Joanna Nestorowicz Gilbert Mbara International Trade Policy prof. J.J. Michałek

F A C T O R P R I C E E Q U A L I Z AT I O N A N D TA R I F F S

Why do factor prices vary across countries? This question is of fundamental importance to governments pursuing policies for development through international trade openness and integration. Economic theory suggests two powerful mechanisms promoting factor price convergence across economies — goods trade and factor mobility. In the case, when markets are highly integrated, and factors of production are more mobile, then factor price equalization is more likely to occur internationally. Such a scenario would be beneficial for the world eceonomy allowing for specialization according to comparative advantage assumptions. Yet there are costs imopsed on trade of goods and factor mobility, which increase the costs of international flows of goods and capital, not allowing for pareto optimality in the world trade game.

1. Introduction Factor price equalization is a standard assumption in most trade theory models. However, empirical observations of factor price equalization have been hard to come by. For instance, the factor price equalization hypothesis has been rejected in important contributions to the trade literature such as Trefler (1993), Cunat (2001), and Bernard, Redding, Schott and Simpson (2002) amongst others. Theoretically, Mokhtari and Rassekh (1989) note that in conducting an empirical test, one must note that theorems need not be tested since they are results whose validity can be proven mathematically. The rejection of the factor price hypothesis in empirical tests should therefore not be surprising in spite of the existence of a variety of explanations for the rejection of the hypothesis such as multiple cones of diversification, quality specificity of factors and increasing returns to scale (see. Feenstra, 2004; Trefler, 1993). Mokhtari and Rassekh, following Samuleson (1971), hence argue that instead of a direct test of the hypothesis, one can test its implications in the sense that a move from autarky to free trade will induce a reduction in the factor price differentials amongst trading partners. They test this implication in their paper. 1

The movement from autarky to free trade usually has various consequences. It has been shown that in the two-factor two-country model of international trade, an increase in the price of output of a factor that is used more intensively results into a more than proportionate increase in the price of the factor (Feenstra, 2004). The analogous is also true, in that a fall in the price of the product results into a more than proportionate fall in the price of the factor used more intensively in its production [we formally show this result in section 2 below]. Conversely, an increase in the price of a product results into a fall in the price of the factor used less intensively in its production, again within the two-by-two model. Such changes then have distributional consequences in the economy. If the changes in prices are induced by a movement from autarky to free trade, then we can think of trade policy instruments such as tariffs and quotas as tools at the disposal of a government to help diminish the negative distributional effects of trade. For instance, a restriction such as an export quota on a labour-intensive good ensures that the price of the good does not increase in the home market and result into fall in the price of capital. In this paper, we attempt to relate the failure of convergence of factor prices across trading regions or countries as being partly driven by the existence of tariffs. Specifically, we will hypothesize that variation in factor prices is partly explained by the levels of tariffs. Furthermore as tariffs increase, the variation in factor prices increases. We devise a simple regression method based on the work of Mokhtari and Rassekh (1989) to test our hypothesis. The paper is structured as follows: section 2 presents the theoretical background and results; section 3 presents the suggested empirical tests. 2. Theoretical Results 2.1 Factor Price Equalization The existence of factor price equalization is a mathematical artefact that results from the maximization of factor prices from a given set of output prices (Ruffin, 1988). Ruffin describes a model in which there are r types of productive factors and n goods. Each factor can be described as ‘labour’ with other interpretations possible. The home country is endowed with Li units of type i labour and the foreign country Li* units. The total amount of type i labour is then denoted Ti. Letting aij denote the amount of type i labour needed to produce a unit of good j. The amount of good j produced by Lij units of type i labour is then given by Xij=Lij/aij. Since good j can be produced by many labour types, total good j production would be Xj = ΣiLij/aij which is a representation of the Ricardian r x n as a special case of the Heckscher – Ohlin (H-O) model. Under the Ricardian 2

framework, there is an efficient reallocation of production factors [each type of labour or ‘country’ in the traditional analysis]. The production set generated by the endowment and technology of each type of labour is then Yi={(Xij): ΣjaijXij≤ Li}. With identical technologies worldwide, a similar set defines the production set generated by the world supply of each type of labour. In the transformation surface consists of points where each type of labour is completely specialised, lines where one type of labour produces more that two goods and higher dimension planes where each labour type produces more than two commodities. Then the dual of the fact that transformation surfaces between countries have flats, lines and points that are geometrically similar is that changes in product prices cause parallel realignments in the product produced by each type of labour among the different countries. Given a set of world prices, labour is located to the activity where it has a comparative advantage in each country. The wage rate of type i labour, Wi, is determined by the allocation that maximises its earnings over all possible activities. If Pj is the world price of good j, then Wi=max j (Pj /a ij). Universal factor price equalisation (FPE) occurs as long as each country faces the same prices.

