società italiana di economia pubblica - c/o dipartimento di scienze politiche e sociali dell’università di pavia

XX

V

CO

NFER

EN

ZA

LA FINANZA PUBBLICA NEI SISTEMI MULTILIVELLO COORDINAMENTO, CONCORRENZA E DISCIPLINA FISCALE

Pavia, Aule Storiche dell’Università, 26 - 27 settembre 2013

THE DETERMINANTS OF LOCAL TAX SETTING:

EVIDENCE FROM THE REFORM OF THE ITALIAN MUNICIPAL PROPERTY TAX

CORRADO POLLASTRI AND ALBERTO ZANARDI

1

The determinants of local tax setting:

Evidence from the reform of the Italian municipal property tax

Corrado Pollastri

ISTAT –Italian National Institute of Statistics and IFEL

Alberto Zanardi

University of Bologna and Econpubblica-Bocconi University

Preliminary draft

Abstract

This paper investigates the determinants of local tax setting in the case of the reform of the Italian municipal property tax on real estate. The municipal property tax was radically reformed in 2012 as part of the fiscal consolidation package adopted by the Italian government. Using a cross-sectional dataset on all Italian municipalities, this paper shows that the institutional profiles of the reform significantly affected the tax rate setting by municipalities, together with factors more traditionally discussed by literature (socio-demographic, economic and political variables, tax interactions). In particular we find that the pre-reform regime - specifically the tax rate previously set by the municipality and the limitation the central government imposed on local taxing power -, the cuts in central government transfers that came with the reform, and the uncertainties in actual disposable resources resulting from the reform played an important role in shaping local fiscal decisions.

JEL Classification: H71, H77

Keywords: Keywords: local taxation, tax rates, tax reform

2

1. Introduction1

As pointed out by an extensive economic literature2 the decisions of local governments

regarding own taxes are the result of a large array of factors. These include both the socio-

economic and demographic characteristics of local jurisdictions, such as population

structure by age, per capita income, unemployment rate, etc., and the political features of

local governments (political stance, electoral margin and fragmentation). External factors

such as the amount of vertical grants stemming from the central government and the tax

policies carried out by neighbouring local governments also usually plays an important role.

This paper aims to investigate the role of these determinants with reference to the case

of the reform of the Italian municipal property tax on real estate, the Imposta municipale

unica – IMU, carried out in 2012. The municipal property tax has been radically reformed

as part of the fiscal consolidation package the Italian government adopted in December

2011 to achieve the balanced budget target by 2013 and to shift the tax burden from capital

and labor income towards consumption and property: the tax base has been enlarged, the

standard tax rate set at national level has been increased and additional scope for

autonomous fiscal effort (in terms of setting the tax rate above / below the standard

national level) have been allowed to municipalities even if a part of the total tax yield has

been assigned to the central government.

The range of objectives and the number of measures included in the reform make a

complex institutional framework where each municipality has actually taken the choice to

set the tax rate of IMU above or below the standard national level. In this case the

characteristics of the tax reform may have severely biased decisions taken by local

governments. This paper therefore differs from previous studies on local tax setting which

typically analyse observed local rates decisions within a general setting, where the local tax

structure is relatively stable and not affected by wide-ranging reforms. This paper highlights

that it is critical to adequately consider the institutional framework when analysing local

fiscal policy.

The paper is organised as follows: section 2 gives a brief description of the reform of

the municipal property tax implemented in 2012; the empirical model is presented in

1 We gratefully acknowledge the Ifel Scientific Direction for the collaboration, and Christian Mongeau (Cefip-Roma3) and Federico Belotti (Ceis-TorVergata) for their valuable advices. Any opinions expressed here are those of the authors alone. 2 See, for example, the review in Delgado at al. (2011).

3

section 3; section 4 describes the dataset used for the estimation; the estimation results are

discussed in section 5 and section 6 concludes.

2. The reform of the Italian municipal property tax

Prior to the reform implemented in 2012, the revenues from property tax on real estate

were around €9.6bn (2011) and were the main tax resource for the Italian municipalities.

