An ex-post analysis of Italian fiscal policy on renovation

Transcription

1 An ex-post analysis of Italian fiscal policy on renovation Marco Manzo, Daniela Tellone VERY FIRST DRAFT, PLEASE DO NOT CITE June 9 th 2017 Abstract In June 2012, the share of dwellings renovation costs that can benefit tax credit increased from 36% to 50% and the total amount of renovation expenses that may benefit the fiscal relief went from euros to euros. This work aims to contribute to the debate on tax credits effects providing an ex-post analysis of 2012 change in the Italian tax relief on building renovation. In particular, the tax credit efficiency is analyzed in terms of economic and fiscal additionality trying to evaluate the policy effect on both the increase in dwellings renovation probability and the decrease in tax evasion. The policy change seems to have mainly fostered aged persons, retired people, and those with a higher yearly income and able to afford non-durable non-consumption goods. On the opposite, it seems that the policy change has not had any significant effects on the renovation likelihood of inactive persons, with a relatively high share of food expenditure, i.e. the class of less wealthier people. Keywords: Residential Sector, Tax Credit, Dwelling Renovation Policy, Italy JEL Classification: D12, H31, E62 The views expressed herein are those of the authors and not necessarily reflect those of the Italian Ministry of Economy and Finance. Italian Ministry of Economy and Finance, Department of Finance. marco.manzo@mef.gov.it Italian Ministry of Economy and Finance, Department of Finance. daniela.tellone@mef.gov.it 1

2 1 Introduction Since 1998, the Italian governments introduce yearly tax incentive to renovate building in order to achieve different goals like the recovery and the energy upgrading of buildings, provide support to the building sector and, last but not least, fight the submerged economy. Additionally, tax credit for buildings renovation has been modified several times and changes concern many aspects like the kind and the total amount of renovation expenses that may enjoy the tax credits and the tax credit share. The tax credit on building renovation has been made permanent since 2012 (law no. 214/2011). Moreover, in June 2012, the share of dwellings renovation costs that can benefit tax credit increased from 36% to 50% and the total amount of renovation expenses that may benefit the fiscal relief went from euros to euros. Behind this change of policy there is the government will both to stimulate further the building sector and to make the contrast to the submerged economy stronger. Empirical studies concerning the effects of tax credits on renovation activities are quite limited. This lack of attention is probably due to the few policies in this area, while tax incentives for energy-saving housing renovations are much more widespread. Therefore, the empirical literature on the latter subject is much more developed: many empirical studies have analyzed tax incentive energy renovation policies in United States. In particular, early studies carried out in this field were mainly based on meta-analysis. Later in time, econometric researches (Dubin and Henson [3], Walsh [5]) do not find a strong or significant link between tax incentives and energy renovations. On the opposite, Hasset and Metcalf [4] using a logit model on panel data, have been the first to observe the positive effects of tax incentive on energy renovations. In Italy, Alberini et al [1] developed an empirical study in order to analize the effect of energy saving tax credit on the renovation probability. In particular, based on cross-section data for the period from 2004 to 2009, and using both a linear probability model and a probit model, the authors study the impact of energy tax credits on the likelihood of replacing both door/windows and heating systems. They shows that tax credits are effective in stimulating the door/windows renovation, while the evidence for heating systems entails either that policy has no effect or that free riding is an important issue. However, the authors underline that due to the low substitution rate of heating systems in the 2

3 sample the phenomena may not be properly estimated. In another empirical study, Berton and Cavallari [2] study the effect of a tax credit change in 2003 on building renovation. The authors find a positive but limited effect of the fiscal measure. Accordingly, in the literature it seems to be no agreement concerning tax credits effects on building renovation. This work aims to contribute to the debate on tax credits effects providing an ex-post analysis of the 2012 Italian fiscal policy change. In particular, the tax credit efficiency is analyzed in terms of additionality aiming to evaluate the policy effect on both the increase in dwellings renovation probability and the decrease in tax evasion. 2 Dataset and selected variables The analysis is carried out using a dataset that matchs: administrative data (income tax and cadastral information), EU-SILC data (European Union - Statistics on Income and Living Conditions) and HBS data (Household Budget Survey). The administrative data relate to 2009, 2011, and 2013 tax years, they concern the renovation cost to which tax credits apply, the taxpayer annual income etc. Cadastral data provide information on buildings. Data from EU- SILC survey and HBS refer to the years 2010, 2012 and 2014; both survey are carried out by ISTAT. The EU-SILC Survey contains many additional information on taxpayers socioeconomic characteristics, as well as the dwelling year of construction, monthly cost of house management, etc. The HBS data deal with household consumption behavior instead. Integration between administrative and EU-SILC data is based on fiscal code whereas the matching with HBS is statistical. The sample consists of taxpayers some of who may have undergone renovation works during the period analyzed and many have benefited from the tax credits. Furthermore in EU-SILC, individuals are ask whether they have undergone dwellings renovation in the last year also. The possible discrepancy between the EU-SILC statement about renovation and the income tax statements it is used as a measure for fiscal evasion. The explanatory variables can be distinguished in two groups, describing: i) 3

