged linear combination of all variables in (2) over time and denote it by EC t-1.

6 Bahmani-Oskooee and Ratha, International Journal of Applied Economics, 4(1), March 2007, We then replace the linear combination of the lagged level variables by EC t-1 and re-estimate the model for each case after imposing the optimum number of lags. A negative and significant coefficient obtained for EC t-1 will reflect the fact that the adjustment is toward equilibrium. Furthermore, as argued by (Bahmani-Oskooee and Ardalani 2006), the negative and significant coefficient of EC t-1 could also reflect cointegration among the variables. The results of this exercise that are reported in Table 3 clearly support the adjustment toward equilibrium as well as cointegration in all cases. 4. Summary and Conclusion Due to adjustment lags, the dynamic short-run path of the trade balance after a devaluation or depreciation is said to follow the well-known J-Curve phenomenon. 3 The literature on the J- curve is vast and includes testing the phenomenon for almost every country for which data is available. However, some countries have received more attention than the others. Sweden is one such country for which previous research has provided mixed and sensitive results depending on the methodology employed. In this paper we consider the trade balance between Sweden and her 17 trading partners and employ bounds testing approach to cointegration and error-correction modeling to investigate the impact of real depreciation of Swedish krona on her bilateral trade balance with each of her partners. The advantage of using bounds testing approach is that the short-run and the long-run effects of depreciation are inferred simultaneously and in one step. Once the method is applied to the models, the results reveal that real depreciation of krona has short-run effects on the bilateral trade balance between Sweden and 14 of her trading partners. The short-run effects support the J- curve phenomenon only in the results for Austria, Denmark, Italy, Netherlands, and the U.K. In most cases, however, the short-run effects do not last into long-run. Endnotes Mohsen Bahmani-Oskooee, Center for Research on International Economics and Department of Economics, University of Wisconsin-Milwaukee, Milwaukee, WI 53201, bahmani@uwm.edu; Artatrana Ratha, Economics Department, St Cloud State University, St Cloud, MN 56301, aratha@stcloudstate.edu. 1. The ratio is said to be a unit free measure of the trade balance. It also measures the trade balance in real or nominal terms (Bahmani-Oskooee, 1991). 2. Note that this approach has already been used by Bahmani-Oskooee and Brooks (1999) to estimate the bilateral J-curves between the U.S. and her six major trading partners. 3. For more on adjustment lags and the J-Curve see Bahmani-Oskooee (1985).

7 Bahmani-Oskooee and Ratha, International Journal of Applied Economics, 4(1), March 2007, References Bahmani-Oskooee, M Devaluation and the J-Curve: Some Evidence from LDCs, The Review of Economics and Statistics, 67(3), Bahmani-Oskooee, M Determinants of International Trade Flows: The Case of Developing Countries, Journal of Development Economics, 20(1), Bahmani-Oskooee, M Is there a Long-Run Relation between the Trade Balance and the Real Effective Exchange Rate of LDCs? Economics Letters, 36(4), Bahmani-Oskooee, M. and F. Niroomand Long-Run Price Elasticities and the Marshall-Lerner Condition Revisited, Economics Letters, 61(1), Bahmani-Oskooee, M. and T. J. Brooks Bilateral J-Curve between US and Her Trading Partners, Weltwirtschaftliches Archiv, 135(1), Bahmani-Oskooee, M. and A. Ratha The J-Curve: A Literature Review, Applied Economics, 36(13), Bahmani-Oskooee, M., G. Goswami, and B. Talukdar The Bilateral J-Curve: Australia versus her 23 Trading Partners, Australian Economic Papers, 44(2), Bahmani-Oskooee, M. and Z. Ardalani Exchange Rate Sensitivity of U.S. Trade Flows: Evidence from Industry Data, Southern Economic Journal, 72(3), Bahmani-Oskooee, M. and A. Gelan Black Market Exchange Rate and Productivity Bias Hypothesis, Economics Letters, 91(2), Cheung, Y-W. and K. S. Lai Finite-Sample Sizes of Johansen s Likelihood Ratio Tests for Cointegration, Oxford Bulletin of Economics and Statistics, 55(3), Hacker, R. S. and A. Hatemi-J Is the J-Curve Effect Observable for Small North European Economies? Open Economies Review, 14(2), Hatemi-J., A. and M. Irandoust Bilateral Trade Elasticities: Sweden versus her Major Trading Partners, American Review of Political Economy, 3(2), Irandoust, M., K. Ekblad, and J. Parmler Bilateral Trade Flows and Exchange Rate Sensitivity: Evidence from Likelihood-Based Panel Cointegration, Economic Systems, 30(2), Magee, S. P Currency Contracts, Pass-through, and Devaluation, Brookings Papers on Economic Activity, 1,

