Intergovernmental Grants as Signals and the Alignment E ect: Theory and Evidence

Transcription

1 Intergovernmental Grants as Signals and the Alignment E ect: Theory and Evidence Emanuele Bracco, Ben Lockwood, Francesco Porcelli, Michela Redoano Lockwood (Warwick) Alignment E ects 1 / 34

2 Introduction Three key stylized facts: Vertical imbalances: across countries, over the period subnational expenditures account for 30-40% of total public expenditures on average However only 40-50% of the funding comes directly from taxes levied by subnational governments. Transfers from upper levels of government account for over 30% of subnational tax revenues data from WB decentralization database, Shared responsibilities between national and local governments in the provision of complex and important services such as health, education For example, in many countries, e.g. UK, national government sets the school curriculum, marks exams, but local governments build and run schools, hire teachers Party-political local government (mayors, councillors have party a liations) Lockwood (Warwick) Alignment E ects 22 November / 33

3 Introduction This paper explores the implications of these three facts for the design and impact of intergovernmental grants Vertical imbalances and shared responsibilities create an inference problem for rational but incompletely informed voters How to assign "credit" for good performance in the supply of local public goods between the local incumbent, the local challenger, and the national incumbent? Lockwood (Warwick) Alignment E ects 22 November / 33

4 Introduction National governments can exploit this inference problem to maximize support for their own parties this behavior generates a (positive) alignment e ect in grants in turn, this generates negative alignment e ects in taxes, positive alignment e ects in expenditure, incumbency advantage Intergovernmental grants as signals, rather than bribes, as in standard political economy theory of grants (Cox and McCubbins, (1986), Dixit and Londregan(1996)) Lockwood (Warwick) Alignment E ects 22 November / 33

5 Roadmap Related literature Theoretical framework RD estimates of alignment e ects on grants, taxes, expenditure, incumbency advantage for a new data-set of Italian municipalities The ypaper e ect revisited Lockwood (Warwick) Alignment E ects 22 November / 33

6 Related Literature Large literatures on political agency, and on inter-governmental grants, but very little on how intergovernmental grants are determined in a political agency framework Grants hardly mentioned in e.g. Persson and Tabellini(2000), Political Economics, Besley(2007), Principled Agents? Empirical literature on alignment e ects on grants: Levitt and Snyder (1995), Sole -Olle and Sorribas-Navarro (2008), Arulampalam et.al.(2009), Brollo and Nannicini (2012) but: nothing on alignment e ects on local taxes and expenditure theoretical frameworks lack convincing microfoundations Large literature on incumbency advantage: Lee (2008), Lee, Moretti, Butler (2004), Ferreira, Gyourko (2009), Alesina, Rosenthal (1989), Snyder (1990), Pelzman (1992), Levitt (1996) and Ferraz, Finan (2008), but nothing on incumbency advantage and alignment Also link to the ypaper e ect literature: Gramlich (1973), Knight (2002), Dahlberg(2008). Lockwood (Warwick) Alignment E ects 22 November / 33

7 Theory Two periods t = 1, 2, local governments (municpalities) i = 1,..n, one central government Two political parties, L, R W.l.o.g, central government is L, M A is the set of aligned municipalities i.e. also L Output of local government i is where: g it = f (τ it + TR it, e it ) + z it + φz it + u it z it is incumbent mayor s ability, Z it incumbent central government s ability, φ 0 ability spillover parameter (shared responsibility) e it is mayor s e ort τ it, TR it : local tax revenue and transfer from centre. Lockwood (Warwick) Alignment E ects 22 November / 33

8 Payo s Voter j in municipality i has payo where g it d(τ it ) X j d(.) is the cost of tax revenue to the voter, captures distortions and compliance costs (d(0) = 0, d 0, d 00 > 0, d(τ) > τ, τ > 0) X j U[ 1/2ζ, 1/2ζ] is an ideological preference for the incumbent. Local politicians are both benevolent and o ce-motivated, with payo s λ(g it d(τ it )) e it + p it, λ > 0 where p it is the probability of re-election (p i2 = 0) Lockwood (Warwick) Alignment E ects 22 November / 33

