Foreign trade and its relation to competition and productivity in Turkish industry

policy. Rteewet and Euteen AS**e & &~ ( -WOH J(4Tn ftf'tHS i ~~Trade Polic Country Economics Department The World Bank February 1991 WPS 604 Foreign Trade and Its Relation to Competition and Productivity in Turkish Industry Faezeh Foroutan Trade liberalization has improved productivity in the industrial private sector - especially tradables - but not in the public enterprises. Improved productivity of the latter has come from other sources. The Poicy, Rummk *n Extaa Affain Com,pkx distuibutes PRE Woaking Papers to diserninstathe tudinp of woik in progrep and to aemnunge Om exchsnp of idea among Bank saff and aU Ohmn intrsed in development isrle. Th papws carry the names of the audhon, rdef only their views, and thould be used arn cited accordingly. The findings. iterpveutiom, and conclusions are the authort ownL They shauld not be asuibued to the World Dank, its BoArd of Dinecor, it management, or any of its nionber countia.  Trade Policy WPS 604 This paper- a product of the Trade Policy Division, Country Economics Department- is part of a larger effort in PRE to study the effect of trade liberalization on the performance and conduct of the industrial sector. Copies are available free from the World B*ik, 1818 H Street NWV, Washington, DC 20433. Please contact Sheila Fallon, room NIO-041, extension 37942 (58 pages). Trade liberalization and more exposure t private sector. In the public sector, deeper trade intemational competition generally benei,ted penetration seems to have lowered the price-cost Turkish industry in the 1980s. But the effect of margin in the public enterprises that were above- intemational competition appears to have been average ih capital intensity-, but had no impact on felt mainly in the private sector- especially in productivity. tradable industries. Improved productivity in the public enter- In the first half of the 1980s, intemational prises appears to be more related to changes in competition decreased the price-cost margin and other areas - probably the reform of manage- increased the growth rate of productivity in the ment. The PRE Working Paper Series disseminates the findings of work under way in the Bank's Policy, Research, and External AffairsComplcx. An objective of the series is to get these rmdings outquickly, even if presentations are less than fully polished. The findings, interpretations, and conclusions in these papers do not necessarily represent official Bank policy. Produced by the PRE Dissemination Center  TABLE OF CONTENTS Page No. I. Introduction 1 II. Background on Turkey 4 a. Import Liberalization 5 b. Export promotion 8 III. Industry-Level Analysis 10 a. Data Preparation 10 b. Analysis of price-cost margins 15 c. Analysis of Productivity 23 d. Comparison With Earlier Findings 33 e. Conclusions 36 Tables (1-28) 37 Bibliography 57 I wish to thank James Tybout, Jaime de Melo, Arvind Panaga:iya for helpful comments and Kaniz Siddique for assistance with the data.  1 I. INTRODUCTION This paper, which is part of a larger research projectl examines how, if at all, the reform of the trade regime in Turkey during the 19809 has affected the performance and competitiveness of the Turkish industrial sector. Of course, reform of the foreign trade regime has coincided with reforms in other areas of the economy, namely the tax system, the state economic enterprises, and the financial markets. Therefore, it is possible that some of the results one attributes to changes in the foreign trade regime of Turkey are indeed a fruit of trade reform as well as reforms in these other areas of the economy. However, to the extent possible, an attempt shall be made here to isolate the effects of changes in the trade regime from other changes in the economy. It is often argued that a more liberal trade regime is desirable because other than static gains deriving from allocative efficiency, a more liberal trade regime gives rise to gains stemming form price efficiency and/or productive efficiency. Price efficiency is addressed in the literature on the relation between trade and market structure.2 The essence of the argument is that l.See James Tybout, December 1987. 2.See for example Caves (1985), Krugman (1986) and Dixit and Norman (1980, c;ap. 9'  2 in domestic markets characterized by entry barriers, exposure to international competitian reduces the market power of domestic producers. This affects both their production and pricing decisions leading to lower mark-ups and higher output levels. Trade reduces monopolistic distortions. Productive efficiency can be thcught of as arising from scale efficiency and technical efficiency.3 There are several reasons why a less protectionist trade regime increases s-.le efficiency. First, because trade enlarges the domestic market which might be too small for the efficient production of goods that show increasing returns to scale. Second, because in domestic markets characterized by excess profits4 due to protection and free entry, trade reduces excess profits and discourages entry by small inefficient firms.5 A more liberal trade regime is also thought to contribute to greater technical efficiency for a variety of reasons. First, greater competition from abroad forces domestic firms to adopt newer, more efficient technology that reduces x-inefficiency and waste. Second, in the case of developing countries, freer trade eases the constraint imposed by the availability of foreign exchange and hence enables a country to import foreign capital goods that embody a more advanced technology than domestic capital goods. Third, a more open economy 3.Corbo et al. (1988). 4.Excess profits disappear of course once entry occurs. 5.Ibid, and the bibliography quoted there.  3 results in faster absorption of technological progress.6 Despite the common sense appea3. of these arguments, their theoretical basis is not very robust. Indeed under the right set of assumptions some of these results can be reversed.7 It is therefor,; ultimately an empirical issue to establish whether trade liberalization leads to greater productive efficiency and lesser monopolistic distortions or not.8 The present study tests the empirical relevance of these arguments to Turkey's experience of trade liberalization of the 1980s. .~Ating in 1980-81, the Turkish government introduced a series of major foreign trade liberalization measures that significantly increased wile share of import and export trade in Turkish industrial output and the degree of exposure of domestic firms to international competition. The remainder of the paper is divided into two parts. Part iI provides a brief background on trade and other economic reforms in Turkey in the 1980s. Part III provides an analysis of industry-level relationships between trade liberalization, firms' competitive behavior as measuLed by changes in price-cost margins, and efficiency gains as 6.See Nishimizu and Robinson (1984), Nishimizu and Page (1987), Romer (1989) and Edwards (1989). 7.See for example Rodrik (1988) and Bhagwati (1988). 8.For a recent survey of empirical literature linking the growth in total factor productivity to changes in the trade regime, see Havrylyshyn (1989). For a recent survey of empirical literature linking economic growth to trade policy see Edwards (1989).  