Fiscal Policy and Economic Development

Contributor Notes

Authors’ E-Mail Addresses: amourmouras@imf.org; prangaza@iupui.edu

This paper offers possible explanations for three generally observed facts about fiscal policy and development: (F1) The relative size of government increases as an economy develops, (F2) The rise in government and taxation are associated with rising or constant economic growth rates, and (F3) Today's developing countries have larger government sectors than did today's developed countries at similar stages of development. The explanations for these facts are based on the structural transformation from traditional (mostly agricultural) to modern (industrial and post-industrial) production, rising public infrastructure investment, and less representative governments in many of today's developing economies.

Abstract

This paper offers possible explanations for three generally observed facts about fiscal policy and development: (F1) The relative size of government increases as an economy develops, (F2) The rise in government and taxation are associated with rising or constant economic growth rates, and (F3) Today's developing countries have larger government sectors than did today's developed countries at similar stages of development. The explanations for these facts are based on the structural transformation from traditional (mostly agricultural) to modern (industrial and post-industrial) production, rising public infrastructure investment, and less representative governments in many of today's developing economies.

I. Introduction

There are several interesting empirical connections between economic development and fiscal policy. First, the relative size of government trends upward as an economy develops—a phenomenon that is known as Wagner’s Law www.logo. This feature is exhibited in Figure 1, which gives the historical ratio of government purchases to GDP, averaged over 11 currently developed countries.2

Figure 1.
Figure 1.

Selected Industrial Countries: Government Purchases (share of GDP)

Citation: IMF Working Papers 2008, 155; 10.5089/9781451870138.001.A001

Source: Tanzi and Schuknecht (2000, Table II.1). The plot is the average government purchase share for 11 currently developed countries (Australia, Canada, France, Germany, Japan, Netherlands, Norway, Spain, Sweden, United Kingdom and the United States).

Second, economic growth rates have typically risen, or at least remained constant, in the face of strong upward trends in taxation and government expenditures. Figure 2 presents the growth rates in productivity per worker for the 11 countries in Figure 1. The rising growth rates seem to contradict the widespread notion that taxation is detrimental to growth (see, for example, Stokey and Rebelo (1995) and Lindert (2004)).

Figure 2.
Figure 2.

Growth Rate in Productivity per Hour Worked (Annualized Percent)1/

Citation: IMF Working Papers 2008, 155; 10.5089/9781451870138.001.A001

Sources: Madison (1987, Table A-5) and van Ark (1996, Table 1).1/ Growth rate is the annualized value over the previous 30 years. From 1870 to 1960, for 6 developed countries (France, Germany, Japan, Netherlands, United Kingdom, and United States) and from 1960 to 1990, for all 11 countries in Figure 1.

Finally, the relative size of government differs significantly across countries at similar levels of development. In particular, currently low-income countries have higher tax rates and larger governments than currently high-income countries did at a similar stage in their economic development.3 Table 1 confirms that tax rates and government purchase shares in today’s developing countries are not much below those of today’s developed countries and well above the purchase shares of today’s developed countries in 1870.

Table 1.

Government in Developed and Developing Countries

article image
Sources: Government purchase share for 1990 from Jha (2007, Table 9). Government purchase share for 1870 from Tanzi ans Schuknecht (2000, Table II.1). Tax rates for 1996 to 2001 are from Gordon and Li (2005)

We present a single model of development that offers possible explanations to all three observations. The model contains two sectors of private production. There is a modern sector, where technologies based on physical capital are used, and a traditional sector that uses land-intensive technologies. The exogenous force for economic development is technical progress that is biased toward the modern sector. As a consequence, the economy experiences a structural transformation as it develops—the traditional sector shrinks and the modern grows. There are three types of economic agents in the model. First, working households supply labor to each sector. Second, landowners hire labor and operate the traditional technology to produce goods. Finally, a government sets fiscal policy to maximize its own welfare and the welfare of the two types of private sector households.

We focus on the early stages of growth when the government spending is dominated by purchases of both consumption and investment goods. We ignore the growth of government transfers and the welfare state that have kept the relative size of government growing in the latter stages of development.4 In the model, the government chooses an income tax rate and decides how much of the tax revenue to consume itself, how many consumption services to provide to the private sector, and how much to invest in public infrastructure. The two private sectors of the economy are affected differently by the government. The traditional sector is harder to tax. Typically, large fractions of the labor employed in the traditional sector receive wages via in-kind payments and much of the traditional sector’s output is consumed without being traded and recorded on a regular basis.5 The traditional sector also benefits less from the government purchases because traditional production relies less on public infrastructure, and because traditional producers typically reside in rural areas where government consumption services are harder to deliver.

We argue that rising tax rates are a consequence of the structural transformation from traditional agriculture to industry. Raising tax rates is costly to the government because higher tax rates reduce the tax base by encouraging workers to remain in the traditional sector. We show that the size of this negative tax base effect is proportional to the size of the traditional sector. Thus, at early stages of development, when the traditional sector is large, increasing the tax rate is more costly because it causes larger reductions in the tax base. As the exogenous force of technological change reduces the size of the traditional sector, the marginal cost of taxing falls, and tax rates begin to rise. The economic transformation causes the tax base and the tax rate to rise over time.

The rising share of the public sector does not lead to lower economic growth because part of the tax revenue is used for public infrastructure investment. A rise in public infrastructure investment will increase economic growth. Empirical evidence suggests that public investment is subject to diminishing returns, similar to investment in private physical capital. Other things constant, diminishing returns imply that growth rates fall over the course of development. However, the diminishing returns of a given unit of public investment can be offset by a rise in the amount of investment. If sufficiently greater investment takes place each period, then growth rates need not fall despite diminishing returns. The combination of a rising tax rate and a rising tax base allows for this possibility. Our calibrations show that for a typical history of the economic transformation, and the associated rise in tax rates, that economic growth rates can rise or remain roughly constant for extended periods of time.

This is not to say that taxation is necessarily good for the growth of today’s developing countries. We explain the relatively high tax rates of currently developing countries by the absence of institutions that give weight to working households in the private sector. Obviously, in countries where the ruling elite place a relatively low weight on the welfare of the private sector, tax rates will be relatively high. The model also predicts that tax rates will be high in countries where landowners in the traditional sector are favored by the government. This is because high tax rates encourage workers to stay in the traditional sector, which lowers wages there and increases the profits of traditional producers. In either case, high tax rates limit the pace of the structural transformation from traditional to modern production. As result, economic growth will be slower than in economies where tax rates are not as high. The model predicts that countries with less democratic governments or governments that favor wealthy land owners will experience high tax rates and slow growth.

We also show that allowing governments to borrow to finance public investment leads to higher tax rates than would otherwise be the case. However, the ability to borrow is unlikely to explain the relatively high tax rates of today’s developing countries. During their early stages of development, many of the Western Offshoot countries borrowed more heavily than today’s developing countries.

II. Related Literature

Our paper is related to several literatures including those that focus on the growth of government, taxation and the traditional sector, and the connection between land inequality and development.

