Francisco José Veiga Forecast Errors in Prices and Wages: The Experience with Three Programme Countries

Transcription

1 C O N Ó M I C A S 44 N O T A S Francico Joé Veiga Forecat Error in Price and Wage: The Experience with Three Programme Countrie Manuel correia de Pinho / Maria Manuel Pinho Eforço Fical em Portugal: Uma Avaliação no Período Patrícia antune / Pedro lope Ferreira / lara n. Ferreira A Utilização da Experiência de Ecolha Dicreta na Valoração de Etado de Saúde Wei-Bin Zhang A Small Open Economic Growth Model with Imported Good, Tourim, and Term of Trade Imprena da UnIverIdade de CoImbra CoImbra UnIverIty pre FaCUldade de economia da UnIverIdade de CoImbra

2 (Página deixada propoitadamente em branco)

3 N O T A S C O N Ó M I C A S 44

4 Propriedade Faculdade de Economia Univeridade de Coimbra Av. Dia da Silva, Coimbra Apoio Eta revita é financiada por Fundo Nacionai atravé da FCT Fundação para a Ciência e a Tecnologia no âmbito do projeto PEt-OE/EGE/UI0031/2014 Número avulo Portugal 7,5 Outro paíe 15 Condiçõe de ainatura bienal Portugal: 20 Outro paíe: 40 ISSN ISSN Digital X DOI Depóito Legal 64932/93

5 Direcção Editorial (Editor) Paulino Teixeira (Faculdade de Economia, U. Coimbra; Francico Veiga (Ecola de Economia e Getão, U. Minho; fjveiga@eeg.uminho.pt) Joana Pai (ISEG, Univeridade de Liboa; jpai@ieg.utl.pt) Joé Varejão (Faculdade de Economia, U. Porto; varejao@fep.up.pt) Editore Aociado (Aociate Editor) António Afono (ISEG, Univeridade de Liboa; aafono@ieg.utl.pt) Ócar Afono (Faculdade de Economia, U. Porto; oafono@fep.up.pt) João Ferreira do Amaral (ISEG, Univeridade de Liboa; famaral@ieg.utl.pt) João Soua Andrade (Faculdade de Economia, U. Coimbra; jaa@fe.uc.pt) Manuela Arcanjo (ISEG, Univeridade de Liboa; marcanjo@ieg.utl.pt) Miguel St. Aubyn (ISEG, Univeridade de Liboa; mtaubyn@ieg.utl.pt) Rui Baptita (DEG-IST, U. Liboa; rui.baptita@tecnico.uliboa.pt) Pedro Brinca (Nova School of Buine and Economic; pedro.brinca@novabe.pt) Anabela Carneiro (Faculdade de Economia, U. Porto; anacar@fep.up.pt) Filipe Coelho (Faculdade de Economia, U. Coimbra; fcoelho@fe.uc.pt) João Paulo Cota (Faculdade de Economia, U. Coimbra; jpaulo@fe.uc.pt) Roa Branca Eteve (Ecola de Economia e Getão, U. Minho; rbranca@eeg.uminho.pt) Pedro Lope Ferreira (Faculdade de Economia, U. Coimbra; pedrof@fe.uc.pt) Joé Rui Figueira (DEG-IST, U. Liboa; figueira@tecnico.uliboa.pt) Francico Louçã (ISEG, Univeridade de Liboa; flouc@ieg.utl.pt) Jorge Braga de Macedo (Nova School of Buine and Economic; jbmacedo@novabe.pt) Joé Pereirinha (ISEG, Univeridade de Liboa; pereirin@ieg.uliboa.pt) Emeralda Ramalho (ISEG, Univeridade de Liboa; eramalho@ieg.uliboa.pt) Joaquim Ramalho (ISCTE, Intituto Univeritário de Liboa; Joaquim.Joe.Ramalho@icte.pt) Miguel Portela (Ecola de Economia e Getão, U. Minho; mangelo@eeg.uminho.pt) Miguel Torre Preto (DEG-IST, U. Liboa; miguel.preto@tecnico.uliboa.pt) Pedro Nogueira Ramo (Faculdade de Economia, U. Coimbra; pnramo@fe.uc.pt) Joé Rei (Faculdade de Economia, U. Coimbra; (jrei@fe.uc.pt) Miguel Soua (Faculdade de Economia, U. Porto; moua@fep.up.pt)

6 (Página deixada propoitadamente em branco)

7 Sumário Artigo Francico Joé Veiga Forecat Error in Price and Wage: The Experience with Three Programme Countrie 7 Manuel Correia de Pinho / Maria Manuel Pinho Eforço Fical em Portugal: Uma Avaliação no Período Patrícia Antune / Pedro Lope Ferreira / Lara N. Ferreira A Utilização da Experiência de Ecolha Dicreta na Valoração de Etado de Saúde 47 Wei-Bin Zhang A Small Open Economic Growth Model with Imported Good, Tourim, and Term of Trade 65

8 (Página deixada propoitadamente em branco)

