Keywords: Fiscal institutions, fiscal federalism, fiscal adjustment, budget deficits

Transcription

1 Fiscal adjustment with vertical fiscal imbalance: Empirical evaluation of administrative fiscal federalism in Norway *) Jørn Rattsø, Department of Economics, Norwegian University of Science and Technology, N-7491 Trondheim, Norway Abstract Fiscal institutions are important for policy outcomes. Recent research has taken benefit of the comparable institutions at the US state level to investigate the effects of budget rules and political structure. The present study attempts at extending the evidence to the administrative fiscal federalism in Norway. In the Scandinavian countries, decentralized government is part of the welfare state, and decentralized spending authority for welfare services is combined with centralized financing. The designed vertical fiscal imbalance and associated mandating restrict the room to maneuver for decentralized government and different fiscal adjustments are expected. In an intertemporal setting, it is shown that investment level is the main response variable to shocks. The fiscal performance also is related to ideological orientation and strength of the political leadership. The result is consistent with political economy studies of fiscal policy, where public investment is identified as a key macroeconomic policy parameter. The rigidity of current spending and revenue is in accordance with the US evidence that revenue response to shocks is constrained by regulations. Keywords: Fiscal institutions, fiscal federalism, fiscal adjustment, budget deficits Date: August 1999 *) The project is financed by the Norwegian Research Council. I am grateful to Morten Bostad and Carl Jacob Midtun for data collection, Bjørg Langset and Per Tovmo for econometric assistance, and Kåre Johansen for econometric advice. Comments at seminars in Trondheim, Umeå and Uppsala and at the European Public Choice Society are appreciated.

2 1 1. Introduction The design of fiscal federalism has received attention with reform in many countries (China, South Africa, old Soviet-Union) and the establishment of EU. While the conventional Musgrave-Oates-Tiebout model represents a productive starting point, in practice decentralization is not limited to local public goods and benefit taxes. The recent overviews of McKinnon and Nechbya (1997) and Shah (1994) acknowledge that there are benefits to both decentralization (spending, accountability) and centralization (taxation, externalities, equity concern). Most countries consequently end up with a mix of local autonomy and central control, and empirical evaluation of alternative assignments is of interest. A related literature addresses the effects of fiscal institutions, motivated by political bias towards excessive public spending and deficits. Roubini and Sachs (1989) started a series of studies of how economic and political institutions can account for differences in patterns of public sector spending and deficits across OECD-countries. The approach includes an evaluation of alternative institutions such as balanced budget rules, nominal fiscal targets and centralized budget processes. The institutional and political homogeneity of the US states has allowed an integration of studies of fiscal institutions and fiscal federalism. Interesting econometric evidence has been presented by Alt and Lowry (1994), Bayoumi and Eichengreen (1995), Bohn and Inman (1996), Holtz-Eakin (1988) and Poterba (1994). Poterba (1997) summarizes the empirical regularities regarding budget institutions and characteristics of state politics. He shows how variation in anti-deficit rules and tax limitation laws is associated with variation in fiscal actions, and that characteristics of state politics influence spending growth and debt accumulation. Here we extend the evidence to a different mode of fiscal federalism. By international standards, the control regime of the US states is quite decentralized. The combined decentralized spending and centralized financing in the unitary Scandinavian welfare states offers an alternative setting. The decentralization of redistributive welfare services has motivated mandating, strict regulation and grant financing, with following vertical fiscal imbalance. While the discussion in the US centers around competitive federalism versus cooperative federalism (Kenyon and Kincaid, 1991), the welfare state

3 2 arrangement is more of an administrative federalism. The design of the two systems looks very different, but they seem to move in the same direction. While mandating and tax limits are on the increase in the US, the Scandinavian countries attempt at allowing more local discretion through consolidation of matching grants into block-grants and extending local freedom to set fees and tax rates. The analysis of economic responses to fiscal shocks and the handling of budget deficits is well suited for international comparison. As stated by Poterba (1994), fiscal shocks are natural experiments that can be used to learn about government behavior. He finds that both tax limits and anti-deficit rules influence local behavior tax limits reduce tax response to deficits and strong anti-deficit rules increase spending response. In terms of variation in political conditions, he finds that single-party states, as opposed to divided government, raise taxes and cut spending by greater amounts following a deficit shock. This article investigates the responses to shocks in decentralized government in Norway. If circumstances otherwise are not too different, the centralized Scandinavian model is expected to bring responses similar to the most regulated cases in the US. The analysis allows a comparison of shock adjustments in different systems of fiscal federalism. Given the centralized control in Norway, it is natural to look beyond current spending and revenues as adjustment variables. We develop an explicit intertemporal framework and include the investment level as a key response variable. Investment as an important part of fiscal adjustment is recognized in the applied macroeconomics literature. Roubini and Sachs (1989, p. 108) write that in periods of fiscal consolidation capital expenditures are the first to be reduced, often drastically. On the empirical side, a general error-correction model is applied to separate between short run and long run effects. We address both the generation of shocks and the responses to shocks. The first step is to assess the importance of shocks for local governments. Access to budget data for 25 local governments in a county in Norway over 11 years allow us to measure shock as the gap between the ex ante budgeted and the ex post actual current surplus The development of the economic environment of the 25 municipalities during is laid out in section 2. The local unemployment rate and the private income growth characterize