A. Growth of Government

Most of the literature on the size of government focuses on the sharp rise in social spending that occurred in developed economies after World War II. Lindert (2004) argues that the rise was the consequence of a transition to a fuller democracy with more widespread political voice. In contrast, our explanation for the growth of government is a purely technical one that proceeds and complements the effects of growing political participation—the structural transformation moved economic activity into more formal modes of production that are easier to tax.

Explaining why the growth of government and taxation does not slow economic growth is easier for us than for those who focus on social spending because in our model at least some of the rise in taxes is spent on productive public infrastructure. Lindert argues that some types of social spending are also productive (e.g. spending that supports continuity in the professional careers of women or that increases the health of children and working- aged adults). He discusses two additional reasons why the growth of government does not slow economic growth that we do not include. First, as the government becomes larger, policymakers take more care to tax in ways that minimize distortions (e.g. the taxation of capital and marginal income tax rates have both declined over time). Second, some policies have reduced employment (welfare, unemployment compensation, and public pensions) and in doing so have removed workers with below average productivity from the workforce, resulting in higher labor productivity.

B. Taxation and the Traditional Sector

Several studies have found a strong negative correlation between the relative size of the agricultural sector and the relative size of government, other things constant (Burgess and Stern (1993), Peltzman (1980), Stotsky and WoldeMarian (1997), and Tanzi (1991)). In fact, the studies find that the relative size of the agricultural sector is more closely correlated with the relative size of government than are other indicators of development, such as income per capita. These empirical studies provide motivation for the development of our structural model.

There is a related literature where the concern is that taxation and other government policies push economic activity underground (see the review by Schneider and Enste (2000, 2002) and Schneider (2007)). In this literature, the paper closest to ours is Loayza (1996). He looks at how government policies contribute to the informal sector and how the informal sector affects long-run economic growth in an endogenous growth model. His theoretical and econometric analysis indicates that the informal sector reduces both public infrastructure investment and economic growth—as in our analysis. However, while Loayza focuses on the urban informal sector and long-run balanced growth, we focus on the rural informal, or traditional, sector and the structural transformation. To do this, we build a framework that allows for transitional growth, where the size of the traditional sector, economic growth rates, and tax rates all change over time. Our interest is in the interaction between transitional growth and the setting of tax rates when the traditional sector is difficult for the government to tax. We find that the reduction in the tax base, when tax rates are raised, is directly proportional to the relative size of the traditional sector. In this sense, tax evasion is more of a problem for developing economies—even on the margin. When the traditional sector is relatively attractive in general, it gives households a “legal” way to avoid taxes. When the traditional sector is not generally attractive, households and firms will remain in the modern sector and must illegally avoid taxes or create more complicated legal ways of avoiding taxes. 6

C. Land Inequality and Development

Our paper is also related to a growing literature that suggests that land inequality may hamper growth. Land inequality may reduce growth through several mechanisms, including through its effect on economic and political institutions, influence over agricultural policy, credit market development, and support for public schooling (see the survey by Erickson and Vollrath (2004)). A common feature of these mechanisms is the attempt by politically powerful interests (such as landowners) to retain a low-cost work force in agriculture by limiting the options of workers outside of agriculture. Hayashi and Prescott (2006) review patterns of Japanese economic development. They argue that the Japanese miracle did not take place until after World War II because of barriers that kept agricultural employment constant throughout the prewar period. They develop a two- sector neoclassical growth model in which the resulting sectoral misallocation of labor creates disincentives for capital accumulation that accounts well for the depressed output level in interwar Japan. Hayashi and Prescott explain the existence of the labor barrier by appealing to the prewar Japanese tradition of patriarchy that forced the son in each family who was designated as heir to stay in agriculture.7 In our paper, landowners may be able to maintain a low-cost work force by supporting high tax rates on modern production sectors where incomes are easier to identify and tax. High taxes levied on modern production techniques act as a barrier, favoring traditional agriculture, especially when production and payment methods there are informal. Workers avoid high tax rates by staying in the traditional sector, driving down wage rates there to the benefit of landowners.

III. The Model

We adapt a standard overlapping-generations model to include two sectors of production (traditional and modern), two household types (workers and landowners in the traditional sector), and a government that chooses fiscal policy based on its preference for its own consumption and the welfare of the two household types. The purpose of the model is to show how economic growth, in particular the economic transformation from traditional to modern production methods, interacts with a country’s fiscal policy. To simplify the exposition, and to focus attention on public capital accumulation as a source of economic growth, we assume the economy is small and open to international flows of private capital.

A. Production

There is one good in the economy that is produced in two different sectors of production as in Hansen and Prescott (2002). The modern sector is comprised of standard neoclassical firms that hire capital and labor to maximize profit. The technology is a Cobb-Douglas production function

Yt=Ktα(DtLt)1α,(1)

where Y denotes output, K is the physical capital stock, DL is the effective labor supply, and α is the capital share parameter. The effective labor supply is comprised of a labor-productivity index, D, and the amount of raw labor allocated to industry, L.

The neoclassical firms operate in perfectly competitive markets and the standard profit-maximizing factor-price equations for the rental rates on human and physical capital are

Wt=(1α)Dtktα(2a)
1+r=(1σt)αktα1,(2b)

where kK/DL, r is the internationally determined after-tax return to capital, σt is the domestic income tax rate on the return to physical capital, and the rate of depreciation rate on private physical capital is assumed to be 1.

The same good is produced in the traditional sector. Production in the traditional sector is based on methods that depend on land and not on physical capital. Each traditional producer owns a fixed plot of land. There is no formal market for land in the traditional sector, instead each plot of land is passed from one generation of landowners to the next generation.8 Landowners possess a production technology given by

Ot=Atft1α,(3)

where A is TFP in the traditional sector and f is the input of raw labor.

Traditional producers compete with modern-sector firms in the market for labor. The labor share in traditional production is assumed to be the same as in the modern sector so that the aggregate labor share does not vary with the stage of development, as is suggested by empirical evidence (e.g. Gollin (2002)). The demand for labor by traditional producers equates the marginal product of labor to the traditional sector wage, W˜t,

W˜t=(1α)Atftα.(4)

The pre-tax traditional-sector wage rate will, in general, differ from the pre-tax modern sector wage because the government can tax income generated in the modern sector more easily than in the traditional sector.

B. Households

There are two private sector household types—workers and landowners. Working households derive their income from supplying labor to the modern and traditional sectors during the first period of their lives. They retire in the second period. Workers choose consumption when young, Cty, and when old, Ct+1o, to maximize utility, lnCty+βlnCt+1o+γlnCt, subject to a lifetime budget constraint, Cty+Ct+1o/1+r=ωt, where β and γ are positive preference parameters, ωt is after-tax wage income, and Ct is the flow of government goods and services provided per household (e.g. police protection and public water supply).9 After-tax wage income is the sum of income earned from each sector, ωtlt(1σt)Wt+(1lt)(1εσt)W˜t, where l is the fraction of work effort supplied to the modern sector and ε is a parameter that lies between zero and one. If the government can tax the entire wage paid to traditional workers then ε = 1 and if the government cannot tax traditional wages at all then ε = 0. The consumption demand functions of workers are

Cty=ωt1+β(5a)
Ct+1o=β(1+r)ωt1+β.(5b)

Traditional landowners have the same preferences as workers. The same amount of government purchases per household is provided to the traditional sector as to the modern sector. However, traditional sector households are assumed to derive less benefit from government services because they are geographically more dispersed than households employed in the modern sector.10 To capture the lower benefit of government services, we write the traditional sector utility function as lnC˜ty+βlnC˜t+1o+γln(εcCt), where 0 < εc < 1.