2.2 Relative Prices and Factor Rewards The Samuleson-Stopler theorem basically describes the relation between relative prices of products and output prices. We use the theorem following Jones’s (1965) approach to describe what happens when there is a change in prices induced by the movement from autarky to free trade. Consider a world with two factors, labour and capital; two goods, good 1 and 2; and two countries, domestic and foreign. Within each country there is perfect competition and free market entry in each industry. The two last conditions imply that profits are equal to zero. Following Feenstra (2004), the zero profit conditions can be described using prices and costs as: Pi=C i (w, r); where Pi is the price, Ci is the cost, w is the cost of labour, r is the cost of capital and i=1, 2. We can now investigate how the changes in the prices affect the return to factors. Differentiating the zero profit function we have: dpi=ail dw + aik dr. This implies that dpi/pi=wail/ci * dw/w + raik/ci * dr/r. This result follows from the fact that pi = ci . Define

and

as the cost shares of

each factor in the production of good i and the percentage changes in the prices of goods (dpi/pi), labour (dw/w) and capital (dr/r) as

,

and

respectively. Then the percentage changes in the

prices of goods and factors can be written as:

3

, i=1, 2. Writing out the equations for the two goods the changes in factor prices can be specified as: , where

=

. Assuming industry 1 is more

labour intensive than industry one so that its cost shares exceed that of industry and . Now supposing that the price of good 1 decreases, so that

then we

can solve for changes in factor prices. For a demonstration we only show the changes in the price of labour. , since

is negative. If industry 2 is

capital intensive, it also follows that there will be a more than proportionate fall in the return to capital if the price of good 2 falls. A plausible narrative that can be used to explain why tariffs may be imposed it to consider wages in the US garment manufacturing sector to those in Vietnamese sector. A movement towards more open trade will mean that there is a fall in the price of garments in the US while there is a rise in the prices of garments in the Vietnamese market. The more labour intensive garment industry in Vietnam therefore has a more than proportionate increase in its wages. In the US side, there is a more than proportionate fall in wages in the garment-manufacturing sector. But since garment manufacturing is a relatively more capital intensive industry in the US, the higher competition from imports also implies a fall in the profits of firms in the industry hence a fall in the return (price) of capital. The imposition of a tariff then ensures that prices are high enough in the domestic market so that the fall in the return to capital is less or does not occur at all. Alternatively, the imposition of tariffs means that the real change in the product prices for the Vietnamese firms is smaller [assuming they absorb most or the entire tariff] hence factor price equalization does not occur. This analysis suggests that the levels of tariffs imposed can explain part of the variation in factor prices.

4

3. Testing the Effects of Tariffs A plausible approach to testing factor price equalization has been suggested by Mokhtari and Rassekh (1989). We follow their approach with some modifications to suit the context of our analysis. Denote real exports, real imports and real gross domestic product in US Dollars of country i at time t by

,

and

respectively, where i = 1 … N and t= 1 … T. The degree of

openness for each country may be measured by

. This ratio measures the

trend over time of trade openness. Overall openness over time can then be computed as . A plot of TOPt can then be used to measure the extent of trade openness over a time, a crucial occurrence factor for the occurrence of factor price equalization. For the FPE theorem to hold, it is assumed that the trading countries share the same technology [i.e. identical production functions] and that their relative factor endowments are not too different. To determine the extent to which the assumptions hold, Mokhtari and Rassekh make the following computation:

where K and L denote real capital stock in US Dollars and

employment. The same will be done by us. By computing the mean and standard deviation across countries for each year, the coefficient of variation (CVKL) can be obtained. If CVKL is converging, then this implies that technologies are converging hence factor price equalization is possible. Furthermore, if CVKL is falling over time, then factor endowments are also converging. Wages can then be used as a measure of the price of labour, specifically average wages in a single sector like manufacturing (Ruffin, 1984). Similar to the generation of coefficient of variation of capital per worker, CVKL, the coefficient of variation of wages, CVW, can be generated. Following Mokhtari and Rassekh, a fall in the coefficient of variation of wages over time indicates wage convergence. FPE occurs if there is a negative significant correlation between CVW and TOP. The above variables can then be used to test the effect of tariffs on FPE. This can be typically done using a regression model. The model is specified as follows: ,

where ut is the residual, CVW, CVKL and TOP are as defined before while Tarr. is the overall level of tariffs and E is the employment level. E is included in the model to capture the larger variation in 5

wages that may be accompanied by the higher employment rates. The regression is run on logarithms in order to normalise the data. The equation is then estimated using ordinary least squares. We expect the parameters to take the following values: b1 > 0, b2 < 0, b3 > 0 and b4 > 0. A negative b2 implies FPE while a positive b3 implies that tariffs prevent the realisation of FPE. 4. Data Source The data for the analysis was obtained from the OECD Database. This database includes measures of capital (Gross capital stock volume; from the OECD STAN database for Industrial Analysis) and employment (total annual employment in ‘000), what gave the capital per worker ratio. Above that we have used data on real hourly average wages. In order to calculate the trade openess ratio we have used export (exports in goods value s.a., in billions of US dollars) and import (imports in goods value s.a., in billions of US dollars) data. GDP was exctrcted from the OECD database as well, at the annual level, in millions of constant 2000 Prices in a national currency. The tariff data has been obtained thanks to kindness of prof. Michałek and his data base, which containes average tarrif data based on unweighted averages for all goods in ad valorem rates, or applied rates, or MFN rates whichever data are available in a longer period for a given country. The sources of the data are: WTO, IDB CD ROM database and Trade Policy Review - Country Report, Various issues, 1990-2003; UNCTAD, Handbook of Trade Control Measures of Developing Countries - Supplement, 1987 and Directory of Import Regimes, 1994; World Bank, Trade Policy Reform in Developing Countries since 1985, WB Discussion Paper #267, 1994, The Uruguay Round: Statistics on Tariffs Concessions Given and Received, 1996; and World Development Indicators, 1998-2003; OECD, Indicators of Tariff and Non-Tariff Trade Barriers, 1996 and 2000; and IMF Global Monitoring Tariff Data. Due to the difference in the time since various measures for different countries available, we perform the analysis for the countries included in the OECD database, namely: Australia, Austria, Belgium, Canada, Czech Republik, Denmark, Finland, France, Germany, Greece, Hungary, Iceland, Ireland, Italy, Japan, Korea, Luxembourg, Mexico, The Netherlnds, New Zeland, Norway, Poland, Portugal, Slovak Republic, Spain, Sweden, Switzerland, Turkey, United Kingdom and the United States.

6

5.Results (from Stata) After inporting the data, we have reached up to 15 complete observations based on the aquired data. An estimation of the model presented above resulted in the output presented below: . edit (5 vars, 15 obs pasted into editor) . reg ln_cvw ln_cvkl ln_top ln_e ln_tarr Source | SS df MS Number of obs = 15 ------------- +------------------------------------------------ F( 4, 10) = 13.65 Model | 1.39654898 4 .349137244 Prob > F = 0.0005 Residual | .255747115 10 .025574712 R-squared = 0.8452 ------------- +----------------------------- Adj R-squared = 0.7833 Total | 1.65229609 14 .118021149 Root MSE = .15992 -------------------------------------------------------------------ln_cvw | Coef. Std. Err. t P>|t| [95% Conf. Interval] ------------- +-------------------------------------------------------------------------------------------ln_cvkl | -5.00037 9.191737 -0.54 0.598 -25.48084 15.4801 ln_top | .280721 .7423757 0.38 0.713 -1.373395 1.934837 ln_e | 6.089321 1.995975 3.05 0.012 1.64201 10.53663 ln_tarr | .0289453 .516853 0.06 0.956 -1.122675 1.180566 _cons | -71.66583 22.66849 -3.16 0.010 -122.1744 -21.15729 ---------------------------------------------------------------------------------------------------------------

Estimated equation: ln_cvw = -71.67 - 5.0 ln_cvkl + 0.28 ln_top + 6.09 ln_e + 0.03 ln_tarr.

