Political Budget Cycle and Government s Arrears

Transcription

1 Political Budget Cycle and Government s Arrears Marco Buso Luciano Greco Luigi Moretti June 2017 Preliminary and Incomplete Draft Abstract In this paper, we investigate how payments on public investment arrears are influenced by the political cycle. We first develop a theoretical model where local politicians, who imperfectly observe their skills along the political cycle, take public finance decisions, face elections and term limits. Each year, incumbent politicians set the level of commitments for current and investment expenditures, taxes and loans; then, they decide the level of payments on committed expenditure (including arrears). Testable predictions suggest a political budget cycle where taxes, commitment (investment) expenditure, and payments are used strategically by incumbent politicians to influence electors perception of their skills. We test such predictions using a dataset of Italian municipalities for the period Our empirical findings confirm that politicians tend to implement fiscal consolidation (i.e., increase of taxes and reduction of loans, commitment expenditure and payments) early after taking office; along the political cycle, commitment investment expenditure and taxes grow until the pre-electoral year, then dramatically shrink in the election year. We show that incumbent politicians regulate the timing of expenditure implementation by steadily increasing payments (on arrears) along the political cycle, reaching the maximum in the election year. Keywords: Municipal elections, Local fiscal policy, Term limits. JEL classification: D72, D78, E62, H72, H76. dsea and CRIEP, University of Padua. dsea and CRIEP, University of Padua. Université Paris 1. 1

2 1 Introduction A large literature has investigated in the last three decades political budget cycles, building on previous contributions on political business cycles (e.g., Alesina and Paradisi (2014)). Theoretical contributions (e.g., Rogoff and Siebert (1988); Rogoff (1990); Shi and Svensson (2006)) have provided an interpretation of such behaviors based on a signaling problem in a framework affected by asymmetric information between elected officials who benefit of better information, take unobserved actions and have private agendas and citizens who try to elicit relevant information about politicians quality (e.g., honesty, competence, etc.) to improve their own welfare by elections. Cross-country and within-country empirical investigations have forged a consensus about the general relevance of the political budget cycles, and about the role of some key variables in some specific contexts (e.g., local property taxes in Italy). However, research is still needed to understand which possible theories could explain marked differences in size, timing and relevant variables of political budget cycles. Recent contributions have, for example, highlighted the relevance of political cycles in the case of Italian municipalities (e.g., Cioffi et al. (2012); Alesina and Paradisi (2014); Repetto (2015)). In this paper, we focus on the role of payments (i.e., cash expenditure) and, in particular, of payments on public investment arrears. The latter are residual commitments for investment expenditures that have been decided in previous years and not yet implemented. The extant literature has not yet focused on this fiscal policy tool which, nevertheless, proves to be quite relevant to understand the dynamics of the political budget cycle. The discrepancy between expenditure commitments and payments allows the local governments to de facto issue a form of public debt (i.e., arrears). Though previous analyses have highlighted the impact of the political cycle on taxes (Alesina and Paradisi, 2014), the behavior of expenditure has been more difficult 2

3 to identify. This is partly determined by sluggishness of current expenditure commitments (e.g., public employment, debt service, etc.), which explains our interest in investment expenditure commitments. Another explanation is that, when fiscal rules require to balance the budget and limit public debt as it is the case for subnational governments in many countries like Italy, the reduction of taxes during election years requires a corresponding restraint on (investment) expenditure commitments. Thus, we can show that the size of the political budget cycle is underestimated if we exclude payments (i.e., implemented expenditure) dynamics from our analysis and we focus on taxes (and expenditure commitments). These considerations explain our focus on the impact of political cycle on payments on public investment outstanding commitments (including arrears). In the following we first develop a theoretical model where local politicians face term limits and take public finance decisions (Section 2). We model a fiscal policy decision process where incumbent politicians decide the level of commitments for current and investment expenditures, taxes and loans, at the beginning of each fiscal year. Then, during the fiscal year, the incumbent politicians decide the level of payments on committed expenditures, including investment arrears. Incumbent politicians are interested in their rents (and thus in their political careers), are risk neutral and can be elected in the local office for a maximum of two terms by citizens. As usual in political budget cycle models, citizens have to infer the quality of politicians by perceived effects of fiscal policy (e.g., taxes and expenditure payments). At the equilibrium, risk-averse politicians tend to accumulate fiscal resources early in their political cycle, and to invest them (by increasing payments and reducing taxes) as far as exogenous uncertainty unfolds in order to avoid to send bad signals to the citizens along the political cycle. In Section 3, we test the theoretical predictions using a dataset of Italian municipalities for the period Our empirical findings confirm that politicians tend to implement fiscal consolidation 3