Traditional landowners derive their first period income from the after-tax residual income generated by traditional production, (1εσt)(OtW˜tft)=(1εσt)αW˜tft1α, where the equality is established by using (4). This income may be interpreted as a combination of land rents and compensation for the landowner’s time.

The labor demand and consumption demand functions for traditional household are

ft=[(1α)AtW˜t]1α(6a)
C˜ty=1εσt1+βαW˜tft1α(6b)
C˜t+1o=β(1+r)(1εσt)1+βαW˜tft1α.(6c)

C. Equilibrium

We consider equilibria where both sectors are operating. This requires that (i) workers be indifferent about supplying labor to each sector, (ii) landowners choose to operate a traditional business rather than becoming workers themselves and (iii) the labor market clears each period.

To satisfy condition (i), the after-tax wage rate in each sector must be equal, (1σt)Wt=(1εσt)W˜t. This condition gives us

W˜t=[1σt1εσt]Wt.(7)

To satisfy condition (ii), the maximum utility of landowners, V˜t, must be greater than or equal to that of workers, Vt,

V˜tVt.(8)

Finally, to satisfy condition (iii), the demand for labor in the traditional sector must equal the supply to that sector,

N˜ft=Nt(1lt),(9)

where Nt is the number of working households, N˜ is the number of landowners, and lt is the fraction of workers in the modern sector. We assume that the number of land plots and landowners are constant over time but that the population of workers grows at the exogenous rate, n.

The equilibrium conditions lead us to our first proposition (proofs of the propositions are found in the Appendix).

The traditional sector will operate as long as (1lt)m˜t1αα(1εc)γ, where m˜tN˜/Nt.

The proposition states that the traditional sector will operate as long as the ratio of workers to landowners in the traditional sector is sufficiently high. When this ratio is high, running a traditional business generates a relatively high level of residual income per landowner. The ratio can remain high, even when the fraction of labor in the traditional sector is low, provided that the absolute number of workers is high relative to land and landowners. Throughout our analysis, we assume that the ratio of workers to land is sufficiently high to prevent owners from abandoning their land.

D. Government

We take a “reduced-form” approach to the formation of policy. We model the government as any other economic agent—by specifying its preferences, constraints, and objectives. There is no deep model of the politics that determine how the government is chosen and how their policies are influenced by voters and interest groups. Instead we take as given the politics of a country that determine the “reduced-form” preference parameters of the government. The parameters dictate the government’s concern with the welfare of the general population, of possibly different household-types, and of households that make up the government itself.

Government officials have preferences defined over their own consumption, ctg, and the welfare of the two private-sector household types. We assume that the current government cares about the government as an on-going institution (i.e. they care about the future operations of the government and the welfare of future government officials) and the welfare of the country’s future citizens. We also consider the special case where there are no distinct government officials. In this case, the optimal government policy is directed by the private households, at no cost, with possibly different political power of different household-types. Even in the absence of explicit officials, we assume that the government continues to consider the policy effects on future generations.

The preferences that determine government policy are given by the intertemporal utility function,

t=0βt{ϕglnCtg+ϕlnVt+ϕ˜lnV˜t},

where ϕg, ϕ and ϕ˜ are constant preference parameters. We think of the government’s preference parameters as functions of exogenous political institutions and de facto political power of the private sector households. Countries with less democratic institutions, and fewer “constraints on the executive,” will tend to have governments that place smaller weights on the welfare of the private sectors households as a whole (low values for both ϕ and ϕ˜), or perhaps that give disproportionate influence to wealthy landowners (a high value for ϕ˜ and a low value for ϕ). By varying the preference parameters we can capture a government of the public interest (the “helping hand”) or a government of its own private interests (the “grabbing hand”)

The single period government budget constraint is

CtgNtg=σt[Wtlt+εW˜t(1lt)]Nt+σt[(YtWtLt)+ε(OtW˜tft)N˜]Gt+1Nt+1Ct(N˜+Nt+1)

where Ntg is the number of government officials, possibly zero, and G is government capital per worker. We interpret the government capital broadly to include both physical capital formed by government infrastructure investments and human capital formed by public education investments Note that, as with private capital, we assume that government capital has a depreciation rate of one, so that government investment in period t equals the stock of government capital per worker in period t+1. This assumption is particularly convenient given our broad interpretation of public capital because, as in the case of human capital, it implies that new public investments must be made in each generation of workers. Note also that we are not considering government debt at this point, a topic that is discussed in a later section.

The benefit of government capital has not been identified as yet. The labor productivity index, D, is a function of the disembodied level of technology, E, and government capital per adult worker and is given by

Dt=Et1μGtμ,(10a)

where 0 < μ < 1 is a constant parameter. We assume that E progresses at the exogenous rate q. This specification of the impact of government capital is similar to Aschauer (1989) and Clarida (1993). Similarly, TFP in the traditional sector is

At=(εgGt)μ(e˜tEt)1μ,(10b)

where εg ≤ 1, since public capital is assumed to be less useful to traditional production in general (as in Loayza (1996)), and where e˜t is the relative state of technology in traditional production. We leave the behavior of e˜t unrestricted for the moment, but we will eventually view it as declining with development (i.e. the pace of technological change is more rapid in the modern sector than in the traditional sector).

To solve the government’s problem, we first rewrite the government budget constraint to incorporate the private sector equilibrium. The government accounts for the general equilibrium interactions of the private sector when setting its policies. This means that the government not only accounts for the household responses given by (5) and (6), but also the equilibrium response given by (7) and (9). Substituting the various features of the private sector equilibrium into the budget constraint and rearranging allows us to rewrite the government budget constraint as

Ct=Et(1+m˜t){ltσwtσt1α(1+q)(1+n)gt+1mtgctg},

where mtgNtg/Nt, ltσlt+ε(1σt)1εσt(1lt) is the effective labor input for the government’s tax base, gt = Gt/Et and ct = Ct/Et are the detrended values for the government capital stock and government services provided per worker, and wtWtEt=(1α)gtμ[(1σt)α1+r]α1α is the detrended formal sector wage. For simplicity, in the case where government policy is set by distinct government officials, we assume that the number of government officials rises at the same rate as the population of workers, so that mtg is a constant throughout. Note that a sufficient condition the government’s problem to be well-defined, is β(1 + q) < 1.11

After substituting the expression for Ct into the government’s objective function, the problem for the government is to choose sequences for cg, σ, and g to maximize utility over an infinite horizon. The associated first order conditions are

ϕgctg(ϕ+ϕ˜)γmgct(11a)
βμgt+1{γ1ασt+1lt+1σwt+1ct+1+(1+β)}=γ(1+q)(1+n)ct(11b)
(ϕ+ϕ˜)γwtltσct(1α){1+ltσσtσtltσα1ασt1σt}(11c)=(1+β){ϕ+ϕ˜(1σt)(1α)+ϕ˜α[ε1εσt1(1σt)(1α)]},

Equation (11a) is the first order condition for consumption by government officials. The marginal benefit, the left-hand-side, is the increase in utility of government officials from the rise in their consumption. The marginal cost, the right-hand-side, is the fall in utility of private sector households from the reduced consumption of government services.