4 the dwelling characteristics and ii) the socio-economic condition of the taxpayers. In particular, concerning the variables describing the dwelling characteristics, we have: construction year of the home, dwelling type (single house, terraced house, flat in a building with less than 10 dwellings, flat in a building with 10 or more dwellings), cadastral category (A1 up to A8), square meters, number of rooms, and finally the dwelling location in the Italian territory divided into macro areas (Islands and South, Center, North East and North West). The taxpayer socio-economic condition is describe in the second group of variables, and we have: taxpayer s age, education degree (primary school, lower and upper secondary school, graduate degree, postgraduate degree), basic activity status (at work, unemployed, in retirement or early retirement or has given up business, other inactive person), ability to keep home adequately warm (related to the affordability of this kind of expenses), number of years of occupation of the dwelling (duration), estimate potential monthly market rent of dwelling, monthly costs connected with the households living in the house, and additional information reported in the income tax like tax on dwelling. 3 Econometric Approach For the econometric analysis, a key issue to assess the effects of tax credit concerns the fact that it is not possible to use a counterfactual approach: it is not possible to identify taxpayers who cannot benefit from the tax credit given the universal nature of the policy. As a matter of facts, the renovation tax credit potentially affects all taxpayers, specifically all those who have undergone a building makeover. Thus the econometric analysis estimate the 2014 year effect which captures the change in tax credits (from 36 to 50 percent), using repeated and independent cross-sections. Another key issue concerns the period under analysis, which immediately follows the 2008 financial crisis. It is well known that the crisis has had a significant impact on both incomes and expenses of the households. Therefore many control variables are included in the model specification in order to better isolate the policy effect. In particular there are variables which allow for taking into account job position (EU-SILC Survey), taxpayers income (income tax) and 4

5 consumption 1 (HBS Surveys) changes. The last key issue we identify, relates to the match between the renovation costs and the dwellings. Since the 2012 income tax, it is possible to punctually perform that match. In previous years, if a taxpayer owned more than one building and used the tax credit, it was not possible to identify which building has been subject to the renovation. For this reason, it was possible to use but he information on taxpayers that own only their main residence: for this group it is always possible to correctly match the renovation expenses with the dwelling. The policy evaluation is divided into three parts. Firstly, taking the makeover information from the income tax, it is estimated the effect of tax credit change on the probability of renovation (economic or fiscal additionality). In particular, we use a restrict dataset composed only by those persons who have undertaken a dwelling renovation between 2010 and Then, we compare the informations about renovations between those from EU-SILC with those from income tax. Accordingly, it is estimated the effect of the tax credit change on probability of using the fiscal relief among those individuals who have ever made a renovation (fiscal additionality). Finally, it is estimated the effect of the tax credit change on the following decisions: 1) to do not renovate, 2) to renovate and to do not declare the expenses and 3) to renovate and to use the fiscal relief. Using a multinomial model we try to estimate both the economic and the fiscal additionality. The economic or fiscal additionality and the fiscal additionality are estimate thanks to models for binary variables (like probit and logit). In particular the policy impact on the renovation probability is estimated using a binary choice model where the dependent variable assumes value: i) one in the renovation case and ii) zero in the non-renovation case. The variable of interest is given by the 2014 dummy, refering to the year in which the policy is fully at work. Specifically, three econometric models are estimated: the linear probability model (LPM), the probit model, and the logit model. Two model specifications (A and B) are estimated, depending on the macro location variable been included or not. 1 The consumptions variables concern both consumption goods and non-durable non consumption goods. 5

