بررسی آزادسازی مالی: یک مدل تعادل عمومی با انتخاب شغل محدود
|کد مقاله||سال انتشار||مقاله انگلیسی||ترجمه فارسی||تعداد کلمات|
|28601||2004||39 صفحه PDF||سفارش دهید||18150 کلمه|
Publisher : Elsevier - Science Direct (الزویر - ساینس دایرکت)
Journal : Journal of Development Economics, Volume 74, Issue 2, August 2004, Pages 269–307
The objective of this paper is to assess both the aggregate growth effects and the distributional consequences of financial liberalization as observed in Thailand from 1976 to 1996. A general equilibrium occupational choice model with two sectors, one without intermediation and the other with borrowing and lending is taken to Thai data. Key parameters of the production technology and the distribution of entrepreneurial talent are estimated by maximizing the likelihood of transition into business given initial wealth as observed in two distinct datasets. Other parameters of the model are calibrated to try to match the two decades of growth as well as observed changes in inequality, labor share, savings and the number of entrepreneurs. Without an expansion in the size of the intermediated sector, Thailand would have evolved very differently, namely, with a drastically lower growth rate, high residual subsistence sector, non-increasing wages but lower inequality. The financial liberalization brings welfare gains and losses to different subsets of the population. Primary winners are talented would-be entrepreneurs who lack credit and cannot otherwise go into business (or invest little capital). Mean gains for these winners range from 17% to 34% of observed, overall average household income. But liberalization also induces greater demand by entrepreneurs for workers resulting in increases in the wage and lower profits of relatively rich entrepreneurs, of the same order of magnitude as the observed overall average income of firm owners. Foreign capital has no significant impact on growth or the distribution of observed income.
The objective of the paper is to assess the aggregate, growth effects and the distributional consequences of financial liberalization and globalization. There has been some debate in the literature about the benefits and potential costs of financial sector reforms. The micro credit movement has pushed for tiered lending, or linkages from formal financial intermediaries to small joint liability or community groups. But a major concern with general structural reforms is the idea that benefits will not trickle down, that the poor will be neglected, and that inequality will increase. Similarly, globalization and capital inflows are often claimed to be associated with growth although the effect of growth on poverty is still a much debated topic.1 Needless to say, we do not study here all possible forms of liberalization. Rather, we focus on reforms that increase outreach on the extensive domestic margin, for example, less restricted licensing requirements for financial institutions (both foreign and domestic), the reduction of excess capitalization requirements, and enhanced ability to open new branches. We capture these reforms, albeit crudely in the model, thinking of them as domestic reforms that allow deposit mobilization and access to credit at market clearing interest rates for a segment of the population that otherwise would have neither formal sector savings nor credit. We take this methodology to Thailand from l976 to l996.2 Thailand is a good country to study for a number of reasons. First, Thailand is often portrayed as an example of an emerging market, with high income growth and increasing inequality. The GDP growth from 1981 to 1995 was 8% per year, and the Gini measure of inequality increased from 0.42 in 1976 to 0.50 in 1996. Second, Jeong (1999) documents in his study of the sources of growth in Thailand, 1976–1996, that access to intermediation narrowly defined accounts for 20% of the growth in per capita income while occupation shifts alone account for 21%. While the fraction of non-farm entrepreneurs does not grow much, the income differential of non-farm entrepreneurs to wage earners is large and thus small shifts in the population create relatively large income changes. In fact, the occupational shift may have been financed by credit. Also related, Jeong finds that 32% of changes in inequality between 1976 and 1996 are due to changes in income differentials across occupations. There is evidence that Thailand had a relatively restrictive credit system but also liberalized during this period. Officially, interest rates ceilings and lending restrictions were progressively removed starting in 1989.3 The data do seem to suggest a rather substantial increase in the number of households with access to formal intermediaries although this expansion (which we call a liberalization) begins 2 years earlier, in 1987. Finally, Thailand experienced a relatively large increase in capital inflows from the late 1980s to the mid 1990s. Our starting point is a relatively simple but general equilibrium model with credit constraints. Specifically, we pick from the literature and extend the Lloyd-Ellis and Bernhardt (2000) model (LEB for short) that features wealth-constrained entry into business and wealth-constrained investment for entrepreneurs. For our purposes, this model has several advantages. It allows for ex ante variation in ability. It allows for a variety of occupational structures, i.e. firms of various sizes, e.g., with and without labor, and at various levels of capitalization. It has a general (approximated) production technology, one which allows labor share to vary. In addition, the household occupational choice has a closed form solution that can easily be estimated. Finally, it features a dual economy development model which has antecedents going back to Lewis (1954) and Fei and Ranis (1963), and thus it captures several widely observed aspects of the development process: industrialization with persistent income differentials, a slow decline in the subsistence sector, and an eventual increase in wages, all