4 (i.e., increase of taxes and reduction of loans, commitment expenditure and payments) early after taking office; along the political cycle, commitment investment expenditure and taxes grow until the pre-electoral year, then dramatically shrink in the election year. We show that incumbent politicians regulate the timing of expenditure implementation by steadily increasing payments (on arrears) along the political cycle, reaching the maximum in the election year. 2 A Theory of Political Budget Cycle [THE THEORETICAL MODEL IS VERY PRELIMINARY] An important innovation of our analysis, with respect to extant literature on political business cycle, is the focus on the specific functioning of the government budget cycle. As in the real world, the budget cycle is divided in periods (say, fiscal years); each period starts with the approval of the (provisional) budget and ends with the approval of the next (provisional) budget. We also distinguish fiscal policy decisions of local politicians on the basis of the accounting criterion and the timing during the budgetary period: accrual or commitment basis: at the beginning of the budgetary period, the local government approves a budget that formally authorizes the appropriation of forecasted public revenues and commits public money to be spent in different fields up to specified amounts (i.e., committed expenditures); cash basis: during the budgetary period, the government cashes in revenues and pays committed expenditures, independent of the budgetary period (current or previous) when commitments were decided. Decisions taken on commitment and cash basis (both as regards public revenues and expenditures) may diverge for a number of reasons (e.g., revenue forecasts may be 4

5 partly wrong, investment expenditures take time to be implemented and time overruns may materialize). In this paper, we particularly focus on the mismatch between expenditure commitments and payments that typically gives rise to arrears. In the real world, arrears arise because the expenditure cycle is longer than one period, for several reasons. The crucial implication of this is that payments implemented in one specific budgetary year may be larger or smaller than expenditures committed in the same period. They are smaller if part of the expenditures committed at the beginning of that period are (for any reason) cashed out in future periods, thus forming new arrears. Conversely, payments are larger than expenditures committed at the beginning of a specific period, when part of the stock of past arrears (i.e., expenditures committed in past periods) are cashed out during the considered period. We formalize a model of political budget cycle where politicians perform career concern, aim at rent extraction and are uncertain about their competence (or skill), though they progressively observe better signals of it with respect to citizens who to decide whether to re-elect them (e.g., Rogoff and Siebert (1988); Rogoff (1990); Shi and Svensson (2006)). Without loss of generality, we assume that the political cycle is made by three phases, corresponding to local government budget periods (or years) 1 : A after the election period, the elected government (or mayor): manages the municipal budget that was approved by the previous government, before the election period (e.g., affecting the flow of payment on committed expenditures); learns if a random exogenous shock anticipates the election period; then approves the main budget variables (i.e., tax revenues, expenditure commitments, loans) for the next budgetary period; 1 The model can be easily extended to include an arbitrary number of budget periods. For example, in Italy the normal political cycle at local level includes five budget periods. 5

6 B if no snapshot election takes place, in the period before election, the incumbent mayor manages the municipal budget that she approved in the previous period and approves the main fiscal variables for the election period; E in the election period be it the natural end of the political cycle or the anticipated one the incumbent mayor just manages the budget that was approved in the previous period, while she can approve the next period budget only in case of re-election. In the more general case, the political cycle is repeated an infinite number of times. We assume that the (commitment) budget constraint of the local government, at the generic fiscal period t, is given by: τ t = i t (1) where: τ t are the total tax revenues levied by the considered local government at time t; i t are the committed investment expenditures at the same period. For the sake of simplicity, we abstract from current expenditures (which are, to some extent, more persistent than investment ones) and from loans 2. As above argued, we make the realistic assumption that the local government determines during the generic budgetary period the part of committed expenditures (both current and past) that has to be transformed in payments. 3 Therefore, we introduce the government budget constraint on cash basis at the generic fiscal period t: p c t + p a t = r t + g t θ t (2) 2 In the extended form of the model considered in the Appendix, we show that our main results are unaffected by this simplifying assumption. 3 For the sake of simplicity, we assume that tax revenues are all cashed in during the current period. 6

7 where: p c t are the payments decided by the local government during the period t out of current committed expenditure; p a t are the payments decided out of the stock of accumulated arrears; r t is the part of such payments that is transformed in private rent for the local politician; g t is the amount of public investment good that is provided to the local community transforming the residual part of payments; and θ t is the randomly distributed politicians productivity that determines how many units of public good can be obtained for each unit of monetary payment. In particular, with probability λ (or 1 λ), θ t is distributed according to the cumulative function F (θ t θ) (or F (θ t θ)), such that E(θ t ) = θ (or E(θ t ) = θ < θ). In each period, two additional equations link the dynamics of arrears to committed expenditures and payments: i t = p c t + a t (3) A t = A t 1 + a t p a t (4) The incumbent politician maximizes the following inter-temporal utility: W t = r t + δe(w t+1 ˆθ) for t {A, B} (5) W t = r t + δπe(w t+1 ˆθ) for t = E (6) where: δ is the intertemporal discount factor of the politician; ˆθ is the politician s estimate of her own productivity (possibly based on Bayesian updates of private signals obtained in previous periods); π is the probability that the representative citizens vote for the incumbent during the election period. [MORE ASSUMPTIONS AND ELABORATION OF THE MODEL HERE] From our model we obtain the following main predictions: Proposition 1 (Political Budget Cycle) After the election year, the fiscal pol- 7