6 The following linear probability model (LPM) is estimated: y = α + X β + D γ + ɛ where the dependent variable is dichotomous and takes value zero if the taxpayer has not renovate and one otherwise; the X-matrix is related to the characteristics of both taxpayer and dwelling; D is the vector of annuals dummies and ɛ represents the error. The objective of the analysis is to test the null hypothesis that the 2014 dummy variable coefficient is different from zero. In the LPM model, with some limitations, the probability of the dependent variable is assumed to be a linear function of the explanatory variables. These limits can be overcome by estimating logit and probit models, as in the following formulation: P (y = 1 X, D) = G(α + X β + D γ) where the function G(z) takes values strictly between zero and one for all z real numbers. Specifically, in the probit model, G(z) is the standard cumulated normal distribution function, whereas in the logit model corresponds to the logistic function. 4 Results 4.1 Economic Additionality The results of the three models are show in Table 1. There is considerable consistency of the results obtained with the different models and the explanatory variables have (almost) always the expected sign. It may be remarked that while the 2012 dummy coefficient is never significant, the 2014 dummy coefficient is always positive and significant, pointing out how the fiscal relief change has been effective in inducing taxpayers to renovate. In particular, the probability to renovate in 2014 increases by 21 percentage points according the LPM model, and between 27 and 29 percentage points according probit and logit models estimates. In the following, we will briefly overview the impact of the explanatory variables that are statistically significant. In particular, it can be noted that owning a house built after the 1990s reduces the probability of renovation; on the 6

7 opposite, if the dwelling was built before the 1970s, the likelihood of renewal increases. The decrease in the isolation degree of the dwelling significantly increases the probability of makeover and the maximum value is observed for flat in buildings with more than ten dwellings. In addition, there is an inverse relationship, albeit weakly significant, between the size of the dwelling and the probability of restructure: as the number of rooms and the square meters increases, the likelihood of renovate decreases. Finally, all cadastral categories of dwellings have a lower probability to renovate with respect to A1 category. Analyzing socio-economic variables of individuals, it can be remark that the education degree has a positive impact on the probability of makeover. In addition, the coefficients value increases with the degree of education. Regard to the basic activity status and taking as a reference group the persons at work it is possible to see that for unemployed and other inactive persons the probability of makeover is lower, while for people in retirment the probability is higher. All models also show a negative relationship between the number of years of dwellings ownership and the probability of renovate (duration). The taxpayer age is in quadratic form and the relative coefficients show that the probability of renovate will increase, but less than proportionally with age. All the variables that may be related with the wealth (the ability to keep home adequatly warm, potential month rent, monthly costs connected with the house, and tax on dwelling) positively affect the probability of makeover. All the explanatory variables that try to capture the economic cycle are statistically significant. In particular, an increase of yearly income and non-durable non-consumption goods increase the likelihood of makeover. On the contrary, increasing household consumption good reduce the likelihood of makeover. It should be pointed out that non-durable non-consumption goods expenditure identifies a better family condition, while higher ratio between non-durable nonconsumption goods and income identifies less wealthier families, with a lower chance to renovate. The results therefore seem to suggest that the economic well-being positively influenced the probability of restructuring. Finally, in terms of location, the probability of makeover is highr in the North- East, followed by North-West, Center, South and Islands. Table 1: Renovation probability 7