The tax was levied on all residential properties and real estate except for owner occupied

dwellings. For most real estate the tax base was calculated by multiplying cadastral rents by

105 and the standard tax rate was set at 0.4% Each municipality had the power to increase

the rate by up to 0.7%, the result of which meant that the actual average tax rate was

0.61%. In 2008 the power the municipality had to increase tax rate has been suspended by

national government. This limitation is likely to have forced local authorities to raise less

revenue than they would prefer.

As mentioned in the introduction, the 2012 reform has radically redesigned the structure

of the municipal property tax. As a part of the effort to shift the national tax mix from

labour and business income to consumption and property, the municipal property tax has

been increased by raising the cadastral multiplier to 160 for most categories of real estates,

by including the owner-occupied dwellings in the tax base and by setting the standard tax

rate at 0.76% (0.4% for owner-occupied houses).3 Each municipality can change the rate by

+/-0.3% (+/-0.2% for owner-occupied houses). The result is that total yield has risen to as

much as €23.7bn of which €19.9bn corresponds to the standard tax rate and €3.8bn to the

additional fiscal effort determined at local level. About 85% of total tax yield can be

ascribed to dwellings other than owner-occupied houses.

Revenue from municipal property tax is only partially acquired by municipalities as the

central government is given half of total yield excluding revenues corresponding to owner-

occupied houses and to the autonomous fiscal effort. In total €9bn has been assigned to

the central government even if the entire tax has been collected at local level.

Table 1 reports the joint distribution tax rate set by municipalities for owner-occupied

houses (+/-0.2% from the standard rate = 0.4%) and other dwellings (+/-0.3% from the

3 For equity reasons, a basic tax deduction of €200 is granted to owner-occupied dwellings. For the years 2012 and 2013 the basic deduction is increased by €50 for each child aged up to 26 years, up to a maximum of €400.

4

standard rate = 0.76%). It is likely that redistributive concerns have prevented the

municipalities from increasing the standard rate on owner-occupied houses. The result of

this is that the power the municipalities have of setting a tax rate different from the

standard level has been mainly exerted on the taxation of other dwellings: out of 8,012

municipalities considered, as much as 65% left the tax rate for owner-occupied houses

unchanged at standard level while 27.7% decided to raise it and 7.2% to reduce it. On the

contrary, in the case of other dwellings the percentage of municipalities which confirmed

the standard tax rate is 42.6%. 55.6% increased the rate and only 1.8% provided for a

reduction from the standard rate. As the variation of local tax rates is mainly due to the tax

rate on dwellings other than owner-occupied houses this is where this paper will focus its

of local tax settings.

Table 1 – Joint distribution of municipal choices for tax rates

The reform of municipal property tax has dramatically changed the backdrop to how

the local tax is set. The main relevant profiles are as follows:

1. At the aggregate level, for all municipalities, the total revenues provided by the new tax

at a standard rate is greater than the tax yield collected by municipalities before the

reform at total tax rates (inclusive of fiscal effort). However a compensating mechanism4

4 Accomplished by assigning part of the tax yield of the new municipal property tax to the central government and, on the other hand, by partially cutting the transfers the central government gives to municipal level.

tax rate for owner-occupied houses / tax rate for other dwellings

reduction up to 0.3%

reduction up to 0.2%

reduction up to 0.1%

standard tax rate (=0.4%)

increase up to 0.1%

increase up to 0.2%

increase up to 0.3%

total

reduction up to 0.2% 28 13 13 107 27 46 44 278

reduction up to 0.1% 6 11 22 82 49 78 47 295

standard tax rate (=0.4%) 8 14 23 3018 820 854 486 5223

increase up to 0.1% 0 4 3 144 607 543 246 1547

increase up to 0.2% 1 0 0 61 142 212 253 669

Total 43 42 61 3412 1645 1733 1076 8012

0.5 0.5 0.8 42.6 20.5 21.6 13.4 100.0

5

is introduced to ultimately leave the municipalities with the actual amount of tax yield

they acquired before the reform and to assign the additional revenues to the central

government.5 This also occurs at the level of individual municipality6. This result of

revenue invariance for the municipalities has two relevant implications which influence

the way the new tax rate depends on pre-reform regime:

• On the one hand, for municipalities in the aggregate, the compensating mechanism

implies that scope for local variations in tax rates (above/below the standard tax rate)

has been “rebased” to a level higher than established before the reform. This makes

the autonomous fiscal effort more costly in terms of political consensus for local

policy-makers. This cost is greater for the municipalities who had lower rates in the

past. From a political point of view, low rates in the past tend to induce local

governments to set lower tax rates.