8 icy is tight (i.e., high taxes, low loans, low commitments for investment expenditure, low payments on arrears). Along the political cycle, commitments for investment expenditure increase peaking in the pre-electoral year. In the electoral year, taxes, loans and commitments for investment expenditure are squeezed. Payments on investment expenditure arrears grow along the whole political cycle peaking in the election year. Proposition 2 (Term Limit) The political budget cycle affects the fiscal policy in the same way in both terms. However, the fiscal policy is tighter in the first term. In particular, commitments for investment expenditure and payments on arrears are higher in the second term. 3 Empirical Evidence from Italian Municipalities 3.1 Data and variables definitions In Italy, municipalities are the smallest administrative units and they provide public goods and services in several policy areas, such as local transport, local police, culture and recreation, land management and environment (waste disposal, water, sewage), nursery school and complementary education services, and registry services. About half of the total government investment expenditure is managed by municipalities. For instance, municipalities manage the outsourcing (through competitive auctions, to private suppliers) of about 50% of the public works (such as road works and building constructions). 4 Annually, each municipality is obliged to transmit its accounting and financial report to the Ministry of Interior (the so-called Certificati di Conto Consuntivo ). This source of information allows us to have a clear picture of the financial situation 4 See, for the years of our analysis, the Annual Reports (Relazione annuale) of the Italian Authority for the Supervision of Public Contracts (AVCP); documents available at and 8

9 of the municipalities for each year for both the revenues and the expenditures sides of the budget. In particular, in our empirical analysis we will focus on the payments on arrears for investment commitments (Payments on arrears) that represents the amount of real euros per capita that the municipalities pays in a given year out of the stock of non-paid past investment expenditure commitments (i.e., Stock of arrears). Annual payments on arrears in our sample are on average about 271 euros per capita, and they represent less than one-third of the stock of arrears. We will also analyze other budget variables to have a broader picture of the effects of the political cycle on budgetary choices, so to highlight the role of payments on arrears. In particular, we will look at (i) Investment expenditure commitments, which represents the amount of euros per capita that a municipality, in a given year, decides to allocate to new investment projects (on average, in our sample, 396 euros per capita); (ii) Current expenditure, which represents on average an higher share of the budget (on average 577 euros per capita) and are less volatile than the investment expenditures. On the revenue side, we will consider the effects of the political cycle on (iii) Tax-revenues (on average 240 euros per capita), (iv) Non-tax revenues (on average 143 euros), and (v) new Loans that are more volatile than the previous two sources but are smaller on average amount (77.5 euros per capita). Since 2000, in Italy, in each municipality, the mayor and city s council are elected for a five-year term and each mayor has a limit of two consecutive terms. Elections do not take place on the same day for every municipalities, so in a given year, different municipalities can find themselves at a different point of their political cycle. Coviello and Gagliarducci (2012) and Repetto (2015) provide narrative analysis based on historical and institutional changes for considering the election dates of Italian municipalities as exogenous. 5 Data for political cycle at the municipality level are from the Italian Ministry 5 See also Alesina and Paradisi (2014). 9

10 Table 1: Sample summary statistics (Real euros per capita). Period VARIABLES Mean SD Payments on arrears Investment expenditure comm Current expenditures Tax revenues Non-tax revenues New loans Stock of arrears Share of payments over stock of arrears Share of payments over investment comm State transfers Av. Taxable income Population Dummy variables: Year of election One year after election Two years after election Three years after election Four years after election First term Second term of Interior. For each municipality we observe the election day and the name of the elected mayor, so that we can construct a measure of political cycle (Cycle defined as the years from the next election) and whether the mayor is at her/his first or second term (Term; in the latter case, s/he cannot be reelected in the following election). 6 Figure 1 shows the year of election of the councils that governed the Italian municipalities during our period of analysis. During , on average, each year, in our sample, there are about 1,200 municipalities holding elections for new mayors and city councils, with a concentration of elections in years 2004 and Note that: (i) Municipalities below 15,000 inhabitants have a different electoral system than larger municipalities; in fact, a single ballot system is in place to municipalities with less than 15,000 inhabitants, while a dual ballot system is in place above that threshold (see, among others, Barone and de Blasio (2013)). (ii) We also observe the name and the political affiliation of the city councilmen; this can allow us to construct measures of the political fragmentation of the city council and measures of the political orientation of the mayors and its majority (although the wide diffusion of independent civic coalitions that limit the possibility of identifying the political affiliation of the mayor or council majority for the whole sample of municipalities). In this preliminary version of the paper, we do not explore these features and their potential effects on the political budget cycle and payments on arrears. A deeper analysis will be available in future versions of the paper. 10