8 LPM A LPM B Probit A Probit B Logit A Logit B (.58) (.35) (.1) (.23) (.1) (.26) ***.21***.91***.91*** 1.5*** 1.5*** (5.0e-256) (5.2e-250) (0) (0) (0) (0) After *** -.083*** -.25*** -.29*** -.44*** -.52*** (3.5e-09) (2.8e-10) ( ) (1.6e-06) ( ) (9.2e-07) *** -.1*** -.34*** -.37*** -.59*** -.65*** (5.6e-14) (1.9e-14) (2.4e-08) (1.6e-09) (3.4e-08) (1.7e-09) *** -.063*** -.19*** -.2*** -.32*** -.35*** (2.0e-06) (1.7e-06) (.0013) (.00067) (.0021) (.00078) ** -.023* (.022) (.055) (.41) (.55) (.49) (.6) **.099*.19**.17* (.13) (.13) (.043) (.059) (.042) (.064) ***.036***.18***.16***.31***.27*** (.00087) (.0022) (.00067) (.0023) (.00081) (.0029) **.028**.14***.12**.25**.21** (.011) (.025) (.0095) (.028) (.01) (.032) ***.027**.17***.12**.3***.19** (.0016) (.03) (.0018) (.035) (.0025) (.047) Before * ** **.086 (.061) (.66) (.026) (.35) (.035) (.42) Terraced ** **.085* house (.61) (.48) (.013) (.11) (.01) (.074) Flat in.059***.065***.28***.32***.52***.58*** build. with < 10 houses (1.3e-22) (5.1e-27) (0) (0) (0) (0) Flat in.2***.2***.69***.73*** 1.2*** 1.3*** build. with > 10 houses (3.5e-173) (2.5e-180) (0) (0) (0) (0) A2 -.26*** -.23*** -.65** -.56** -1.1** -.97** ( ) (.00021) (.011) (.029) (.013) (.031) A3 -.27*** -.25*** -.71*** -.63** -1.2*** -1.1** ( ) ( ) (.0058) (.014) (.0069) (.015) 8

9 A4 -.31*** -.27*** -.91*** -.77*** -1.6*** -1.4*** (2.1e-06) ( ) (.00041) (.0028) (.00046) (.0028) A5 -.31*** -.26*** -1.1*** -.89*** -2*** -1.7*** (6.5e-06) ( ) (.00062) (.0057) (.00069) (.0046) A6 -.26*** -.22*** -1.1*** -.9*** -2*** -1.7*** (.00012) (.00095) (.00066) (.005) (.00073) (.0053) A7 -.27*** -.25*** -.65** -.6** -1.1** -1** ( ) (.00006) (.012) (.019) (.014) (.022) A8 -.2* (.078) (.1) (.16) (.2) (.15) (.2) square meters *** ** *** ** *** ** & rooms (1.2e-06) (.032) (2.1e-07) (.033) (4.1e-07) (.049) Lower Secondary.049***.047***.22***.21***.37***.36*** School (4.3e-15) (7.7e-14) (8.9e-16) (7.3e-15) (1.5e-14) (1.5e-13) Higher.091***.099***.34***.38***.57***.65*** Secondary School (9.7e-37) (6.1e-44) (0) (0) (0) (0) Graduate.13***.14***.41***.49***.68***.82*** (1.7e-37) (4.8e-47) (0) (0) (0) (0) Postgraduate.13***.15***.37***.47***.61***.8*** (3.5e-12) (1.8e-16) (2.4e-10) (4.4e-16) (1.4e-09) (4.2e-15) Unemployed -.024* * * -.2 (.064) (.12) (.092) (.15) (.083) (.12) In retirement.034***.03***.14***.13***.25***.23*** (5.4e-06) ( ) (4.1e-07) (5.0e-06) (2.5e-07) (2.7e-06) Inactive person -.029*** -.022*** -.16*** -.12*** -.28*** -.21*** (.00017) (.0038) (9.0e-06) (.00093) (8.5e-06) (.0012) Duration *** *** *** *** *** *** ( ) (.00025) ( ) (.00012) ( ) (.00043) Age.0065***.0075***.024***.03***.044***.054*** (4.9e-08) (2.8e-10) (1.4e-06) (3.6e-09) (5.0e-07) (1.7e-09) Age sq *** *** *** *** *** *** ( ) (4.9e-07) ( ) (2.1e-07) (8.6e-06) (8.3e-08) 9

10 can keep.037***.021***.19***.1***.34***.19*** house warm (1.5e-08) (.0018) (9.3e-08) (.0037) (6.1e-08) (.0026) Potential.089***.062***.39***.28***.66***.49*** monthly rent (2.9e-62) (3.0e-26) (0) (0) (0) (0) Monthly.017*** *** ***.048 cost ( ) (.1) ( ) (.11) ( ) (.14) Tax on.024***.024***.11***.11***.18***.19*** dwelling (6.6e-188) (1.2e-191) (0) (0) (0) (0) Yearly Income.0096***.0086***.054***.052***.1***.097*** (1.3e-78) (1.1e-62) (0) (0) (0) (0) Consumption -.017*** -.012** -.051** ** goods (.0032) (.034) (.014) (.14) (.016) (.14) non durable.026***.019***.085***.055***.14***.093*** non cons. goods (2.9e-08) ( ) (1.8e-06) (.0023) (4.1e-06) (.0029) Center.022***.19***.35*** (.00037) (1.5e-11) (1.9e-11) North East.12***.56***.99*** (1.4e-73) (0) (0) North West.11***.49***.87*** (4.5e-55) (0) (0) cons -.84*** -.69*** -5.9*** -5.4*** -10*** -9.5*** (2.2e-21) (5.3e-15) (0) (0) (0) (0) N p < 0.10, p < 0.05, p < Fiscal Additionality Another point of interest concerns the effectiveness of the policy change in reducing tax evasion. This analyse tries to isolate the fiscal additionality from 10