contributing to growth with changing inequality. Our extension of the LEB model has two sectors, one without intermediation and the other allowing borrowing and lending at a market clearing interest rate. The intermediated sector is allowed to expand exogenously at the observed rate in the Thai data, given initial participation and the initial observed distribution of wealth. Of course in other contexts and for many questions one would like financial deepening to be endogenous.4 But here the exogeneity of financial deepening has a peculiar, distinct advantage because we can vary it as we like, either to mimic the Thai data with its accelerated upturns in the late 1980s and early 1990s, or keep it flat providing a counterfactual experiment. We can thus gauge the consequences of these various experiments and compare among them. In short, we can do general equilibrium policy analysis following the seminal work of Heckman et al. (1998), despite endogenous prices and an evolving endogenous distribution of wealth in a model where preferences do not aggregate. We use the explicit structure of the model as given in the occupation choice and investment decision of households to estimate certain parameters of the model. Key parameters of the production technology used by firms and the distribution of entrepreneurial talent in the population are chosen to maximize the likelihood as predicted by the model of the transition into business given initial wealth. This is done with two distinct microeconomic datasets, one a series of nationally representative household surveys (SES), and the other gathered under a project directed by one of the authors, with more reliable estimates of wealth, the timing of occupation transitions, and the use of formal and informal credit. Not all parameters of the model can be estimated via maximum likelihood. The savings rate, the differential in the cost of living, and the exogenous technical progress in the subsistence sector are calibrated to try to match the two decades of Thai growth and observed changes in inequality, labor share, savings and the number of entrepreneurs. As mentioned before, this structural, estimated version of the Thai economy can then be compared to what would have happened if there had been no expansion in the size of the intermediated sector. Without liberation, at estimated parameter values from both datasets, the model predicts a dramatically lower growth rate, high residual subsistence sector, non-increasing wages, and, granted, lower and decreasing inequality. Thus financial liberalization appears to be the engine of growth it is sometimes claimed to be, at least in the context of Thailand. However, growth and liberalization do have uneven consequences, as the critics insist. The distribution of welfare gains and losses in these experiments is not at all uniform, as there are various effects depending on wealth and talent: with liberalization, savings earn interest, although this tends to benefit the wealthy most. On the other hand, credit is available to facilitate occupation shifts and to finance setup costs and investment. Quantitatively, there is a striking conclusion. The primary winners from financial liberalization are talented but low wealth would-be entrepreneurs who without credit cannot go into business at all or entrepreneurs with very little capital. Mean gains from the winners range from 60,000 to 80,000 baht, and the modal gains from 6,000 to 25,000 baht, depending on the dataset used and the calendar year. To normalize and give more meaning to these numbers, the modal gains ranges from 17% to 34% of the observed, overall average of Thai household income. But there are also losers. Liberalization induces an increase in wages in latter years, and while this benefits workers, ceteris paribus, it hurts entrepreneurs as they face a higher wage bill. The estimated welfare loss in both datasets is approximately 115,000 baht. This is a large number, roughly the same order of magnitude as the observed average income of firm owners overall. This fact suggests a plausible political economy rational for (observed) financial sector repressions. Finally, we use the estimated structure of the model to conduct two robustness checks. First, we open up the economy to the observed foreign capital inflows. These contribute to increasing growth, increasing inequality, and an increasing number of entrepreneurs, but only slightly, since otherwise the macro and distributional consequences are quite similar to those of the closed economy with liberalization. Indeed, if we change the expansion to grow linearly rather than as observed in the data, the model cannot replicate the high Thai growth rates in the late 1980s and early 1990s, despite apparently large capital inflows at that time. Second, we allow informal credit in the sector without formal intermediation to see if our characterization of the dual economy with its no-credit sector is too extreme. We find that at the estimated parameters it is not. Changes attributed to access to informal credit are negligible. The rest of the paper is organized as follows. In Section 2, we describe the LEB model in greater detail. In Section 3, we describe the core of the model as given in an occupational choice map. In Section 4, we discuss the possibility of introducing a credit liberalization. In Section 5, we turn to the maximum likelihood estimation of seven of the ten parameters of the model from micro data, whereas Section 6 focuses on the calibration exercise used to pin down the last three parameters, matching, as explained, more macro, aggregate data. Section 7 reports the simulations at the estimated and calibrated values for each dataset. Section 8 performs a sensitivity analysis of the model around the estimated and the calibrated parameters. Section 9 delivers various measures of the welfare gains and losses associated with the liberalization. Section 10 introduces international capital inflows and informal credit to the model. Finally, Section 11 concludes.