8 Bahmani-Oskooee and Ratha, International Journal of Applied Economics, 4(1), March 2007, Pesaran, M. H., Y. Shin, and R. J. Smith Bounds Testing Approaches to the Analysis of Level Relationships, Journal of Applied Econometrics, 16(3),

9 Bahmani-Oskooee and Ratha, International Journal of Applied Economics, 4(1), March 2007, Appendix Data Definition and Sources All data are quarterly form the period 1980I-2005IV and come from the following sources: a. Direction of Trade Statistics of IMF (CD-ROM). b. International Financial Statistics of IMF (CD-ROM). Variables TB i Y i Y S = Sweden s trade balance with her trading partner i defined as the ratio of Sweden s exports to i over her imports from i (exports and imports are collected from source a). = Index of real GDP of country i (collected from source b). = Index of real GDP of Sweden (collected from source b). REX j = Bilateral real exchange rate between the Swedish Krona and trading partner i s currency. It is defined as (P i *E i )/ P S where P i is the CPI in country i, P S is the CPI in Sweden, and E i is the nominal exchange rate between Krona and trading partner i s currency defined as number of krona per unit of i s currency. Thus an increase in REX j is a reflection of real depreciation of Krone relative to j s currency.

10 Bahmani-Oskooee and Ratha, International Journal of Applied Economics, 4(1), March 2007, Table 1. Sweden s Trade Shares with Her Trading Partners in 2004 Trading Partner i Exports (millions of dollars) Imports (millions of dollars) Trade share (%) Australia Austria Canada Denmark Finland France Germany Italy Japan Netherlands New Zealand Norway Portugal Spain Switzerland United Kingdom United States INDUSTRIAL COUNTRIES WORLD Note: Together these 17 countries accounted for 74% of Sweden s trade with the Rest of the world.

11 Bahmani-Oskooee and Ratha, International Journal of Applied Economics, 4(1), March 2007, Table 2. The Result of F-Test for Cointegration Among the Variables of Bilateral Trade Balance Model between Sweden vis-à-vis her Trading Partners Trading Partner i Australia Austria Canada Denmark Finland France Germany Italy Japan Netherlands New Zealand Norway Portugal Spain Switzerland United Kingdom United States Calculated value of F-Statistic F(8,1,7,12)=6.63 F(3,7,3,0)=1.72 F(4,0,1,0)=3.48 F(6,6,10,0)=3.78 F(2,5,2,0)=4.07 F(1,0,0,6)=10.27 F(2,0,3,0)=4.25 F(5,9,0,0)=2.46 F(8,3,7,12)=7.25 F(12,12,12,9)=5.15 F(12,11,10,12)=8.23 F(12,11,11,12)=9.11 F(2,1,3,0)=1.53 F(7,7,3,9)=5.50 F(2,0,3,0)=2.98 F(1,12,12,12)=10.93 F(4,11,0,12)=3.19 Notes: a. F (8,1,7,12) is the calculated F statistic when 8 lags are imposed on Δ LogTB, 1 lag on Δ Log Y S, 7 lags on ΔLog Y i, and 12 lags on Δ Log REX. b. At the 10% level of significance, the upper bound critical value of F test is This comes from Pesaran et al. (2001, Table CI: Case II, p. 300).