Equation (11b) is the first order condition for the choice of gt+1. The marginal benefit, is made up of two parts. The first part gives the rise in next period’s tax base from greater government investment in the current period. The second part is the rise in welfare of the next generation of both private households due to the rise in labor productivity associated with the increase in government capital. The marginal cost, the right-hand side, gives the loss in current period private government services caused by using more current tax revenue for investment.

Equation (11c) is the first order condition for the choice of σt. The marginal benefit is rise in government services financed by the increased tax revenue. The direct benefit of a higher tax rate is mediated by a reduction in the tax base, as workers reduce their labor supply to the modern sector when it is taxed more heavily than the traditional sector (ltσt0) and as private physical capital intensity falls with taxation (α1ασt1σt).

The right-hand-side gives the marginal cost of taxation. The first term on the right-hand- side captures the lost in welfare from reducing the after-tax wages of workers. The more complicated second expression on the right-hand side captures the effect of taxation on traditional producers. If the traditional sector is taxed as heavily as the modern sector (ε = 1), then this expression vanishes, since labor costs in the traditional sector will not be lowered by a rise in taxes. However, if taxes chase workers into the traditional sector because they are less heavily taxed there (ε <1), then traditional wages will fall and the traditional producer is made better off—lowering the private marginal cost of labor taxes.

IV. Explaining Fiscal Policy

The introduction mentions three general facts about fiscal policy and development.

(F1) The size of government increases as an economy develops.

(F2) The rise in government and taxation are associated with rising or constant economic growth rates.

(F3) Today’s developing countries have larger government sectors than did today’s developed countries at similar stages of development.

This section uses the model of the previous section to offer possible explanations for each fact. The explanations are based, in part, on the following two propositions. The first proposition determines a closed-form transition equation for public capital accumulation. We simplify notation a bit by restricting parameters values to satisfy the condition, ϕg+γ(ϕ+ϕ˜)=1, i.e. the weights placed on government consumption by public officials and private households sum to one.

For any value of ε and any tax rate sequence, the transition equation for public capital accumulation is
gt+1=B(1+q)(1+n)σtltσ[(1σ)α1+r]α1αgtμ,(12)where B=βμ[1+(1+β)(ϕ+ϕ˜)1+βμ(1+β)(ϕ+ϕ˜)]<1.

Proposition 2 says that detrended government investment is simply a constant fraction of the tax base.

Given Proposition 2, we next characterize the optimal tax rate sequences in two extreme cases, where the traditional sector is fully taxed (ε =1) and where the traditional sector is untaxed (ε=0).

(a) For ε =1, σt=σ=1α1+(1B)(1+β)ϕ.

(b) For ε =0, the optimal income tax sequence satisfies1 1>σt(e˜t,ϕ,ϕ˜)>0, with σte˜t<0,σtϕ<0,σtϕ˜>0.

Assuming full ability to tax the traditional sector (ε =1), as in part (a), implies that the optimal income tax rate is constant. This leads to the prediction that tax rates are constant over the course of development, a prediction clearly contradicted by the data. However, when the traditional sector cannot be taxed (ε =0), as in part (b), then the optimal income tax changes over time. In particular, if the pace of technology in the modern sector exceeds that in the traditional sector, i.e. e˜t decreases over time, then the income tax rate rises over time.

We now relate the results from Propositions 2 and 3 to the three fiscal policy facts. The intuition for the result that σt/e˜t<0, from part (b), rests on the fact that taxation becomes less costly as the traditional sector shrinks. This offers a possible explanation for (F1). As shown in the Appendix, the amount by which the tax base falls, as tax rates rise, is increasing in the size of the traditional sector. The marginal cost of increasing the tax rate is high when the traditional sector is large, and declines over time as the traditional sector shrinks. Thus, during the structural transformation from traditional to modern production, the economy experiences rising tax rates. Combined with the fact that the share of labor in the modern sector rises, the relative size of government unambiguously increases.

The second development fact, (F2), states that there is a rise in the size of government and taxation over periods when economic growth rates are increasing or at least constant. The presence of a traditional sector that is hard to tax is also important in explaining this fact. To see why, suppose that ε =1 and the tax rate is fixed over time. In this case, the tax base grows solely due to growth in government investment. With public capital accumulation subject to diminishing returns (μ < 1), output, the tax base, and government capital all grow at a decreasing rate (see (12).

Consider the same economy with the exception that ε =0, the traditional sector is untaxed. Now both the tax base and the tax rate grow over the course of development. This causes government investment, which is a constant fraction of the tax revenue, to grow as a fraction of total output and income. The increasing rate of government investment works against the diminishing returns to investment, and may result in a constant or even rising growth rate. We calibrate the model in the section VI with a conservatively low estimate of μ, yet we find that the rise in the rate of government investment more than offsets diminishing returns, causing growth rates to gradually rise.

Thus, (F2) may also simply be a by-product of the economic transformation. The economic transformation increases the tax base, leading to a rise in tax rates. The rise in both the tax base and tax rates increase the rate of government investment over time, which helps to offset or mediate the diminishing returns associated with a growing stock of public capital.

The results from Proposition 2 offer two possible explanations for (F3)—the relatively high tax rates and government shares in today’s developing countries. A third explanation for (F3) is discussed in the next section. First, governments that place less weight on the private sector’s welfare will set higher tax rates. As noted by Glaeser, LaPorta, Lopes de-Silanes, and Schliefer (2004), almost all developing countries after World War II were dictatorships. A lack of institutional constraint on executive power is likely to be correlated, although not perfectly so, with relatively low regard for private sector welfare. Acemoglu, Johnson, and Robinson (2005) and Acemoglu, Johnson, Robinson, and Yared (2006) provide evidence that Europe and its Western offshoots had institutions that placed constraints on executive power early on in their development, where currently developing countries do not. Thus, the relatively high tax rates of currently developing countries could stem from a government with relatively few institutional constraints on its power—modeled here as relatively high values of ϕg and relatively low values of ϕ.

A low value for ϕ also implies a low value of B, i.e. a low investment share of the government budget. The economy then suffers not only from high tax rates, but also from low levels of investment. These consequences of a lack of institutional constraint on the executive may slow the structural transformation and economic growth and permanently lower the steady level of worker productivity.