11 Summary statistics in Table 1 shows that about one-fifth of the municipalityyear observational units are distributed quite uniformly across the five years of the political cycle. The same Table also shows that about three-fifth of our observations have a mayor at his/her first term. Figure 1: Years of elections in the municipalities in our sample In this analysis, we focus on the 6,700 municipalities belonging to the fifteen ordinary regions; we do not consider the about 1,400 municipalities of the remaining five regions since the latter enjoy a larger degree of legislative and financial autonomy and respond to different regulations in many fields. We focus on the decade , although not all variables are available in all municipality-year observations. Our sample of municipalities have on average a population of about 7,000 inhabitants and an average taxable income of about 14,000 euros per capita. Municipalities differs also in other socio-economic dimension, such as their geographical location. This does not only implies a North-South differential in the country in terms of social capital, financial development, presence of criminal activities, and functioning of the institutions in general, but there is also a relevant within-region variation in these and other characteristics (such as the composition of the population or the touristic or non-touristic vocation) which likely affect the budgetary choices. In the Section 11

12 3.3, we present the approaches we take to deal with these sources of heterogeneity. 3.2 Suggestive evidence In this paragraph, we present descriptive evidence on the relationship between political cycle and payment on arrears and other expenditures (Figures 2-4) and revenues budgetary outcomes (Figures 5-7). We present the evidence for each year of the first and second term of the mayor political cycle, showing average and median values as well as deviations from the year after the elections. We denote with election the year the city-council and mayor are elected. Given that the municipal elections are usually held in Spring, the budgetary choices for the election year are the results of decisions taken by both the mayor and city-council holding office before the elections and the newly elected ones. In particular, the level of expenditure commitments, taxes and loans are set by the incumbent politicians (at the beginning of the fiscal year), while the payments (including the payments on arrears) are decided both by the incumbent politicians (in the first part of the election year) and by the politicians winning the elections (in the second part of the election year). We denote the first year after election with +1 and this is also taken as the base year for the differences shown in Figures 4 and 7. As this is the first year (and farthest from the next election), according to our theoretical predictions, the new elected mayor and city-council implement in this year the tighter fiscal policy over the political cycle. The last year the mayor and city council autonomously format the budget for the whole year is the fourth year after elections (i.e., +4, which is also expected to be the year before the next elections). 7 Evidence from these figures are in line with the theoretical predictions presented 7 Note that in these figures the differences for the years in the second term are taken with respect to the first after the re-election. 12

13 in Section 2, as payments on arrears seem to follow a political cycle and so the other expenditure and revenue outcomes do. In particular, investment expenditure gradually increase from the year of election to the year before elections (when they reach their maximum), both in the first and second term. Taxation gradually increases from the year of elections but it reduces in the year before elections, while loans gradually increase from the year of election. Payments on arrears are high in the year of elections, decrease in the following year, but gradually increase up to reach their maximum in the year before elections and stay high in the year of elections. 13

14 Figure 2: Expenditures (mean) Figure 3: Expenditure (median) Expenditures (differ- Figure 4: ences) Figure 5: Revenues (mean) Figure 6: Revenues (median) Figure 7: Revenues (differences) 14

15 3.3 Estimated equation The reduced form empirical model that we estimate using different approaches to analyze the relationship expenditure and political cycle looks as follows: y mt = α + βcycle mt + γt erm mt + δcycle mt xt erm mt + ɛ mt. (7) All monetary variables are expressed in per-capita values at constant prices. In our main model specification, the dependent variable y denotes the amount of payments on arrears for investments in year t by municipality m. We also alternatively use other dependent variables: e.g., payment speed (on investment new commitments) i.e., the share of new investment commitments that are actually paid within the year, arrears disposal rate i.e., the share of outstanding investment arrears that are actually paid within the year, expenditure (commitments), and revenue variables i.e., tax and non-tax revenues, and loans. Cycle is one of our key variables of interest and it is operationalized as a set of dummy variables, each one representing the number of years from the past elections. As described in Sections 3.1 and 3.2, the dummy variable Election year indicates that in a given year municipal elections have been held in a given municipality and, thus, the budgetary choices for that year have been taken both by the former mayor and city-council (governing the city in the months of the year before the elections) and the new elected mayor and city-council (governing the city in the months of the year following the elections). We take as base category for this set of dummy variables the first year after the election. We decided to base our political cycle definition on the years from past elections instead of the years from the next elections to reduce endogeneity problems associated with possible choices of incumbent politicians leading to premature end of the term. Term is the other key variable and indicates whether the mayor is in her first or 15