• On the other hand the compensating mechanism guarantees to each municipality the

past yield, irrespective of the past tax rates: leaving new rates at the standard level

would allow the municipality to get the same revenue as before the reform. Thus,

from a budgetary point of view, the decision of changing tax rates should be

independent from the past tax rate. However, there is another way in which the level

of past rates may affect new ones. As mentioned before, 2012 reform restored the

municipalities back to their power to increase tax rates after a long period of

suspension. As a consequence we expect that in those municipalities that in the pre-

reform regime had already set tax rates at the maximum level imposed by the central

government(0.7%) have strong incentives to increase tax rates after the liberalization

provided by 2012 reform (Revelli, 2010).

2. The reform of the municipal property tax has been preceded and matched up with

severe cuts in central government transfers to municipalities. This greatly affects the

total resources available to local authorities. Fiscal consolidation packages adopted by

the Italian government for 2012 included cuts to local municipalities measuring up to

€3bn. This decrease resources is likely to affect local tax setting.

5 Accomplished by assigning part of the tax yield of the new municipal property tax to the central government and partially cutting the amount transfered from the central government to the municipal level. 6 By means of a system of inter-governmental transfers.

6

3. The way in which the co-habitation in the municipal property tax of a central

government component and a local component produces a disincentive for local

authorities to tax rates reductions. As a matter of fact, since the share of the tax yield

assigned to the central government is determined by applying the standard tax rate to a

large part of total tax base, any x% reduction in the tax rate implies a 2 x% reduction in

tax revenue for the local authorities that have decided on that reduction. More generally,

the co-habitation compromises the capacity of the local property tax to enhance the

‘electoral accountability’ of the local government.

Given the complex institutional setting in which the decision on local tax rate is taken, a

thorough consideration of these institutional profiles is fundamental in the analysis of the

new Italian municipal property tax.

3. The econometric model

In order to evaluate the effects of different factors on the choice of tax rates by local

authorities we must first define a tax rate-setting equation. In this equation the dependent

variable is the level of local tax rate determined by local government. The explanatory

variables include a number of socio-demographic, economic and political factors that

traditionally are expected to influence the tax setting decision of a municipality.

Following relevant literature, we take also into account the possible effects of tax

interactions across local jurisdictions (tax mimicking) on local tax-setting. The existence of

tax mimicking can be related to tax competition (policy makers mimic the tax policy of

their neighbours for fear of tax base mobility), yardstick competition (fiscal choices made

in nearby jurisdictions provide a benchmark for local policy makers to be re-elected) or

social interactions (politicians belonging to the same party interact with each other to draw

inferences about party preferences). We allow for this possible horizontal spatial

dependence by resorting to the appropriate specification and estimation procedure based

on spatial econometrics (LaSage and Pace, 2009). In particular, we estimate a spatial lag

model that includes amongst the explanatory variables, a weighted average of the local tax

rates determined by using a spatial weight matrix.

7

Once we have estimated the base tax-setting model, we augment it by a set of

institutional factors, specifically related to the 2012 Italian reform in order to test their

impact on the actual tax rates decision at local level.

The estimated tax-setting base equation can be written as follows:

iiiii TWXt (1)

where:

the index i refers to the municipalities

it represents the property tax rate set by municipality i (the ith element of vector T)

iX represents the vector of socio-demographic, economic and political variables

iW denotes the ith row of vector of the spatial weight matrix W

Equation (1) is estimated by the Maximum Likelihood technique in order to overcome

the issue of the endogeneity (spatial simultaneity) of tax rates decisions across local

authorities.