11 the economic one. Specifically, the fiscal additionality concerns two cases: (a) in absence of any variation in the tax credits, the cost of makeover would still be made but would not have been included in the income tax (fiscal but not economic additionality); b) the tax credit increase has fostered the makeover and the inclusion in the income tax (fiscal and economic additionality). The analyse is run by comparing the data from the EU-SILC survey with those form the income tax: if someone reported in the EU-SILC that she had carried out a renovation but she did not benefit from the tax credit, then it can be a possible compliance gap. However, this kind of analyses presents some limitations and caution must be use in interpreting results: there can be multiple distortions due to measurement errors, response or recall bias. Concerning the econometric strategy, in order to estimate the reduction in submerged economy, the sample is restrected taking into account only those people that have undergone at least one renovation in the reference period. In this case the binary dependent variable takes value one if this renovation has not benefit the tax credit and value zero otherwise. Also in this case the explanatory variable of interest is the 2014 year dummy. In order to have a full comparison, three econometric models are estimated: the linear probability model (LPM), the probit model, and the logit model. For each model, there are two specifications (A and B) depending on whether or not the location is included (Islands and South, Center, North East and North West). The estimated results are display in Table 2. As in the previous paragraph, there is a remarkable results consistency among the models and all explanatory variables have the expected sign. It can be remarked that the 2014 tax credit change decreases the probability of not benefit the fiscal relief. In other words, the probability of not using the fiscal relief has fallen by about 26 percentage points, in all three estimated models. Now, we briefly discuss the impact of statistically significant explanatory variables. It can be remarked that the probability not to use the fiscal relief increases if the dwelling has been built in the 1990s and 1980s and the monthly managment costs are higher. The explanatoty variables that have a positive impact 11

12 on the use of the fiscal relief are: the ownership of flat in building with more than ten dwellings and higher education degrees. Inactive persons have a higher chance of not using the fiscal relief with respect to persons that are at work. All the variables that identifies a higher wealth (ability to keep home adequately warm, estimated potential monthly rent, tax on dwelling, yearly income) positively affect the use of the fiscal relief. Lastly, the location of the dwelling affect the use of the fiscal relief: the probability is higher if the dwelling is located in the North West of Italy. Table 2: Probability not to declare into the income tax LPM A LPM B Probit A Probit B Logit A Logit B * (.73) (.56) (.15) (.11) (.11) (.075) *** -.26*** -1.1*** -1.1*** -2.2*** -2.2*** (2.8e-102) (1.5e-99) (0) (0) (0) (0) After ** -.048* -.21* * -.34 (.049) (.088) (.071) (.12) (.075) (.11) (.31) (.16) (.32) (.18) (.25) (.15) ***.077***.27**.29***.53***.55*** (.0098) (.0059) (.013) (.0075) (.0056) (.004) **.052**.2*.2**.37**.36** (.036) (.037) (.053) (.05) (.034) (.038) (.34) (.28) (.39) (.31) (.31) (.27) (.94) (.77) (.88) (.91) (.99) (.83) (.41) (.28) (.53) (.36) (.48) (.35) (.87) (.5) (.95) (.58) (.96) (.62) Before **.15.22*.29.39** (.12) (.036) (.18) (.053) (.14) (.048) Terraced house *.11**.15*.19** 12