4 3 the economic conditions and they are contrasted with the growth of local public current spending, investment and revenue. The shocks are defined and identified in section 3. The shocks are compared to the development of spending, revenue, current deficit and overall savings. The intertemporal model framework applied to understand local government adjustment to shocks is outlined in section 4. The empirical formulation of the analysis is discussed in section 5. Section 6 estimates how surplus shocks account for adjustments in current spending and revenue and investment. Asymmetric effects of positive and negative shocks and interaction effects of shocks and political characteristics are investigated. The long run determinants of fiscal balance are derived in section 7. Concluding remarks are summarized in section Local public sector in economic turbulence Decentralized government in the Scandinavian countries increased its importance in the economy with the expansion of the welfare state after WWII. Presently local and county governments represent about 2/3 of all government service production or in the order of % of GDP. The local public sector is responsible for the provision of the main welfare services such as kindergartens, primary school through high school, primary health care and hospitals, and care for the elderly. In addition, the local public sector provides more typical local public goods such as infrastructure and planning. A basic principle of the welfare state is to arrange equality and standardization of welfare services across the country, and national politicians are held accountable for its performance. Decentralization of the welfare state is consequently puzzling. The lack of clear guidelines for the design of decentralization with redistributive services and the related problems of heterogeneity and aggregation of preferences explain why countries experiment with different forms of local autonomy and central control (see Rattsø, 1998). While welfare services are decentralized under strict central government mandating, the financing of decentralized government is strongly centralized. Vertical fiscal imbalance results, with associated problems of accountability and local democracy. Partly the centralized financing represents instruments of macroeconomic control and partly the desire to equalize

5 4 the provision of welfare services. The welfare services are to be distributed independent of the economic conditions of the households and the localities involved. The centralized financing also is the result of inadequate tax bases for local taxation. Property taxes and user charges hardly anywhere account for the necessary % of GDP to finance welfare spending, and regulated income tax revenue sharing adds necessary funds. Income tax is less suitable as a benefit tax, and in particular when it is shared between local, county and central governments. Since income levels available for taxation are very different in urban and rural areas, a comprehensive tax equalization system de-links the relationship between local income tax rate and tax revenue. Local governments operate under legal and economic constraints set by the central government. All municipalities in Norway must obey the same rules, and three aspects should be noticed: - The centralized system of financing implies that about 80 % of the revenues consist of regulated income tax revenue sharing and lump sum grants. The general grants include tax equalization grants and spending needs grants. The rest of the revenues relate to matching grants, property tax in urban areas and from power production, and fees. Property tax and fees offer a limited opportunity of local discretion. - The Local Government Act requires budget balance of current revenues and expenditures including interest payments and debt service. Actual deficits are allowed to be carried over, but must be paid in an agreement with the central government within two years. - Investment can be financed by loans. The debt accumulation is constrained by formal approval of loans and the budget balance requirement whereby current spending inclusive of interest payments and debt service must be financed by current revenues. The 25 municipalities of the Sør-Trøndelag county have seen current revenues steadily increasing from on average 1990-NOK per capita in 1983 (USD 2.500) to 1990-NOK in 1993 (USD 4.000). Only very few of the municipalities have seen a decline in real current revenues during any year of this period. However, the municipalities have different economic structures and income levels. In 1990, current revenues per capita varied from NOK to NOK The richest are small and their revenues are based on equalizing grants and/or electric power production.

6 5 The Norwegian economy has had a turbulent period after 1980 compared to the earlier post- WWII development. The economy has experienced serious unemployment and economic fluctuation related to oil price shock and banking crisis. The turbulence created problems for most economic agents in the country, including local governments. The economic development presented in Table 1 shows the fluctuations as described by average private income growth (INCG) and unemployment rate (UNEMP) in the municipalities studied. The unemployment rate was gradually reduced during the boom, and reached a low 1.7 % in The economy moved into a cyclical downswing starting in 1987 and ending about 1993, and the unemployment rate was steadily rising in this period. The growth of private incomes has had a more erratic pattern. Fairly high growth continued through 1989, when the oil-driven recession started influencing private incomes. The local government current spending growth (EXPG) in the region under study has been relatively unaffected by the fluctuations of the overall economy. The average real spending growth, leaving out the reform year 1988, is above 3 % per year. Real investment growth (INVG) is expected to be more responsive to cyclical swings, and the boom brings with it strong investment growth with a lag. The downswing after 1987 also is reflected in reduced investment with a lag. The most pronounced aspect of the investment series is the large shifts from year to year, even in these averages. Rattsø (1999) shows how local sector investment in this period is part of a long time trend of slowdown and stagnation. Table 1 Local business cycle, spending and revenue Given the centralized system of financing in Norway, revenues are fairly independent of the local private economy. Using the national control of income tax revenue sharing and grants, the central government tries to modify the tendency for pro-cyclical income and expenditure growth in local governments. Since Johansen (1965, ch. VIII), the government has been well aware of the destabilizing role of the balanced budgets in the local public sector. Local governments therefore have been used as an active part of counter-cyclical policies. Borge and Rattsø (1999) show that systematic counter-cyclical revenue growth has been implemented over a long period. In the current dataset, the low current per capita revenue