A second way to generate a high-tax economy is to alter the composition, rather than the level, of the government’s concern for the private sector. Formally, we can raise ϕ˜ and lower ϕ, but keep the sum of ϕ and ϕ˜ constant. Raising ϕ˜ and lowering ϕ, increases the government weight on traditional producers relative to workers. Traditional producers prefer high taxes because it lowers the wage and labor costs in the traditional sector. Thus, tax rates will be higher in the economy with relatively high ϕ˜. To view this as a possible explanation of (F3), one needs to explain why today’s developing countries would favor traditional producers more than today’s developed economies did during their historical development.

In the vast majority of countries, the most important type of traditional production is in labor/land-intensive agricultural production. The concentration of rural land ownership, however, differs significantly across countries at similar stages of development. For example, the United States and Canada had relatively egalitarian distributions of land compared to countries in Latin America, where the ownership of land was concentrated among a small fraction of the rural population. When land ownership is concentrated in the hands of a few, one might expect that the wealthy landowners would have significant political influence that shifts the favor of government in their direction. Thus, high tax rates may appear in developing countries with concentrated landownership and political regimes that fail to limit the disproportionate influence of the wealthy. To the extent that this combination is commonplace among developing countries, we have offered a second explanation for (F3). The next section offers a third explanation.

V. Government Debt

In this section we consider what happens when government debt is allowed. With government debt, private consumption of government services can be written as

Ct=Et(1+m˜t){ltσwtσt1α+(1+q)(1+n)bt+1(1+q)(1+n)gt+1mgctg(1+r)bt},

where bt is detrended government debt per worker. If the government is unrestricted in their choice of bt, then along with (11) we get the additional first order condition

ct+1ct=β(1+r)(1+q)(1+n),(13)

a standard Euler condition for the path of consumption. In a model where the developing economy has the same β, q, and n as the rest of the world, it is reasonable to insist that the world interest satisfy the condition 1 + r = (1 + q)(1 + n)/β. However, the world interest rate conditions implies that private consumption of government services in all economies, developing economies included, is constant over time, i.e private consumption of government services is always at its steady state value.

This counterfactual prediction stems from the unrealistic assumption that the government of a developing country can borrow with no restrictions from the developed world at the going interest rate. A more reasonable assumption is that the government is allowed to borrow to finance government capital formation, but not consumption. Under this borrowing constraint, we have bt+1 = gt+1. Substituting the constraint into the expression for government consumption above and solving the government problem generates the following proposition.

For any value of s and any tax rate sequence, public capital accumulation satisfies
gt=B^σt1+rltσ[(1σt)α1+r]α1αgtμ,(14)

where B<B^=μ[1+(1+β)(ϕ+ϕ˜)1+μ(1+β)(ϕ+ϕ˜)]<1. Further more, optimal taxes have the same form as in Proposition 3, except that B is replaced by B^, implying that the optimal tax is higher, other things constant, than in Proposition 3.

The proposition tells us that constrained borrowing does not affect the qualitative conclusions of the previous section. The analysis that was conducted in the previous can be repeated and would go through without qualitative changes when borrowing is allowed.

However, quantitative outcomes are affected by borrowing. Tax rates are higher, at any given value of lt. Borrowing allows the payment of government investment to be postponed, which lowers the cost of public investment. This raises investment expenditures, including future interest payments, and the taxes needed to finance them. However, it is unclear whether detrended public capital (net of interest expenses) will be higher or lower, since one can show that B/(1+q)(1+n)>B^/(1+r) — the rate of public investment out of a given tax base is higher without borrowing. Finally, note that (14) is not a difference equation. For given tax rates and a given rate of investment out of tax revenues, borrowing allows the government to “jump” to its steady state (conditional on given tax rates) rather than “transition” toward it. This implies that under borrowing, investment responses will be more sensitive to changes in the tax base.

In summary, granting the government access to borrowing raises tax rates for a given level of development. If poor countries today are able to borrow more heavily than developing countries did historically, then government debt is a third possible explanation for (F3). However, this is not generally true. During the late 19th century and early 20th century, the Western offshoots (Australia, Canada, Argentina, United States, and New Zealand), as well as the developing economies of Sweden and Norway, borrowed heavily from Europe (Van den Berg (2004, p. 500). The access to funds for government, in itself, does not explain the difference in tax rates across today’s and yesterday’s developing economies.

VI. Calibration Experiments

In this Section, we calibrate the model to gauge the quantitative importance of the link between the structural transformation and fiscal policy.

A. Calibrating the Model

The calibration is based on the historical patterns observed in Figures 1 and 2. We interpret a period of the model to be 30 years. The exogenous rate of technological change (q) is set to match, along with the endogenous contribution of government investment, the average annualized growth rate of labor productivity over the period 1870 to 1990, 2.27 percent for the 11 countries from Figure 2. The population growth rate (n) was set to an annualized rate of 1 percent. This is similar to the population growth rates at the end of the 20th century. We do not attempt to account for the higher rates of population growth rates observed at the beginning of the century. The international aftertax rate of return to capital (r) is set to an annualized rate of 7 percent, similar to estimates of the return to physical capital found in Caselli and Feyrer (2007). The parameter β is set equal to (1+n)(1+q)/(1+r), so that the detrended growth of government consumption in the steady state is zero. The parameter α is set to its standard value of 1/3.

The output elasticity for public capital (μ) is set to 0.30. This is somewhat less than the value estimated by Aschauer (1989) and Clarida (1993). However, the value (1−α) μ = 0.20 is in the middle of the range surveyed by Glomm and Ravikumar (1997). Also remember that we have broadened the interpretation of government investment to include public school expenditures. From this perspective, μ is also the wage elasticity of public school expenditures. If one holds student’s time constant, then a wage elasticity of 0.30 is too high (Manuelli and Seshadri (2005) is the only study we know that estimates a wage-elasticity this high). However, much of the rise in public school spending over the course development is associated with rising school enrollment and rising attendance during the school year. In this case, where student’s time and expenditures rise together, the estimated wage elasticity is much higher, typically 0.50 or greater (e.g. see Rangazas (2002) and Manuelli and Seshadri (2005)). Based on this reasoning, we also consider a case with μ is set to 0.40.

To make the initial calibration more parsimonious, we set ϕg to zero. In a subsequent calibration experiment we consider the effect of increasing ϕg. The parameters ϕ and ϕ˜ were set to match the initial government purchase and government investment shares in 1870 of 0.045 and 0.025. The government purchase share is from Figure 1. The government investment share is the sum of the shares of public expenditures on education and public investment in physical capital for the same 11 countries as Figure 1 (Tanzi and Schuknecht (2000, Table II.5 and II. 13)). The value of γ is set to maintain the restriction, ϕg+γ(ϕ+ϕ˜)=1.

