CHAPTER V THE GREAT CONTRACTIONS IN RUSSIA, THE BALTICS, AND THE OTHER COUNTRIES OF THE FORMER SOVIET UNION: A VIEW FROM THE SUPPLY SIDE

Abstract

The early years of transition from a command to a market-based economy in central and eastern Europe have typically witnessed considerable output drops, at least according to official national statistics. Output contractions were even more pronounced in the Baltics, Russia, and the other countries of the former Soviet Union. By 1997, growth had resumed in the vast majority of transition countries. Even though a number of them faced renewed output declines in 1998, mainly associated with the financial crisis in Russia, it appears that the bulk of the transitional recession is now over in most cases. With the benefit of hindsight, this study attempts to put the contractions endured in the Baltics as well as in Russia and the other countries of the Commonwealth of Independent States (CIS) in perspective, examining them from several angles.

The early years of transition from a command to a market-based economy in central and eastern Europe have typically witnessed considerable output drops, at least according to official national statistics. Output contractions were even more pronounced in the Baltics, Russia, and the other countries of the former Soviet Union. By 1997, growth had resumed in the vast majority of transition countries. Even though a number of them faced renewed output declines in 1998, mainly associated with the financial crisis in Russia, it appears that the bulk of the transitional recession is now over in most cases. With the benefit of hindsight, this study attempts to put the contractions endured in the Baltics as well as in Russia and the other countries of the Commonwealth of Independent States (CIS) in perspective, examining them from several angles.

A first and obvious question arises as to their magnitude. On many scales, it seems to be very large. While this impression can be corroborated in a number of ways, quantification has been, and remains, a major challenge because of the poor quality of available statistics. In fact, because of these uncertainties, any point estimate is likely to be misleading. Bearing this caveat in mind, a second issue is how to evaluate the wel-fare consequences of these contractions. All too often, the magnitude of the output decline is equated, at least implicitly, with that of the wel-fare loss. On reflection, however, it turns out that there is no straightforward, one-to-one, relationship between the former and the latter, as welfare measures reflect the type of social wel-fare function used.

Next, the stylized features of the contractions in the Baltics and the CIS countries are described and contrasted with the experience of selected central European transition economies. While many studies tend to focus on the depth of the contractions, their length and breadth are also highlighted here, and performance is set against countries’ longer-run growth record.

The contractions have been ascribed a variety of causes in the literature, including the dislocation of traditional domestic and international links, fiscal retrenchment, and credit crunches, with the relative importance of these factors differing across countries.1 This study does not consider such a broad range of factors, but instead focuses on changes in inputs and the evolution of productivity to interpret the observed decline in output, in line with earlier work byEasterly and Fischer (1994). It also sheds light on the reallocation of inputs across sectors.

Background Considerations

Starting in the late 1980s or early 1990s, official measures of output and value added declined precipitously in the Baltics, Russia, and the other countries of the CIS. The size of the actual contractions, however, is hard to gauge.

On the one hand, part of the falling activity reflected the replacement of the previous incentive to report the (over-)fulfillment of plan targets by the incentive to avoid the scrutiny of the tax and other authorities (Koen, 1994). Anecdotal evidence of the dynamism of underground businesses abounds, at least in some sectors. In the early years of transition, statistical offices generally did not have at their disposal the tools required to track the volume of economic activity under the new conditions. Erring on the conservative side, they typically preferred to base their published estimates on what was still reported to them rather than on their best estimate of aggregate output. Some alternative indicators of overall economic activity, most notably electricity consumption, suggested that the official real GDP estimates were substantially downward-biased (Gavrilenkov and Koen, 1995; and Dobozi and Pohl, 1995).

Later on, several of the statistical offices, recognizing those problems, started to incorporate survey-based evidence and even guesstimates of hidden activity into their GDP series.2 They also revisited the series they had published thus far and implemented some substantial upward revisions, in particular in Russia, Kazakhstan, and Lithuania (Appendix 5.1).3 The GDP series as revised for the initial transition years are indeed more in line with what is suggested by independent estimates carried out from the demand side of the national accounts (in the case of Russia) or based on physical output statistics (in the case of Kazakhstan), and they continue to show a massive output decline. In addition, alternative indicators are not necessarily good proxies for actual GDP (Appendix 5.2), and not all biases need go in the same direction.4 Finally, the authorities sometimes face political incentives to publish higher rather than lower GDP figures.5

On balance, it is hard to believe that output did not contract massively in the early years of transition, even if the magnitude of the actual decline may be somewhat overstated by the official statistics, especially where no efforts were deployed to take into account new and underground economic sectors. Further output declines beyond the initial transition years are generally measured more accurately in the official numbers as statistical agencies made progress in implementing market-based methodologies and the overall economic environment became more stable.