9 Payo s, Contd Incumbent central government s payo is: λ (g it i2m d(τ it )) + i2m L p it C (TR it ) i2m where: central government cares about re-election of aligned municipalities (term i 2ML p it ) C () is a strictly convex cost of providing grants. Lockwood (Warwick) Alignment E ects 22 November / 33

10 Order of Events Period 1 1 Grants TR i1 are allocated by central government 2 Mayors choose e i1, τ i1 3 Voters observe g i1, τ i1 and vote for incumbent vs. challenger Period 2 1 Grants TR i2 are allocated by central government 2 Mayors choose e i2, τ i2 Lockwood (Warwick) Alignment E ects 22 November / 33

11 Information structure Neither mayor nor voters know z i1, zi1 c in period 1, BUT: rational voters will interpret g i1, τ i1 as signals of incumbent and challenger ability z i1, zi1 c rational incumbents (mayor, central) will anticipate this, and exploit it to get re-elected Key assumption: aligned incumbents have positively correlated abilities (Geys & Vermeir(2008)) cov(z it, Z t ) = µσ 2 Z, cov(zc it, Z t ) = 0, i 2 M A cov(zit, c Z t ) = µσ 2 Z, cov(z it, Z t ) = 0, i 2 M N This information structure: creates uncertainty in voter s minds about which level of government is responsible for good performance generates information spillovers for local incumbent and challenger Lockwood (Warwick) Alignment E ects 22 November / 33

12 Signalling In period 1, signal extraction problem of the voters (infer z i, zi c from g i, τ i ) can be solved to give zi e, z c,e i, voter expectation of incumbent and challenger ability in period 1. De ne the non-predictable part of g i as s i = g i f (τ i + TR e i, e e i ) where TR e i, e e i are voter expectations of unobservable TR i, e i In equilibrium, expectations are rational: TR e i = TR i, e e i = e i Lockwood (Warwick) Alignment E ects 22 November / 34

13 Signalling: Aligned case: z e i s i = θ A = σ 2 a + φµσ 2 A σ 2 a + φ(φ + 2µ)σ 2 A +, σ2 u z c,e i s i = 0 Higher s i gives no credit to local challenger; 0 < θ A < 1, credit shared between local and national incumbents θ A decreasing in φ i.e. greater shared responsibility means local incumbent gets less credit θ A increasing in µ when φ > φ i.e. in the presence of shared responsibility, greater ability correlation can bene t local incumbent Lockwood (Warwick) Alignment E ects 22 November / 34

14 Signalling Non-aligned case: z e i s i = θ N = σ 2 a σ 2 a + φ 2 σ 2 A + σ, u z c,e i s i = θ c N = φµσ 2 A σ 2 a + φ 2 σ 2 A + σ u Credit shared between local and national incumbents as before (θ N decreasing in φ) Credit to local incumbent strictly lower than with alignment i.e. θ N < θ A But now some credit to local challenger: θ c N > 0, increasing in shared responsibility φ, as long as µ > 0 Link to Arulamapalam et al (2009): they just assume θ A = 1, θ N xed at some value less than 1 Lockwood (Warwick) Alignment E ects 22 November / 34

15 Equilibrium Equilibrium conditions: TR A, TR N, τ A, τ N, e A, e N solve (λ + ζθ A )f e = 1, f r = d 0, i 2 M A (λ + ζ(θ N θ c N ))f e = 1, f r = d 0, i 2 M N (λ + ζθ A )f r = C 0 (TR i ), i 2 M A λf r = C 0 (TR i ), i 2 M N Lockwood (Warwick) Alignment E ects 22 November / 33