16 last (second consecutive) term. In case the mayor is at her/his second consecutive term is not re-eligible for third consecutive term. It is operationalized with a dummy variable taking the value 1 if the mayor is at her/his Second term, the value 0 otherwise (i.e., in the first term). The interactions between second term and the dummy variables that indicate the years from past elections capture the differential behavior in the second term with respect to the first term for a given year from the past elections. The error term ɛ captures all factors that influence the outcome but that are not captured by the model specification and consists of the following: (i) municipalityspecific time-invariant effects, (ii) municipality-specific time-varying effects, and (iii) time-varying macro effects that influence all municipalities. To reduce omitted variables problems and deal with (i), we follow two alternative approaches. The first approach consists of augmenting the model specification (7) with a set of municipality-level control variables (M ) that aim to control for the constituency s structural characteristics. In particular, we control for the municipality being a touristic location (proxied by the number of per-capita bed places in tourist accommodations), the location being in a mountainous area, the extension of the existing road network in the municipality, the share of young and old population, a measure of education of the population, a measure of social capital (i.e., the number of per capita non-profit associations) 8, and province-fixed effects to control for factors that influence municipalities operating in contexts with similar socio-institutional qualities (such as crime, and effectiveness of the judicial system) and levels of economic and financial development. The second alternative approach to reducing municipality-level time-invariant omitted variables involves the inclusion of municipality-fixed effects in the model specification (7). This approach fully captures the cross-sectional variability and allows us to exploit the within-municipality 8 Data for all these variables come from the Italian National Institute of Statistics - ISTAT 16

17 variability. To deal with (ii) we augment the model with a set of MT mt control variables, such as average taxable income, the amount of transfers from the central government, the population density, and the categories of population size. 9 In some specifications, we also control for the Stock of arrears in the municipality at the beginning of the year. To deal with (iii) we include year-fixed effects. In some specifications, we also include region-year effects so to capture within region-year variations. All in all, the fact that for historical and institutional reasons (see Coviello and Gagliarducci (2012) and Repetto (2015)), the Italian municipalities hold elections in different years allows us to include both municipality-fixed effects (and thus to exploit within municipality variations in terms of political cycle), but also year-fixed effects (and thus to compare municipalities at different points in the political cycle for a given year). 3.4 Main estimation results To estimate the augmented equation (7), we employ different estimators that will allow us to capture different dimensions of the variability of our data and to deal with different concerns about omitted variables. In Table 2, we report our main estimation results for the relationship between payments on arrears and political cycle. In columns 1 and 2, we employ a pooled-ordinary least square estimator (pooled-ols) that includes the controls at the municipality level (M ), the municipality-year level (MT ), and province-fixed and year-fixed effects. The estimation results in column 3 and 4 are from the within-group estimator (i.e., municipality-fixed effects), which allows us to exploit the time dimension of our data. 9 We include dummy variables for populations below 1,000 inhabitants, between 1,000 and 5,000, between 5,000 and 15,000, between 15,000 and 200,000, and above 200,

18 Estimation results show that the coefficients of the dummy variables for the years from past elections on the payments on arrears are positive, stastically significant and similar across estimation approaches. Keeping the year after the election as the base year, the level of payments on arrears gets significantly higher as new elections approach. Furthermore, estimations show that in the year of elections the payments on arrears is even higher than the year before the elections with respect to the base year. This result holds both when we estimate with an OLS and with a within-group estimator, and both when we do not and do allow the coefficients for the Cycle dummy variables to vary between first and second Term. In particular, during the second term the level of payments on arrears is on average higher than in the first term and, in particular, this applies to the year of elections when the mayor starts her/his second consecutive term. These results confirm the predictions on the payments on arrears from our theoretical model. To put forward solid interpretation on the reasons/mechanisms through which the incumbent political actor increases the payments for arrears as elections approach, we need also to have a broader picture of the relationship between the political cycle and other budget outcome variables. To this end, in Table 3, we show the effect of the cycle on the some other sources of expenditures and revenues. In particular, we study the investment expenditure commitments (columns 1-2), current expenditures (columns 3-4), tax-revenues (columns 5-6), non-tax revenues (columns 7-8), and loans (columns 9-10). Estimation results show that the levels of investment expenditure commitments vary with the political cycle both in general and when comparing the first and second term. Investment expenditures increase as elections are approaching and reaching in the year before elections the highest positive difference with respect to the base year. The relationship between current expenditure and the political cycle follows a similar path, but it is worth to note that differences at different years of the cycle 18

19 Table 2: Main results: Political cycle and payments on arrears Column (1) (2) (3) (4) Estimator OLS OLS Within Within Dependent var. Payments on arrears Election year *** *** *** *** (2.753) (3.166) (2.599) (2.949) 2-y post election * ** (2.477) (3.092) (2.355) (2.823) 3-y post election *** *** *** *** (2.671) (3.452) (2.429) (3.157) 4-y post election *** *** *** *** (3.122) (3.885) (2.817) (3.593) Second term *** *** *** *** (2.511) (3.859) (2.339) (3.849) Election year x Second term (5.079) (4.752) 2-y post election x Second term *** *** (4.804) (4.513) 3-y post election x Second term *** *** (5.352) (5.183) 4-y post election x Second term *** *** (6.051) (5.802) State transfers 0.766*** 0.766*** 0.136*** 0.137*** (0.052) (0.052) (0.049) (0.049) Av. taxable income 0.011*** 0.011*** 0.012*** 0.012*** (0.002) (0.002) (0.003) (0.003) Municipality-level controls YES YES YES YES Year-fixed effects YES YES YES YES Province-fixed effects YES YES NO NO Municipality-fixed effects NO NO YES YES Observations 63,989 63,989 67,056 67,056 R-squared Notes: Robust standard errors clustered at the municipality level are in parentheses. Inference: *** p<0.01, ** p<0.05, * p<0.1. Municipality-level controls include: in columns 1-4, population size categories, and density of population; in columns 1-2: share of young population, share of old population, number of non profit organizations per capita as a proxy of social capital, index of tertiary educated people, number of bed places in tourist accommodations, area of the municipality, km of roads within the municipality, and dummy variables for the municipality being located in a mountainous area. are much more pronounced for investment expenditures, which are generally more flexible than current expenditure. Furthermore, the largest negative differences with respect to the year after the elections for both variables is in the year of elections. However, this negative difference is lower in the year of elections if the mayor starts her/his consecutive mandate. Similarly, for the revenue side variables, we observe a cycle. In particular, municipalities tend to increase the revenues when elections are still far thus contributing to the accumulation of commitments for investment expenditure (i.e., payment capacity), then they decrease taxation in the year before elections and in the election 19