In addition to the practical obstacles standing in the way of a comprehensive and precise measurement of activity, a serious theoretical conundrum deserves to be highlighted. The quantification of changes in real GDP relies on the prices used to aggregate developments in enterprises and branches. The choice of prices matters a lot for the end result in a period of highly unstable relative prices. Depending on the issue under consideration, it may be more sensible to use actual or notional market prices, current or base-period prices. No weighing scheme is intrinsically superior to all others when aggregating observations. Moreover, when notional prices are used (e.g., “world” prices instead of distorted, actual prices), it must be recognized that quantities would have been different had that set of prices prevailed. Thus, even if information on quantities and prices were perfect, it can be argued that there would still be no single, “true” real GDP series.

The scope for divergence across estimates can be significant, as illustrated by the example of the Russian economy, which experienced a seven-year contraction over the period 1989-96. The cumulative drop in real GDP in Russia during the period amounts to 45 percent if volumes are aggregated at 1989 prices. If prices of the previous year are used instead, it amounts to 42 percent. If 1996 prices are used, possibly on the grounds that those are more relevant because closer to market levels, the cumulative drop is yet smaller, at 41 percent.6

Assumptions regarding the size and dynamism of activities in the shadow economy that are not captured in the official statistics also affect the size of the decline.7 A conservative assumption could be that it represented 10 percent of official GDP in 1989 and expanded at 2 percent per annum. This would be in line with a perception that official numbers capture much of the unrecorded activity, or that there are limits to the speed at which it can develop. An alternative assumption would be that the shadow economy represented 20 percent of official GDP in 1989 and expanded at 5 percent per annum. In the context of the Russian example, the first assumption implies that the shadow economy represented 18 percent of official GDP by 1996 (using a weighing scheme based upon the 1989 shares) and the second that it represented 38 percent of official GDP by that date (a range well within the range of estimates that are commonly cited for the Baltic and the CIS countries).8 The conservative view then implies that actual GDP declined by 35 percent between 1989 and 1996, while the alternative view implies a smaller cumulative drop of 23 percent. Relying on electricity consumption as a proxy for actual GDP would lead to yet another range of estimates (Appendix 5.2).

Not only is the magnitude of the contraction difficult to pin down, but its implications for wel-fare are also complex, even abstracting from distributional considerations.9 Part of the output decline may have been welfare-enhancing. This could apply to the termination or downsizing of some types of military or prestige investment, for instance. It could also apply to some heavy, energy-intensive, and polluting industries producing unsophisticated intermediate outputs for which value-added measured at market prices would be negative. Assigning a social value to such (dis)investments or recompiling value-added at undistorted prices involves a fair degree of discretion, however, over and beyond access to data that may not be available.10 Another set of welfare effects disregarded by conventional output measures but important in transition economies are those associated with enhanced access to markets by consumers. The disappearance, or at least lessening, of rationing and queuing, as well as the increased variety on offer are tangible benefits that are not directly reflected in real GDP measures. Again, the size of such offsetting welfare gains depends on the type of social welfare function used, but it can be consequential (Roberts, 1997). In the subsequent sections, only output movements as captured by the official data are considered, and the distributional consequences and welfare implications of the output decline are left aside.

Selected Stylized Features

Notwithstanding the numerous caveats out-lined in the previous section, official national accounts appear to broadly reflect the main characteristics of the underlying output movements during the transition. In view of the poor quality of the data, the analysis in the remainder of the paper is, however, limited to an overview of the main features of the contraction and to a study of the relationship between outputs and inputs at both the aggregate and sectoral levels, data on which are generally of comparable quality.11

The length of the contraction varied considerably across countries, but on the whole was much longer in the Baltics and CIS countries than in central Europe. Looking at the transition through 1997, the duration of contractions ranged from four to nine years or more, with a median of six years (Table 5.1). In contrast, the contractions in the six comparator countries of central Europe lasted between two and five years, with a median of four years.

Table 5.1.

Contraction Length

(In years, based on observations through 1997)

article image
Sources: Yearbooks of national statistical offices; CIS, Statistical Committee; CMEA Yearbooks.

Contraction from peak year, as given in Table 5.2.

Based on the revised series.

Indicative only, since pretransition real output series are generally for net material product.

0utput declined anew following the first recovery.

Activity rebounded modestly in 1990 following a sharp drop in 1989, but GDP contracted again in 1996.