Finally, an important feature of the model is the share of labor in the modern sector, lt. For the calibration experiment we interpret the modern sector as non-agriculture and the traditional sector as agriculture. Maddison (1994) contains estimates of the share of labor in agriculture for the 8 of the 11 countries and Cipolla (1974) for 10 of the 11 countries from Figure 1. For 1900, the average is 44 percent from Madison and 43 percent from Cipolla. Maddison also contains data for 1990, where the average for the 8 countries is 5 percent. In the simulation we assume that the decline in the agricultural share is linear from 44 percent in 1900 to 5 percent in 1990.

Cipolla (1974) also has estimates for 6 of the countries around 1850, with an average agricultural share of 57 percent. The average share for these 6 countries was 42 percent in 1900 (very close to the average for the larger sample of 10 countries). Based on this data, we used 55 percent as an upper bound estimate of the agricultural share in 1870 and 50 percent as a lower bound estimate. We match the historical shares exactly by setting the relative TFP in agriculture, e˜t, so that the model’s predictions about the labor share matches the data. Table 2 summarizes the parameter calibration.

Table 2.

Calibrated Parameters

article image

B. Historical Growth Experience of Currently Developed Countries

Using the calibration above, we now need only set the initial government capital intensity. The initial government capital intensity was set so that the intensity in 1870 was the same as that in 1840, i.e. the economy was in a quasi-steady state at the beginning of the simulation. Setting government capital intensity below this value would only increase the growth explained by the model, so our approach is conservative.

Figures 3 to 5 compare the model’s predictions for the government purchase share, the government investment share, and the growth rate in labor productivity with the actual data. The lower is the initial share of labor in the modern sector, the greater is the growth in the tax base and the greater is the rise in the government purchase and investments shares. Over the period from 1870 to 1990, the model explains between 50 and 75 percent of the historical rise in government purchase shares and between 60 and 85 percent of the historical rise in government investment shares. The model also predicts gradually increasing growth rates in worker productivity that match the actual growth rates closely until the final period, 1960 to 1990, where actual growth rates jump. The model does not capture the “catch-up” growth of France, Germany, United Kingdom, and Japan after they recover from the destruction of World War II. However, the model predicts gradually increasing growth rates for more than a century. The model does not generate the typical diminishing growth rates of neoclassical models because the rise in the government investment share out of GDP more than offsets diminishing returns to investment.

Figure 3.
Figure 3.

Predicted (Dashed) versus Actual (Solid) Government Purchase Share

Citation: IMF Working Papers 2008, 155; 10.5089/9781451870138.001.A001

Notes: Long-dashed is for upper bound estimate of 1870 fraction in agriculture and short-dash is for lower bound estimate.
Figure 4.
Figure 4.

Predicted (Dashed) versus Actual (Solid) Government Investment Share

Citation: IMF Working Papers 2008, 155; 10.5089/9781451870138.001.A001

Source: Government investment share is the sum of the shares of public expenditures on education and public investment for the same 11 countries as Figure 1 based on data from Tanzi and Schuknecht (2000, Table II.5 and II. 13). Notes: Long-dashed is for upper bound estimate of 1870 fraction in agriculture and short-dash is for lower bound estimate.
Figure 5.
Figure 5.

Predicted (Dashed) versus Actual (Solid) Labor Productivity Growth Rates

Citation: IMF Working Papers 2008, 155; 10.5089/9781451870138.001.A001

Notes: Long-dash is for upper bound estimate of 1870 fraction in agriculture and short-dash is for lower bound estimate.

Given our calibration strategy, raising μ does not change the portion of the rise in government purchases and investment shares that are explained by the model; other parameter values change to maintain the shares in the initial period, and that is sufficient to maintain the paths in the government shares. Raising μ only changes the portion of the labor productivity growth that is explained by government investment rather than exogenous TFP. With μ set to 0.30, government investment explains between 1/4 and 1/3 of a percent of labor productivity growth per year from 1870 to 1990 (depending on the initial value of the share of labor in the modern sector). If μ is raised to 0.40, the annualized growth rate explained by government investment is between 0.40 and 0.50 percent per year.

C. Growth of Currently Developing Countries

At the end of the 20th century many developing countries had not yet reached the level of prosperity achieved by the currently-developed countries in 1870. This is evidenced not only by low levels of worker productivity of currently developing countries, but also by the large shares of their workforces in traditional agriculture.

We use calibration experiments to examine two possible explanations for the large share of labor in traditional agriculture: (1) low TFP in the modern sector and (2) fiscal policy that is unfavorable to growth: high taxes and low government capital. We target the initial share of labor in agriculture to be 75 percent, rather than the 50 to 55 percent targeted in the historical experiment above.

In the first experiment, we keep all parameters the same but raise the initial value of e˜ until the initial share of labor in agriculture is 75 percent. The rate of decline in e˜ is kept the same as in the historical experiment. In the second experiment, we change the weights the government places on its own welfare and the weights placed on working households and traditional producers. More specifically, we raise the weights on the welfare of the government officials and traditional producers, and lower the weight on working households, until taxes and government capital drive the share of labor in traditional agriculture to 75 percent. The path of e˜ is identical to that of the historical experiment.

We simulate the implied growth paths for the tax rates, the share of labor in the traditional sector, and the labor productivity growth rates for both the low-TFP and the high-tax developing countries. The results are given in Figures 6 to 8, where the paths are compared to the simulated paths for the currently-developed countries from Figures 3 to 5.

Figure 6.
Figure 6.

Tax Rates

Citation: IMF Working Papers 2008, 155; 10.5089/9781451870138.001.A001

Notes: Rich country (solid line), Poor country due to low TFP (long-dash), and Poor country due to high taxes (short dash).
Figure 7.
Figure 7.

Share of Labor Force in Modern Sector

Citation: IMF Working Papers 2008, 155; 10.5089/9781451870138.001.A001

Notes: Rich country (solid line), Poor country due to low TFP (long-dash), and Poor country due to high taxes (short dash)
Figure 8.
Figure 8.

Labor Productivity Growth Rates

Citation: IMF Working Papers 2008, 155; 10.5089/9781451870138.001.A001

Notes: Rich country (solid line), Poor country due to low TFP (long-dash), and Poor country due to high taxes (short dash).

Figure 6 indicates that tax rates in the low-TFP developing country begin below those of the 1870 tax rates of currently-developed economies. After a few periods, the low-TFP country’s tax rates converge to those found on the historical path of the currentlydeveloped economies. The high-tax developing country starts with tax rates slightly above the 1990 tax rates of developed countries, and then tax rates rise dramatically over the course of development reaching a rate in excess of 50 percent after 5 periods (120 years).

The consequence of the high-tax rates is made clear in Figure 7. The growth of the modern sector grows fast in the low-TFP country, quickly converging to the historical growth path of labor shares in the modern sector along the historical growth path of today’s developed countries. However, the modern sector labor shares grow very slowly in the high-tax economy, remaining below 40 percent after 5 periods. There is a two-way causation between the structural transformation and the size of government: exogenous factors that increase the pace of the structural transformation (e.g. technological progress) cause the relative size of government to increase, and exogenous factors that increase the size of government (selfish government officials or politically powerful landowners) slow the structural transformation.