4 measured by changes J total factar productivity. Due to the lack of aDpropriate data, the same analysis at plant level as well as the analysis of the patterns of entry and exit into and out of each industry is postponed until such data become available. However, unlike other countries which are part of this research project, Turkish data differentiate public from private enterprises. Public enterprises in Turkey account on average for a quarter of industrial output although their share varies greatly from one industry to another (see Tables 2-5). The availability of separate data for priv.ate and public sectors highlights important differences that have historically existed between public and private firms and permits a test of whether the reforms of the 1980s have narrowed these differences over time. II. BACKGROUND ON TURKEY9 Beginning in 1980, after two years of economic recession that followed the foreign debt crisis of 1977, Tur-;y adopted a stabilization program that represented a radical break with her traditional inward- loo:cing development strategy. Liberalization of foreign trade and payments was a key feature of the stabilization program. The trade liberalization measures introduced in the 1980s aimed at shifting the economy towards an export-led growth by dismantling the set of complex and highly restrictive rules and regulations that had in 9.This part relies heavily on the following sources: Baysan and Blitzer (1985), Foroutan (1987), Milanovic (1986), and the World Bank Report No. 5365-TU (1984)  5 the past regulated Turkey's transactions with the outside world. These measures comprised four important steps: -a gradual shift from non-tariff barriers to%. ds tariffs or tariff type restrictions; -a gradual reduction in the height and the variability of trade taxes, at least until 1985, when the surge in the so-called levies may have reversed the trend; -deliberate promotion of exports; and -devaluation of the real exchange rate. II.a Import liberalization The principal thrust of import liberalization was carried out in two separate steps, namely in 1981 and December 1983-January 1984. The main aspects of import liberalization in 1981 were the abolition of quotas and the increase in the number of goods which could be legally imported. ,n order to appreciate fully the ra of reforms introduced in the 1981 import regime, it is necessary to recall that imports were classified according to three lists: -Liberalized List I (LLI), consisting of all goods that could be freely imported; -Liberalized List II (LLII), comprising all goods whose importation required a license; -Quota list (QL),. comprising all items whose imports were subject to a quota. The importation of a good that did not appear in any of the above mentioned lists was prohibited altogether. According to one set of  6 estimates 10 imports in the LLI, the LLII and the QL amounted to 17, 70 and 12 percent of the to-al value of imports in 1980 respectively. In the 1981 import regim3, two sets of reforms were introduced. Firstly, the QL was abolished. Approximately one third of items in the QL were transferred to LLI and the remaining quota items were transferred to LLII. Secondly, approximately 200 tariff positions, equivalent to 3.8 percent of the value of imports in 1980, which belonged to the license list were transferred to the LLI. The reform ot the import lists was accompanied by a number of other administrative reforms and the lowering of other taxes on imports such as the stamp duty from 25 to one percent, and guarantee deposits on imports from 25-1O percent to 7.5-15 percent. The new 1981 import regime remained in full force without major changes until December 1983, when a new set of far reachi.ng liberalization measures were adopted by the Government. The reform of the import regime enacted in December 1983-January 1984, represented a major break with the past. The measures introduced in this period affected three aspects of import policy, namely the redefinition of the import lists; t,riff vs. non-tariff barriers; and the height of import barriers. Import lists. The most significant feature of the 1984 import regime consisted of the transition from a "positive list" to a "negative list". Under the new regime, all commodities that were not explicitly prohibited could be imported. This was in sharp contrast to the previous lO.See World Bank Report 5365-TU of August 1984.  7 regime which allowed the importation only of the those commodities that were explicitly mentioned in the Free or License Lists. Thus, now two import lists, the Prohibited List, and the License List, replaced the previous LLI and LLII. The Prohibited list initially included some 219 tariff positions consisting mostly of consumer goods. By May 1985, the Banned list was for all practical purposes abolisihd and the goods on the list were transfer,.ed either to the License List or became freely importable. Similarly, The License List initially comprised 369 items which accounted for 28 percent of 1984 imports. By 1987, the number of goods on the List was reduced to 111 or about 18 percent of total imports. In 1988, the list was further cut to 33 items. However, the 1983/84 import regime introduced a new list, the so-called 'levy" or "Fund' list. Goods on this list pay a specific dollar denominated surcharge in addition to custom duties and other trade taxes. The levy proceeds are channeled to the so-called extra- budgetary funds. The levies were initially supposed to serve two purposes. First, they were to finance social projects, such as mass housing, by taxing the importation of luxury goods. Second, the levies were to provide temporary protection to domest!c industries that produced goods competing with those imp. .ts that were no longer subject to QRs. However, the scope of levies has since expanded far beyond the stated original intentions. In 1984, levies were applied to 200 goods and the implicit average tariff equivalent rate of levies stood at 2 percent. In 1987, more than 570 items were subject to levy with a tariff equivalent rate of 6.1 percent on average. Simultaneously with the steps just mentioned, the 1983/84 reforms resulted in a large scale raticnalization of tariff rates. The  overall import weighted average tariff rate for goods for which the rates were modified stood at 22.7 percent in 1984, compared to 38.8 percent prior to December 1983. With similar gradual reductions in the tariff rates in subsequent years, the average statutory tariff rate in Turkey is reasonably low today. The overall import v '.ghted average tariff rate in 1987 stood at only 9.5 percent. The actual rate (defined as tariff revenues as a proportion of total imports) was even lower due to widespread exemptions granted to one or the other categories of users. II.b. Export promotion Export promotion constituted another focal point of the 1980 Government program. Export promotion was accomplished through three types of measures: (i) maintenance of a competitive real exchange rate: (ii) provision of direct subsidies; and (iii) simplification of the administrative and bureaucratic procedures. The Government devalued the Turkish lira by more than 50 percent in nominal terms in 1980. At the same time, with few exceptions regarding agricultural inputs, it abrogated the multiple exchange rate system that had existed until then. Moreover, since 1980, the Government has adopted a policy of flexible exchange te. Specifically, since 1981 daily adjustments have been made to prevent the appreciation of the lira. The second type of assistance by the Government to the exporters has come through direct incentives. These consisted of:  g (a) Fxport tax rebate. The rebate was initially to compensate exporters for indirect taxes on their inputs, but included a substantial subsidy element to the extent that the refund exceeded the actual taxes paid. After January 1985, with the introduction of the VAT and given that exports are zero rated, all the rebate is a pure export subsidy. In accordance with the the Government's declared objective of gradual phasing out of direct export incentives, the subsidy conferred by the export tax rebate scheme declined steadily during the period under examination. (b) Duty drawbacks on imported inputs. Although duty free imports for exports do not constitute a subsidy in the GATT definition, they represent an important incentive to exportin3 as opposed to selling in the domestic market. The incentive content of duty drawbacks varies with the ir.port content of exports in each industry. On average, during the 1980s it reached 5 percent of the value of exports for the ianufacturing industry. (c) Access to preferential credit. Credit to exporters at preferential rates of interest constituted an important incentive to exports during the early eighties when there existed a substantial difference between the general short term rate of interest and the rate of interest applied to export credits. However, the importance of this scheme dwindled gradually until it was finally abrogated in 1985. Effective from January 1987, a new export credit regime was instituted. Although no precise estimates on the subsidy content of the new scheme are available, evidence suggests that they were not very important. 