Likewise, the depth of the contractions varied tremendously across countries and was in general distinctly larger than in central Europe (Table 5.2). In Uzbekistan, where the contraction was the shallowest, measured real GDP shrank by 19 percent, compared with a 77 percent fall in Georgia at the other end of the spectrum. On an unweighted basis, the contraction averaged some 51 percent,12 two-and-a-half times the 21 percent in central Europe on a comparable basis.

Table 5.2.

Contraction Depth

(Real GDP level in percent of historical peak)

article image
Sources: National statistical offices; CIS, Statistical Committee.Note: Through the late 1980s or the early 1990s, real net material product series are typically used as a proxy for real GDP for Russia, the Baltics, and the other countries of the Former Soviet Union. These data are shown through the end of the contraction, or if it had not ended, through 1997.

Higher, revised series.

Output declined anew following the first recovery.

An alternative way to gauge the depth of the contractions which is fairly popular in some transition countries is to compute how far back the contraction threw the level of economic activity. On official measures, GDP typically reverted to levels witnessed two decades or more earlier,13 again over twice as deep in the Baltics and CIS countries as in central Europe, which “regressed” by only about one decade.

The depth of the contractions also varied substantially across sectors. For the fifteen successor countries as a group, overall (unweighted) out-put dropped by 46 percent between 1990 and 1997. The decline was more pronounced in industry and transport and communication, where production fell by over half, and most spectacular in the construction sector, where it shrank to around one third of the level observed in 1990, mirroring the collapse in investment spending. In contrast, production in agriculture and trade fell by around one third. Much sharper variations across sectors have been reported in a number of individual countries.

In most of the countries under consideration, industry has been the largest sector of the economy, prompting the question of the extent to which the contraction in GDP tracks the collapse of industrial output. On an unweighted basis, the peak-to-trough decline in gross industrial output on average exceeds the fall in measured GDP only marginally, while the length of the industrial contractions is broadly comparable, also ranging from four to at least nine years with a median of six years (Tables 5.3 and 5.4). In contrast, the depth of the industrial contractions in the comparator central European countries is almost twice that of the decline in measured GDP, although similar in length. Two non-mutually exclusive conjectures offer themselves. Since output in industry is probably less difficult to monitor than in other sectors, this evidence might be taken to suggest that the GDP contractions are overstated in the Baltics and CIS countries to the extent industry might be expected to show larger declines than other sectors.14 It may also be that the environment in these countries was less conducive to the development of sectors other than industry than in central Europe, thus limiting their cushioning potential.

Table 5.3.

Depth of the Contractions in Industrial Output

(Gross industrial output level in percent of historical peak)

article image
Sources: National authorities; CIS, Statistical Committee.

Temporary rebounds were officially recorded in 1993 and 1996.

Table 5.4.

Length of the Contractions in Industrial Output

(In years, based on observations through 1997)

article image
Sources: Yearbooks of national statistical offices; Statistical Committee of the CIS.

Temporary rebounds were officially recorded in 1993 and 1996.

Output declined anew following the first recovery.

Most studies of output developments have focused on countrywide indicators. At a more disaggregated level, the evolution of output has varied as much from one region to another as from one country to another. In fact, some regions experienced a relatively shallow and brief recession, while in others output continued to contract well after the start of the national recovery, falling much deeper than the average trough.15 Such regional disparities are particularly striking in vast countries such as Russia, Ukraine, and Kazakhstan (Table 5.5).16 To some extent, and like at the nationwide level, the differences across regions reflect, among other factors, the diversity in the stance of local policies, as documented by Berkowitz and Dejong (1997) in the case of Russia.

Table 5.5.

Fall in Industrial Output Across Regions During the First Half of the 1990s1,2

article image
Sources: Regional statistical yearbooks of the national statistical offices and authors’ computations.

Between 1990 and 1995 (Kazakhstan), 1996 (Russia), or 1997 (Ukraine). Peak to trough for each region.

All columns in percent except for the last one.

Coefficient of variation.

Two regions, Chechnya and Ingushetia, are not included.

Inputs and Outputs: Contraction Accounting

The remainder of the paper explores the sharp output decline in the Baltics and CIS countries from the angle of the relationship between inputs and outputs. A commonly used approach to link output to inputs and to measure economic performance over time is to set up a growth accounting framework to assess the efficiency with which the factors of production labor and physical capital are employed. To illustrate how the transition process has affected the economic performance of the fifteen transition countries of the former Soviet Union, a growth-accounting exercise is performed for each country individually and for all fifteen as a group. Furthermore, to put the output decline during the transition in perspective and relate it to distortions and misallocations inherited from central planning, the sample period is extended to include the last two decades prior to the Soviet breakdown. The exercise covers the overall economy and the aforementioned six main sectors of the central planning recording system.