In the second step of the analysis we estimate an augmented model where the set of

explanatory variables is increased by a number of reform-specific factors iS that represent

the institutional setting which affects the local tax setting mechanism:

iiSiTiWiXit (2)

4. The Data

The empirical analysis is based on a cross-section data-set for Italian municipalities

which we specifically collected for 2012. This dataset combines data from IFEL (the

research centre for local public finances of the Association of the Italian municipalities),

the Italian Ministry of the Interior, the Italian Ministry of the Economy and the Italian

8

Statistical Office. Estimations are carried out on a dataset that includes 7,898 observations,

corresponding to almost all Italian municipalities (98%).7

The dependent variable of our model is the statutory (standard and autonomous) tax

rate of property tax on dwellings different from owner-occupied houses set by the Italian

municipalities in 2012.8 As mentioned in section 2, this component accounts for about 85%

of total tax yield.

In the econometric analysis of the determinants of the local rates, we include four

blocks of variables as regressors:

Structural characteristics

Political context

Spatial interaction

Reform-specific effects

A thorough discussion of these variables follows.

Structural characteristics

We consider one set of socio-demographic variables that aim at capturing the

expenditure needs of the municipality and one set of economic variables that proxies the

economic resources structure of the municipality as indicators of structural factors which

may influence tax rates.

With the socio-demographic variable, we have to consider that the Italian municipalities

are very heterogeneous in terms of population, from 34 to a maximum of 2.5 million

residents. In order to find systematic differences in taxing behaviour across municipalities

7 Because of their peculiarity the sample excludes the bigger cities (12 municipalities with more than 250,000 residents), the 104 municipalities involved in 2012 earthquake in Emilia Romagna, Lombardia and Veneto and the municipalities located in the autonomous region of Valle d'Aosta (74 municipalities) which have missing values for some of the variables in the sample. 8 The effective rates for 2012 were set by municipalities from December 2011 to 30 October 2012. Our analysis is based on the results of the censual analysis of the municipality resolutions, carried out by IFEL in January 2013

9

of different sizes we include four dummies (DEMO) in the equation which correspond to

different demographic classes.9

Specific dummy variables (RIPT) indicate macro-regions (north, central and south Italy).

We also evaluate the differences in tax rate decisions relating to specific regional

institutional framework. In Italy there are four major autonomous regions which are

characterised by a higher level of fiscal autonomy. The municipalities located in these

regions receive transfers from a specific system. The dummy RSSP catches this effect.

In order to take account of some of the characteristic of real estate composition at local

level that may influence local tax rate decisions, we also include in the model the share of

industrial buildings over total tax base. We verify if this has a positive effect (high rates may

imply high revenues without directly affecting voters) or a zero / negative effect (higher tax

base on industrial building may imply a lesser tax rate, given a fixed amount of fiscal

revenue objective) (KIMD).

Finally, a proxy is included to measure the economic condition of households. The

hypothesis we want to test is that municipalities where residents have a higher income are

more likely to set higher tax rates. To proxy household economic conditions we use the log

of the normalised per capita taxable income (MRGI).

Political context

In order to determine the influence of politics, we focus on the political stance of the

voters in each municipality. This is done by taking the percentage of votes for the main

political parties in the national parliament from the most recent election (Chamber of

deputies).10 11 Parties are categorised into in four groups, centre-left (PPSX), centre-right

(PPD1), autonomist party Lega-nord (PPD2) and other autonomist parties (PPAU).

We also proxy political participation by the electoral turnout (PPAS) in order to control

the effects of more ‘politically active’ environments on local tax policies.

9 As mentioned before, the twelve largest cities (population > 250,000) are not included in the analysis due to their peculiarities compared to other municipalities. All of these cities have their tax rate set to the highest level, regardless of financial and political differences. 10 Ministry of the Interior, Historical elections archive, http://elezionistorico.interno.it 11 It is difficult to draw this kind of information directly from the official databases of Italian municipal governments (mayors, council composition, etc) as over 75% of the elected mayors are recorded as being a candidate for a local party (‘liste civiche’) as opposed to of a national party. These local parties cannot be easily placed in the standard left- right political spectrum.