11 ll.See Turkey, Country Economric Memorandum of August 5, 1988, Report no. 7378-TU.  10 (d) Cash grant financed by one or more of the so-called Extra Budgetary Funds (EBFs). In January 1985, the Government established the Resource Utilization and Support Fund (RUSF) with objective of encouraging exports by granting a 4 percent flat subsidy to all exporters. The cash grant was subsequently reduced to 2 percent and finally discontinued in November of the same year. However, starting in January 1987, the Government reintroduced the EBF subsidy to exports, this time from the Resource Utilization and Support Fund (RUSF). The RUSF subsidy, unlike its predecessor, is granted only to selected products in the form of specific dollar amounts. What has been the outcome of these liberalization/promotion measures? As the discussion in the following pages will show, the outcome has been a substantial increase in the degree of openness of the manufacturing sector in Turkey. In the following sections an attempt will be made to determine to what extent the greater openness of the economy and the increased exposure to foreign competition have affected the performance and conduct of the Turkish industry. III. INDUSTRY-LEVEL ANALYSIS III.a Data Preparation Industry-level data were obtained from Turkey's State  11 Institute of Statistics (SIS)12 and contained observations for the variables described below at the three-digit International Standard Industry Classification (ISIC), distinguished by ownership, namely private versus public firms. The variables are: -labor input defined as total number of production persons. The entry does not distinguish various types of workers, such as by sex, age etc. Also no information is available on the number of working days lost due to strikes or other factors. -total labor cost which comprises wages and other payments made by the employer such as overtime payment, bonuses, employers contribution to retirements funds, etc. -total value of intermediate inputs at current prices. -total value of output at current prices. -beginning-of-year and end-of-year value of inventories of final output and intermediate inputs. -sectoral output price deflators. The original data were subsequently used to derive the value of a number of additional variables necessary for computations. The latter included the stock of capital and industry specific input price deflators. 12.From 1973 to 1982, the data cover all manufacturing firms with 10 or more employees. From 1983 onwards, the data cover only firms with 25 employees or more. The change in coverage is not relevant for some industries with few large firms, but is important for others that comprise a large number of small, family concerns.  12 Capital stock series were constructed by using the perpetual inventory method. A complete description of the sources of data and the methodology is offered in K. Siddique (1989).13 Sector specific input price deflators were computed by using input-output coefficients and output price deflators (For details of the methodology and data requirement see Siddique, op. cit). Data on total value of imports and exports also classified according to ISIC were obtained from the UN Geneva data base. Foreign trade data, expressed originally in US dollars were converted into Turkish lira by using mid-year average exchange rates. Trade figures in local currency wer. then used to derive import penetration and export share series. The next step involved the correction of the data for inflation distortion. This correction was necessary to make stock and flow variables comparable within a year by expressing all stocks and flows in mid-year current prices. Once the data were corrected for the inflation bias, they were expressed in real terms by deflating outputs and inputs by industry-specific output and input price deflators. Most of the data series cover the period 1973-1985, although in the process of data transformation a number of years are lost. As a result our regression analyses in the following sections cover the years 1976 to 1985. The list of industries in our sample is contained in Table 1. In order to study more closely the effect of trade liberalization on 13.See "Data Requirement and Data Availability", chap. 4 of her PhD dissertation.  13 industry in Turkey, the industries were divided into three subgroups: exportables, importables, and non-tradables. In accordance with the methodology followed for other countries in this project, exportable industries were defined as all three-digit industries that showed an increase in exports during the sample period and whose average exports in 1984-85 was at least equal to 25 percent of their output. Industries 321, 322, 331, 332, 362, and 385 met this criterion. Among the remaining industries, importable industries were defined as those industries for which the share of imports in total domestic sales, defined as output minus exports plus imports, in 1984-85 exceeded 25 percent. Industries 351, 372, 382, and 384 fell in this category. All the remaining Lndustries were ranked as non-tradables. The justification for choosing 1984-85 is that those represent the years most favorable to both export promotion and import liberalization in the entire sample period. Because the data refer to only three-digit industry disaggregation, some Lndustries like 371 and 385 show both a high degree of import penetration and export share. However, because an industry is only classified in one category and the exportable industries were chosen first, both of the above were included among exportables. The above remarks make it clear that the classification chosen here, as any other classification that might have been adopted, is only approximate. Tables 6-9 report the average export share and import share for the three group of industries and for all industries pulled together  14 over the period 1976-85.14 The data show that export and import shares increase dramatically in all the three industry grouping, but more so in the tradable industries. The share of exports in the output of exportable industries increases from less than 4 percent in the 1970s to over 42 percent in 1985. Similarly, the share of imports in total domestic sales of the importable industries increases from its lowest level in 1979, 21 percent, to over 40 percent irh 1985.15 14.In the actual computation of the shares as well in subsequent regression analysis, industry 322 was omitted because of inconsistent trade figures. More precisely, during 1980-85, the value of exports exceeded the value of total output by as much as 100 percent towards the end of the sample period. This outcome may be explained by at least two factors: a) Trade data include all firms whereas output data exclude firms with less than 10 employees until 1982 and with less than 25 employees thereafter. Because in textile and clothing industry small establishments are common, output figures which exclude such establishments underestimate the true value of industry's output. b) The policy of subsidizing exports is well known to have induced a certain degree of overinvoicing of exports, especially in the textile and clothing industry. 