7 6 growth (REVG) during 1986 and 1987 may be the result of central government tightening of budgets during the boom. The high revenue growth during , when unemployment rose, also is a counter-cyclical episode. The budgeted revenue growth (REVBG) has been fairly consistent with the pattern of actual revenue growth, reflecting the central government control of both forecasting and realizations. 3. The identification of shocks The documentation above shows that local governments to a large extent are sheltered from business cycles. Accordingly, we expect that the local governments are not much disturbed by economic shocks. As will come clear, economic shocks seem to have a significant economic importance anyway. Shocks are defined by comparing the budget decision and the ex post local government accounts. Shocks related to current revenues and expenditures are identified separately. Consistent with Poterba (1994), shocks are defined as the gap between actual current revenue and expenditure per capita (REV and EXP) and budgeted current revenue and expenditure per capita (REVB and EXPB). The budgeted values refer to the official budget decided before the start of the fiscal year. Poterba was also able to correct for tax and spending revisions during the fiscal year. An alternative measure of shock separating between expected and unexpected revenues based on estimated forecasting equations is suggested by Holtz-Eakin and Rosen (1993) and applied for Norwegian local governments by Rattsø (1999). To compare the two, a longer time series is needed than available for this study. It is probably productive to look at both budgeted and actual economic figures as decision variables of the local government. The budgeted revenue is not necessarily the 'best estimate' of next year revenue given decisions about tax and fee rates. The forecasting and budgeting of revenues and expenditures are part of the political decision making and is influenced by political priorities. The estimated revenue forecast often is decided by voting in the local council, where parties desiring higher spending argue for a higher revenue forecast. The shocks experienced consequently reflect a combination of economic disturbances and policy

8 failure. Important for this study, deviations from the budget are a concern for the economic management and must be tackled. 7 Economic shocks on the revenue side are basically related to the forecasting of income tax revenues and grants. The unexpected component of the shock represents exogenous shifts in local incomes and central government income tax revenue sharing and equalization grants. The revenue shock in local government i, year t is defined as: (1) REVSHOCK it = REV it - REVB it A positive revenue shock implies that the actual ex post current revenues are larger than in the official budget of the fiscal year. In the descriptive Table 2, the revenue shocks are measured as shares of current revenue, and the shares have been positive and very stable at an average of about 6-7 %. The local governments have a conservative estimate of next-year revenues. The central government sees to that the budgeted revenue is realistic. The success of this control is shown by the fact that only one of the municipalities in only one of the years has a negative revenue shock. The stability of the average revenue shock over time is remarkable given the economic fluctuations shown in section 2. Comparing Tables 1 and 2, and to be analyzed below, the cyclical swings seem not to affect the revenue shocks. Either the local governments have good predictions of the business cycles that they take into account. Or more likely, they trust that the central government cushions the economic fluctuations to generate a stable growth of local public revenues. The consistently positive revenue-shock should be understood on the background of central government forecasting. The central government estimates the growth of the local public consumption as part of the National Budgets. The National Budgets are based on policy-motivated low predictions of growth in prices and wages, and thereby they under-estimate the growth of local public consumption. Interestingly, a recent and similar study of economic shocks in Swedish municipalities, Lundberg (1998), shows the same pattern. Table 2 - Local government shocks and deficits

9 8 The gap between budgeted and actual current expenditure is harder to interpret. Most of the gap probably reflects lags in project implementation and adjusted spending plans. The unexpected component includes expenditures affected by changes in economic and demographic conditions. Similar to the revenue side, expenditure shock is defined as: (2) EXPSHOCK it = EXP it - EXPB it The expenditure shock on average is positive (except 1984) and on a rising trend according to Table 2. The standard deviation is about the size of the average. In 233 cases of the 250 studied, local governments underestimate their expenditures. They consistently reach higher spending levels than budgeted. The combined revenue and expenditure shocks add up to a current surplus shock : (3) SURSHOCK it = REVSHOCK it - EXPSHOCK it = (REV it - EXP it ) - (REVB it - EXPB it ) When the budgeted current revenues and expenditures are in balance, the minimum requirement regulated by law, the actual surpluses are equal to the surplus shocks. But they are not. On average, the budgeted current surplus is always positive and contributes to investment financing. The actual current surplus (SUR) on average is about 1/2 the volume of investment. Only 8 out of 275 observations show current deficits. There is no strong business cycle pattern in the deficits, contrary to the findings for US states of Poterba (1994). But the current surplus has been gradually declining over time. The gap between the surplus shock and the surplus in Table 2 has been about 5% of current revenues, reflecting the fact that the local governments have budgeted a steady current surplus on average over the period. Accordingly, the actual surplus has moved with the surplus shock and has a declining trend. During the overall boom , the actual current surplus was much larger than budgeted and nearly all of the municipalities experienced positive surplus shock. When the contraction came in 1988, the surplus shock turned negative on average and