9 Francico Joé Veiga Ecola de Economia e Getão Univeridade do Minho e NIPE fjveiga@eeg.uminho.pt Forecat Error in Price and Wage: The Experience with Three Programme Countrie Francico Joé Veiga ABSTRACT Thi paper evaluate the accuracy of price and wage forecat made under the cope of the Greek, Portuguee and Irih Economic Adjutment Programme (EAP). Beide comparing the quality of forecat for the three EAP, it check if they were le accurate than forecat for other euro area countrie, and compare the European Commiion forecat with thoe of the IMF and OECD. Programme forecat for Greece were, overall, the leat accurate, but thoe made under the econd Greek EAP eem to have gotten cloer to actual value. The reult alo ugget that EC forecat were not le accurate than thoe produced by the IMF or the OECD. Keyword: Forecat error; Price and wage; Economic Adjutment Programme; Greece; Portugal; Ireland. JEL Claification: E37; E31; E6; J3. Acknowledgment: Francico Veiga in thankful for the very helpful comment of Angar Belke, Daniel Gro, Miguel Lebre de Freita, and Antoni Adam. Received for publication: November 16, 2016 Reviion accepted for publication: February 6, 2017

10 Nota Económica Julho '17 (7-24) 1. Introduction Economic outlook were central piece of the Economic Adjutment Programme (EAP) implemented in European Union countrie affected by the overeign debt crii. In fact, thoe economic forecat were ued a a bai for the deign of the policy meaure Greece, Ireland, Portugal, and Cypru were aked to adopt. Given thi central role of economic forecat, it i vital that they are of high quality. Unfortunately, forecat are eldom a accurate a economit and policymaker would deire, and forecat error can potentially lead to inappropriate policie which generate ignificant economic cot. That may have been the cae in Programme countrie uch a Greece and Portugal, where auterity meaure were epecially harh. Thi paper evaluate the forecat performance of the Greek, Irih and Portuguee Programme in the pecific area of price and wage development, uing data between 2009 and Firt, the forecated behaviour of wage and price i compared with the value that actually materialized, checking if forecat error were larger for Greece (where the greatet forecat error eem to have been committed) 1 than for Portugal and Ireland, and/or for the EU countrie in general, and comparing EC forecat error with thoe of the IMF and the OECD. Development in price are proxied by the annual percentage change in the Harmonized Index of Conumer Price (HICP), while development in wage are proxied by annual percentage change in unit labour cot (ULC). 2 The paper i tructured a follow. Section 2 provide a graphical preentation of the forecat and forecat error, regarding price and wage, for the Greek, Irih and Portuguee EAP. Section 3 compare the forecat accuracy of the Programme forecat uing meaure uch a the mean error (ME), the mean abolute error (MAE), and the root mean quared error (RMSE). The following exercie, in Section 4, compare the current year, 1-year ahead and 2-year-ahead forecat of the EC with thoe of other international agencie, uch a the IMF and the OECD. 3 In Section 5, econometric etimation are performed in order to check if error tatitic for the 3 programme countrie are higher than for other euro area countrie. Section 6 conclude the paper. 2. Programme Forecat Of Price And Wage Development Thi ection preent the ucceive programme forecat for price and wage development, until April 2014, and compare them with actual value, taken from the February 2015 releae of the European Commiion AMECO databae. 1 An earlier urvey on the development of wage, price and the real exchange rate in Greece vi-à-vi the projection wa made in Piany-Ferry, et al. (2013), p Wage development were alo evaluated uing compenation of employee per capita. Since the reult are imilar to thoe obtained for ULC, they are not hown here in order to economize pace. 3 EC forecat were obtained in the publication European Economy (everal iue), IMF inflation forecat are from the World Economic Outlook (WEO) databae, and OECD forecat are from the Economic Outlook databae. 8

11 Francico Joé Veiga Forecat Error in Price and Wage: The Experience with Three Programme Countrie 2.1 Greece The ucceive inflation forecat included in the Greek EAP are preented in Figure 1. A vertical bar i included for the initial forecat and for thoe of each of the periodic review, while actual value are hown in a black line and with number next to it. The initial inflation forecat (in May 2010) were generally far from the actual value. Forecat for 2010 and 2011 were well below the value that materialized, while thoe of 2013 and 2014 were above them. Only the initial forecat for 2012 (1.2%) wa cloe to the actual value (1%). Subequent forecat made in 2010 corrected the etimate for that year, but underetimated inflation in 2011 and 2012 and overetimated it in 2013 and In fact, practically all forecat overetimated inflation for thee latter year. Inflation forecat error for the current year, 1-year-ahead and 2-year-ahead are preented in Figure 2. Thee error were calculated a the difference between each forecat and the correponding value obtained from the February 2015 releae of AMECO. 4 A mentioned above, the initial programme forecat (May 2010) were quite far from the final value. Except for the latter review, inflation tended to be underetimated in the forecat for the current year and 1-year ahead. Figure 1: Inflation forecat for Greece Source: European Commiion, European Economy, Occaional Paper and AMECO. Figure 2 to 8 are baed on the ame ource. 4 Contrary to Cabanilla and Terzi (2012), who calculate forecat error uing the firt available etimate for the current year forecat, and the firt ettled etimate for the year-ahead forecat, we preferred to ue the mot recent releae of AMECO, which better reflect final value than the firt available or ettled etimate. The May2013, July year-ahead forecat error and the April year and 2-year-ahead forecat error are in fact forecat correction, a they refer to 2015 and 2016 for which final value are not available yet. That i, in thoe cae the forecat error/correction wa calculated a the difference between thoe programme forecat and the February 2015 AMECO forecat. The ame procedure will be ued in the following figure for forecat error in Greece and in other programme countrie. 9