15.Due to differing coverage of firms of trade and output data, export and import shares may be slightly exaggerated, but the trend is reliable. See footnote 14.  15 III.b Analysis of price-cost margins This section analyzes the relationship between the Turkish industry's price-cost margins and other aspects of her industrial structure for both the private and the public sectors. The price-cost margin is widely used as a measure of gross profitability in an industry. It expresses the gross return to fixed inputs, here assumed to be capital, as a share of the value of output. As a result, variations in the price-cost margin for an industry reflect both variations in net industry profitability and the average capital intensity of the industry in question. It is easy to show that under constant returns to scale, the price-cost margin for a firm equals the mark-up of price over long-run average (and marginal) cost plus the net return to capital and depreciation. If all industries are perfectly competitive (price mark-up is zero) and if the rate of return to capital (r) and the rate of depreciation of capital (d) is the same for all firms in all industries, then price-cost margin for industry i equals (r+d)(Ki/Yi), where K is the total stock of capital and Y is the total value of output in industry i (see Schmalensee 1989). Thus, according to conventional theory a high price-cost margin, or equivalently a high rate of accounting profit in an industry is an indication of market power by at least some firms in that industry. If this is so, an increase in competitive pressure, say through import liberalization, should lower the industry's price-cost margin. However, there is substantial controversy in the literature as to whether high profits Rare to be interpreted as a sign of good or bad performance" (Bresnahan 1989, p.1013). An unconventional interpretation of high profits is that they  J6 are a sign of good performance because they indicate lower costs.16 The truth is probably somewhere in the middle. In the case of Turkish indastry for example, the combination of a capital-output ratio in the public sector which is two to three times higher than that in the private sector and considerably lower or even negative margins in the public sector is probably a good indication of poor performance by public enterprises rather than exploitation of market power by private firms, even though some firms in some industries might enjoy some market power. Here the interest is focused on how changes in the trade regime in Turkey and exposure to international compotition have affected the price-cost margins in various industries. The analysis is done at the three-digit industry level separately for private and public sectors for the period 1976-85. On the whole, there are 22 industries for the public sector and 24 industries for the private sector.17 The basic model to be tested is:18 16. Schmalensee (1985) proposes a test for this alternative interpretation of high profits when working with disaggregated firm- level data. Here, because we are dealing with sectoral data his methodology is not applicable. 17.Industry 314 (tobacco processing) was also excluded from the regression analysis due to lack of complete data for the sample period. 18.For similar models see Journal of Industrial Economics, 1980, Vol. XXIX no.2 (December) dedicated to a symposium on international trade and industrial organization.  17 (1) PCMit - f(IMPit, KORit, Di, Dt) PCMit - Price cost margin for industry i in year t. It is constructed as the value of output of each industry corrected for changes in final output inventories minus the value of intermediate inputs also corrected for changes in input inventories and total labor cost divided by the value of output. IMPit - A measure of trade penetration defined as the share of imports in total apparent consumption for each industry and each year. Because greater competition from abroad is expected to reduce the monopoly power of domestic firms, it is generally expected that this variable has a negative sign.19 KORit = Capital-output ratio for industry i in year t. It is constructed as the end-of-year capital stock of industry i divided by the real value of output of the same industry. Because price-cost margins are affected by industries' capital intensity, it is necessary to include this variable in the regression analysis. Ceteris paribus, one would expect this variable to be positively associated with the margin. 19.It may be argued that because in equation 1 the denominator of variables PCM and KOR is output, import penetration should also be defined as the ration between imports and output. The regression results change very little if this other definition of import penetration is adopted.  18 D, - A set of dummy variables used to control for any omitted industry- specific effects that are likely to be present in every y' r. The number of industry dummy variables differs for public and private sectors. It is equal to the number of industries in which private or public firms are active less one. Dt - A set af nine dummy variables to control for any year-specific effects which are likely to affect all industries. In the analysis of price-cost margins, it is customary to include among the explanatoLy variables an indicator of industry concentration ratio, such as Herfindahl index, on the assumption that firms in more concentrated sectors have greater market power and hence enjoy larger margins.20 This variable is missing in the regression equations estimated here because industry concentration ratios were unavailable. However, because of the presence of industry dummy variables, the exclusion of the concentration ratio variable should be of a lesser loss than if these dummy variables were not included. Overview. Before examining the regression results, it is interesting to look at average capital-output ratios and price-cost margins for the private and public sector. The average figures for all industries over the 1976-85 decade appear in Table 10. It immediately appears that on average the capital-output ratio is much higher in the public sector than in the private sector while public sector's price- 20.For an alternative interpretation of positive correletirn between concentration and PCM see Schmalensee (1985)  19 cost margin is below the private sector's. These results conform to given a priori knowledge of the Turkish industrial structure. For example, it is well known that state enterprises have historically been engaged in the most capital-intensive industries and the pursuit of profitability has not been one of their strengths. The figures in Table 3 also show that during the years of slow economic growth, namely 1978-80, the average capital-output ratio increases in both the private and the public sectors while the price-cost margin decreases. In the subsequent years this trend is reversed. More interestingly, there appears to be a narrowing of the gap between private and public firms due to a gradual but steady cecline in the public sector's capital- output ratio. Comparing the three industry grouping, several interesting results emerge from the data in Tables 11-13. First, although for all the three groups of industries public firms' capital-output ratio is greater than the privat firms', the difference is more pronounced for the non-tradable industries. Second, for both the private sector and the public sector, the average capital-output ratio for non-tradables is higher than for tradables, and within the latter exportables show a higher capital- output ratio than importables. The latter result is of course no indication that Turkish exports are relatively more capital intensive than Turkish imports, because the correct measure of relative factor intensity is the average capital labor ratio and not capital output ratio. The former ratios were computed for export and import competing industries and are also reported in the last two columns of Tables 11 and 12. The data show clearly that Turkish importables are more capital  20 intensive than Turkish exportables irrespective of whether public or private enterprises are taken into account. 