Figure 8 shows that the growth rates of the two developing countries also differ dramatically. The initial labor productivity of the low-TFP economy is 2/3 of the labor productivity of the currently developed countries in 1870. The low-TFP country’s transitional growth rates exceed the growth rates along the historical path of the currently-developed countries. If we imagine the low-TFP country’s initial period as 1870, then the labor productivity of this relatively poor country would converge toward the labor productivity of the currently-developed economies. The opposite is true of the high-tax country. In its initial period, labor productivity is 57 percent of the historical labor productivity of the currently-developed economies in 1870. The high-tax country’s growth rates are below those of the currently-developed countries. Beginning in the same period, the labor productivities of the high-tax countries and the currently-developing countries would diverge over time.

VII. Conclusion

The size of government expands as an economy develops over time. Our results suggest that this tendency is, in part, a natural consequence of the structural transformation. During the structural transformation, an economy shifts from traditional to modern production methods. We show that this causes the marginal cost of taxation to fall and tax rates to rise. This economic explanation for increasing government size complements explanations based on the spread of political voice as democracies develop and mature.

Rising tax revenues, allow the government to increase the level of public investment over time. Increasing levels of investment help to offset the diminishing returns associated with public capital accumulation. Thus, as the public sector expands, the economy’s growth rates do not necessarily decline.

We have argued that rising tax rates and government shares are a natural byproduct of the structural transformation associated with development. However, currently developing countries have high tax rates and government shares relative to their state of development. This may be a result of today’s developing countries having relatively less democratic governments and institutions. In our model relatively high tax rates, resulting from low concern for private sector welfare or disproportionate political influence of landed elites, reduce economic growth by slowing the structural transformation.

Our quantitative analysis suggests that the connections between the structural transformation and the size of government are potentially important. The model can explain 50 to 75 percent of the rise in the government purchase share, and 60 to 85 percent of the government investment share, in developed countries since 1870. The analysis also shows that high tax rates and government consumption at early stages of development can slow the structural transformation and economic growth.

There are several ways to extend the study. First, we have focused on the size of government and the overall level of taxation. However, it is known that the composition of taxation changes with development. Sales taxes, property taxes, and capital income taxes are more prominent at early stages of development relative to labor income taxation (Burgess and Stern (1993) and Gordon and Liu (2005)). Second, we have not examined the possibility that the government can purposely target different amounts of public capital and consumption services to the modern and traditional sectors. Finally, we have kept population growth exogenous. There are important connections between the structural transformation and the demographic transition that could be examined, including the fact that fertility is higher in the traditional sector.

References

  • Acemoglu, D., Johnson, S., and Robinson, J., 2005, “Institutions as the Fundamental Cause of Long-Run Growth,” in Aghion, P. and Durlauf, S., editors, Handbook of Economic Growth, Amsterdam: North Holland.

    • Search Google Scholar
    • Export Citation
  • Acemoglu, D., Johnson, S., Robinson, J., and Yared, P., 2006, “Income and Democracy,” MIT Working paper.

  • Alston, L. and Hatton, T., 1991, “The Earnings Gap Between Agriculture and Manufacturing Laborers, 1925-1941,” Journal of Economic History, 51, 83-99.

    • Search Google Scholar
    • Export Citation
  • Aschauer, D., 1989, “Is Public Expenditure Productive?,” Journal of Monetary Economics, 23, 177-200.

  • Atkeson, A., Chari, V., and Kehoe, P., 1999, “Taxing Capital Income: A Bad Idea,” Federal Reserve Bank of Minneapolis Quarterly Review, 23, 3-17.

    • Search Google Scholar
    • Export Citation
  • Burgress, R., and Stern, N., 1993, “Taxation and Development,” Journal of Economic Literature,” 31, 762-830.

  • Caselli, F., and Feyrer, J., 2007, “The Marginal Product of Capital,” Quarterly Journal of Economics, 122, 535-568.

  • Cipolla, C., 1974, The Economic History of the World Population, Baltimore: Penguin Press.

  • Clarida, R., 1993, “International Capital Mobility, Public Investment, and Economic Growth,” NBER Working Paper 4506.

  • Doepke, M, 2004, “Accounting for Fertility Decline During the Transition to Growth,” Journal of Economic Growth, 9, 347-383.

  • Drazen, A. and Eckstein, Z., 1988, “On the Organization of Rural Markets and the Process of Development,” American Economic Review, 78, 431-443.

    • Search Google Scholar
    • Export Citation
  • Erickson, L., and Vollrath, D., 2004, “Dimensions of Land Inequality and Economic Development,” IMF Working Paper 158.

  • Galor, O., Moav, O., and Vollrath, D., 2006, “Inequality in Land Ownership, the Emergence of Human Capital Promoting Institutions, and the Great Divergence,” Brown University Working Paper, 2006-14.

    • Search Google Scholar
    • Export Citation
  • Glaeser, E., La Porta, R., Lopez-de-Silanes, F., and Schleifer, A., 2004, “Do Institutions Cause Growth?,” Journal of Economic Growth, 9, 271-383.

    • Search Google Scholar
    • Export Citation
  • Glomm, G., and Ravikumar, B., 1997, “Productive Government Expenditure and Long-run Growth,” Journal of Economic Dynamics and Control, 21, 183-204.

    • Search Google Scholar
    • Export Citation
  • Gollin, D., 2002, “Getting Income Shares Right,” Journal of Political Economy, 110, 458-474.

  • Gordon, R., and Li, We, 2005, “Tax Structure in Developing Countries: Many Puzzles and a Possible Explanation,” NBER Working Paper 11267.

    • Search Google Scholar
    • Export Citation
  • Hansen, G., and Prescott, E., 2002, “Malthus to Solow,” American Economic Review, 92, 1205-1217.

  • Hayashi, F, and E.C. Prescott, 2006, “The Depressing Effect of Agricultural Institutions on the Prewar Japanese Economy,” NBER Working Paper 12081 (March). http://www.nber.org/papers/w12081

    • Search Google Scholar
    • Export Citation
  • Jha, R., 2007, “Fiscal Policy in Developing Countries: A Synoptic View,” Australia South Asia Research Centre, Working Paper 2007/01.

    • Search Google Scholar
    • Export Citation
  • Larrabee, Eric, 1988, Commander in Chief: Franklin Delano Roosevelt, His Lieutenants, and Their War, New York: Touchstone.

  • Lindert, P., 2004, Growing Public, Cambridge: Cambridge University Press.

  • Loayza, N., 1996, “The Economics of the Informal Sector: A Simple Model and Some Empirical Evidence from Latin America,” Carnegie-Rochester Conference Series on Public Policy, 45, 129-162.

    • Search Google Scholar
    • Export Citation
  • Maddison, A., 1994, “Economic Growth and Standards of Living in the Twentieth Century, Research Memorandum 576 (GD-15), University of Groningen.

    • Search Google Scholar
    • Export Citation
  • Manuelli, R. and Seshadri, A., 2005, “Human Capital and the Wealth of Nations,” Mimeo, University of Wisconsin.