20 year, and increase loans in the year before elections (probably to compensate higher expenditures). 20

21 Table 3: Main results: Political cycle and other expenditures and revenues Column (1) (2) (3) (4) (5) (6) (7) (8) (9) (10) Dependent var. Inv. exp.comm. Curr. exp. Tax-rev. Non-tax rev. Loans Election year *** *** *** *** ** *** *** (7.320) (6.998) (1.351) (0.935) (0.855) (0.515) (1.046) (0.762) (2.098) (2.032) 2-y post election ** ** 4.639*** 2.936*** 2.198** 1.552*** *** 6.510*** (7.780) (7.326) (1.350) (0.876) (0.899) (0.534) (1.048) (0.714) (2.210) (2.147) 3-y post election *** *** 3.948*** 2.021** 1.714* * *** *** (7.078) (6.870) (1.437) (0.980) (0.921) (0.570) (1.058) (0.806) (2.129) (2.082) 4-y post election *** *** 3.597** 3.058*** *** 2.111* *** *** (7.363) (7.202) (1.626) (1.074) (1.033) (0.623) (1.231) (0.939) (2.314) (2.293) Second term *** (7.284) (7.124) (2.070) (1.386) (1.252) (0.712) (1.572) (1.164) (2.117) (2.168) Election year X Second term *** *** *** 9.381*** (10.457) (10.108) (2.164) (1.490) (1.288) (0.799) (1.669) (1.234) (3.004) (2.930) 2-y post election X Second term (11.086) (10.826) (2.090) (1.375) (1.325) (0.796) (1.580) (1.119) (3.094) (3.067) 3-y post election X Second term * (10.913) (10.667) (2.389) (1.628) (1.474) (0.906) (1.785) (1.316) (3.188) (3.159) 4-y post election X Second term * 3.474*** (11.815) (11.300) (2.754) (1.893) (1.667) (1.037) (2.048) (1.519) (3.459) (3.490) State transfers 1.048*** 0.265*** 0.637*** 0.329*** *** *** 0.108*** *** 0.031** (0.078) (0.090) (0.040) (0.025) (0.026) (0.011) (0.027) (0.015) (0.015) (0.025) Av. taxable income 0.012*** *** 0.008*** 0.010*** 0.003*** 0.004*** 0.005*** (0.003) (0.005) (0.002) (0.001) (0.001) (0.001) (0.001) (0.001) (0.001) (0.001) Municipality-level controls YES YES YES YES YES YES YES YES YES YES Year-fixed effects YES YES YES YES YES YES YES YES YES YES Province-fixed effects YES NO YES NO YES NO YES NO YES NO Municipality-fixed effects NO YES NO YES NO YES NO YES NO YES Observations 63,989 67,056 63,989 67,056 63,989 67,056 63,989 67,056 63,989 67,056 R-squared Notes: Robust standard errors clustered at the municipality level are in parentheses. Inference: *** p<0.01, ** p<0.05, * p<0.1. Municipality-level controls include: in columns 1-10, population size categories, and density of population; in columns 1,3,5,7,9: share of young population, share of old population, number of non profit organizations per capita as a proxy of social capital, index of tertiary educated people, number of bed places in tourist accommodations, area of the municipality, km of roads within the municipality, and dummy variables for the municipality being located in a mountainous area. 21