10 9 for the majority of authorities. The positive revenue shock was about 'normal', but the positive expenditure shock dominated. More puzzling, the actual surplus was smaller than budgeted on average during the upswing The explanation is the dominance of large positive expenditure shocks. And still about half of the municipalities had a positive surplus shock. There is a tendency of decreasing surplus shocks from positive values on average in the mid- 1980s to negative values on average in the early 1990s. The declining surplus shocks are associated with the rising expenditure shocks. The fluctuations of the investment share (INV) were commented in relation to Table 1, and the share has been about % of current revenues on average. The trends and the fluctuations of investments and surpluses add up to the time pattern of local government contribution to national savings (SAV). The savings contribution has been negative on average, and the drop in current surpluses explains a rising (negative) share of revenue. Only low investment in the last part of the period helps modify the absorption of savings from the rest of the economy. The steadily higher negative savings are a concern if it continues over too long a period. 4. Model of shock-adjustment There is no established workhorse model of local government intertemporal decision making to analyze shock adjustment. Gramlich (1978, 1991) offers a simple model based on a community preference function including private consumption, local public consumption and stocks of balances. Given the proper budget constraint, demand functions for current expenditures and revenues and stocks of balances are derived. Bohn and Inman (1996) also develop a desired deficit concept different from the actual, but they are not explicit about the intertemporal context. The most complete dynamic local fiscal model available is Holtz- Eakin and Rosen (1993), but this requires a dataset for a long period to reveal intertemporal preferences. Alt and Lowry (1994) propose a different model inspired by the tax smoothing literature. They assume that expected revenue is set as a share of permanent revenue, taking into

11 10 account any unexpected deficit. Expected expenditures are set equal to expected revenue plus adjustments for transitory business cycle factors and unexpected deficit. In this way, revenue sources only affect revenues directly and business cycle factors only influence expenditures directly. It follows that they assume different revenue and expenditure decisions. We propose a two-period framework to spell out the dynamic aspects of shock adjustment. Assume a two period model (periods 1 and 2) with local government consumption spending EXP 1 and EXP 2 financed by revenue REV 1 and REV 2. The evaluation of utilities and costs will be discussed below. At this stage, the local government enjoys utility of consumption spending, U 1 (EXP 1 ) and U 2 (EXP 2 ), assumed to be additively separable over time, and with standard properties (U i > 0, U i < 0). The deadweight loss of distortionary taxes in each period is assumed to be a convex function of the revenue level. The tax technology is consistent with tax smoothing over time. The utility cost of revenues is represented by V 1 (REV 1 ) and V 2 (REV 2 ), with V i > 0 and V i > 0. The dynamics of the investment process will depend on the flexibility of adjusting the capital stock. It is typically assumed that the relationship between the desired stock of real capital and the investment level is influenced by adjustment costs and that the unit cost of new capital increases with the size of the investment level (see Holtz-Eakin and Rosen, 1993). Investment INV in the first period produces production capacity for period 2, and government consumption in period 2 is assumed to be determined by this investment priority, EXP 2 (INV), with EXP 2 > 0 and EXP 2 < 0. To investigate the effects of a shock in period 1, we introduce a shock parameter α in the social cost function, V 1 (REV 1, α). Compared to the empirical study, it is simpler here to investigate a revenue shock, which is a key component of the surplus shock. A positive revenue shock in period 1 can be interpreted as a cyclical rise in the tax base, and the marginal cost of generating revenue REV 1 is reduced, with V iα < 0. When the time preference rate and the interest rate are set to zero, the intertemporal budget constraint is (4) EXP 1 + EXP 2 + INV REV 1 REV 2 = 0