Only one coefficient, that of employment is significant (t stat. 3.05) suggesting that the main employment levels can esplain variation in wages across countries. The R squared 0.85 suggests that the 85% of the variation in wages is explained by this model. Furthermore the high F-statitsic indicates the overall significance of the regression. The insignificance of the tarrif coefficient indicates that at least for the data set considered, the average tariff levels do not explain the variation in wages. It may be that taking wages sectorally, and similarily tariffs on goods and services in certain sectors, would bring more consistant results. The reasons for such an outcome of our study could be in the approach used, which does not include non-tarrif costs into the eqation. Such costs may include: transportation costs, transactions costs of enforcing trade etc. According to Anderson, Wincoop (2004):

7

A variety of ad hoc trade cost functions have been used to relate the unobservable cost to observable variables. Plausibly, cost falls with common language and customs, better information, better enforcement and so forth. Willingly, due to unsatisfactory results, we have estimated another equation, yet by introsucing the GDP per capita measure as a proxy for capital stock volume. The choice of ammending this specific variable was due to the largest gaps in the data as far as the capital stock variable was considered. The results, despite more observations in he final regressiin produced even less satisfactory results: . edit (7 vars, 25 obs pasted into editor) . reg ln_cvw ln_cvgdp ln_top ln_e ln_tar Source ------------- Model Residual ------------- Total

| SS df MS +----------------------------- | .041121131 4 .010280283 | .115509496 20 .005775475 +----------------------------- | .156630627 24 .006526276

Number of obs = 25 F( 4, 20) = 1.78 Prob > F = 0.1725 R-squared = 0.2625 Adj R-squared = 0.1150 Root MSE = .076

-----------------------------------------------------------------------------ln_cvw | Coef. Std. Err. t P>|t| [95% Conf. Interval] ------------- +---------------------------------------------------------------ln_cvgdp | 1.133291 .5234107 2.17 0.043 .0414751 2.225106 ln_top | .3016337 .1934313 1.56 0.135 -.1018569 .7051243 ln_e | -1.004763 .669202 -1.50 0.149 -2.400694 .3911675 ln_tar | .0795251 .1462053 0.54 0.593 -.2254538 .3845041 _cons | 13.46069 9.184143 1.47 0.158 -5.697099 32.61847 ------------------------------------------------------------------------------

Large standard errors suggest problem of multi-collinearity. There are wide confidence intervals and insignificant t-statistics of the coefficients of variables under consideration. Only coefficient of variation of GDP is significant, although the p-value of the whole estimation does not allow for significant conclusions. Results plotted ona graph are presented in the appendix. Using such techniques as ridge results was not successful either. The conclusion we derive from this lesson is that our model, despite the fact that we have derived it from trade literature, is insufficient in explaning the contemporary trade, tariff and factor proce equalization patterns.

8

6. Further observations The scatter plot for tariff levels and does confirm the effectiveness of the WTO vision of decreasing the overall level of tariffs in subsequent years:

Tarrifs 15.0 12.5 10.0 7.5 5.0 1989

1991

1993

1995

1997

1999

2001

2003

Supporting yearly data has been presented in the appendix.

The decrease in tarrifs would be as it is assumed by the WTO or other trade agreements. This finding can be also confirmed by depicting a larger data base, as follows: Tarrifs_High Income OECDs (10) NA NA NA NA NA NA NA Tarrifs_High Inc. Non-OECD (14) Tarrifs_Middle Income (86)

Tarrifs

Tarrif Level (%)

40 30 20 10

19 8 19 1 8 19 2 8 19 3 8 19 4 8 19 5 8 19 6 8 19 7 8 19 8 8 19 9 9 19 0 9 19 1 9 19 2 9 19 3 9 19 4 9 19 5 9 19 6 9 19 7 9 19 8 9 20 9 0 20 0 0 20 1 0 20 2 0 20 3 0 20 4 05 26

0

Time

The GATT was sign in 1947 and led to the decrease in trade restrictions. As of now, there are over 100 countries that participate in this agreement. The interpretatiion of the grpah presented above

9

we may say, the a decrease in tariffs most significant among the mid-income countries. Such policies may be resulting formthe strong prssure of he world economy to become more and more inclusive, the price of what is increased economic activity of emerging economies, due to decrease in average tariff levels. The coefficients of variation for other variables considered n the final estimated equasion look in the following way:

10.0

7.5

5.0

2.5 CVW CVKL TOP

0 1989

1991

1993

1995

1997

1999

2001

2003

Supporting yearly data has been presented in the appendix.