10

We also measure the proximity of the percentages of the two main coalitions in the

national elections (centre-left and centre-right) as a proxy of the political strength of the

leading party in the municipal council (PMAJ). 12 We expect low strength to negatively

affect the ability of the government to set high tax rates.

Finally, we verify the existence of a political-cycle-effect (the closer an election is, the

lower the tax rate) by including time (in years) to the next local election among the

regressors (ELEZ).

Spatial interaction

In order to capture the existence of spatial interactions on tax rates, we introduce into

the model a spatially-weighted average of the tax rates of other municipalities (WALQ).

This comes from the product between spatial weight matrix W and the vector of tax rates.

In this preliminary draft we use a weight matrix, W, in which the elements are the

normalised reciprocal of the aerial distance (in kilometres) among municipalities.13 We set

the distance equal to infinite in the case of municipalities further than 100km, and to 1 in

the case of neighbouring municipalities.14

Reform-specific effects

This latter group of variables includes several factors specifically connected to the 2012

reform. We consider three subsets of factors which may potentially influence local tax rates

decisions:

the tax rates of municipal property tax before the reform

the transfers cuts accomplished by the central government

12 As the absolute value of the percentage difference. 13 We will extend the analysis verifying the performance of different specifications of the weight matrix. In particular we will test for the maximum range of tax mimicking, varying the distance over which we set the weight to zero. Alternatively we will limit the number of municipalities with non negative weights to a fixed number (k neighbours method). We will also test if the mimicking effect is better caught by a weight matrix based on real distance, road distance or time of travel (data taken from Google Distance API). In future developments we will also test if tax mimicking occurs only amongst similar (in terms of dimension) municipalities., The hypothesis that it does seems reasonable in context of yardstick competition. 14 The aerial distance matrix is built from municipality coordinates taken from Google Geocoding Application Program Interface, and the adjacent neighbours pattern is drawn from the Italian municipality Shapefile developed by Istat (www.istat.it/it/archivio/44523).

11

the actual process of the reform implementation.

We can identify two specific sources of influence from earlier property tax rates on local

fiscal efforts which are derived from the effects of the compensating mechanism described

above. One is based on ‘political’ factors and the other on ’financial’ factors:

former tax rates may influence current fiscal efforts because in municipalities with low

pre-reform tax rates the fiscal burden for the taxpayers is higher than in others (even at

standard rate)

the municipalities who have reached the upper bound of ICI rate are more likely to set a

optimal level of property tax yield above the former level.

We introduce these factors into the model using two variables, the actual level of pre-

reform property rate (ICIP) and a dummy variable (ICIM) which is equal to one when the

pre-reform property tax rate is already set at the top level.

We also consider that the property tax reform is part of a broader consolidation package

that includes transfers cuts for local governments. We consider the impact of these cuts on

tax decisions, including in the model the per capita 2012 transfers cuts (RRIS).

Finally, it must be stressed that the actual implementation of the reform has been

strongly biased by the uncertainty of tax yield estimates and the corresponding cuts in

compensating central government .15 This uncertainty may have led local governments to

be prudent when setting tax rates in order to prevent financial unbalances.

To account for this we include two additional explanatory variables in the model:

the percentage difference between the actual IMU yield and the most recent estimate

(SCGT)

a dummy variable that identifies the municipalities which experienced a large downward

revision (more than -30%) of the former property tax yield (SICI).

15 Transfer cuts (equal, as mentioned, to the difference between IMU and former property tax) are carried out in the first part of the year on the basis of an estimated projection of potential IMU yield at standard rate, before the payment of the first instalment (at standard rate) by the taxpayers. The IMU estimates, as the amount of compensation cuts of central government transfers, have been revised three times during 2012, in March, August and October. For non marginal groups of municipalities these projections show great variability, and for some the projections differ significantly from the actual IMU yield at the standard rate. The IMU yield at the standard rate has to be estimated when the tax rate differs from its basic level. The official estimation is still unavailable, we refer to provisional estimation carried out by IFEL.