The exercise is based upon the assumption that output is produced according to a neoclassical production function of the following form,

Y(t)=A(t)F[K(t),L(t)],(1)

where A is an index of total factor productivity (TFP), K is the capital stock, and L measures the inputs of labor. Differentiation of equation (1) with respect to time yields, after division by Y,

gY=gA+ηKgK+ηLgL,(2)

where gY, gA, gK, and gL, are the rates of growth of Y, A, K, and L, respectively, and where ηK = (K/Y).(∂F/∂K) and ηL = (L/Y).(∂F/∂L) denote the elasticities of output with respect to capital and labor, respectively. Further differentiation of equation (2) with respect to time gives

gY/t=gA/t+(ηK/t).gK+ηK.(gK/t)+(ηL/t).gL+ηL.(gL/t),(3)

which allows one to decompose changes in the growth rate of output period by period.

This type of accounting relies on underlying assumptions about the nature of the production function that is specified in equation (1).17 In the absence of information on factor prices that would allow one to approximate the elasticities of output with respect to capital and labor, the computations below assume that these elasticities are constant over time and add up to 1.18 The computed changes in TFP should be interpreted as residuals that reflect, in addition to biases due to methodological assumptions and measurement errors, a wide range of factors affecting the efficiency with which inputs are used.19 The remainder of the paper focuses on analyzing the fall in growth during the transition period relative to pretransition growth, in line with equation (3). This is a valid approach as long as the biases and errors in the computations are not subject to a regime shift following the transition. Furthermore, since no corrections are made for variations in hours worked and capacity utilization, or, more generally, for quality changes and factor obsolescence, this approach also implies that, except for reductions in reported employment and investment, the impact of the transition will be reflected entirely in changes in estimated TFP.

The computational results are summarized in Table 5.6, which reports average annual output and input growth rates, factor contribution rates, and TFP growth rates by country and by sector for the periods 1971-90 and 1991-97. TFP growth turned sharply negative during the transition in all countries and sectors, after having fallen to close to zero in the last two decades of central planning.20 The drop was especially pronounced in conflict-torn countries (Armenia, Azerbaijan, Georgia, Moldova, Tajikistan) and in Ukraine, where the output fall continued throughout the sample period. Negative values for TFP growth during the transition are also observed on a sectoral basis, mostly so in construction, industry, and transport and communication.

Table 5.6.

Output and TFP Growth, Period Averages

article image
article image

Based on the revised aggregate growth rates.

Assuming the growth rate of employment in the trade sector was the same in 1997 as in 1996.

To test the robustness of these computations, alternative calculations are presented based on labor productivity growth rates, which do not depend on assumptions about the nature of the production function and about output elasticities with respect to inputs of factors nor on particular data construction procedures for the capital stock variable. The results, presented in the last column of Table 5.6, show a broadly comparable pattern for productivity measures based upon either TFP or labor productivity. According to both measures, in all countries and sectors, productivity fell during 1991-97 by on average more than 6 percent a year. Labor productivity growth rates were slightly less negative than TFP growth rates in all countries reflecting the more rapid expansion of capital inputs over the period. At the sectoral level, however, this was not the case in agriculture, trade, and services. In these three sectors, labor productivity fell by more than TFP, reflecting increases in sectoral employment.

The implications of the productivity decline in the early transition years can be put in perspective by comparing its contribution to the overall output fall to that from reductions in inputs of capital and labor. In most countries and sectors, factor contribution was negative during the transition, reflecting reductions in employment and investment. Total employment in the 15 successor states as a group fell by more than 12 percent in 1991-97, with sharper reductions in the Baltic countries (by around 20 percent) and in Russia (by 15 percent), while it actually expanded in Turkmenistan and Uzbekistan. On a sectoral basis, employment in the trade and services sectors rose significantly, was broadly stable in agriculture, but shrank by over 35 percent in industry and by over 40 percent in construction. The contribution of capital accumulation also was reduced sharply, reflecting the investment collapse during the transition. By 1997, real investment fell short of its 1990 level in 14 of the 15 countries and in all six sectors,21 with the average shortfall amounting to around 70 percent. In most countries and sectors, investment fell to well below replacement rates and the growth rate of the capital stock turned negative.