12 11 Net social welfare of the fiscal policy is (5) W = U 1 (EXP 1 ) + U 2 (EXP 2 (INV)) - V 1 (REV 1, α) - V 2 (REV 2 ) Inserting the budget constraint and maximizing the net welfare, we reach the following simple first order condition: (6) U 1 = U 2 EXP 2 = V 1 = V 2 In this setup, the marginal utility of local government spending must be equal in the two periods, the marginal costs of local government revenue must be equal over time, and the marginal utility of spending must equal the marginal cost of revenue. The marginal utility of consumption spending is equal to the marginal utility of investment spending, given the production function assumed for period 2. In the special case of identical preferences and identical social costs across periods and assuming that the marginal product of investment spending is equal to 1, local government spending is constant over time, EXP 1 = EXP 2. The consequences of a temporary revenue shock in the first period can be derived by differentiating the first order conditions: dexp 1 / dα = V 2 Z V 1α 1/D > 0 dinv/ dα = U 1 V 2 V 1α 1/D > 0 dexp 2 / dα = EXP 2 dinv/ dα > 0 drev 1 / dα = [(V 2 - U 1 ) Z + U 1 V 2 (EXP 2 + 1)] V 1α 1/D > 0 drev 2 / dα = U 1 Z V 1α 1/D < 0 d(rev 1 dexp 1 )/ dα = [U 1 V 2 (EXP 2 + 1) - U 1 Z ] V 1α 1/D > 0

13 12 d(rev 2 dexp 2 ) / dα < 0 where D = U 1 (V 1 + V 2 ) Z - U 1 V 1 V 2 (EXP 2 + 1) > 0 and Z = U 2 EXP 2 + U 2 EXP 2 < 0 The positive revenue shock reduces the marginal costs of generating revenue REV 1. When marginal social costs of revenue goes down, net welfare can be increased by raising local government spending and investment in period 1. When the revenue shock is temporary, the relative marginal costs of revenue across periods are changed, and the revenue is shifted from period 2 to period 1. The first period current surplus goes up, since part of the revenue raising in this period finances the higher investment. In the following period, the current surplus is reduced since current revenues go down and current expenditures go up. A permanent revenue shock reducing marginal costs of revenue in both periods will increase spending and revenue in both periods. The other component of a surplus shock, the expenditure shock, may work similar to the revenue shock, but with opposite sign. A negative expenditure shock can be understood as a negative shift in the marginal utility of expenditure in the first period. Then current spending goes down, while investment goes up. Revenue raising will be reduced in both periods. It follows that current surplus in period 1 is raised, while current surplus in period 2 goes down. The budget constraint assumes that the government can choose a time path of current expenditure and investment spending using financial markets to cover deviations from the current income stream. This is a crucial assumption. According to law, Norwegian local governments must balance current spending inclusive of interest payments with current revenue. They are allowed to choose whatever investment level given this restriction and typically about half of investment is financed by loans. Current surplus on average finances the other half, but also current deficits are observed and are allowed to be carried over. Consequently it seems realistic that the local public sector has been able to take advantage of the flexibility assumed by the above budget constraint. Poterba (1994) analyzes same-period responses to shocks. With strict budget balance rules,

14 13 local governments are forced to raise revenues and/or reduce expenditures in the same period as a negative surplus shock. In our intertemporal understanding, a negative surplus shock may lead to immediate reduction in current revenues. A reduced revenue base increases the marginal costs of funds. In the future, revenues will be increased and expenditures will be reduced, similar to the same-period adjustments in the balanced budget case. The dynamics of the adjustments are analyzed below. 5. Empirical setup The first order conditions of the optimization model describe the behavior of each local government given the budget constraint, production function and revenue technology. The empirical model is consistent with this theory, but adds a richer description of the underlying economic and social conditions. The independent variables representing present and future conditions can be divided into three groups: The economic factors include regulated income tax revenue sharing and lump sum grants, private incomes, and economy-wide economic performance. Various local characteristics, mainly demographic, are included as preference variables. In an environment of full information and stability and no constraints, the demands are satisfied and the current deficit always is desired. Needless to say, the desired deficit does not necessarily reflect the preferences of the voters and is not expected to satisfy general criteria of efficiency. The political institutions form the outcome. Actual implementation of local government priorities will reflect economic disturbances, which is represented by the surplus shock (SURSHOCK ). The dynamics can embody adjustment inertia and the responses to the surplus shock may imply asymmetry and multiplicative effects. The econometric formulation assumes an error correction model that allows for a separation between temporary and permanent effects. The demand variables are dependent of current and past economic, demographic and political conditions. Given the limited time span of the data, we have not experimented with expected and permanent values. The surplus shock is introduced as a separate variable affecting the actual values, and interaction effects are