16 Main Results Alignment e ect on grants. If e ort and revenue are independent, (f er = 0), aligned jurisdictions are assigned more grants by the central government: TR A > TR N. Intuition: TR i " =) g i " =) p i ", which bene ts central government only if i is aligned Alignment e ect on local taxes. If e ort and revenue are weak substitutes, (f er 0) aligned jurisdictions impose lower taxes than unaligned ones: τ A < τ N. Intuition: alignment =) (1) TR i " =) τ i # (2) e i " =) τ i # Alignment e ect on local expenditures. If f er = 0, aligned jurisdictions set higher expenditures than unaligned ones: τ A + TR A > τ U + TR U. Alignment e ect on incumbents re-election. If some voters are non-fully rational (retrospective) and base their vote only on g, then the incumbent is more likely to be re-elected if he is aligned Lockwood (Warwick) Alignment E ects 22 November / 33

17 Background Information on Italy Italian Institutional Framework 8,101 municipalities (comuni) [ 7,391 < 15,000 inhabitants]. Comuni are ruled by a local government (giunta), headed by an elected mayor (sindaco). Mayors: are elected directly by residents, following the 1993 reform, which established di erent rules for large (>15,000) and small municipalities (<15,000); stay in power for ve years and are subject to a two-term limit; nominate and dismiss the members of the executive committee (giunta); are in charge of land management and environment, local transport, local police, culture and recreation, education; have some discretionary powers on how to raise scal revenue; are supported by a coalition of parties. There are two main party coalitions in Italy: centre-left and centre-right. Lockwood (Warwick) Alignment E ects 22 November / 32

18 . Distribution G,&2' of YKDK municipalities 80+(/0&$(05" by5= mayor s '2'#(05"+ coalition &9,203"'-,"- "5(,203"'- %$"0#0H,20(0'+ M/'3/'++05" +,%H2'NK 9',/ 1203"'- f5( 1203"'- G5(,2 '2'#(05" S'"('/J/03C( S'"('/J2'=( G5(,2 S'"('/J/03C( S'"('/J2'=( G5(,2 FaaV E Ya Ya YD E YD af Faaa E F_P F_P be E be FaP DEEE E _Y _Y DV E DV bd DEEF YF E YF E Y_ Y_ VY DEED P_ E P_ E PP PP FEV DEE_ DF E DF E _D _D P_ DEEY Y_ E Y_ E FFU FFU FbE DEEP DE E DE E _a _a Pa DEEb E YP YP _U E _U VD DEEU E YE YE YD E YD VD G5(,2 FUV _E_ YVF DEa DVb YaP aub Lockwood (Warwick) Alignment E ects 22 November / 31

19 Background Information on Italy Local Government Financing Two main sources of municipalities revenues : -transfers from upper levels of government (mainly ordinary grants from central government); -own tax revenues. Ordinary Grants (fondo ordinario): -account for about 1/5 of municipalities revenues (but huge variability); -no formula or hard rule on grants: a lot of path dependence (spesa storica). Own tax revenues: -mainly from a Property Tax (ICI) (about 1/2 of own tax revenue), -also from Tax on Waste Disposal (TARSU) and other small taxes (ex. Income tax and electricity surcharge). -Large autonomy of setting ICI and TARSU, within some boundaries Lockwood (Warwick) Alignment E ects 22 November / 32

20 Our Data Italian data on comuni with population over 15,000 for the period Electoral data (municipal elections) are from the Interior Ministry; Economic and demographic data are from ISTAT; Public Finance data (Conti Consuntivi) are from the Ministry of Finance. Variables: TR it are per capita transfers from fondo ordinario. τ it is per capita tax revenue. r it = Tr it + τ it are per capita expenditures. AL it is an alignment dummy MA it margin of alignment, is the margin of victory of the mayor with sign + if aligned, sign if non-aligned. Controls are: population, under 15 and over 65, income, altimetric zone, %of self employed, illiterate, graduates, unemployed and service sector workers, left local and central governments. Lockwood (Warwick) Alignment E ects 22 November / 32

