آزمون برای صاف کردن مالیات در یک مدل تعادل عمومی رشد

This paper constructs, estimates and tests a general equilibrium model of endogenous growth and optimal fiscal policy. Income tax revenues finance government consumption and production services, with the latter generating long-term endogenous growth. A key result from this model is that benevolent policymakers find it optimal to keep the income tax rate constant over time. Despite its popularity amongst theorists, there have been no formal econometric tests of this type of general equilibrium models. We find that data from 22 OECD economies uniformly reject the model over the period 1960–1996.

Over the past decade, general equilibrium models of economic growth have increasingly been employed to study the role of fiscal policy in the growth process. The main idea (see Barro, 1990) is that some government-provided services enhance the productivity of private firms. For this reason, at the aggregate level, there are no diminishing returns, and hence the economy is capable of long-term (endogenous) growth. In this framework, given that government services are financed by distortionary taxes, it is particularly important to identify the optimal level of government expenditures and the associated optimal tax rate. In this context, for particular specifications of technology and preferences, Barro (1990) showed that the optimal income tax rate is constant.1 The intuition is well-known: since tax policy is distorting, the fiscal authorities find it optimal to allocate this policy over time to avoid further intertemporal distortions. Basically, this means that the tax rate should change only if there are unanticipated shocks, i.e. tax rates should not be state-contingent. This is a form of the classic tax-smoothing result.2 There have been numerous papers that build upon Barro's (1990) setup to address various theoretical issues in economic growth (see for example Barro and Sala-i-Martin, 1992, Alesina and Rodrik, 1994, Benhabib and Velasco, 1996, Glomm and Ravikumar, 1994, Glomm and Ravikumar, 1997 and Devereux and Wen, 1998). However, surprisingly, and despite its influence and popularity amongst theorists, there has been no formal testing of this class of general equilibrium models. This paper solves, and formally tests, a relatively general version of Barro's heavily cited general equilibrium model of endogenous growth, public (production and consumption) services and optimal policy, in which policymakers find it optimal to keep the tax rate constant over time. The paper is organized as follows. In Section 2, we set up an endogenous growth model, in which a benevolent government3 chooses a path of distorting income tax rates to finance the provision of public services. We enrich the basic setup by making the reasonable assumption that the government uses the collected tax revenues to finance both public production services (which provide production externalities to firms) and public consumption services (which provide direct utility to households). The government acts as a Stackelberg leader vis-à-vis households and firms. We solve for Markov policy strategies, and hence Markov-perfect general equilibria, in which optimal policy is time consistent. We obtain an exact closed-form general-equilibrium solution, which consists of behavioral relations for private consumption, private capital, government production services, government consumption services and the income tax rate. Our results show that it is optimal for policymakers to keep the tax rate—as well as, the output shares of private consumption, private capital, government production services and government consumption services—constant over time. Our closed-form analytical solution enables us, in Section 3, to test the cross-equation restriction(s) implied by the interaction between optimizing private agents and optimizing fiscal authorities. Our empirical testing is conducted by using annual data from all OECD economies, where full data sets are available, over the period 1960–1996. We find that the data resoundingly reject the empirical validity of the model.4 It therefore appears that the class of general equilibrium models based on Barro's (1990) seminal paper, in which it is optimal for policymakers to keep the tax rate constant over time so as to smooth out its distorting effects on growth, are not supported by the data. Thus, previous testing based on partial equilibrium models has over-favored the tax-smoothing hypothesis of policymaking. Our findings are consistent with the findings of Jones et al. (1993) and Chari et al. (1994) amongst others for the US. Finally, in Section 4, we discuss our conclusions and related research. How is our work related to the relevant literature? There are at least three strands. First, there is an empirical literature that uses regression analysis to investigate how growth is affected by the structure of public expenditure (for example government consumption vs. government production services) and the associated public finance decisions (see for example Devarajan et al., 1996 and Kneller et al., 1999). However, in these papers there is no testing of theoretical cross-equation restrictions and also the government's actions are treated as exogenous. Second, it is well-known that Real Business Cycle (RBC) models have also incorporated fiscal policy in general equilibrium setups (see for example Christiano and Eichenbaum, 1992, Baxter and King, 1993, Jones et al., 1993, Chari et al., 1994, McGrattan, 1994, Stokey and Rebelo, 1995 and Ambler and Paquet, 1996). These models are assessed with the use of calibration techniques to check their ability to match the observed moments of the data. Here, by contrast, we use traditional econometric techniques to estimate and test the model. What we therefore do is complementary to the RBC literature and has the advantage that we maintain a clear link between theory and econometrics.5 Third, there has much empirical interest in tax smoothing in the context of partial equilibrium models (see for example Sargent and Velde, 1999 and Serletis and Schorn, 1999).6 However, there has been no formal econometric testing of the general equilibrium renditions of these models, and in particular the cross-equation restriction(s) implied by the interaction between optimizing private agents and policymakers.7