15 14 analyzed with respect to the political descriptives. Adjustment inertia are represented with lagged dependent variables and by the clear separation between short run and long run effects. Symmetry in the response to surplus shocks is investigated by separating between positive and negative surplus shocks. To account for adjustments related to the stock of debt, lagged net interest payments (INT) are included. The economic conditions are described by two variables. The local governments receive regulated income tax revenue sharing and lump sum grants from the central government (EXOREV). The balancing of the income tax rate applicable in the revenue sharing and the volume of grants is an issue of business cycle policy. Typically the revenue sharing tax rate is reduced and the grant share increased during booms, so that municipalities have about the same revenue development whether they are dependent on tax revenues or grants. It follows that a separation of revenue sharing and grants effects have little meaning. Matching grants have played a minor role during the period studied, but they have been increasing after a reform consolidating them into a block grant. To account for business cycle factors, we have included data about the local unemployment rate (UNEMP). The unemployment rate captures economic instability that may affect the implementation of budget policies. Preferences for local public services are assumed to vary with demographic factors. Since most of the services are directed towards specific age groups, in particular the young and the old, the age composition of the population is assumed to be an important demand factor. Given the low mobility between municipalities, we take the age composition of the population and the population size (POP) as exogenous. The age composition is described by the share of the population below 15 years of age (YOUNG) and the share above 67 years old (OLD). Voter preferences are channeled through the political system, and characteristics of the political structure are expected to affect the local public resource use. The formulation allows a comparison with the US studies of divided government. Kalseth and Rattsø (1997) suggest several measures of political strength in the Norwegian system of multi-party proportional representation, and they are shown to influence administrative spending. Here political strength is measured by a Herfindahl index of party fragmentation in the local council. The

16 15 variable HERF has 1 as the highest value when only one party is represented in the council. To measure the ideological orientation of the political leadership, the share of socialist representatives in the local council (SOCSHA) is included. Dummy variables for socialist share above 50 % (DSOCSHA) and Herfindahl index above 0.7 (DHERF) are used to investigate interaction effects (due to limited data). The data base covers 275 observations, 25 municipalities in 11 years. All economic variables (except the unemployment rate) are measured in 1990-NOK and per capita terms. The political variables reflect the 4-year term of the local council, and elections have been held in 1983, 1987 and No formal tests of the time series properties of the variables are reported. It is assumed that the surplus shock is stationary and that the key local decision variables current expenditure, current revenue and investment are non-stationary, but I(0) on first difference form. The variables describing local economic conditions, private income and unemployment rate, are treated as I(1), although the unemployment rate in the very long run is expected to be stationary. The political and demographic characteristics are assumed to be stationary. In the long run they can only influence the level of the local decision variables. It should be noticed that both the current surplus and the overall savings are allowed to be nonstationary, basically implying that financing of investments develops with the per capita investment level. The error correction form of the basic version of the model is: REV, EXP, INV = α0 + α1 DEPENDENT -1 + α2 SURSHOCK -1 + α3 EXOREV + α4 EXOREV -1 + α5 UNEMP + α6 UNEMP -1 + α7 OLD -1 + α8 YOUNG -1 + α9 INT -1 + α10 POP -1 + α11 SOCSHA -1 + α12 HERF -1 The short run effects can be read directly from the variables on first difference form, and the short run effect of a change in exogenous revenue is reflected by α3. The permanent effect of higher exogenous revenue is estimated by -α4/ α1. The estimation method takes into account community specific effects by including dummy

17 16 variables for each municipality, the fixed effect model. The formulation utilizes only time series variation in the data. Time dummies are introduced to control for changes in other factors over time. When a fixed effect model includes lagged values of the dependent variable, OLS-estimates are biased even if the residuals are white noise. To obtain consistent estimators, we use the Generalized Method of Moment (GMM) developed by Arellano and Bond (1991). The model is first-differenced to remove the individual fixed effects. The GMM estimators are calculated using the program package DPD written in Gauss, as documented by Arellano and Bond (1988). One-step estimates robust to cross-section and time-series heteroscedasticity are reported. If the error terms in the basic version of the model are serially uncorrelated, then the first-differenced error terms are a MA(1)-process implying negative first order serial correlation and absence of second order correlation. The models reported have acceptable dynamic properties. A WALD χ 2 -test of joint significance of all included variables is documented. 6. Fiscal adjustments to shocks According to the descriptive statistics in section 3, there is no obvious relationship between economic fluctuations in the local private economy and the measured shocks to the local public economy. An econometric investigation (not reported) confirms that revenue shocks, expenditure shocks and surplus shocks are independent of short run changes of private incomes and unemployment. The result is consistent with an understanding of the local public sector as insured against macroeconomic shocks. Central government tax revenue sharing policy, tax equalization grants and spending needs grants break the link between the local private and the local public economy. Poterba (1994) finds that the unemployment rate explains only a small part of the fiscal shocks in US states. Gaps between budgeted and actual revenue and expenditure and thereby budgeted and actual surplus seem to be determined in the area of planning and forecasting. There is systematic and persistent underprediction of revenues and expenditures. The net surplus shock effect of this underprediction varies over time and space. Since the underprediction of revenue on average is stronger, positive current surplus shocks dominate. Shocks measured in this way are not