The investment collapse had additional negative repercussions, as it accelerated the aging of the capital stock. Reflecting the slowdown of investment spending in the second half of the 1980s and its reduced effectiveness,22 capital stock obsolescence had already begun to set in before the transition (Akopian, 1992). In Russia, for instance, the average age of plant and equipment had increased to 10.8 years in 1990 from 8.4 years in 1970. With investment collapsing in most countries, the capital obsolescence problem became more severe in the course of transition.23 In Russia, the average age of plant and equipment increased further to 14.9 years in 1996, by which time the volume of investment had fallen to less than one-fourth of its 1990 level. Since the production function in equation (1) does not incorporate vintage effects, efficiency losses associated with increasing capital obsolescence are reflected in the TFP growth estimates in Table 5.6.

An analysis of productivity developments on a year-by-year basis offers additional insights (see Table 5.12). A distinct V-shaped pattern in TFP growth emerges. TFP growth was generally sharply negative in the early years of the transition but turned positive in most countries, in some cases very significantly so, by 1995-96, indicating that part of the initial sharp productivity decline was temporary, with production factors being less than fully used. Factors such as trade disruptions and disorganization related to the breakdown of central planning played an important role during this initial period (Blanchard and Kremer, 1997). Similarly, the rapid increase in TFP in countries and sectors where output growth turned positive again more recently likely reflects a recovery in the rate of factor utilization.

Sectoral Reallocation

The output decline during the transition is to a large extent accounted for by TFP losses. These losses in turn reflect a range of factors.24 One important factor is the sectoral reallocation of inputs, which is considered to be one of the key dimensions of the transition process (Blanchard, 1997). Since the output and input data used in this study are available on a broad sectoral basis, the contribution of changes in the sectoral composition of inputs to aggregate TFP growth can be quantified.

Following Bernard and Jones (1996) and Cameron and others. (1997), aggregate TFP can be expressed as a weighted sum of sectoral TFPs, where the weights are ratios of an index of inputs in each sector to an aggregate index of inputs.25

A=Y/(KαL1α)=ΣjωjYj/(KjαLj1α),

where

ωj=(Kj/K)α(Lj/L)1α(4)

and j indexes sectors. Accordingly, the change in aggregate TFP can be decomposed into a productivity change and a share effect. Between year (t-1) and year t

ΔA/A=1/A{ΣjΔTFPj.ωjt1within-sectoreffect+ΣjΔωj.TFPjt1}Shareeffect(5)

The first effect measures the contribution of productivity changes within each of the six sectors, and the second one the contribution of changes in sectoral composition to aggregate TFP growth, which will be positive if resources are reallocated from lower to higher-productivity sectors.

Share effects computed according to Equation (5) for both the pretransition and the transition periods are presented in Table 5.7. For the 15 countries as a whole, the share effect accounts for around 8 percent of the change in TFP in the early transition years. Broadly comparable share effects are found in the 15 countries individually. Since the initial transition years were characterized by negative TFP growth rates, these results point to a productivity reducing re-allocation of resources. The share effect is relatively small, however, indicating that sectoral input reallocation did not have a major impact on productivity.

Table 5.7.

Contribution of Sectoral Reallocation to TFP Growth

(Percent)

article image

TFP growth was negative in this period.

The productivity effect stemming from sectoral input reallocation during transition merits further examination. The sectoral composition of inputs changed noticeably during the transition. The shares of agriculture, trade, and services in aggregate employment rose, while those of industry and, in particular, construction fell.26 Within industry, the share of electricity and other energy branches in total sectoral employment increased, while that of machine-building decreased. The sectoral composition of the capital stock also shifted, with industry and transport and communication gaining in importance and agriculture declining.27 Since there was also reallocation of inputs during the pre-transition period, for the contribution of reallocation to aggregate TFP to be different during the transition, the reallocation process had to change.

To examine whether this has been the case, the following approach is adopted. For each sector and country, annual sectoral employment growth rates relative to the aggregate employment growth rate as well as annual sectoral investment shares are computed, according to data availability. Next, the averages of these employment growth rates and investment shares, respectively, are computed for the pretransition and transition periods. Finally, a bootstrap methodology is applied to test the hypothesis that there was no difference between the pre-transition and transition averages.28 The hypothesis is rejected if, based upon the frequency of the outcomes for 1,000 replications, there is a probability of 1 percent or less that a random drawing from the sample corresponding to the whole period would produce a difference between the two sub-periods as large or larger than the observed one.