21 Third, compatible with capital-output ratios, for both the private sector and the public sector the price-cost margin is higher in the non-tradable industries than in the tradable industries. However, within the tradable industries, private firms do not show any appreciable difference between the price-cost margin between importables and exportables. For the public sector on the other hand, the price-cost margin is systematically higher in the exportable industries than in the importable industries. This result, which runs contrary to the observed pattern of capital-output ratio in the two groups of industries, indicates that private firms are more or less equally profitable in the two sectors, whereas public firms in import competing industries are less efficient and less profitable than public enterprises in the export sectors. In fact, the difference between private and public firms in terms of both average capital output ratio and average price-cost margin is the least in the exportable industries. Public firms also seem to perform better in the non-tradable industries than in the importable industries. This result may be attributed to the fact that public firms in import competing industries have historically had a high degree of protection, even after the liberalization of the 1980s, which has allowed them to continue to operate at very low level of efficiency and 21.Notice that in the presence of intermediate inputs, the correct measure of Capital-labor ratio is the direct capital-labor ratio for gross output and the total (direct plus indirect) capital-labor ratio for net output. See Derr (1979).  21 profitability. Higher margins for private firms in non-tradables than in the tradablos seem to be mostly due to differing capital-output ratios in the two groups of industries. Regression Results. The regression analysis of price-cost margins was done by pooling all industries, but maintaining the distinction between private and public sectors. The results are reported in Tables 14 and 15. Four models are specified for both the private sector and the public sector. In all four models the variable for trade penetration is the share of imports in total domestic consumption. Model 1 and 2 differ in in that model 2 includes one additional variable, KOR.IMP, which allows for the effect of import penetration to vary with the degree of capital intensity of an industry. If this variable is significant, it implies that import penetration affects relatively more the capital intensive industries. Models 3 and 4 are exact replicas of models 1 and 2 with the exception of the exclusion of the industry dummy variables. The first columns of Tables 14 and 15 report the results of Model 1. For the private sector, both import penetration and capital output ratio have the expected sign although the latter is not significant. For the public sector, import penetration has the expected sign but is not significant. The capital cutput ratio has a negative sign but is also insignificant. The results are not greatly modified for the private sector in model 2. For the public sector, however, variable KOR.IMP is significant and negative, indicating that import penetration has reduced the margins in the public sector in the most capital intensive industries.  22 A glance at the regression results of models 3 and 4 for both sectors reveals that most of the explanatory power of the basic model derives from industry dummy variables. When these are excluded the adjusted R2 for both the private and the public sectors is greatly reduced whilp generally the level of significance of the remaining variables increases. When industry dummy variables are excluded, import penetration variable for the private sector becomes insignificant while the variable KOR.IMP in model 4 becomes significant and negative, indicating that import penetration has indeed exerted a more powerful impact on those industries that were more capital intensive. For the public sector, on the other hand, it is now the import penetration variable that becomes significant. Variable KOR continues to have a negative sign. Summarizing, it appears that in Turkey greater exposure to international trade has exerted some effect on market power of firms both in the private ano the public sectors. For the private sector, when industry specific effects are allowed for, greater impcrt penetration appears to affect all industries in the same way; when industry specific effects are excluded, greater import penetration appears to affect the margins only in the capital-intensive industries. In both case! the impact of trade penetration on the margins is quite small indicating most likely that Turkish private manufacturing industry did not enjoy a significant market power even prior to the 1980s reforms22 . In the public sector, when industry dummy variables are included, import discipline appears to have affected relatively more the capital intensive industries. Vice versa when these variables are excluded, it is the import penetration variable rather than the interactive term to 22.See Foroutan (1990) forthcoming.  23 become significant. Also in the public sector the coefficient of the iriport penetration variable is relatively small, although higher than in the private sector. The negative sign for the variable capital-output ratio in the public sector conforms to the earlier finding that public enterprises are more concentrated in capital intensive, non-profitable industries. Finally, the significance of industry dummy variables and time dummy variables is a clear indication that studies that do not include these variables suffer from a serious shortcoming.23 III.c Analysis of Productivity The relation between total factor productivity growth in Turkish industry and its trade policy during the 1976-85 period is examined in this section. Total factor productivity growth is a measure of industry performance that proxies the efficiency with which factors are used. TFP growth is computed as the rate of growth in real gross output minus a weighted average of the rates of growth of individual inputs where the weights are the average share of each input in total value of output over the period t-1 to t, namely: (2) TFP.= Q. - it Li + SK. K.t+ SM i it Tit it it 'it it it ) Q is the the real gross output, SL, SK, and SM refer to the 23.See for example the references cited in the previous note.  24 average share of labor, capital, and raw naterials in total value of output respectively and are defined as: (3) SLt 1/2 {(PktLit /p it +(ti1iti it-i-)) (4) SKiT 1(2 {(P tKiK/PiQit) + (P K K,tu/PitQit) ( ) it / {( it it/ itQt (it_1 Ltl lit-1 it-1) (5) Hir1/2(P P+(H H /P I itMit/it (t /P t1it1iPit-Qit-1) A ^ over a variable indicates the growth rate of the relevant variable from period t-l to period t. Subscript i refers to industry and superscript L, K, and M refers to labor, capital and intermediate inputs, respectively. Pi indicates the price of output in industry i. Overview. In the first stage of the analysis, the observation period is divided into two subperiods. The first subperiod covers the five- year term up to 1980 and is characterized b.y the crisis of the late 1970s including recession, high inflation and an inward looking trade regime. The second subperiod from 1981 to 1985 includes the period of radical reform in economic policy marked by a signifir:.nic liberalization of foreign trade, especially in the form of removal of QRs on imports and substantial promotion of exports. The summary results appear in Table 17. During 1976-80, for both private and public enterprises, the average TFP growth was