12

Table 2 gives an overview of the explanatory variables used in the econometric model.

13

Table 2 - Summary statistics

Municipalities (%)

Mean 5th percentile95th

percentileReference Up to 2000 43.73

1 From 2000 to 5000 26.882 From 5000 to 10000 14.573 From 10000 to 60000 13.674 From 60000 to 250000 1.14

Territory RIPT Reference South 55.001 North 12.672 Center 32.32

Autonomous region RSSP Reference No 83.382 Yes 16.62

Industial buildings tax base (%) KIMD Scale 0.28 0.04 0.59Log of average taxable income MRGI Scale -0.09 -0.7 0.35

PPAS Scale Turnout (%) 0.81 0.89 0.69PPSX Scale Centre left 0.4 0.23 0.59PPD1 Scale Centre right 0.36 0.21 0.53PPD2 Scale Lega nord 0.12 0.00 0.37PPAU Scale Other autonomists 0.02 0.00 0.03PMAJ Scale Closeness 0.58 0.04 1.76

Future elections (year) ELEZ Scale 1.75 0.00 4.00SPATIAL EFFECTS Spatially weighted IMU tax rate WALQ Scale 8.45 7.6 9.6

Former property tax rate ICIP Scale 4.2 3.00 5.5Reference No 69.32

1 Yes 30.68Per capita transfers cuts in 2012 (hundreds of euro)

RRIS Scale0.57 0.25 1.10

% Difference between IMU projections and actual yield

SCGT Scale-0.06 -0.14 0.00

Reference No 30.681 Yes 30.68

Former property tax estimation reduction greater than 30%

SICI

POLITICAL CONTEXT

Class of residents DEMO

Voters (%) at national parliament (Camera 2008)

Former property tax rateat maximum level

ICIM

REFORM - SPECIFIC EFFECTS

STRUCTURAL

14

5. The results

Table 3 shows the estimation results for the base model (1) and the augmented model

(2). We first look at the base model results in order to assess the effects of the variables

more traditionally discussed by literature on tax rates decisions.

Table 3 – Estimation results

Structural, political and spatial factors included in the model fit well the data. Statistical

significance is beyond 5% for almost all the variables except for PPAS (electoral

turnout).Demographic structure (DEMO) positively affects tax rate: smaller towns (up to

2,000 residents, our reference category) set lower rates, that rise monotonically in upper

demographic classes. For the largest cities in our sample (60,000 to 250,000 residents) the

estimated tax rate reaches a level of +1.05 points (per mil) above the smallest towns.

Variable

Constant 4.74 (*) 3.87 (*)[DEMO=1,00] 0.17 (*) 0.14 (*)[DEMO=2,00] 0.57 (*) 0.49 (*)[DEMO=3,00] 0.76 (*) 0.63 (*)[DEMO=4,00] 1.05 (*) 0.87 (*)[RIPT=1,00] 0.21 (*) 0.24 (*)[RIPT=2,00] 0.37 (*) 0.29 (*)[RSSP=1,00] 0.39 (*) 0.29 (*)KIMD -0.24 (*) -0.14 (**)MRGI 0.21 (*) 0.10PPAS 0.30 0.29PPSX 0.50 0.61 (**)PPAU -6.15 (*) -4.40 (*)PPD1 0.71 (*) 0.68 (*)PPD2 -0.66 (**) -0.61 (**)PMAJ 0.08 (*) 0.07 (*)ELEZ 0.05 (*) 0.04 (*)

SPATIAL WALQ 0.27 (*) 0.23 (*)ICIP - 0.21 (*)ICIM - -0.17 (*)RRIS - 0.12 (*)SCGT - -0.64 (*)SICI - 0.26 (**)

(*) Significant 1%; (**) Significant 5%

Base Model Augmented model

STRUCTURAL

POLITICAL

REFORM - SPECIFIC EFFECTS

15

In the southern part of Italy (reference category for RIPT) the estimate shows lower

rates. A coefficient comparing test shows that the coefficient corresponding to central Italy

(0.37) is significantly higher than that one for northern Italy. The municipalities located in

autonomous regions (reference category for RSSP are the other regions) set tax rates lower

than those located in ordinary regions.