18 17 necessarily unexpected, but they do require an adjustment to fiscal imbalance. Estimation results of the demand model for current revenues, current expenditures and investment expenditures are reported in Table 3. The consequences for current deficit and overall savings are derived. The econometric formulation assumes that the surplus shock in period t influences the change of revenue and spending from t to t+1. The results imply that surplus shocks should be included to have a full understanding of local fiscal determinants. A surplus shock has no statistically significant effect on current revenue or expenditure in the following year. Current revenue and spending are held independent of shocks. Our interpretation is that local governments under the model of administrative federalism and vertical fiscal imbalance have little room to maneuver in the short run. Regulation and mandating allow little flexibility in short run resource allocation. Poterba (1994) estimates a same-period revenue adjustment of 45 % of the shock in US states and find that the revenue side takes most of the adjustment. Lundberg (1998) identifies no revenue adjustment to shock in Swedish municipalities, and concludes that the expenditure side absorbs the shock with reductions of about 40 øre per krone deficit shock. Current revenue responds to short run changes of exogenous revenues: (7) REV = 1.07 EXOREV Approximately all of a change in exogenous revenue ends up as change in current revenue, in a close to 1:1 relationship. Higher tax revenue sharing and grants from the central government are not offset by other revenue instruments. This confirms the designed vertical fiscal imbalance. Current spending seems to be driven by long run demand factors, to be discussed. Since business cycles are taken into account in the central government determination and balancing of tax revenue sharing and grants, separating out cyclical effects are impossible in the Norwegian system. Table 3 Fiscal adjustment to shock

19 18 When short-run response in current revenue and expenditure is absent, shocks are channeled into intertemporal fiscal adjustment. Investment responds to revenue and expenditure surprises compared to the budget. Surplus shocks have a clear and negative effect on investment. A deficit shock reduces investment by about 1/3 of the shock in the following year. The short run investment determinants are: (8) INV = 0.35 SURSHOCK EXOREV Short run exogenous revenue growth is channeled to investment. About 25 øre of every krone in higher exogenous revenue end up as investment. Investment is independent of variation in unemployment, and consequently there is no sign of countercyclical local public policy. We have investigated whether the effects of the shocks are asymmetric, that is whether a positive surplus shock has a different effect from a negative shock (higher deficit than budgeted). Results are reported in Table 4. The results confirm the independence of both current revenue and expenditure of both positive and negative shocks. However, asymmetry seems to be important for investment. A positive surplus shock leads to a stronger increase in investment (about 40 øre) than the reduction in investment following a negative shock (about 30 øre). There is consequently an upward investment bias in the response to shocks. This asymmetry is inconsistent with both Poterba (1994) and Lundberg (1998), who both find stronger responses to deficit shocks compared to surplus shock. They have not investigated the reaction of investment. But the results are consistent with public investment analyses separating between expected and unexpected revenue. Both Holtz-Eakin and Rosen (1993) and Rattsø (1999) find that unexpected revenue, based on a forecasting model, increases local public investment. Table 4 - Asymmetric fiscal adjustment and politics The lack of short-run responses to shocks limits the investigation of political effects. The US background is the Poterba (1994) result that single party states raise taxes and cut spending by greater amounts to deficit shocks. In the multi-party proportional representation system in Norway, the fragmentation of the local council measures the same kind of decision problems

20 19 as divided government in the US. The fragmentation measured by the Herfindahl-index (dummy variable) has no effect on the investment response to shock. Table 4 reports an important interaction effect between shocks and ideology for investment. While a nonsocialist majority council (dummy variable) has an investment response close to 50 % of a shock, socialist majority councils adjust investment spending by only 11 %. Positive surplus shocks dominate, and socialist controlled councils seem to put much less emphasis of channeling these into investment. The limited dataset has not allowed a combined check of asymmetry and political interaction effects. Current deficits are the result of current revenue and spending decisions. It follows that we can derive determinants of current deficits from the estimates in Table 3. The short run relationship can be seen directly from equation (7). Since short-run changes in exogenous revenue influences current revenue about 1:1 and current expenditure is not affected, a positive shift in exogenous revenue will reduce the deficit by the same amount. The adjustment rule can be understood as smoothing to a long run spending plan (Barro, 1979), but is an extreme result in our intertemporal model with endogenous spending and revenue. Since the surplus shock in general has no effect on current revenue and spending, the actual surplus is independent of surplus shocks. The local government contribution to the national savings pool is the result of both current surplus and investment. Combining the three equations of Table 3, we can calculate the short run determinants of overall savings: (9) SAV = SURSHOCK EXOREV A surplus shock reduces the local government savings in the following year. Since surplus shocks increase investment, they leave less savings for the rest of the economy. Higher exogenous revenue imply higher current revenues and higher investment. The net effect is positive, and more exogenous revenue stimulates savings.