negative. However, the decline in productivity in the public sector (-7.5 percent per year) was stronger than the fall in productivity in the private sector, (-4.1 percent per year). During the second subperiod from 1981 to 1985, both these trends are reversed. The growth in T?P in both private and public enterprises turns positive, but  25 the TFP growth of public firms surpasses tha. of private enterprises (5.7 percent in the former and 3.4 percent in the latter). Looking at individual results (Table 16), among the 26 industries listed, only 4 registered positive TFP growth within the private sector and 3 within the public sector during 1976-80. On the other hand, in the next subperiod, only five industries within the private sector and five industries within the public sector show negative TFP growth. In the second stage of the analysis, the observation period was divided into three subperiods, namely 1976-78, 1979-80, and 1981-85, in order to isolate the effect of the debt crisis years which coincide with the second subperiod. The debt crisis of 1979-80 led to a substantial fall in output, a three-digit rate of inflation, and a considerable fall in the value of foreign trade. Other than TFP growth rates for private and public industry in the three aforementioned subperiods, the rate of growth for output, labor, capital and intermediate inputs were also computed. This was done for all the industries as a whole as well as for the three industry groupings, namely exportables, importables and non-tradables. The purpose of calculating growth rates for the latter variables was to see the relation of each individual variable with the the TFP growth pattern. As expected, output growth rate is strongly correlated with the TFP growth for both public and private firms (see Table 18). Summary results of the computations appear in Tables 19-22. Considering all the industries, during the first subperiod, 1976-78, productivity growth is modestly positive in the private sector and modestly negative in public sector. In the second subperiod, 1979-80, the years of economic crisis, productivity growth is strongly negative, as is the output growth. Despite the decline in output in this period, employment and the  26 stock of capital continue to rise both in the private and the public sectors, indicating the difficulty in adjusting input to output growth in the short to medium run. Finally, during the thi d subperiod, 1981-85, productivity growth is positive and higher in the public than in the private sector. The higher productivity growth in the public sector Is due to a high increase in output accompanied by a very modest increase in inputs. It appears that the economic reforms of the 1980s, especially the reform of state economic enterprises including their increased autonomy in terms of resource management, is responsible for the improved productivity performance of the public firms. Comparing the results for the three industry groupings, again one can detect a number of interesting results. First, for the private sector, the rate of growth of total factor productivity during the 1980s following the liberalization of foreign trade, is much higher for exportable industries (5Z) and importable industries (4.82) than for non-tradables (1.9Z). This result can be interpreted as a clear indication that exposure to international competition has had a positive impact on productivity growth of private enterprises. In fact, the contribution of TFP growth to output growth for the private sector during the period 1981-85 in exportable industries, importable industries, and non-tradable industries is respectively equal to 262, 302, and 172. The same pattern is not true for the public sector. The average productivity growth of public enterprises during the 1980s is higher in non-tradable sectors (6.52) than in the tradable sectors (4.32 in exportable industries and 4.72 in importable industries). Also, in relative terms, TFP growth for public enterprises contributes 922 to output growth in non-tradable industries higher than 30? in importable industries, but lower than 1592 for  27 exportable industries. These results again indicate that the reform of public enterprises in Turkey during the 1980s must have had a positive effect on productivity at least as important as exposure to foreign competition. Second, despite what one may expect, trade liberalization does not appear to have had a negative impact on labor employment. During the 1980s, employment in the private sector as a whole increases by 5.12 and decreases very slightly in the public sector (-0.7Z). However, because the share of public enterprises in total industrial output is only around 25 percent (see Tables 2-5), this means an overall increase in the level of labor employment in industry. More interestingly, employment in the private sector not only increases in exportable industries, it does so also in import competing and non-tradable industries. In the public sector employment during the 1980s increases only in the importable industries, but it declines in the exportable and non-tradable industries. The trend in employment in public enterprises most likely reflect the state of overmanning prior to the 1980s reform. The overall trend in employment in the Turkish industrial sector appears to confirm Choksi, Michaeli and Papagergiou's (forthcoming) conclusion that trade liberalization does not appear to have affected negatively labor employment in the developing countries.24 Regression results. The next step in the TFP analysis consisted of relating the observed trend in the TFP growth in Turkey to changes in the trade regime. For the regression analysis all three industry grouping were 24.This outcome is all the more outstanding remembering that the data used in this paper tend to underestimate the growth ir labor employment for the period 1983 to 1985. See footnote 14.  28 pulled together but the distinction between private and public was retained. Two sets of regressions were implemented. In the first, the following equation was estimated for both the private sector and the public sector: (6) TFP - f(Qit, mit, Di, Dt) Growth in total factor productivity is related to the rate of growth in output (Q), the rate of growth in import penetration (m) measured as the share of imports in total domestic absorption25 , and a set of industry and year dummy variables. Output growth is intended to capture scale effects and variations in capacity utilization. If economies of scale are present and/or if production capacity is underutilized, output growth is expected to have a positive sign. Import penetration rates are intended to capture the effect of exposure to foreign competition on efficiency. If exposure to foreign competition forces domestic firms to maximize production efficiency, the growth in import penetration is also expected to be pesitively related to TFP growth. As with the regression analysis of price- cost margins, industry and year dummy variables are included to control for industry-specific and time-specific factors. The estimation results of equation 6 for the private sector for the periods 1976-77 to 1984-85 are reported in Table 23. The two models in Table 23 differ only in that Model 2 does not include the industry dummy 25.If export penetration ratio is also included in the same equation or if it replaces import penetration, its coefficient is never significant for either the private or the public sectors.  