The share of tax base referred to industrial buildings (KIMD) has a negative impact on

tax rates. It appears that a large share of industrial buildings spur local government to lower

rates rather than inducing them to exploit the presence of ”big non voter” taxpayers by

setting high tax rates.

The personal taxable income per capita (MRGI) has a significant and positive impact on

tax rates: the greater the ability of the taxpayer to pay, the higher the rate set by local

governments.

On the political side, the results show negative and significant relation between the

percentage of voters for autonomist parties and the tax rate (PPAU other autonomists and

PPD2, Lega Nord). The centre-right political stance (PPD1) is correlated to higher tax rates

whereas the impact of centre-left (PPSX) is not statistically significant.

The closeness measure (PMAJ) has a positive impact on tax rates. This verifies the

hypothesis that weak majorities encounter more difficulties in raising tax rates. We also see

some evidence of a political-cycle in setting tax rates, the further away the election (ELEZ),

the higher the rates.

The spatial mimicking effect (WALQ) is also significant and shows the expected

positive sign. The tax policy of neighbours within an aerial radius of 100 kilometres tends

to positively affect local government tax rates.

Table 3 shows in the second column the estimated results of the model (2) where the

base model has been augmented by including among the regressors some reform-specific

factors. We can observe that the augmented model preserves sign and the significance of

the base model coefficients with the exception of the taxable per capita income (MRGI).

The coefficients of all institutional factors that are included are statistically significant

and show the expected sign. In particular:

The positive coefficient of ICIP (the tax rate before reform) underpins the hypothesis

that because of the higher burden (even with new rates at standard level) suffered by

16

residents of municipalities with low former property tax rates, local governments tends

to keep the new rates lower than in others municipalities. We have to consider that the

compensating mechanism that caused a mismatch between actual rates variation and

revenue variation entails an handicap for the municipalities who had lower rates of the

former property tax. Due to this, municipalities would have suffered a revenue loss if

they had restored a similar tax reduction as before the reform. The observed lesser

number of municipalities that set lower new rates respect to what happens in the past

regime, seems to suggest that the potential revenue loss have discouraged them to set

low tax rates as done in the pre-reform regime. As a matter of fact before the reform

more than 20% of municipalities set rates at least one point below median value,

whereas only 1,6% do that after the reform.

The positive coefficient of ICIM (reference category: former property tax rate below the

maximum level) shows, as expected, that municipalities in which the tax rate was at the

maximum level before the reform tend to exert more fiscal effort.

The impact of cuts in central government transfers (RRIS) on tax rate setting is positive

and significant. There is evidence that the transfers cuts, combined with the

strengthening of the local tax, have produced a shift of the tax burden from central to

local governments.

The results for SCGT (percentage difference between actual yield at the standard rate

and the most recent projection) and for SICI (dummy variable equals to 1 for the

municipalities that suffered a cut of intergovernmental transfers of more than 30% of

the estimated yield of re-reform tax) suggest that the process of reform implementation,

particularly the uncertainties in projections at the basis of the compensating mechanism,

explains the observed increase in tax rates.

o The negative (and significant) sign of variable SCGT shows that the higher the

projection (and so the higher the cuts in transfers from the central government)

compared to the actual yield, the higher the tax rate set by the municipality.

o The positive (and significant) sign of variable SICI shows that this group

(approximately 60 municipalities) significantly raises tax rates in order to compensate

for the loss caused by the estimation reassessment.