21 Summary Statistics Means and standard deviations by MA Lockwood (Warwick) Alignment E ects 22 November / 31

22 Estimation Strategy Regression Discontinuity Design where: Y i,t = α + γal i,t + f (MA i,t )ϕ + β 0 X i,t + λ t + v i,t Y = TR, τ, r f () is a p-th degree polynomial of MA, including interaction with the alignment dummy. X it is a matrix of controls, λ t are time dummies and v i,t an error term Expect γ > 0 if Y = TR, r, and γ < 0 if Y = τ We also implement the local linear regression approach, which restricts the sample to observations in the interval MA it 2 [ h, +h], where h is an optimally chosen bandwidth. Lockwood (Warwick) Alignment E ects 22 November / 32

23 Main Results: Alignment e ect on grants Regression results Dependent Variable: Grants (per capita) No With With controls & Coefficient on Alignment Controls Controls Fixed Effect Observations Linear regression (all sample) 0.555*** 0.138*** 0.134*** 3546 [0.0324] [0.0218] [0.0314] First order polynomial (RD Sample) 0.263*** 0.257*** 0.265*** 3141 [0.0663] [0.044] [0.0572] Second order polynomial (RD Sample) 0.462*** 0.323*** 0.289*** 3141 [0.099] [0.0652] [0.0816] Third order polynomial (RD Sample) 0.571*** 0.414*** 0.432*** 3141 [0.128] [0.085] [0.108] Forth order polynomial (RD Sample) 0.475*** 0.364*** 0.365*** 3141 [0.157] [0.137] Local Linear regression (h) 0.326*** 0.403*** 0.439*** 856 [0.120] [0.121] [0.0930] Local linear regression (half h) ** 0.301** 324 [0.183] [0.182] [0.137] Local linear regression (double h) 0.319*** 0.348*** 0.303*** 1750 [0.0792] [0.0838] [0.0643] Lockwood (Warwick) Alignment E ects 22 November / 31

24 Main Results: Alignment e ect on grants Graphical Analysis Lockwood (Warwick) Alignment E ects 22 November / 32

25 Main Results: Alignment e ect on local tax revenue Regression results Dependent Variable: Tax Revenue (per capita) No With With controls & Coefficient on Alignment Controls Controls Fixed Effect Observations Linear regression (all sample) *** * ** 3540 [0.0140] [0.0125] (0.0104) First order polynomial (RD Sample) *** *** 3135 [0.0272] [0.0173] ( ) Second order polynomial (RD Sample) *** *** *** 3135 [0.0390] [0.0252] (0.0137) Third order polynomial (RD Sample) ** ** *** 3135 [0.0509] [0.0323] (0.0188) Forth order polynomial (RD Sample) *** 3135 [0.0616] [0.0394] (0.0227) Local Linear regression (h) ** ** *** 772 [0.0596] [0.0392] [0.0238] Local linear regression (half h) [0.0765] [0.0515] [0.0362] Local linear regression (double h) *** *** *** 1580 [0.0419] [0.0289] [0.0197] Lockwood (Warwick) Alignment E ects 22 November / 31

26 Main Results: Alignment e ect on local tax revenue Graphical analysis Lockwood (Warwick) Alignment E ects 22 November / 32

27 Main Results: Alignment e ect on expenditures Regression results Dependent Variable: Expenditures (per capita) No With With controls & Coefficient on Alignment Controls Controls Fixed Effect Observations Linear regression (all sample) * [ ] [ ] [ ] First order polynomial (RD Sample) ** *** *** 3613 [0.0188] [0.0132] [ ] Second order polynomial (RD Sample) *** 3613 [0.0268] [0.0191] [ ] Third order polynomial (RD Sample) * 3613 [0.0342] [0.0241] [0.0130] Forth order polynomial (RD Sample) ** 3613 [0.0413] [0.0297] [0.0163] Local Linear regression (h) * ** 721 [0.0411] [0.0304] [0.0156] Local linear regression (half h) [0.0577] [0.0420] [0.0208] Local linear regression (double h) ** 1470 [0.0284] [0.0211] [0.0110] Lockwood (Warwick) Alignment E ects 22 November / 31