22 Overall, these results confirm our theoretical predictions. They indicate that when the elections are still far, municipalities tend to increase their future expenditure capacity, increasing the revenues and keeping low the expenditures. The year before elections, municipalities tend to reach their highest level of expenditures (particularly, investment expenditure) and lower taxation, while they increase loans. While, the year of elections, the investment expenditures, taxation, and loans are all lowered. In this framework, the payments on arrears play an additional role as they are a particular budgetary tool that allows the incumbent politicians to adapt the timing of expenditure implementation, thus increasing the overall level of actual expenditures (i.e., payments) also during the year of elections without the need of additional increase in revenues (e.g., taxation or loans) which as seen in the theoretical setting involve additional political costs. In fact, the budget coverage for the investment expenditures that have not been paid (i.e., the arrears) is made in the year the investment commitments are taken. Incumbents can thus keep high the payments on arrears also in the year of elections, so to compensate private firms that have executed public works for the municipalities (i.e., reduce their level of trade debt). Finally, these mechanisms are present also in the second term, although some particular differences exist and are planned to be studied much more in detail in future versions of the paper. 3.5 Robustness checks In this section, we present several robustness to assess the validity of our main results and interpretation. In particular, we introduce additional time-varying covariates to our empirical model specification to reduce omitted variable problems. First, we introduce the stock of arrears measured at the end of the previous year so to interpret the effect of the political cycle on the payments on arrears for 22

23 a given level of stock. Estimation results in Table 4 show that our main results are confirmed. This latter control is particularly helpful to reduce the influence of the effects that higher investment commitments decided in the previous years can have on the payments for arrears. In particular, we want to exclude that higher investment commitments taken when elections approach lead in the following year to higher payments on arrears for investment expenditure. If this were the case, our interpretation that incumbent politicians use payments on arrears strategically would not be correct, but it would be simply driven by higher investment commitments. Estimation results for our outcome variables (in columns 1,3,5,7,9) confirm our main results. Actually, the inclusion of this variable tend to increase both the empirical model predictive power and the fit between our predictions, descriptive results, and estimated coefficients (particularly, when we look at the differences between first and second term). Secondly, we introduce region-year fixed effects so to capture those characteristics that are common to municipalities within the fifteen regions we study in each given year. Estimation results in columns (2,4,6,8,10) show that our main results are confirmed. 23

24 Table 4: Robustness checks: Controlling for additional time-varying factors Column (1) (2) (3) (4) (5) (6) (7) (8) (9) (10) (11) (12) Estimator Within Dependent var. Paym. arrears Inv. exp.comm. Curr. exp. Tax-rev. Non-tax rev. Loans Election year *** *** *** *** *** *** *** *** *** ** *** *** (2.702) (2.781) (6.940) (7.032) (0.938) (0.950) (0.518) (0.521) (0.760) (0.779) (2.031) (2.053) 2-y post election ** ** 2.777*** 1.903** 1.520*** 0.998* *** 6.947*** (2.556) (2.574) (7.186) (7.226) (0.877) (0.891) (0.535) (0.535) (0.714) (0.734) (2.141) (2.186) 3-y post election *** *** *** *** ** *** *** (2.786) (2.853) (6.794) (6.876) (0.981) (1.009) (0.571) (0.576) (0.805) (0.824) (2.082) (2.138) 4-y post election *** *** *** *** 2.389** 3.014*** *** *** *** (3.078) (3.115) (7.105) (7.360) (1.082) (1.110) (0.625) (0.626) (0.942) (0.983) (2.293) (2.339) Second term *** *** *** *** (3.163) (3.141) (7.201) (7.082) (1.380) (1.364) (0.711) (0.699) (1.162) (1.158) (2.172) (2.158) Election year X Second term 9.299** 8.342** *** *** ** *** 9.423*** (4.177) (4.145) (10.007) (9.824) (1.485) (1.466) (0.799) (0.786) (1.231) (1.222) (2.925) (2.919) 2-y post election X Second term *** *** * (4.019) (4.005) (10.674) (10.559) (1.370) (1.370) (0.796) (0.785) (1.116) (1.119) (3.058) (3.057) 3-y post election X Second term *** *** ** 1.715* (4.310) (4.271) (10.597) (10.455) (1.624) (1.614) (0.906) (0.889) (1.316) (1.317) (3.153) (3.151) 4-y post election X Second term ** *** *** 2.823*** (4.813) (4.799) (11.358) (11.271) (1.893) (1.882) (1.038) (1.012) (1.520) (1.525) (3.489) (3.454) State transfers 0.076* 0.076* 0.283*** 0.275*** 0.327*** 0.352*** *** *** *** ** (0.043) (0.044) (0.092) (0.091) (0.025) (0.025) (0.011) (0.011) (0.015) (0.015) (0.025) (0.026) Av. taxable income *** 0.004*** 0.003*** *** 0.004*** (0.002) (0.003) (0.006) (0.006) (0.001) (0.001) (0.001) (0.001) (0.001) (0.001) (0.001) (0.001) Stock of arrears 0.172*** 0.171*** *** *** 0.008*** 0.006*** 0.002*** 0.001* 0.005*** 0.004*** *** *** (0.005) (0.005) (0.007) (0.007) (0.001) (0.001) (0.000) (0.000) (0.001) (0.001) (0.001) (0.001) Municipality-level controls YES YES YES YES YES YES YES YES YES YES YES YES Year-fixed effects YES YES YES YES YES YES YES YES YES YES YES YES Municipality-fixed effects YES YES YES YES YES YES YES YES YES YES YES YES Region-year-fixed effects NO YES NO YES NO YES NO YES NO YES NO YES Observations 67,056 67,056 67,056 67,056 67,056 67,056 67,056 67,056 67,056 67,056 67,056 67,056 R-squared Notes: Robust standard errors clustered at the municipality level are in parentheses. Inference: *** p<0.01, ** p<0.05, * p<0.1. Municipality-level controls include: in columns 1-12, population size categories, and density of population. 24