21 20 7. Long run determinants of fiscal balance The long run determinants of current revenue and expenditure involve exogenous revenues, demographics and politics. Since the time series is only 10 years, no strong conclusions can be drawn about the long run structure. The estimates imply an elasticity of current revenue with respect to exogenous revenue of about 1.7, presumably reflecting the increased role of earmarked grants. Current expenditure over time is driven by the age composition of the population, with high shares of young and old leading to higher spending. The result is consistent with the investigation of demographic effects by Borge and Rattsø (1995). To be expected, long run investments are determined by the same demand factors influencing current expenditure. In particular the share of old in the population has a clear effect on investment spending, probably reflecting the capital intensity of institutional care for the elderly. A high level of debt holds back both current and investment spending. The lack of income effects on investment is at odds with the documentation of clear short and long run effects of revenue by Rattsø (1999). Changes in local political characteristics have systematic effects on local government revenue generation. According to Table 3, socialist orientation drives up current revenues, while political strength holds down the revenue level. The effects are consistent with the understanding that ideological dominance and ability to hold back interest group pressure matter. This understanding of the importance of the political structure in Norwegian local governments was first documented by Kalseth and Rattsø (1997) in an analysis of administrative spending. The Herfindahl index measures party fragmentation of the local council, and less fragmentation is interpreted as more political strength. The size of the effect is of economic importance. Current revenue increases by more than 100 NOK per capita for each percentage point increase in the socialist share. The effect of strength is harder to interpret quantitatively, but one standard deviation increase in strength raises current revenue by about NOK 1500 per capita. The effects are fairly consistent with the long time series analysis of revenue growth by Borge and Rattsø (1999). Looking at the coefficients of the variables representing socialist orientation (SOCSHA) and strength (HERF), we see that each of them has about the same quantitative impact on current

22 21 spending and revenue (although the spending effects are not significant at 10 % level). Socialist orientation means higher current level of activity and strength holds back the level of current revenue and spending. The net effect on the surplus is negligible. If we emphasize significance at the 10 % level, socialist influence means higher revenue and thereby larger surplus, while strength is associated with lower revenue and therefore lower surplus. The identification of political characteristics as important determinants of fiscal behavior is in accordance with several recent international studies. The main conclusion of Poterba (1994) is that single party control of states is able to hold down deficits and they adjust more quickly. Clingermayer (1991) is in agreement with Poterba - states with divided government have higher deficits. In the Norwegian context, Borge (1995) documents that strength reduces deficits in a study covering all municipalities. Another aspect is analyzed by Baber and Sen (1986), who find that there is a stronger positive effect of deficit in election years in states with divided government. Contrary results are provided by Bohn and Inman (1996). They find no evidence that divided government or party labels have a significant effect on deficit behavior. The importance of ideology has been investigated in other studies. Borge's (1995) Norwegian study implies a significant and positive effect of socialist influence on actual deficits. His interpretation is that socialists use deficits to expand local government spending in a situation with largely exogenous revenues. In neighboring Sweden, Lundberg (1998) concludes that non-socialist local governments adjust faster to deficit shocks than socialist. This is consistent with his finding that expenditure reduction takes most of the adjustment. Other studies have shown socialism to be associated with deficits. In a broad analysis of OECD countries, Alesina, Cohen and Roubini (1993) and Roubini and Sachs (1989) document a positive relationship between public deficit and socialist influence at the national level. Also the responses to the actual deficits are analyzed (not reported), and they allow a comparison with conflicting evidence regarding the pattern of fiscal adjustment to actual deficits. In studies of the US local public sector, Alt and Lowry (1994) conclude that current revenue is the main adjusting factor, while Bayoumi and Eichengreen (1995) and Bohn and Inman (1996) emphasizing regulation constraints find that current spending takes most of the

23 22 change. Our results indicate that investment also take the adjustments to actual deficit in Norway. The size of the effect is a bit higher than the responses to the surplus shock. Given that the budgeted surplus is fairly stable, we expect the effects of actual surplus and surplus shocks to be of similar size. The long run equilibrium relationships between current surplus, savings and economic and political variables can be derived from Table 3. In the long run context, the current surplus should be interpreted in terms of investment financing. Exogenous revenue clearly increases the surplus over time, but does not affect the investment level. It follows that higher exogenous revenue is associated with less external financing of investment. To derive the effects of political variables on savings, we also have to take into account the investment behavior. Socialist influence increases savings by holding down investment. 8. Concluding remarks The analysis has addressed fiscal adjustment under administrative federalism in Norway. Experiences of shocks and responses to shocks in the local public sector are investigated. A dataset covering budgets and accounts of local governments has been available, and shocks are defined as deviations from budgets in the current fiscal balance. An interesting aspect of this centralized system is the lack of a clear relationship between economic fluctuations in the local private economy and the shocks to the local public economy. Local governments systematically and persistently underpredict revenues and expenditures. Since the underprediction of revenue on average is stronger, positive current surplus shocks dominate. The central government insures the local public sector against macroeconomic shocks to avoid the destabilizing effects of the balanced budget regulation. Conservative forecasts in the national budget lead the locals to underpredict their activity. While the studies of shocks in the US states discuss the balance between current revenue and spending adjustment, this centralized system of fiscal federalism seems to hold current revenue and spending independent of shocks. Mandating and regulated revenue sources reduce the room to maneuver, and shocks are channeled into investment response. A deficit