28 Main Results: Alignment e ect on expenditures Graphical analysis Lockwood (Warwick) Alignment E ects 22 November / 32

29 Turnover of Incumbents Regression Discontinuity Design I i,e = α + γal i,e + f (MA i,e )ϕ + β 0 X i,e + λ t + v i,e where I i,e is a dummy equal to 1 if the incumbent has been re-elected and 0 otherwise in jurisdiction i, and e are the election years theory suggests that γ > 0 Lockwood (Warwick) Alignment E ects 22 November / 32

30 Turnover of Incumbents Polynomial Controls Incumbents at Incumbents in terms order their first mandate of political parties no polyn. no 0.30 (0.03) 0.24 (0.03) 1st no 0.51 (0.05) 0.43 (0.04) 2nd no 0.58 (0.05) 0.50 (0.05) 3rd no 0.63 (0.06) 0.39 (0.11) 4th no 0.73 (0.06) 0.50 (0.07) no polyn. yes 0.24 (0.04) 0.33 (0.06) 1st yes 0.60 (0.17) 0.55 (0.12) 2nd yes 0.55 (0.20) 0.64 (0.13) 3rd yes 0.54 (0.25) 0.68 (0.14) 4th yes 0.35 (0.21) 0.64 (0.15) Observations R-squared (1) Lockwood (Warwick) Alignment E ects 22 November / 32

31 Turnover of Incumbents Lockwood (Warwick) Alignment E ects 22 November / 32

32 The Flypaper E ect Flypaper e ect - a grant from federal government to local government raises spending of that local government by a greater amount than an equivalent increase in local income (crowding in) Our model theoretically predicts a ypaper e ect, because local taxes are distortionary, so TR " implies τ # and thus a lower excess burden of taxation, =) higher expenditure Studies nding a ypaper e ect may be biased upwards because grants are endogenous and may positively correlated with local preferences for public goods (Knight(2002)) Possible strategies: study non-manipulable formula-based grants e.g. Dahlberg et al (2008)) nd instruments for unobserved preferences (Knight(2002): uses level of state representation on the Congressional Transportation Committee to instrument state grants from the Federal Highway Aid Program) Following Knight(2002) and Hilber et al (2011), we use the alignment dummy to instrument grants in a ypaper regression Lockwood (Warwick) Alignment E ects 22 November / 32

33 The Flypaper E ect (1) (2) (3) (4) (5) OLS OLS FE 2SLS First Stage 2SLS Second Stage Dependent Net of Grant Net of Grant Net of Grant Grants Net of Grant Variable Expenditures Expenditures Expenditures Expenditures Grant *** *** *** ** [0.055] [0.084] [0.052] [0.269] Real Estate Income (per capita) *** *** ** * ** [0.015] [0.015] [0.040] [0.168] [0.042] Total Decleared Income (per capita) *** *** *** [0.002] [0.002] [0.008] [0.003] [0.009] Population *** *** *** [0.000] [0.002] [0.001] [0.002] Population squared *** *** * *** [0.000] [0.000] [0.000] [0.000] Population below 15 (%) *** * *** ** [3.049] [1.240] [0.649] [1.705] Population over 65 (%) *** [1.974] [0.701] [0.405] [0.743] Electoral cycle [1.835] [0.922] [0.506] [0.945] Municipality Left ** [10.173] [8.083] [3.323] [8.742] Central Government Left *** *** *** [19.101] [14.479] [3.960] [15.794] Alignment *** [1.916] Observations 3,546 3,546 3,546 3,538 3,538 R-squared Year dummies no yes yes yes yes Municipality Fixed Effect no no yes yes yes Lockwood (Warwick) Alignment E ects 22 November / 31 32

34 Conclusions This paper presented a new theoretical framework for studying intergovernmental grants as signals in an environment with shared responsibilities and political parties Theoretical predictions (some of them new) taken to a new Italian data-set, generally supported by the evidence Future work: endogenise theory to allow for upper-level elections? Lockwood (Warwick) Alignment E ects 22 November / 32