29 variables. The results of Model 1 indicate that growth in output is the most powerful explanatory variable in this model. However, growth in import penetration also has a signifi:ant albeit modest effect on the TFP growth in the private sector. The results of Model 2 indicate that unlike in the price-cost margin regressions, industry dummy variables are insignificant. The negative sign on year dummy variables indicate that, relative to 1976-77, productivity growth was lower in all the subsequent periods. Table 24 contains the regression results of equation 6 for the public sector. Except for the effect of import penetration, the results are very similar to those for the private sector. Here, however, growth in import penetration has a negative albeit statistically insignificant coefficient. This result indicates that exposure to greater competition from abroad has exerted no significant effect on the productivity performance of public enterprises. This outcome confirms the earlier findings that the productivity growth in public firms was more affected by their internal reform, including the freedom of the management to determine the level of employment and set output prices, than by exposure to foreign competition. Moreover, as noted earlier, some public enterprises, especially in import competing industries, continued to enjoy a relatively high level of protection through non-tariff barriers. The second set of regressions are based on a standard demand decomposition model. To see the rationale for demand side decomposition let us start with identity (7) which states that total domestic output either satisfies domestic demand, which is equal to total demand (D) minus imports (H), or is exported (X). (7) Qt - Dt - Mt + Xt  30 Given the derinition of import penetration ratio, mt-Mt/Dt, with a little manipulation it is easy to show that: (8) Qt= Dt - smmt + sx where Sd - (l-m)D/Q : share of domestic Aemand in total output sm - mD/Q : share of imports in total output 8x = X/Q : share of exports in total output Thus total output growth, G, can be decomposed into three components: (9) G = Xl - X2 + X3 where Xl, X2, and X3 represent respectively output growth due to domestic demand growth, import penetration growth, and export demand growth. With this demand decomposition it is possible to test whether the growth of output originating from domestic demand, exports, or import penetration affects productivity in a different manner. Thus the next set of regressions for both the private and the public sectors in Turkey is based upon the estimation of the following equation: (10) TFP = f(Xl, X2, X3, Di, Dt) If higher exports or higher import penetration, or in other words greater openness, do not affect productivity, one expects the coefficients of variables Xl, X2 and X3 to be exactly the same, except for the negative sign on X2. On the other hand, if higher import penetration or higher  31 exports do have a positive effect on productivity beyond their impact on domestic output, then one would expect to find a higher coefficient for X3 than for Xl and a lower absolute value for the coefficient of X2 than for Xl. The estimation results of equation 10 for the private and public sector are reported in Tables 25-26. Because separate trade data for the public and private enterprises are unavailable, Xl-X3 are computed by summing together private and public sector output for each industry. The regression results for the private sector are contained in the first part of Table 25. There appears to be no significant difference between the coefficionts of variables Xl-X3 indicating that import or export growth exert no significant effect on productivity performance of the private sector. This tesult appears to contradict the earlier finding that greater import penetration affected positively, albeit modestly, the productivity performance in the private sector. The contradiction disappears if in equation 10 the rate of growth of imports, m, is included as an additional explanatory variable in equation 10 (model 2 in Tables 25 and 26). If import share is constant, variables X2 and m, would be perfectly collinear. However, because import share is not constant, the model is identified. As with equation 6, this new variable has a positive and statistically significant coefficient. This result means that the growth in import penetration affects positively the pr'Juctivity growth in the private sector, but that the F-test fails to detect any significant difference between the various components of demand on productivity. The reason is the following. Let us assume that the true relation between productivity and various components of demand is given by:  32 (11) TFP - Po + plXl + p2X2 + p3X3 + A4m Since m-X2/sm, equation 11 can be rewritten as: (12) TFP - GO + alXl + a2X2 + a3X3 where 'o-Po ai'Pi 92 (P22P4 18m) 13-P3 When A is positive, the absolute value of @2 is smaller when growth in import penetration is excluded from the regression. However, if P4/sm is small, the F-test rejects the hypothesis that a2 and a1 are statistically different from each other and growth in import penetration appears to have no significant impact on productivity. The estimation results of equations 11 and 12 for the public sector sector appear in Table 26. The results are similar to those for the private sector except for the coefficient of X3. Also in accordance with previous findings, when growth in import penetration is included as an independent variable, its coefficient is negative but not significant. These results once more show that trade policy has not had a great impact on the productivity of Turkish public enterprises. Notice that the fit of equations 11 and 12 for the public sector is not very good because the components Xl, X2, and X3 are computed from total output and the share of the public sector in total output is low.  33 III.d Comparison With Earlier Findings To my knowledge, for Turkey the relation between trade policy and industry conduct as represented by the evolution of price-cost margins has not been addressed in the English language literature. There exist, however, some studies on the relation between trade policy and productivity performance in Turkey. Here a brief description of the main findings of these studies is presented and an attempt is made to compare the results of earlier studies with those obtained in the present paper. A first study evaluating the effect of trade policy in Turkey on the productivity of Turkish industry is that by Krueger and Tuncer (1980) (henceforth referred to an KT). KT compute the rate of growth of TFP in the Turkish manufacturing industries during 1963-1976 separately for the private sector and public sector. They subdivide the period covered into four subperiod, namely 1963-67, 1967-70, 1970-73, and 1973-76. The first and the third subperiod represent years of relatively more liberal trade policies whereas the second and the fourth subperiod rzpresent years of severe, import restriction. KT find that, compared to some developed countries, TFP growth was on the whole modest in Turkey over the entire period (2.1Z) considered. More interestingly, they find that TFP growth slowed down considerably during period of stringent import restriction. KT conclude that their findings provide evidence that "productivity growth might uniformly be more rapid during periods of liberalization of the foreign trade regime than during periods of severe foreign exchange shortage' (ibid p.4]. Thus, although the period covered by the KT study differs from that in the present paper, the inference about the positive effect of trade liberalization on TFP growth in industry confirms the findings reported here.  34 Another finding by KT which is also similar to that of this paper is the relatively more rapid rate of TFP growth in the public sector. Although results differ greatly from one industry to another, KT find that on average the TFP growth in the public sector (2.652) for the period 1963-76 outpaces the TFP growth in the private sector (1.84Z). Despite several attempts, KT are not able to provide a satisfying explanation for this seemingly paradoxical result. They are, however, able to infer something about the absolute level of efficiency in the two sector. KT find that in general pub.ic enterprises had a higher level of labor and capital input per unit of output. On