نتیجه گیری انگلیسی
From the welfare numbers presented, there seems to be a lot at stake in credit liberalizations. Even by our most conservative estimates there is a group of low wealth talented households who have much to gain, period by period, in income and wealth. On the other hand, the estimates reveal a group of entrepreneurs who have much to lose, period by period, in income and wealth, particularly if one takes into account the growth in wages. We do not push here any particular any number as the most compelling, because the numbers do vary and do depend on the dataset used. Indeed, the larger point is that welfare gains and losses are sensitive to the presumed, estimated micro underpinnings of the economy. If there were more substantial intermediation, then variations associated with further liberalization would matter less. Indeed, if there were more substantial intermediation, then the impact on dynamics of (endogenous) changes in the wealth distribution would matter less, as in Krusell and Smith (1998). But the micro data reject such presumed underpinnings, making welfare gains, potential losses, and the dynamic aspect of liberalization more substantial. Still, the surprisingly large order of magnitude of these gains and losses suggests the need for further refinements along a number of dimensions, to see if the magnitude survives somewhat more realistic specifications. One refinement has to do with labor and the labor market. The labor of the model here is uniform with respect to productivity, that is, every laborer earns the same wage. This flies in the face of much empirical work mapping wage differentials to skills differentials and acquired human capital. More generally, earnings inequality contributes to overall inequality, and this might be salient, as in the work of Ricardo Paes de Barros in Brazil, for example. Jeong and Townsend (2000) document the success and failure of the LEB model in explaining inequality movements in Thailand, comparing it to an extended version of Greenwood and Jovanovic (1990). Though the LEB model at the maximum likelihood parameter estimates does surprisingly well, it is clear that wage differentiation is needed in models and in model based empirical work. In the model of occupation choice of Evans and Jovanovic (1989), for example, unobserved heterogeneous skills influence both wage earnings and the profits from entrepreneurship. This might suggest that entrepreneurs worry less about increased wages as they potentially exit to become part of a skilled work force. Second, one could endogenize access to credit as in the Greenwood and Jovanovic (1990) model with transactions costs. This would slow down the growth of financial infrastructure and would rationalize some of the limited participation that we see but there would be no Pareto-improving policy intervention. However, Townsend and Ueda (2001) in an extended version of the Greenwood and Jovanovic model draw the conclusion that restrictive financial sector policies may have nevertheless slowed down entry into the financial sector below the endogenous rate. Thus financial sector liberalization, allowing intermediation at its otherwise endogenous value, is also associated with welfare gains, as in this paper. Although occupation choice models should allow more endogenous financial sector participation, they illustrate well as they stand the fact that there may be welfare losses for some sectors of the population, not just gains, to liberalization. This offers a political economy rational for the apparently restrictive policies that we observed in Thailand. Third, the imagined industrial organization of the Thai economy is relatively simple. In the model here fixed setup costs are allowed to vary across potential entrepreneurs, as if drawn from a quadratic cumulative distribution, but the quadratic production function mapping labor and capital into output is uniform across potential entrepreneurs. This delivers the occupation partition diagram and variation over time in the size distribution of firms, a function of the wage and the endogenous distribution of wealth. Indeed, an industrial organization literature starting with Lucas (1978) begins the same way, at least in spirit. He postulates an underlying distribution of personal managerial talent and then studies the division of persons into managers and employees and the allocation of productive inputs across managers. This has implications for secular changes in average firm size. This point is revisited by Gollin (1999). Here the distribution of size and profits among firms is driven by self-financing, an endogenous and evolving distribution of wealth, and differential access to credit. That is, high set up costs and limited credit can limit the use of real physical capital in the standard part of the production function or can impede entry entirely. Likewise, some portion of end-of period profits is passed on to subsequent time periods if not subsequent generations. This could be an explanation for some of the serial correlation in size, profits, and employment that is seen in actual data, even when shocks in the form of set up costs are independent and identically distributed over time and households. There would be implications for the cross-sectional dispersion in growth rates. Indeed, it seems that in Thailand larger firms, and those with financial access, may have grown faster than smaller ones as the credit market expanded in the late 1980s. Recent theoretical work is beginning to readdress earlier the supposed facts of firm growth and survival in the context of endogenous limited financial contracts. See Albuquerque and Hopenhayn (2001) and Cooley et al. (2000). More generally both the industrial organization and credit market literature need to be brought together. Existing empirical work has documented relationships between investment and the balance sheet, for example, but much of this work is somewhat atheoretic, documenting that the world is not neoclassical but leaving us wondering what the impediments to trade really are. The general equilibrium models of Banerjee and Newman (1993), Piketty (1997) and Aghion and Bolton (1997) take different stands on those underpinnings but collectively make the point that growth and inequality can be related to imperfect credit markets. That of course was our starting point here. Indeed, in related work Paulson and Townsend (2001) use the Townsend-Thai data to estimate via maximum likelihood methods not only the LEB model featured here, but also collateral based lending as in model of Evans and Jovanovic (1989) (EJ for short), for example, and also incentive-based lending as in the mechanism design literature of Aghion and Bolton (1997) and Lehnert (1998) (ABL for short). Observed relationships of entrepreneurship, investment, and access to credit as functions of wealth and talent suggest that the ABL model fits the micro Thai data best, but the EJ model fits well for those with relatively low levels of wealth and those in the Northeast, while LEB, the model here, is a close contender. This suggests that a calculation of the welfare gains and losses to financial intermediation based on these other models would be worthwhile, though the average and modal estimates here should not be rejected out of hand. It does seem plausible, however, that the dramatic gains near the wealth equal set-up costs or 45° line would be vulnerable to alternative specifications. The growth and inequality literature relying on each of these underpinnings presupposes, as in the LEB model here, either an overlapping generations model with a bequest motive or a simplistic, myopic solution to the household savings problem. More needs to be done to make the models dynamic. Coupling households with firms and modelling the firms' inter-temporal decision problems will require more work, but again, given the preliminary results here, that work would appear to be warranted.