STUDYING INVESTMENT PATTERNS IN RUSSIAN AGRICULTURE
RAUSHAN BOKUSHEVA, IRINA BEZLEPKINA AND ALEXANDER KUPAVYCH Leibniz Institute of Agricultural Development in Central and Eastern Europe (Germany)
[email protected] Business Economics, Wageningen University (The Netherlands)
[email protected]. Agricultural Economics, Lomonosov Moscow State University (Russia)
[email protected]
Paper prepared for presentation at the joint IAAE- 104th EAAE Seminar Agricultural Economics and Transition: „What was expected, what we observed, the lessons learned."
Corvinus University of Budapest (CUB) Budapest, Hungary. September 6-8, 2007
Copyright 2007 by Raushan Bokusheva, Irina Bezlepkina and Alexander Kupavych. All rights reserved. Readers may make verbatim copies of this document for non-commercial purposes by any means, provided that this copyright notice appears on all such copies.
1
ABSTRACT The study analyzes the investment behavior of Russian farms during the period of economic stabilization that followed the financial crisis of 1998. While the literature primarily uses accelerator and adjustment costs models to describe farms’ investment behavior in the context of transition countries, our study reveals that error-correction formulation of the investment equation is more suited to analyzing Russian farms’ investment patterns. In addition, by distinguishing between different categories of farms, we reveal significant differences in the investment patterns, not only across farms both with and without access to external finance, but also across mature and newly-established farms, owner- and employee-managed farms, and farms with high and low managerial competence, respectively. Keywords: dynamic investment models, error-correction investment equation, Russia.
2
1
INTRODUCTION
Empirical evidence from transition countries shows that farms’ investment activity seriously receded during the initial reform period. Indeed, the decline in investments at the beginning of the transition process was caused primarily by radical changes in agricultural product and factor markets. Price liberalization and the related deterioration of the terms of trade seriously affected farms’ output and consequently reduced demand for machinery and other production inputs. Initial production cuts, and thus, reduced demand for investment, were inevitable for the economic recovery of transition countries, and are in line with the Schumpeterian concept of creative destruction (SCHUMPETER 1952). Moreover, it was to be expected that in the early economic recovery period producers’ efforts would be targeted at reducing inherited inefficiencies of input use rather than expanding production inputs, be they capital or labor (HAVRYLYSHYN et al., 1998). At the same time, farms’ investment activity was strongly affected by high uncertainties relating to demand for farms’ output. During the socialist period, farms were not concerned about marketing their products, but in the initial transition phase they were forced to search for new markets for their products – markets which barely existed in most transition countries during the early 1990s. The absence of an appropriate market infrastructure at the beginning of transformation caused high uncertainties concerning demand and prices for agricultural products, which not only seriously affected farms’ investment, but endangered the very existence of many farm businesses. This made farms uncertain about the development of output markets and caused a high discount rate on their investment. However, as transformation progresses, farms are reconsidering their investment decisions. In addition, the collapse of state financial and supply mechanisms accompanied by subsidy cuts lessened agricultural enterprises’ access to external finance and input markets. Prior to reform, investment funds were allocated to producers according to credit plans approved by national and regional authorities; in accordance with these plans regional supply services delivered machinery and other inputs to agricultural enterprises. Though farms in many transition countries could benefit from soft budget constraints (SBCs) at the beginning of transition, the introduction and enhancement of financial reforms during transition has forced agricultural producers to adjust their financial policies to the prevailing financial market rules. In the course of reforms, the tough reality has become evident; the low profitability of agricultural production critically limited access to external financing. Reductions in investment levels during the early transition period have inevitably caused the depletion of farms’ capital stock. Therefore, it is obvious that farms’ investment rates will exceed actual capital depreciation rates if they restart investment after a long period of under investment. Capital widening will be also necessary to incorporate technological change. Based on these facts we can conclude that there are several additional factors that have determined the investment behavior of farms during transition compared with those investment models describing firms’ investment behavior in the established market economies. At the same time, as will be shown in the paper, empirical studies on farm investment behavior in transition countries (LATRUFFE 2004; ZINYCH and ODENING 2007) extensively apply general model specifications without adjusting them to transition circumstances, thereby a priori exterminating the effect of transition on firm investment pattern. This calls into question both the theoretical consistency as well as the empirical validity of the results obtained in these studies. In this paper we discuss and test alternative model specifications regarding their suitability for modeling the investment behavior of farms 3