According to this procedure, in the 15 successor states as a group, the average growth rate of employment fell significantly during the transition in industry and construction, while it increased in agriculture and trade (Table 5.8). The same pattern is observed in individual countries in all cases where changes are found to be significant, with the rare exception of agriculture in Estonia. Somewhat different results obtain for investment (Table 5.9). The average growth rate of investment during transition was significantly lower in agriculture—in contrast with the finding for employment growth—and construction and industry, while it was significantly higher in the housing, education, and services sector. The downward shift in investment growth in agriculture is observed in all 15 countries and may be related to the break-up of the collective farm system with its centralized investment decisions. The computations for the other two sectors are inconclusive at the group-wide level, but indicate significant changes in individual countries, reflecting, for instance, additional investment efforts in natural resources extraction or telecommunications projects.

Table 5.8.

Total Employment Shifts1

article image

+ Denotes a significant shift towards the sector; - denotes a significant shift away from the sector; 0 denotes the absence of a significant shift in either direction.

Table 5.9.

Total Investment Shifts1

article image

See footnote 1 to Table 5.8.

More disaggregated data on the composition by major branch of the inputs used in industry shed some light on the factors underlying input shifts for the sector as a whole. For the successor states as a group, the growth rate of employment increased significantly in the electricity, fuels, metallurgy, and food industries, whereas it dropped significantly in machine-building (in part reflecting the sharp reduction in military procurement) and light industry (Table 5.10). The starkest contrast is between electricity and machine-building, with employment growth increasing significantly in the former branch in each individual country and falling significantly everywhere in the latter branch. As regards investment, the results tend to show less of an inflection, although the machine-building branch again stands out as one from which resources are most clearly diverted (Table 5.11).

Table 5.10.

Industrial Employment Shifts1

article image

See footnote 1 to Table 5.8.

Table 5.11.

Industrial Investment Shifts1

article image

See footnote 1 to Table 5.8.

On the whole, these findings are consistent with a pattern of sectoral reallocation of labor inputs during the transition away from the old state firms in construction and industry toward new small-scale activities in agriculture and trade and toward service activities (including public services), and of scarce investment resources from heavy industry to infrastructure projects in the energy and transport and communications sectors. This would indicate that the negative productivity effect stemming from sectoral input reallocation during the early transition could be related to the direction of the reallocation process, as resources appeared to have moved to lower productivity occupations.29

Conclusions

The depth, length, and breadth of the output contractions in the Baltic and CIS countries following the breakdown of central planning have been massive, even when compared with the experience in other transition countries in central Europe. To explore the causes of these particularly sharp contractions, the analysis has put output developments in these 15 countries in a longer-run perspective, linking the initial transition years with the last two central planning decades, with a focus on the role of factor input and productivity changes.

The analysis is based upon official statistics, with some minor corrections. These data suffer from various serious weaknesses. Moreover, the quality of the series for the transition period varies considerably across countries, as some expended more effort than others to revise initial estimates and strengthen collection and reporting procedures. However, alternative output proxies suggested in the literature appear to be of limited use to overcome the deficiencies of the official statistics. Even so, the magnitude of output and input changes is such that the main qualitative insights derived from the official data are unlikely to be affected by even major measurement errors.

The Baltic and CIS countries started out with economies that had exhausted their growth potential, as reflected in the fall in total factor productivity growth to near zero in the last two decades under central planning. Capital obsolescence and economic distortions inherited from central planning contributed to further declines in total factor productivity to significantly below zero early in the transition. The output collapse was further associated with pronounced reductions in the inputs of capital and labor, as investment spending in particular was cut deeply in the early transition years. Some of the effects of transition-related factors have been temporary, however, which helps explain the distinct V-shaped pattern observed for output and productivity as the transition unfolded.

A more detailed analysis shows that this sharp fall in investment spending, to levels substantially below what would be needed for capital renewal, has reduced both the volume and the efficiency of the capital stock. At the same time, although perhaps with the exception of the Baltics, sectoral input reallocation failed to raise productivity, as labor made redundant in industry and construction has tended to move into small-scale activities in agriculture and trade and into public services, while investment in new industrial activities has been minimal. The data used here can shed no further light on the causes underlying the investment slump and the observed pattern of sectoral resource reallocation. More research drawing on additional evidence is called for on the incentives to invest, restructure, and reallocate during transition.

Appendix 5.1. Data Description30

National Accounts

The series for output, labor, and the capital stock are built up starting from the data set used by Easterly and Fischer (1994), covering 1970-89. The data have been cross-checked against series on the same variables provided by the Statistical Committee of the CIS for 1980-90, and some minor corrections introduced. Output and capital data for the pretransition period are level data and are expressed in “comparable prices, “a price concept that is considered not to fully adjust for inflation. An adjustment has been applied following Unites States Congress, Joint Economic Committee (1990), Kellogg (1990), and Noren and Kurtzweg (1993). Labor data refer to total employment numbers, unadjusted for variations in hours worked.