  • Peltzman, S., 1980, “The Growth of Government,” Journal of Law and Economics, 23, 209-287.

  • Rangazas, P., 2002, “The Quantity and Quality of Schooling and U.S. Labor Productivity Growth (1870 to 2000),” Review of Economic Dynamics, 5, October 2002.

    • Search Google Scholar
    • Export Citation
  • Ruggles, S., 2001, “Living Arrangements and the Well-being of Older Persons in the Past,” Population Bulletin of the United Nations, 42/43, 111-161.

    • Search Google Scholar
    • Export Citation
  • Schneider, F., and Enste, D., 2000, “Shadow Economies: Size, Causes, and Consequences,” Journal of Economic Literature, 38, 77-131.

  • Schneider, F., and Enste, D., 2002, The Shadow Economy: An International Survey, Cambridge, UK: Cambridge University Press.

  • Schneider, F., 2007, “The Size of the Shadow Economies of 145 Countries all over the World: First Results on the Period 1999 to 2003, Journal of Population Economics, 2007, 20(3), 495-526.

    • Search Google Scholar
    • Export Citation
  • Stokey, N., and Rebelo, S., 1995, “Growth Effects of Flat-Rate Taxes,” Journal of Political Economy, 113, pp. 519-550.

  • Stotsky, J., and WoldeMariam, A., 1997, “Tax Effect in Sub-Saharan Africa,” IMF Working Paper, Wp/97/107.

  • Tanzi, V., 1991, “Structural Factors and Tax Revenue in Developing Countries: A Decade of Evidence,” Mimeo, International Monetary Fund, Washington DC.

    • Search Google Scholar
    • Export Citation
  • Tanzi, V., and Schuknecht, 2000, Public Spending in the 20th Century,” New York: Cambridge University Press.

  • van Ark, B. 1996, “Issues in Measurement and International Comparisons of Productivity—An Overview,” Research Memorandum GD-28, University of Groningen.

    • Search Google Scholar
    • Export Citation

Appendix

Proposition 1. Substitute the demand functions, given by (5) and (6), back into the utility functions for workers and landowners to form the indirect utility functions Vt and V˜t. Substitute the indirect utility functions into (8) to rewrite (8) as (1+β){ln(α1α)+γlnεc+ln(ft)}0. Use 9) to eliminate ft and solve (8) for the inequality condition in the proposition.

Proposition 2. “Guess” that the solution for the transition equation has the form indicated in the proposition, with B an undetermined coefficient that is independent of gt+i. Substitute the guess into (11b) and use (11a) to verify that (11b) is satisfied for the value of B indicated in the proposition.

Proposition 3. First, substitute (12) into (11c) to get

11B1σt{1α1ασt1σt+ltσσtσtltσ}(A1)=(1+β){ϕ+ϕ˜(1α)(1σt)ϕ˜α[1(1α)(1σt)ε(1εσt)]}

Next, comparative statics reveals that

ltσσt1ltσ=(1ltσltσ) [(1εσt(1σt)(1α)(1α)ε)]1α1ε(1εσt)2(A2)

where

ltσ=lt+ε(1σt)1εσt(1lt)(A3)=1[1ε1εσt]π˜tft=1[1ε1εσt]π˜t[(εg)μe˜t1μ(1εσt)[(1σt)α1+r]α1α(1σt)]1α.

Setting ε = 1, allows one to use (A1) to solve for the constant tax rate given in part (a) of the proposition. Setting ε = 0, allows (A1) to be written as

1σt=[1ltlt1α+α1α+(1+β)(1B)1α(ϕ1ααϕ˜)]11σt.(A1)

The left-hand-side is strictly decreasing in σt and can be interpreted as the marginal benefit of taxation. Using (A3), the right-hand-side is strictly increasing in σt and can be interpreted as the full marginal cost of taxation, with the reduction in the tax base now included on the cost-side. A sketch of the marginal benefit and marginal cost shows for any e˜t0 generating a positive marginal cost of taxation, that there is a unique interior optimal tax rate. Finally the marginal cost of taxation is increasing in e˜t and ϕ, and decreasing in ϕ˜ for 1> σt >0, completing the proof of part (b).

Proposition 4. “Guess” the solution for the transition equation has the form indicated in the proposition, with B^ an undetermined coefficient that is independent of gt+i. Substitute the guess into the first order condition for gt+i and verify that the first order condition is satisfied for the value of B^ indicated in the proposition. The form of the first order conditions for tax rates are the same as in Proposition 3. The proof of Proposition 3 goes through with B^ substituted for B. That σ is higher than in Proposition 3 when ε = 1 is immediate. When ε = 0, the marginal costs of σt is shifted down when B is replaced with B^, which causes σt to be greater for any value of lt.

1

For useful comments and suggestions on an earlier draft, we thank our colleagues in the European Division of the IMF Institute and participants at a seminar in the Economics Department of Kyung Hee University, Seoul, Korea.

2

On Wagner’s Law, see Weil (2005, Figure 12.3, p. 342). He shows the pattern of rising government shares for the UK, US, Japan, France, and Sweden during the 20th century.

3

Gordon and Li (2005) show that statutory tax rates differ very little across developing and developed economies in direct comparison today. Tax revenue as a share of GDP in developing countries is somewhat smaller than in developed countries today, but the tax shares are well ahead of the historic tax shares of today’s developed economies at similar stages of development.

4

See Lindert (2004) for a historical analysis of the rise in social transfers.

5

This was true even in the United States well into the twentieth century (see, for example, Alston and Hatton (1991)).

6

For today’s developing countries, Stotsky and WoldeMarian (1997) provide empirical evidence of an inverse relationship between the tax share of GDP and the share of labor in agriculture, where most of the production is based on traditional methods. We argue that the causation behind this correlation can run in both directions.

7

According to historians, military careers in pre-World War II Japan were among the few avenues for upward social and economic mobility available to ambitious young men from the countryside. Many midrank officers came from the countryside. For the political consequences of the limited upward social and economic mobility in prewar Japan, see Larrabee (1988).

8

Land is passed from one generation of landowners to the next either because of the absence of a land market in the early stages of development or a preference of landowners to keep the land in the family. Similar assumptions are common in the literature. See, for example, Drazen and Eckstein (1988), Doepke (2004), and Galor, Moav, and Vollrath (2006).

9

For simplicity, we assume that government consumption services are consumed in the first period of life only.

10

For example, the government could provide the same number of police officers and water facilities per household in the traditional sector. However, the geographic dispersion of these households implies that delivering the government services is more difficult, resulting in lower benefits.

11

The government objective function contains terms of the form βtln[(1 + q)t ], which must converge to zero as t increases in order for the objective function to be well-defined. Given that (1 + q)t > ln[(1 + q)t ] we have βt ln[(1 + q)t] < βt (1 + q)t =(β (1 + q))t, which converges to zero if β (1 + q)< 1.

Fiscal Policy and Economic Development
Author: Mr. Alex Mourmouras and Mr. Peter Rangazas