Following Mokhtari and Rassekh, a fall in the coefficient of variation of wages over time indicates wage convergence. Increasing variation in wages resulting form our research unfortunately does not support this notion. 7. Conclusions Trade costs are large when broadly defined to include all costs involved in getting a good from producer to final user. Both international trade costs and local distribution costs are very large and together dominate the marginal cost of production. Trade costs also vary widely across countries. On average developing countries have significantly larger average trade costs. In specific sectors or infiant industries of one of the trade partners, the barriers might be even higher, imposeing greater costs.

10

The “missing globalization puzzle” defined by Coe et al.(2002) might be the reason why not incorporating distance into equaions measuring tade cost effects may be a significant drawback and result in a crucial impact on the data. Yet current litaerature does not give a clear answer to the question, which factors should we prioriztize in measuring the effect of barriers to trade on factor prices. The factor price equalization theorem implies that workers in different countriesat the end of the day will earn the same wages, even if international labor migration does not take place. The commodity trade substitutes factor mobility in this case. Moreover the theory implies, that economies, where labor is the abundant factor do not have to encourage migration (export people) inorder to raise wages due to reduction in the supply of labor. It seems, according to theory, that liberalizaing trade, especially in the domain of the labor intensive products would increase the wages in that secotr and be beneficial for the standard of living. In reality though it seems that factor prices have not equalizes. In this paper we have made an attempt to show that tarrifs have an impact of factor price equalization by artificially seting the payoffs in the game. Unfortunately the insignificance of the coefficient of the “tariff” variable does not make it legitimate for us to derive any conclusions or real policy results. However other literature does seem to show a tendancy towards factor price equalization under the circumstances of trade liberalization.

11

Appendix Graph 1. Year 1989 1990 1991 1992 1993 1994 1995 1996 1997 1998 1999 2000 2001 2002 2003

CVW 2.330124 2.306121 2.387245 2.502619 2.943433 3.436899 4.003884 4.451429 4.828325 5.05179 5.175281 5.251665 5.280408 5.307305 5.317825

CVKL 3.53834 3.53968 3.62274 3.61745 3.61319 3.61945 3.62694 3.63081 3.62669 3.63142 3.63534 3.63560 3.63779 3.63979 3.63493

TOP 0.508954 0.522711 0.5014 0.509133 0.435821 0.460543 0.527033 0.522855 0.496591 0.510269 0.502359 0.473904 0.452294 0.455257 0.52223

Graph 2. Year 1989 1990 1991 1992 1993 1994 1995 1996 1997 1998 1999 2000 2001 2002 2003

Tarrifs 12.22381 11.36303 10.8141 8.464444 8.579321 7.977778 8.28567 9.239625 8.022463 7.883229 8.505597 7.282949 7.006635 6.911966 5.953949

12

References Anderson J. E., Wincoop E., Trade Costs, NBER Working Paper 10480 , 2004. Bernard, A., Redding, S., Schott, P. and Simpson, H. (2002). Factor Price Equalisation in the UK? NBER Working Paper 9052, July 2002. Cunat, A. (2000). Can International Trade Equalize Factor Prices? IGIER, April 2000. Feenstra, R. (2004). Advanced international Trade. Mokhtari, M. and Rassekh, F. (1989). The Tendency towards Factor Price Equalization among OECD Countries. The Review of Economics and Statistics, Vol. 71, No. 4. (Nov., 1989), pp. 636-642. Samuelson, P. (1971). Ohlin Was Right. Swedish Journal of Economics, December 1971, 73, 365-84. Trefler, D. (1993). International Factor Price Differences: Leontief was Right! The Journal of Political Economy,Vol. 101, No. 6. (Dec., 1993), pp. 961-987.

13