This amended data set is supplemented for more recent years by information drawn from national statistical yearbooks (in particular for the Baltics), from various publications of the Statistical Committee of the CIS (notably its CDROM Official Statistics of the CIS Countries, various issues of The Statistical Yearbook of the CIS, The World in Figures (1992) and its fortnightly Bulletin) and from the World Bank’s Statistical Handbook, States of the Former USSR; in a few cases, data have been obtained directly from the central statistical offices (data from these various sources may differ from the estimates used by IMF country desks and appearing in the IMF’s World Economic Outlook or in IMF Staff Country Reports).

The output, labor, and capital stock data for the transition period are constructed to ensure consistency with the 1970-89 series. Output level data for this period are extrapolated from the pretransition period using information on real growth rates. Labor data for the period 1990-97 continue to refer to total employment numbers, which in most cases are broadly consistent with the pretransition numbers (with the possible exception of employment data in the trade and services sectors in 1996-97, which in a number of countries appear to reflect reclassification of occupations). The capital stock data for the transition period are obtained using 1990-97 gross investment data and applying depreciation rates that are imputed from 1970-89 capital stock and investment data and from information in Kellogg (1990).

A number of countries have substantially revised their GDP and sectoral output series, including Russia (World Bank and the Russian Federation, Goskomstat 1995), Kazakhstan (World Bank, 1997), and Lithuania (for the aggregate output series only). In such cases, the revised series are used.

TFP computations, in addition to data on output and inputs of labor and capital, require proxies for the elasticities of output with respect to capital and labor (Table 5.12). For this purpose, observed shares of factor payments in total product are commonly used. Reported factor share data for the 15 countries in the sample are, however, incomplete and, except for the most recent years, do not reflect market-determined factor payments. Share estimates that could approximate elasticities of output with respect to capital and labor can only be obtained on the basis of detailed adjustments and additional information on wages and rentals, as in Bergson (1978). Rather than introducing such adjustments, output elasticities with regard to capital and labor are posited to equal 0.3 and 0.7 respectively, for all countries, sectors, and years.

Finally, the statistical yearbooks of the Council for Mutual Economic Assistance (CMEA) are used for pretransition data on central European comparators when no national yearbooks were available.

Alternative Measures of Output

Electricity consumption data are taken from various publications of the International Energy Agency, including Electricity in European Economies in Transition, 1994; Energy Statistics and Balances of Non-OECD Countries 1993-1994, 1996; Energy Balances of OECD Countries, 1994-1995, 1997; and Electricity Information, various issues. For some countries of central Europe and for the pretransition period, those sources were supplemented by the statistical yearbooks of the CMEA.

Table 5.12.

Yearly TFP Growth Rates

(Percent)

article image
article image

Freight and mail data are taken from the CDROM Official Statistics of the CIS Countries, and from the yearbooks published by the central statistical offices of the Baltics and of the comparator countries.

Appendix 5.2. Alternative Summary Statistics for Aggregate Output

The chaotic nature of transition and the conspicuous deprivation of the statistical agencies have led some observers to give up altogether on any attempt at deriving GDP estimates. Rather than vainly trying to collect, adjust, and aggregate economy-wide data, their line of argument goes, it is safer to use a single, relatively straight-forward proxy for overall activity. While obviously imperfect, such a proxy is more relevant than more elaborate but completely opaque and deficient measures such as the official GDP estimates, they contend. This annex discusses the merits and shortcomings of alternative summary statistics, and concludes on a very skeptical note.

Electricity Consumption

The most popular surrogate measure is electricity consumption. It is sometimes casually claimed that the long-run and even the short-run elasticity of electricity consumption with respect to real GDP is close to unity, even though cross-country empirical evidence points to numerous and significant departures from unity (IEA, 1985).31 Based on this claim, the supporters of the electricity consumption measure have proposed to have it replace official real GDP series (Dobozi and Pohl, 1995), and have used it to estimate the size of the shadow economy (Kaufman and Kaliberda, 1996; Hernández-Cata, 1997; and Johnson and others, 1997). 32, 33 Since electricity consumption was typically much more resilient than officially measured output during the great contractions (Table 5.13), those authors conclude that much of the underground activity is overlooked in the published national accounts.