17

In order to point out the impact of reform-specific factors of tax rates setting, we finally

try to picture a “no-reform” scenario and to determine the tax rates the municipalities

would have set by if they were confronted with this scenario. Substituting into reform-

specific regressors the values corresponding to the “no-reform” scenario16, we can derive

from the fitted values of model (2) a rough measure of overall impact of these effects

(regarding transfers cut and uncertainty effects) on tax rates decisions by municipalities. As

shown in table 4, the un-weighted average of the fitted values of the model (2) in the no-

reform scenario is lower than the one estimated in the base scenario; the transfers cuts and

the uncertainty effect explains about 13% of the fiscal effort17.

Table 4 - Impact of reform-specific effects on tax rates

6. Concluding remarks

This paper investigates the determinants of local tax setting in the case of the reform of

the Italian municipal property tax on real estate. The municipal property tax was radically

reformed in 2012 as part of the fiscal consolidation package adopted by the Italian

government. Using a cross-sectional dataset on all Italian municipalities, this paper shows

that the institutional profiles of the reform greatly affected the tax rate setting by

municipalities, together with factors more traditionally discussed by literature (socio-

demographic, economic and political variables, tax interactions). We find that these reform-

specific effects played an important role in shaping local fiscal decisions. In particular:

16 In this exercise we set the values of the independents in order to picture a scenario in which reform factors do not condition the tax setting process. This "no reform" scenario assumes: no transfer cuts (RRIS=0); no uncertainty (SICI=0; SGT=0). It is more complex to design a counterfactual scenario for the other two reform-specific variables that captures the effect of the previous property tax rates on the new ones, because these effects should hold in any case. In this exercise we left the two variables (ICIP and ICIM) at the actual levels. 17 The weighted average rate is substantially higher (0,93%).

percent ratesUnweighted average rate

Fiscal effort

Base scenario (*) 0.847 0.087No-reform scenario (**) 0.836 0.076

(*) Fitted rates of model (2) with full parameters specification(**) Fitted rates of model (2) with reform effect variables set to zero

18

The mismatch between actual rates variation and revenue variation from the previous

property tax to the new property tax due to the reform design entails a drawback for

the municipalities who had lower rates in the pre-reform regime. Although the political

pressure of the taxpayers residents in those municipalities may produce some lesser

fiscal effort, we observe that the tax rate gap with the other municipalities is now

narrower.

Wider margins of fiscal effort granted by the reform, combined with relevant transfer

cuts from central government, lead municipalities to shift the cuts to local taxpayers.

The reform implementation itself stimulated higher rates. In order to enforce tax

compliance, the central government subordinated the level of transfers to the

achievement of a projected revenue of the property tax by each municipality. That

measure induced some uncertainty on groups of municipalities that, as our model

shows, is associated with higher tax rates. Our simulations show that the transfer cuts

and the uncertainty factors explain about 13% of the total rates dynamic.

This work can be extended in several directions. In particular, further insights are

needed on the mimicking mechanism across municipalities by considering different

specification of the weight matrix. For example, we can evaluate the effect of varying the

maximum distance over which the weight to set to zero; we can test if tax mimicking

mainly occurs amongst municipalities that are similar in terms of demographic size; we can

explore if a leader-follower pattern is relevant in tax mimicking behaviour by considering

the time sequence by which different municipalities have taken their tax rates decisions.

References

Bordignon, M., Cerniglia, F. and Revelli, F. (2004). Yardstick competition in intergovernmental relationships: theory and empirical predictions. Economics Letters, 83, 325-333.

Delgado, F.J., Lago-Penas, S. and Mayor, M. (2011). On the determinants of local tax rates: new evidence from Spain, Working Paper 2011/4, Institut d'Economia de Barcelona (IEB).

Dubois, E., Leprince, M. and Paty, S. (2007). The effects of politics on local tax setting: evidence from France. Urban Studies, 44, 1603-1618.

19

LeSage, J., and Pace, R. K. (2009). Introduction to special econometrics. Chapman and Hall.

Revelli, F. (2010). Tax mix corners and other kinks, Working Paper 2010/50, Institut d'Economia de Barcelona (IEB).

Solé Ollé, A. (2003). Electoral accountability and tax mimicking: the effects of electoral margins, coalition government, and ideology. European Journal of Political Economy, 19, 685-713.