in transition. In analyzing how investment behavior evolved in Russian agriculture during the period of economic stabilization that followed the financial crisis of 1998, we first employ a traditional formulation of the investment equation for optimal capital accumulation in the presence of capital adjustment costs. To evaluate the effect of capital market imperfections, we discriminate between farms operating under two financial regimes: credit constrained and credit unconstrained. Furthermore, we classify the sample of farms into groups according to several technological and organizational characteristics to reveal factors that, in addition to access to external finance, significantly affect investment behavior of Russian farms. Finally, we test whether the investment behavior of Russian farms may be explained by an errorcorrection formulation of a dynamic investment model. The paper is organized as follows. We start Section 2 by providing an overview of theoretical models available for studying investment behavior, and then reviewing the empirical studies of investment patterns in transition countries. The section concludes with motivation for the choice of empirical model specifications. The data and estimation techniques are presented in Section 3. Section 4 presents the research findings, and conclusions can be found in Section 5. 2
MODELING INVESTMENT BEHAVIOR
2.1
Investment model formulations
When investigating firm investment behavior, empirical studies usually apply various formulations of the investment equation. In this section we begin by presenting empirical specifications of those most widely used, but also some recent investment model formulations. A presentation and discussion of the theoretical models used to derive the empirical specifications discussed here is beyond our study objective. Interested readers are referred to the theoretical model descriptions in the literature (ABEL and BLANCHARD 1986, FAZZARI et al., 1988, BOND and MEGHIR 1994, BOND et al., 1997, BLOOM et al., 2005). We start by presenting the common and augmented specifications accelerator and adjustments costs models before presenting an error-correction investment model specification with and without the uncertainty effect. Accelerator model The main idea of the accelerator approach is that the demand for firm investment depends on changes in the firm output. An increasing volume of production demands more capital and thus explains firm investment. Accordingly, the basic accelerator investment equation is defined as follows:
I ∆S = α 0 + α1 + d t + α t + vit K it K it (1)
,
where Iit stands for a gross investment in fixed assets of firm i over period t, Kit is a capital stock (fixed asset) of firm i at the beginning of the period t, ∆Sit = Sit - Sit-1 is a change in sales of firm i in period t relative to period t-1 , dt is a vector of time dummies, αi is a vector of firm dummies and v it is the error term (ABEL and BLANCHARD 1986). Assuming the presence of imperfections in financial markets FAZZARI et al., (1988) formulate the augmented accelerator investment equation. They introduce the cash flow ratio to capital CF as an additional explanatory variable, which enables testing the sensitivity of firm K investment to the availability of internal funds and thus financial market imperfections: 4
I ∆S CF = α 0 + α1 +α2 + d t + α i + vit K it K it K it −1
(2).
Adjustment costs model
In the adjustment costs framework, firms are assumed to maximize the expected net present value of profits over the infinite time horizon, subject to capital adjustment costs. The firm investment behavior is modeled as a dynamic process which describes capital accumulation rates in individual periods. Thus, investment in a particular year is defined not only by sales growth and firm liquidity (in the case of financial market imperfections), but also by firm investment in the previous periods (BOND and MEGHIR, 1994): 2
I I I CF S = β 0 + β1 + β 2 + β 3 + β 4 + d t + α i + vit K it K it −1 K it −1 K it −1 K it −1
(3).
From the theoretical model, it is expected that the coefficient of the lagged investment term β1 is positive and greater than one if the firm’s real discount rate is positive. The coefficient sign of the investment term squared is expected to be negative and greater than one in absolute value. The sign of the coefficient of cash flow should be negative or not significant under the assumption that the firm can raise as much money as it desires at a given cost. A significantly positive cash flow coefficient is primarily interpreted as an indicator of financial constraints; though BOND and MEGHIR (1994) mention that it also may reflect a marginal profitability of investment. A positive sign on the sales ratio to capital implies the presence of imperfect competition in the output market. However, it can also point to the presence of non-constant returns to scale. The adjustment-cost model presented in (3) can be extended by introducing an additional 2
B financial variable : K 2
2
I I I CF S B = β 0 + β1 + β 2 + β 3 + β 4 + β 5 + d t + α i + vit K it K it −1 K it −1 K it −1 K it −1 K it (4), where Bt represents the farm’s borrowing during year t. The specification in (4) allows testing for non-separability between investment and borrowing decisions (BOND and MEGHIR, 1994). In empirical application, the firm’s debt at the beginning of period t is used to proxy the firm’s borrowing. Thus, a negative sign of the coefficient on 1 this term is interpreted as an indicator for tax-bankruptcy costs . If the specifications of the investment equation in (3) and (4) do not provide an adequate explanation of the firm’s investment, empirical analyses distinguish between firms in two different financial regimes. The effect of financial variables on the firm investment is then studied by applying a specification which enables model parameters to differ across firms
1
According to the tax-borrowing costs hypothesis, retained earnings may be preferred to new share issues if the tax system treats capital gains more favourably than income from dividends. In the presence of bankruptcy costs, debt financing becomes increasingly expensive because it raises the probability of bankruptcy (BOND and MEGHIR, 1994).