13 (.24) (.1) (.065) (.02) (.074) (.024) Flat in -.063*** -.065*** -.2*** -.21*** -.34*** -.36*** Build. with < 10 houses (9.3e-06) (4.8e-06) ( ) ( ) (.00011) ( ) Flat in -.1*** -.11*** -.36*** -.37*** -.61*** -.64*** Build. with > 10 houses (2.4e-14) (1.5e-15) (1.9e-13) (2.4e-14) (2.8e-13) (3.4e-14) A (.8) (.54) (.99) (.86) (.98) (.89) A (.87) (.9) (.81) (.93) (.82) (.92) A (.43) (.76) (.59) (.79) (.61) (.79) A (.11) (.22) (.22) (.37) (.31) (.47) A (.14) (.27) (.35) (.47) (.39) (.51) A (.47) (.39) (.81) (.78) (.88) (.85) A8 -.18*** -.19*** (.0089) (.0056) (.) (.) (.) (.) MQ & * 8.3e ** ** rooms (.057) (.64) (.037) (.51) (.033) (.49) Lower Sec *** -.064*** -.21*** -.21*** -.36*** -.37*** School (3.0e-06) (1.8e-06) (.00001) (6.3e-06) (7.6e-06) (4.3e-06) Higher Sec *** -.097*** -.3*** -.34*** -.51*** -.58*** School (2.6e-10) (1.7e-12) (3.1e-09) (2.5e-11) (3.5e-09) (2.5e-11) Graduate -.11*** -.12*** -.4*** -.46*** -.7*** -.81*** (1.9e-11) (7.2e-15) (1.8e-10) (3.0e-13) (1.5e-10) (2.8e-13) 13

14 Post graduate -.092*** -.11*** -.37*** -.46*** -.66*** -.8*** ( ) (3.5e-07) (.00038) ( ) (.00054) (.00002) Unemployed.064* *.19.37*.33 (.084) (.13) (.064) (.11) (.067) (.1) In retirment (.4) (.6) (.49) (.66) (.42) (.61) Inactive.087***.08***.31***.29***.55***.51*** person (1.2e-06) (7.4e-06) (5.5e-07) (3.9e-06) (3.6e-07) (2.3e-06) Duration (.29) (.34) (.28) (.27) (.28) (.29) Age * ** ** ** (.092) (.026) (.11) (.033) (.15) (.044) Age sq ** * * (.12) (.045) (.16) (.069) (.2) (.081) Can keep -.077*** -.058*** -.24*** -.17*** -.4*** -.29*** house warm ( ) (.0014) ( ) (.0034) ( ) (.0037) Potential -.085*** -.062*** -.33*** -.24*** -.56*** -.4*** monthly rent (7.5e-15) (3.0e-08) (4.2e-15) (3.4e-08) (3.3e-14) (1.8e-07) Monthly.087***.094***.31***.34***.55***.6*** Cost (1.4e-23) (3.8e-27) (0) (0) (0) (0) Tax -.034*** -.034*** -.15*** -.16*** -.3*** -.3*** dwelling (3.0e-96) (3.0e-96) (0) (0) (0) (0) Yearly -.019*** -.018*** -.057*** -.056*** -.095*** -.093*** income (4.2e-34) (1.9e-32) (0) (0) (0) (0) Consumption goods.018**.015*.083**.07*.14**.12* 14

15 (.031) (.079) (.024) (.056) (.032) (.068) Non durable non cons. goods (.24) (.69) (.31) (.77) (.27) (.69) Center -.068*** -.24*** -.39*** (2.4e-06) (4.0e-06) ( ) North East -.12*** -.43*** -.74*** (2.3e-19) (0) (0) North West -.11*** -.39*** -.67*** (8.1e-16) (1.8e-14) (3.3e-14) cons.9***.82*** 1.6*** 1.2** 2.8*** 2.1** (3.6e-12) (1.6e-10) (.0056) (.029) (.0074) (.037) N p < 0.10, p < 0.05, p < Multinomial Analyses Table 3: Fiscal and economic additionality Multinomial Probit A Multinomial Probit B Multinomial Logit A Multinomial Logit B 0 = Not renovate = Base category 1 = Renovate and not use of the fiscal relief **.089**.14**.14** (.012) (.016) (.011) (.014) (.2) (.21) (.23) (.26) After *** -.55*** -.77*** -.8*** (1.5e-06) (6.6e-07) ( ) (8.6e-06) ** -.24** (.028) (.028) (.1) (.11) (.28) (.26) (.13) (.12) 15