the other hand, public enterprises were able to purchase material inputs at :s;ubsidized prices. Had these enterprises faced the same cost for their internmediate inputs, their level of efficiency, as measured by the unit ratio of capital, labor, and material input in the public sector relative to that in private sector, wold have been lower than in the private enterprises. Another study examining the relation between trade orientation and productivity growth in Turkey is that by Nishimizu and Robinson (1984). Nishimizu and Robinson analyze the relation between the growth in total factor productivity and trade orientation in four countries: Korea, Turkey, Japan, and Yugoslavia. The analysis for Turkey covers the period 1963 to 1976 for 13 broadly defined industries without distinguishing public from private enterprises. For each industry the rate of growth of TFP is regressed on output growth allocated between export expansion and import substitution. The results show that for 9 out of 13 industries TFP growth was significantly and positively correlated to export expansion whereas for 4 out of 13 industries import substitution was negatively and significantly correlated to TFP growth. During this period Turkey was a  35 relatively closed economy with exports accounting for less than 4 percent of total manufacturing output and imports contributing to only 11 percent of domestic supply of manufacturing output. Total factor productivity growth was therefore modest during this period increasing on average by 1.3 percent per year and contributing to 12 percent of output growth. A third study worth mentioning is the World Bank Audit Report (1988) evaluating SALs IV and V. This report addresses the issue of productivity in the public sector in Turkey during the 1980s. By examining the evolution of labor productivity, defined as real output per employee, during 1982-86, the Report concludes that "improvements in SEE [State Economic Enterprises] profitability are due to price increases, not to efficiency gains" (page 84). This conclusion clearly contrasts with the findings of this paper. A number of reasons explain the contrasting conclusions. First, the results of the above-mentioned report are based on labor productivity as opposed to total factor productivity. Second, and more importantly, the data used in the above report apparently refer to all SEEs whereas the data used here refer only to manufacturing S:Es. Thus both the real output and the number of employees in the present paper are only a fraction of those contained in the above report. However, for the sake of comparison I also calculated labor productivity as defined in the Bank report for both the private sector and the public sector. The results are reported in Tables 26 and 27. The numbers -learly show that after a historic low in 1979-80, labor productiv4ty rises sharply in both the private sector and in the public sector. More interestingly, in conformity with earlier results, labor productivity in the public sector rises faster and its level is above that in the private sector during the 1980s. In summary, the findings in the present paper appear to confirm earlier findings that trade liberalization is associated with improvement  36 in productivity in Turkey's manufacturing industry. Moreover, improvement of productivity in the publiic sector is attributable not only to trade liberaiization but also a host of reforms of public enterprises. III. e Conclusions The results of the analysis contained in the present paper indicate that trade liberalization and greater exposure to international competition have generally had a beneficial effect on the Turkish industry during the 19809. The effect of international competition, however, appears to be significant mainly in the private sector, especially in tradable industries. In the first half of the 1980s, international competition bas decreased the price-cost margin and increased the rate of growth of productivity in the private sector. In the public sector, higher trade penetration appears to have lowered the price-cost margin in the public enterprises that had a higher-than-averag capital intensity. Trade penetration shows no impact on the productivity performance of the public sector. Productivity improvement in public enterprises appears to be more related to other factors, most likely the reform of the management of these enterprises.  37 Table 1 Industrial Classification Code Industrial Activity 311, 312 Food processing 313 Beverage 314 Tobacco 321 Textile 322 Wearing apparel 323 Fur and leather products 324 Leather shoes 331 Wood and cork 332 Furniture & fixture 341 Paper and paper products 342 Printing and publishing 351 Industrial chemicals 352 Other chemicals 353 Petroleum and coal 354 Petroleum derivatives 355 Rubber products 356 Plastic 361 Ceramics and porcelain 362 Glass and glass products 369 Other non-metal products 371 Iron and steel 372 Non-ferrous metals 381 Metal products 382 Non-electric machinery 383 Electric machinery 384 Transport equipment 385 Professional equipment  38 TABLE 2 AVERAGE SHARE OF P2IVATE (ASPR) AND PUBLIC (ASPU) FIRMS IN TOTAL INDUSTRIAL OUTPUT, 1976-1985 O5s YEAR ASPR ASPU 1 1975 0.721674 0.278326 2 1976 0.740659 0.259341 3 1977 0.724289 0.275711 4 1978 0.752052 0.247948 5 1979 0.753372 0.246628 6 1980 0.754654 0.245346 7 1981 0.743426 0.256574 8 1982 0.766957 0.233043 9 1983 0.765165 0.234835 10 1984 0.7689a1 0.231069 11 1985 0.760505 0.239495 TABLE 3 AVERAGE SHARE OF PRIVATE (ASPR) AND PUBLIC tASPU) FIRMS IN TOTAL OUTPUT OF EXPORTABLE INDUSTRIES OBS YEAR ASPR ASPU 1 1976 0.840750 0.159250 2 1977 0.830160 0.169840 3 1978 0.862868 0.137132 4 1979 0.857863 0.142137 5 1980 0.836718 0.163282 6 1981 C.815149 0.184851 7 1982 0.8'4955 0.125045 8 1983 0.857638 0.142362 9 1984 0.870658 0.129342 10 1985 0.818399 0.181601 TABLE 4 AVERAGE SHARE OF PRIVATE (ASPR) AND PUBLIC (ASPU) FIRMS IN TOTAL OUTPUT OF IMPORiABLE INDUSTRIES O0S YEAR ASPR ASPU 1 1976 0.768374 0.231626 2 1977 0.696982 0.303018 3 1978 0.750455 0.249545 4 1979 0.775547 0.224453 5 1980 0.774497 0.225503 6 1981 0.766003 0.233997 7 1982 0.777076 0.222924 8 1983 0.783287 0.216713 9 1984 0.788850 0.211150 10 1985 0.790460 0.209540  39 TABLE 5 AVERAGE SHARE OF PRIVATE (ASPR) AND PUBLIC (ASPU) FIRMS IN TOTAL OUTPUT OF NON-TRADABLE INDUSTRIES OBS YEAR ASPR ASPU 1 1976 0.664083 0.335917 2 1977 0.664565 0.335435 3 1978 0.688629 0.311371 4 1979 0.682873 0.317127 5 1980 0.692831 0.307169 6 1981 0.685197 0.314803 7 1982 0.696513 0.303487 8 1983 0.699822 0.300178 9 1984 0.698908 0.301092 10 1985 0.708396 0.291604 TABLE 6 TRADE PENETRATION RATIOS 1976-85 OBS YEAR AXSHARE AMSHARE 1 1976 1.8941 15.4643 2 1977 1.5995 14.7307 3 1978 1.6387 10.9462 4 1979 1.6954 10.1034 5 1980 2.9405 12.4574 6 1981 6.8825 13.0226 7 1982 10.0012 13.6607 8 1983 9.1189 13.3348 9 1984 13.3050 17.0171 10 1985 19.0618 18.3276 AXSHARE=SHARE OF EXPORTS IN TOTAL OUTPUT AMSHAREuSHARE OF IMPORTS IN TOTAL APPARENT DOMESTIC CONSUMPTION. DEFINED AS OUTPUT+IMPORTS-EXPORTS  40 TABLE 7 AVERAQE SHARE OF EXPORTS (AXSHARE) AND IMPORTS (AMSHARE) IN TOTAL OUTPUT OF EXPORTABLE INDUSTRIES ;N % TERMS 08S YEAR AXSHARE AMSHARE 1 $976 3.9293 f9.7575 2 1977 3.6639 19.8307 3 1978 3.8270 14.4596 4 1979 3.8087 11.9783 5 1980 6.3960 17.0996 6 198$ t5.2136 18.5571 7 1982 23.5270 19.4807 a 1983 20.7087 18.8979 9 1984 29.7332 22.2022 10 1985 42 2294 25.3638 TABLE a AVERAGE SH4ARE OF EXPORTS (XSHARE) AND IMPORTS (mSHARE) IN TOTAL OUTPUT OF IMPORTABaLE INOUSTRIES IN % TERMS o8s YEAR AXSHARE ANSHARE 1 1976 1.8771 35.5776 2 1977 1.6279 32.4762 3 1978 1.2386 24.0558 4 1979 1.1427 21.0782 5 1980 2.4416 25.2792 6 198t 4.4357 29.8413 7 1982 7.1163 32.516t 8 1983 7.4535 31.2197 9 1984 9.5345 38.2476 tO i985 14.0271 40.8900 TABLE 9 AVERA&Q SHARE OF EXPORTS (AXSHARE) AND IMPORTS (ANSNARE) 1% TOTAL OUTPUT OF NON-TRADABLE INDUSlRIES IN % TERMS OBS YEAR AXSHARE AMSHARE 1 s976 0.95559 S.S3856 2