In this paper, we have constructed and tested a general equilibrium model of endogenous growth, public services and optimal policy, in which benevolent policymakers find it optimal to keep the tax rate constant over time. Despite its popularity and influence among theorists, data from 22 OECD countries uniformly reject the empirical viability of the class of models based on Barro's (1990) seminal paper. Thus, in contrast to the findings from the partial equilibrium studies cited above, our results suggest that the policy recipe to keep the tax rate flat over time (so as to smooth out its distorting effects on growth) does not hold in general equilibrium settings in which private agents and policymakers endogenously react to each other. As with almost any empirical exercise in hypothesis testing, rejection of a model implies that some, or all, of the assumptions required in the theoretical model are somehow at odds with the data.25 In the class of models examined in this paper, the major assumptions include: (i) logarithmic consumer preferences; (ii) Cobb–Douglas technology at the firm's level; (iii) fully rational and long-sighted agents (private agents and policymakers) who all share the same objectives; and (iv) benevolent governments can achieve their objectives. In a general equilibrium context, it is extremely difficult to disentangle these assumptions so as to formally assess which one, or combination of them, is causing rejection of the model's cross-equation restrictions. Nevertheless we can speculate. For example, from a theoretical perspective, as we have discussed at the end of Section 2, as long as one focuses on time consistent optimal policies, the tax smoothing result does not change if different functional forms for preferences and technology are employed. Furthermore, from an empirical perspective, it is important to stress that alternative model specifications, which would add more structure of this sort and hence unavoidably imply more (both in number and algebraic complexity) cross-equation theoretical restrictions,26 seem quite unlikely to find more support by the data. Accordingly, we suggest that more data friendly model specifications might be found by allowing for “simplicity” in the form of for example bounded rationality, short-sighted behavior, or policymakers whose behavior is less complex and maybe less efficient than that of a benevolent Stackelberg leader (see for example Simon, 1954 and Gregory, 1989). Bounded rationality (for example rule-of-thumb behavior) is based on results from experimental economics which demonstrate that the main motive for deviating from rationality is indeed an attempt to simplify decision problems (see for example Ellison and Fudenberg, 1993, Rubinstein, 1998 and Lettau and Uhlig, 1999). Short-sighted behavior on the part of private agents and/or policymakers works in the same direction when it implies a less complicated structure than that of the general far-sighted case.27 Finally, it is widely believed that policymakers have their own political agendas, which might be systematically different from those of a benevolent government. Policymakers may try to please a subset of voters to remain in office (see for example Drazen, 2000, chap. 7 for a recent survey, which also emphasizes that partisan and electoral motives cannot be separated). Moreover policymakers may try to extract rents—associated with office-holding per se—for themselves and other interest groups at the voters' expense (see for example Buchanan et al., 1980, Mueller, 1989, Drazen, 2000 and Persson and Tabellini, 1999a)28. This is an important and still growing literature (see the surveys in Drazen, 2000 and Persson and Tabellini, 1999b), but the results—from formal econometric testing of the micro-political foundations of the various models—are still far from conclusive. Given the above considerations, we suggest that more data friendly model specifications, which imply less cross-equation theoretical restrictions, might be found by allowing for bounded rationality, short-sighted behaviour, or non-benevolent politicians. General equilibrium models incorporating such assumptions (see Malley and Philippopoulos, 200129 for a formal attempt in this direction) will hopefully prove to be more useful for analysing the intertemporal interaction between the private economy and fiscal authorities in the context of growth.