25 Finally, in Table 5 we use as alternative dependent variable the share payments on the stock of arrears to further control for the role played by past investment decisions and stock of non-paid commitments. Estimation results in columns 1-3 confirm our main results. In columns 4-6 we use the share of payments on investment commitments decided in the year. We expect that if incumbent politicians strategically increase payments for committed investments when elections approach, they do not only increase payments for the past commitments with pending payments (i.e., arrears), but they are likely to increase the payments on investment commitments decided in the year. In particular, in the year of elections incumbents would like to pay a larger share of investment commitments decided in that year. Estimation results confirm this hypothesis. 4 Conclusion In this project we aim at studying a mechanisms through which the political cycle affects the budget decisions. In particular, we build a framework to understand the role played by the payments on investment expenditure arrears in the transmission of the political economy incentives of the incumbent politicians on the municipalities budgets. We show that, while in the years before elections along the political cycle commitment for investment expenditure tends to increase and taxation to decrease and in the year of elections both drop, payments on investment arrears (i.e., execution of commitments for investment expenditure that have been decided and for which financial coverage has been found in previous years) increase in both these years. [MORE ON INTERPRETATION HERE] Further refinements and more detailed evidence are planned for the future version of the paper. In particular, further attention should be dedicated to the interesting 25

26 Table 5: Other robustness checks Column (1) (2) (3) (4) (5) (6)) Estimator Within Dependent var. Sh. paym./stock arrears Sh. paym./invest. comm. Election year 0.024*** 0.024*** 0.022*** 0.017*** 0.018*** 0.018*** (0.002) (0.002) (0.002) (0.003) (0.003) (0.003) 2-y post election ** (0.002) (0.002) (0.002) (0.002) (0.003) (0.003) 3-y post election 0.009*** 0.009*** 0.012*** *** ** ** (0.002) (0.002) (0.002) (0.002) (0.002) (0.003) 4-y post election 0.020*** 0.024*** 0.027*** (0.002) (0.002) (0.003) (0.003) (0.003) (0.003) Second term 0.008*** 0.010*** 0.013*** 0.006** 0.007*** (0.002) (0.002) (0.004) (0.003) (0.003) (0.004) Election year X Second term 0.010*** 0.009*** 0.012*** (0.003) (0.003) (0.003) (0.004) (0.004) (0.004) 2-y post election X Second term ** ** *** (0.003) (0.003) (0.003) (0.003) (0.003) (0.004) 3-y post election X Second term *** *** *** (0.003) (0.003) (0.004) (0.004) (0.004) (0.004) 4-y post election X Second term ** *** *** (0.004) (0.004) (0.004) (0.004) (0.004) (0.004) State transfers *** 0.000* 0.000* (0.000) (0.000) (0.000) (0.000) (0.000) (0.000) Av. taxable income ** (0.000) (0.000) (0.000) (0.000) (0.000) (0.000) Municipality-level controls YES YES YES YES YES YES Year-fixed effects YES YES YES YES YES YES Municipality-fixed effects YES YES NO YES YES NO Region-year-fixed effects YES YES NO YES YES NO Mayor-fixed effects NO YES NO NO YES NO Observations 67,056 67,056 67,056 67,056 67,056 67,056 R-squared Notes: Robust standard errors clustered at the municipality level are in parentheses. Inference: *** p<0.01, ** p<0.05, * p<0.1. Municipality-level controls include: in columns 1-6, population size categories, and density of population. differences between the first and second term that seem to emerge already from the current analysis. 26

27 References Alesina, A. and Paradisi, M. (2014). Political Budget Cycles: Evidence from Italian Cities. National Bureau of Economic Research NBER Working Papers Barone, G. and de Blasio, G. (2013). Electoral rules and voter turnout. International Review of Law and Economics, 36(C): Cioffi, M., Messina, G., and Tommasino, P. (2012). Parties, Institutions and Political Budget Cycles at Municipal Level: Evidence from Italy. mimeo. Coviello, D. and Gagliarducci, S. (2012). Tenure in Office and Public Procurement. Tor Vergata University, CEIS CEIS Research Paper 179. Repetto, L. (2015). Balance sheet disclosure and the budget cycle of Italian municipalities. mimeo. Rogoff, K. (1990). Equilibrium political budget cycles. The American Economic Review, 80(1): Rogoff, K. and Siebert, A. (1988). Elections and macroeconomic policy cycles. The Review of Economic Studies, 55(1):1 16. Shi, M. and Svensson, J. (2006). Political budget cycles: Do they differ across countries and why? Journal of Public Economics, 90(8-9):