12 Bahmani-Oskooee and Ratha, International Journal of Applied Economics, 4(1), March 2007, Table 3. Coefficient Estimates of Exchange Rate and Error Correction Term Based on Akaike Information Criterion Trading Partner Australia Austria Canada Denmark Finland France Germany Italy Japan Netherlands New Zealand Norway Portugal Spain Switzerland United Kingdom United States Lags of Δ Log REX (1.13) (3.76) 0.23 (0.12) (1.41) 0.01 (0.21) (1.52) (2.36) (2.50) 0.17 (0.52) (2.39) (0.32) 1.45 (3.11) (1.84) (1.35) (2.02) (1.99) 0.54 (1.61) (1.43) 0.31 (1.38) (0.72) 0.80 (2.64) 0.31 (0.95) 0.62 (1.31) (0.10) (1.76) 0.28 (0.40) 1.38 (3.54) 0.42 (1.22) (0.78) 0.33 (1.52) (0.50) 0.49 (1.87) (1.74) 0.94 (1.97) 2.27 (1.78) (0.73) (2.18) 1.18 (2.95) 0.52 (0.48) (3.48) 0.11 (0.52) (3.12) 0.41 (1.46) 0.40 (0.87) 0.07 (0.05) (3.24) 1.18 (1.97) 1.18 (3.17) 0.16 (0.65) (0.03) 0.28 (1.40) 0.11 (2.37) 0.27 (1.07) 0.42 (0.92) (1.48) (0.40) (2.22) 1.00 (2.65) 0.23 (0.67) 0.26 (3.66) 0.50 (2.43) 0.88 (3.58) 1.17 (2.80) (1.83) 0.18 (0.04) 0.72 (1.26) 1.09 (2.88) 0.25 (0.47) 0.12 (1.52) 0.58 (2.45) 0.73 (1.82) (0.21) (1.60) 0.73 (1.45) 1.28 (3.20) 0.16 (0.96) 0.26 (1.04) 0.51 (1.25) 1.37 (1.24) (0.85) 1.54 (3.89) 0.02 (0.06) 0.45 (1.82) 0.03 (0.07) (2.17) (0.77) 1.25 (3.01) (1.96) 0.16 (0.42) (3.63) (0.78) 0.65 (1.68) (1.21) (0.03) ( (2.50) 0.64 (1.89) 1.31 (3.64) 0.62 (1.81) EC(-1) (3.89) (2.82) (3.35) (3.55) (3.61) (6.66) (5.20) (4.05) (4.13) (4.25) (5.54) (4.59) (4.52) (5.00) (3.44) (6.47) (2.02) Note: Figures in parentheses represent absolute values of t-statistic.

13 Bahmani-Oskooee and Ratha, International Journal of Applied Economics, 4(1), March 2007, Table 4. Estimated Long Run Coefficients of the Bilateral Trade Balance Model Trading Partner (country j) Australia Austria Canada Denmark Finland France Germany Italy Japan Netherlands New Zealand Norway Portugal Spain Switzerland United Kingdom United States Log REX j Log Y S Log Y i Intercept 0.05 (0.13) (1.57) 0.39 (0.12) (1.78) 0.12 (1.61) (1.57) (2.52) (3.92) (0.004) (1.57) 0.62 (1.06) 1.08 (2.12) 1.14 (1.59) 0.86 (0.91) (2.27) (9.01) (1.85) (5.18) (2.09) 3.76 (2.13) (5.43) (1.52) 0.29 (0.51) (8.07) 1.38 (2.77) (1.22) 2.24 (2.39) (1.55) 2.08 (1.46) 3.08 (1.90) (2.46) (1.56) 9.86 (6.06) 5.08 (2.05) 8.15 (5.39) 0.75 (2.32) (1.34) 1.65 (3.40) 1.74 (2.63) (0.65) 1.55 (8.91) (0.14) 3.31 (3.42) (2.74) 1.77 (1.67) (2.24) (0.44) 2.14 (2.37) 2.31 (2.36) (4.83) 3.45 (1.59) (5.00) 8.23 (1.96) 7.84 (3.34) 2.40 (2.41) 2.35 (1.43) 0.50 (0.76) 1.45 (3.15) (4.56) (1.65) 0.38 (0.31) 1.49 (0.91) (0.19) (1.49) 6.14 (1.28) (2.63) (7.64) (2.61) Note: Figures in parentheses represent absolute values of t-statistics.