16 ***.28***.44***.46*** (.0058) (.0037) (.0033) (.0022) ***.3***.45***.47*** (.0013) (.0013) (.0019) (.0017) ***.27***.41***.41*** (.0031) (.0038) (.0058) (.0065) ***.35***.53***.53*** (.0003) (.00038) (.00058) (.00068) ***.26***.41***.39** (.0038) (.0082) (.0082) (.013) Before ***.38***.63***.58*** ( ) (.00029) ( ) (.00038) Terraced house.15***.14***.21***.19*** (.00029) (.0012) (.00093) (.0025) Flat in buil..097**.11** with < 10 houses (.032) (.012) (.25) (.14) Flat in buil..48***.5***.56***.59*** with > 10 houses (0) (0) (2.2e-16) (0) A2 -.98** -.94** -1.2* -1.2* (.033) (.038) (.074) (.078) A3 -.96** -.93** -1.2* -1.2* (.036) (.041) (.083) (.087) A4-1.1** -1** -1.4** -1.3* (.016) (.022) (.047) (.055) A (.1) (.14) (.19) (.22) A6 -.92* -.83* (.064) (.092) (.15) (.18) A7-1.1** -1.1** -1.4** -1.4** (.02) (.02) (.046) (.045) A8-11*** -11*** -14*** -14*** 16

17 (0) (0) (0) (0) MQ & rooms * (.065) (.31) (.14) (.37) Lower Sec School (.17) (.21) (.53) (.57) Higher Sec.11**.13*** School (.025) (.0075) (.24) (.12) Degree * (.22) (.08) (.76) (.51) Post graduate (.97) (.61) (.65) (.92) Unemployed (.27) (.21) (.28) (.22) In retirement.15***.15***.19**.19** (.0023) (.0031) (.011) (.012) Inactive person.14**.16***.23***.26*** (.014) (.0049) (.0052) (.0022) Duration ** ** * * (.025) (.043) (.066) (.092) Age (.2) (.14) (.44) (.38) Age sq (.22) (.15) (.44) (.37) Can keep house -.088* -.12** -.18** -.21*** warm (.086) (.018) (.018) (.0055) Potential.16***.094**.17***.09 monthly rent ( ) (.028) (.0063) (.18) Monthly cost.52***.49***.78***.75*** (0) (0) (0) (0) Tax on dwelling * (.9) (.68) (.064) (.11) 17

18 Yearly Income (.11) (.23) (.96) (.78) Consumtion goods (.67) (.48) (.55) (.42) Non durable.059* non cons. goods (.064) (.17) (.14) (.26) Center (.28) (.4) North East.28***.32*** (1.4e-09) (8.1e-06) Nord West.23***.26*** (7.6e-07) (.00042) cons -6.1*** -5.7*** -8*** -7.6*** (0) (0) (0) (2.2e-15) Multinomial Probit A Multinomial Probit B Multinomial Logit A Multinomial Logit B 2= Renovate and use of fiscal relief ** **.058 (.045) (.12) (.043) (.12) *** 1.2*** 1.5*** 1.5*** (0) (0) (0) (0) After *** -.46*** -.51*** -.59*** (1.3e-06) (3.8e-08) (1.9e-06) (3.3e-08) *** -.53*** -.63*** -.68*** (6.8e-09) (4.6e-10) (8.3e-09) (4.1e-10) *** -.25*** -.3*** -.33*** (.0034) (.0019) (.0037) (.0016) (.82) (.97) (.78) (.94) ***.18**.23**.21** (.0093) (.013) (.014) (.021) 18

19 ***.26***.35***.31*** (.00011) (.00042) (.00021) (.00083) ***.22***.3***.26*** (.0011) (.0039) (.0023) (.0084) ***.2**.34***.23** (.00034) (.011) (.0007) (.02) Before ***.14.29***.14 (.0032) (.11) (.0075) (.18) Terraced house.11***.073**.14***.1** (.0034) (.046) (.0031) (.031) Flat in build..39***.44***.52***.59*** with < 10 houses (0) (0) (0) (0) Flat in build. 1*** 1.1*** 1.3*** 1.3*** with > 10 houses (0) (0) (0) (0) A2-1.2*** -1.1*** -1.5*** -1.3*** (.00097) (.0029) (.0024) (.0052) A3-1.3*** -1.2*** -1.6*** -1.4*** (.00047) (.0012) (.0013) (.0023) A4-1.6*** -1.4*** -2*** -1.7*** ( ) (.00016) ( ) (.00032) A5-1.8*** -1.5*** -2.3*** -2*** (.00009) (.0011) (.00017) (.0011) A6-1.8*** -1.5*** -2.3*** -2*** ( ) (.00082) (.00018) (.0013) A7-1.2*** -1.2*** -1.5*** -1.4*** (.00093) (.0015) (.0025) (.0032) A8-1.2** -1.1** -1.4** -1.3** (.028) (.037) (.035) (.047) MQ & rooms *** ** *** ** (7.7e-08) (.025) (1.6e-07) (.037) 19