5
separated in two groups according their financial status. This is possible by defining a dummy variable (Zit) which is equal to zero for financially unconstrained firms and is one for constrained firms (RIZOV 2004). This leads to the following specification of the investment equation: 2
I I I CF S = β 0 + β1 + β 2 + β 3 + β4 K it K it −1 K it −1 K it −1 K it −1 2
I I CF S + β 6 Z it + β 7 Z it + β 8 Z it + β 9 Z it + d t + α i + vit K it −1 K it −1 K it −1 K it −1
,
(5)
where the parameters β1to β4 are related to the group of unconstrained firms and the parameters β6 to β9 describe differences in the effects of the individual explanatory variables across two financial regimes. Error-correction model
BOND et al., (1997) introduce an error-correction formulation of the investment model. The error correction model is based on the assumption that sales and capital are proportional in the long run, while in the short run, the dynamics of the relationship between these two variables may diverge from the optimal path. In this formulation of the investment equation, the relationship between desired and actual capital stock is described as an autoregressivedistributed lag of length two:
k it = ϕ 0 + γ 1 k i ,t −1 + γ 2 k i ,t − 2 + ϕ1 s it + ϕ 2 s i ,t −1 + ϕ 3 s i ,t − 2 + d t + α i + vit , (6) where kit is the logarithm of the fixed capital value Kit for the firm i at the end of the year t, sit is the logarithm of sales Sit for the firm i in the year t. To obtain an error-corrected specification, the authors subtract ki,t-1 from both sides of the equation (6) and rewrite it as follows: ∆k it = ϕ 0 + (γ 1 − 1)∆k i ,t −1 + ϕ1 ∆s it + (ϕ1 + ϕ 2 )∆si ,t −1 + (γ 2 + γ 2 − 1)(k i ,t − 2 − si ,t − 2 ) + (ϕ1 + ϕ 2 + ϕ 3 + γ 2 + γ 2 − 1) s i ,t − 2 + d t + α i + vit
(7).
In this form of the equation, the growth rate of capital stock is a function of both growth rates and levels of information. The first three terms of the equation (3) capture the short run dynamics. The coefficients’ estimates for the last two terms can be used for testing errorcorrecting behavior and constant returns to scale in the long run, respectively. It is expected that the coefficient of the error-correcting term p = γ 2 + γ 2 − 1 shall be negative, indicating that investment is higher when capital stock is lower than its optimal level, and conversely, that investment is lower when capital is over its optimal level. The scale coefficient λ = β + p = (ϕ1 + ϕ 2 + ϕ 3 + γ 2 + γ 2 − 1) (where ϕ = ϕ1 + ϕ 2 + ϕ 3 ) is expected to be not significantly different from zero, which implies that the long-run elasticity of capital to sales is unity. To capture effects which are associated with financial constraints of the firm, equation (7) is Pit augmented with the current and lagged ratios of profits to the fixed capital value , K i ,t −1
6
where Pit is profit of the firm i in the year t. Finally, the investment ratio
I it is employed to K i ,t −1
proxy the net growth in capital ∆k it . The equation to estimate has the following form:
I I it = ϕ 0 + (γ 1 − 1) i ,t −1 + ϕ1 ∆s it + (ϕ1 + ϕ 2 )∆si ,t −1 + (γ 2 + γ 2 − 1)(k i ,t − 2 − si ,t −2 ) K i ,t −1 K i ,t − 2 + (ϕ1 + ϕ 2 + ϕ 3 + γ 2 + γ 2 − 1) s i ,t − 2
Pi ,t −1 Pi ,t −2 P + ϕ 4 it + ϕ 5 + ϕ6 + d t + α i + vit K i ,t −1 K i ,t − 2 K i ,t − 3
(8).
Non-linear model of investment under uncertainty
BLOOM et al., (2005) adjust the error-correction model to analyze the effect of demand shocks and uncertainty on firm investment. They argue that higher levels of uncertainty increase the real options values associated with investment and dis-investment, and thus make firms more cautious in responding to changes in their market environment. The presence of irreversibility and uncertainty causes non-linear dynamics in firms’ investment behavior with an increasing marginal investment response to larger demand shocks (BLOOM et al., 2005, p. 2). Accordingly, uncertainty is introduced in the following way:
I it P = η 0 + η1∆s it + η 2 ∆sit2 + η 3 (σ it ∆sit ) + η 4 ∆σ it + η 5 ∆ it K i ,t −1 K i ,t −1 + η 6 (k i ,t −1 − s i ,t −1 ) + η 7σ i ,t −1 + η 8
Pi ,t −1 K i ,t − 2
,
(9)
+ d t + α i + vit
where σit and ∆σit represent uncertainty and a change in uncertainty of the firm i in the year t, respectively. From the theoretical formulation of the model, it is expected that η1>0 and η3