Relying on overall electricity consumption as a proxy for actual GDP is a bold move, however, given the numerous reasons for divergence between the two series. On the one hand, sources of upward bias include the following:

  • a significant portion of the electricity consumed by enterprises is used for overhead purposes and hence does not fall as much as output;

  • the share of electricity in the energy mix is likely to rise over time as modern, electricity-intensive industrial processes are being adopted (although arguably this may occur mainly once output and investment are recovering rather than during the earlier phases of transition);

  • residential electricity consumption, which typically represented between one tenth and one fifth of total electricity use at the onset of transition, has displayed considerable inertia or has even increased in some countries (e.g., Belarus, Estonia, and Tajikistan), reflecting increased use of portable electric heaters as a substitute for scarce heating oil (e.g., in Moldova) and of other electrical consumer devices (notably, kitchen appliances);34

  • likewise, network losses have seen their share in total consumption increase, often even rising in absolute terms, owing to the deterioration of the infrastructure in a period where investment plummets and funds for maintenance are squeezed.

Table 5.13.

Cumulative Decline in Measured Real GDP and Electricity Consumption

(Percent decline between historical peak and trough of contraction or 1997)

article image
Sources: National statistical offices; CIS, Statistical Committee; CMEA Yearbooks; IEA publications.

During 1990-93.

Electricity consumption increased between 1990 and 1995.

On the other hand, there are also reasons for the evolution of electricity consumption to understate that of activity:

  • the composition of value-added shifts away from traditional heavy, energy-voracious industries to services and other activities that are less energy intensive;

  • the resilience of residential electricity consumption partly results from the expansion of private entrepreneurial activity under-taken from private dwellings due to the lack of business space (e.g., in Ukraine);

  • the rising share of losses partly reflects underground activity;

  • the relative price of electricity typically increases, helping reduce wasteful consumption (although in practice electricity bills are often left unpaid).

When comparing electricity consumption in peak and trough years, one should furthermore take into account weather conditions, since they influence this variable more than they affect GDP. On the whole, it is thus very hazardous to assume any simple relationship between electricity use and real GDP. In the Baltics and the CIS countries as a group, where electricity consumption fell by around 27 percent between 1990 and 1997 compared with a decline in officially reported (weighted) output by 43 percent, the observed significant shift in inputs of labor and capital toward electricity production and away from other industrial activities during the transition further weakens the arguments supporting the existence of any such simple relationship.

The case for caution is reinforced by a look at Finland, which also experienced a major recession in the early 1990s (partly for reasons related to the transition in neighboring Russia). Since the quality of the national accounts data is far superior in Finland, it is safe to assume that measured GDP closely matches actual GDP in that country. The evolution of electricity use in the course of the Finnish recession (Figure 5.1) shows that it can be a very poor proxy for real GDP, overstating it by a considerable margin during a large-scale contraction.35 Indeed, electricity use rose by almost 6 percent in the course of the three-year GDP contraction, with half of the increase accounted for by rising residential consumption.36

Figure 5.1.
Figure 5.1.

Finland: Real GDP and Overall Electricity Consumption

Sources: Statistics Finland; and International Energy Agency.

Freight and Mail

Prior to transition, freight was widely considered to be a sturdy proxy for overall economic activity, both inside the former Soviet Union and by Western analysts.37 Some observers have argued that this remained true in the 1990s, claiming that transportation data are relatively reliable, especially as regards freight.38 The bulk of freight traffic, it is argued, consists of rail and pipeline transport and as such is mostly correctly registered. Even road transportation, the contention goes, can be estimated fairly accurately, thanks to police records of vehicle numbers, and petrol production and sales.

In the case of Russia, there exists a roughly homogeneous series for overall freight, but only from 1991 and only for turnover (i.e., reflecting both volumes and distances).39 It shows a cumulative decline of 38 percent between 1991 and 1996 (which is exactly in line with the official measure of real GDP), followed by a further drop of 3.6 percent in 1997 (in contrast with a 0.8 percent increase in the official measure of real GDP). Interpreting this evidence one way or the other is hazardous, however, not least for several of the reasons already mentioned in the case of electricity consumption.

If in general there were to be a strong correlation between freight and real GDP, several factors would weaken it in the countries under consideration. Some of the new or expanding activities, such as financial services, are presumably less transportation intensive than traditional ones, while others, such as trade, may be more so. In some ways, transportation arrangements were notoriously inefficient in Soviet Union (Holt, 1993; and Strong and others, 1996), implying that part of the decline in freight may not correspond to any decline in value-added. Changes in the relative price of transportation (adjusted for nonpayments) would also have a bearing on turnover, discouraging the haulage of heavy shipments of low value over long distances. Furthermore, the intrinsic reliability of the transportation data is probably overstated by the supporters of this proxy. Even so, the case of Finland suggests that if anything, freight might be no worse a proxy for real GDP than electricity consumption, although again it declined much less than real GDP during the Finnish recession of the early 1990s (Figure 5.2).40