20 Lower Sec..3***.29***.37***.36*** School (1.1e-15) (1.0e-14) (1.5e-14) (1.6e-13) Higher Sec..47***.54***.58***.67*** School (0) (0) (0) (0) Graduate.56***.68***.69***.83*** (0) (0) (0) (0) Post graduate.5***.65***.61***.8*** (1.3e-09) (4.0e-15) (4.6e-09) (1.8e-14) Unemployed (.15) (.25) (.12) (.17) In retirement.21***.2***.27***.25*** (4.3e-08) (6.3e-07) (3.5e-08) (4.4e-07) Inctive person -.19*** -.13*** -.26*** -.18*** (.00012) (.0073) ( ) (.0059) Duration *** *** *** *** (2.7e-06) ( ) (8.7e-06) (.00011) Age.034***.042***.045***.055*** (1.0e-06) (2.4e-09) (4.7e-07) (1.6e-09) Age sq *** *** *** *** ( ) (1.3e-07) (7.6e-06) (6.9e-08) Can keep house.23***.11**.32***.17*** warm (1.8e-06) (.022) (4.7e-07) (.0093) Potential.55***.39***.69***.5*** monthly rent (0) (0) (0) (0) Monthly cost.18***.12***.21***.13*** (1.5e-12) (3.4e-06) (5.3e-11) ( ) Tax on dwelling.15***.15***.18***.19*** (0) (0) (0) (0) Yearly Income.074***.07***.1***.097*** (0) (0) (0) (0) 20

21 Consumption -.067** ** goods (.02) (.18) (.021) (.18) Non durable.12***.079***.15***.099*** non cons. goods (9.6e-07) (.0016) (1.9e-06) (.0019) Center.27***.35*** (1.3e-11) (1.9e-11) North East.8*** 1*** (0) (0) North West.7***.89*** (0) (0) cons -8.3*** -7.6*** -10*** -9.7*** (0) (0) (0) (0) N p < 0.10, p < 0.05, p < Conclusion The 2014 change in tax credit for building renovation it seems to have stimulated the likelihood of makeover both in terms of economic and fiscal additionality. In particular, the policy change seems to have mainly fostered aged persons, retired people, and those with a higher yearly income and able to afford non-durable non-consumption goods. On the opposite, it seems that the policy change has not had any significant effects on the renovation likelihood of inactive persons, with a relatively high share of food expenditure, i.e. the class of less wealthier people. A number of barriers have been identified for energy makeover that prevent consumers from making interventions. Some of them concern the high cost of implementation and information asymmetries (Murphy and Meier, 2011). Such barriers are also present for building renovations. Moreover, consumers give more relevance to actual savings than possible future ones; additionally for renovation income and price elasticity are relatively high (Hausman, 1979). As a matter of facts, the results confirm what could be expected from a theoretical 21

22 evaluation of this fiscal policy: leaving the burden to fully support the initial costs to consumers, the most favored ones are those who can support the initial costs (Du Can et al., 2014). References [1] Alberini, Anna, Andrea Bigano, and Marco Boeri. Looking for free riding: energy efficiency incentives and Italian homeowners. Energy Efficiency 7, no. 4 (2014): [2] Berton, Fabio, and Alessandro Cavallari. Evaluating consumer incentives: the Italian case of building renovations. Politica economica 29, no. 3 (2013): [3] Dubin, Jeffrey A., and Steven E. Henson. The distributional effects of the federal energy tax act. Resources and Energy 10, no. 3 (1988): [4] Hassett, Kevin A., and Gilbert E. Metcalf. Energy tax credits and residential conservation investment: Evidence from panel data. Journal of Public Economics 57, no. 2 (1995): [5] Walsh, Michael J. Energy tax credits and housing improvement. Energy Economics 11, no. 4 (1989):