12 Nota Económica Julho '17 (7-24) Figure 2: Inflation forecat error for Greece Figure 3 preent the forecat and actual value of annual percentage change in unit labour cot (ULC). A happened with inflation, the initial programme forecat were below the actual value of 2010 and 2011 and above actual value in 2013 and Figure 4 how that forecat error for the current year were motly negative (except for the lat 3 review), while for 1-year-ahead and 2-year-ahead were motly poitive (again, with the exception of the lat 3 review). Figure 3: Unit Labour Cot forecat for Greece 2.2 Portugal Figure 5 preent HIPC inflation forecat error for Portugal. Current year forecat error were generally quite mall, except thoe of the 2014 review, which overetimated 10

13 Francico Joé Veiga Forecat Error in Price and Wage: The Experience with Three Programme Countrie Figure 4: Unit Labour Cot forecat error for Greece the inflation rate by almot 1 percentage point. Except for the initial (June 2011) and firt review (September 2011) forecat, inflation forecat for Portugal tended to overetimate inflation. In fact, that wa alway the cae for 2-year-ahead forecat. Neverthele, inflation forecat error for Portugal never reached 3 percentage point, a they did for Greece (ee Figure 2). Figure 5: Inflation forecat error for Portugal Forecat error for the annual change in unit labour cot (ULC) are illutrated in Figure 6. Forecat made in 2011 for the current year and 1-year-ahead overetimated the growth in ULC, while latter forecat underetimated it. Two-year-ahead forecat generally underetimated the change in ULC. Although error are generally higher for ULC than for 11

14 Nota Económica Julho '17 (7-24) inflation, the ame happened for Greece, which had even higher forecat error in ULC than Portugal, in ome occaion exceeding 6 percentage point (ee Figure 4). Figure 6: Unit Labour Cot forecat error for Portugal 2.3 Ireland A pattern imilar to that of inflation forecat for Portugal i oberved for Ireland in Figure 7, a initial current year and 1-year-ahead forecat tended to underetimate inflation, while latter one overetimated it. A for Portugal, 2-year-ahead forecat alway overetimated inflation, and the bigget forecat error were around 1 percentage point. Figure 7: Inflation forecat error for Ireland Forecat error in change in ULC for Ireland, hown in Figure 8, were generally higher than thoe for Portugal, and comparable to thoe for Greece. In fact, change in ULC were ometime underetimated by almot 6 percentage point. 12

15 Francico Joé Veiga Forecat Error in Price and Wage: The Experience with Three Programme Countrie Figure 8: Unit Labour Cot forecat error for Ireland 3. Forecat Performance of The Three Economic Adjutment Programme Thi ection compare the forecat performance of the three EAP (Greece, Portugal, and Ireland), uing three ummary tatitic: the mean error (ME), the mean abolute error (MAE), and the root mean quared error (RMSE) FORECAST ERRORS AND SUMMARY STATISTICS Following Cabanilla and Terzi (2012), and previou tudie, we define the forecat error and the ummary tatitic a explained below. The forecat error i the difference between the forecat and the realization. That i:, =, +1, = +1, +1 +2, = +2, +2, for the current year, for the year ahead, for two year ahead, where y t,t, y t+1,t and y t+2,t are the forecat of y made at period t for t, t+1 and t+2, repectively; and y t, y t+1 and y t+2 are the realization of y in year t, t+1 and t+2, repectively. The programme forecat accuracy i aeed uing three commonly ued ummary tatitic: The mean error (ME), which i the average difference between the forecat and the outturn. Although it ha the problem that negative error may compenate poitive one, not being a good accuracy indicator, it may help ignal biae in the forecat. 5 Thee tatitic were ued by Arti (1996) to evaluate IMF-WEO forecat, and by Keereman (1999), Melander et al. (2007), and Cabanilla and Terzi (2012) to evaluate the accuracy of the European Commiion forecat. Although thee tudie only conidered the current year and the year ahead forecat, we alo analye the 2-yearahead forecat. See alo Öller and Barot (2000), who analyed 1-year-ahead forecat by the OECD and by national intitute of GDP growth and inflation in 13 European countrie. 13

16 Nota Económica Julho '17 (7-24) = 1 =1,, for the current year 1 = 1 =1 1 +1,, for the year ahead = 1 =1 +2,, for two year ahead. The mean abolute error (MAE), which i the average abolute difference between the forecat and the outturn. Since poitive and negative error no longer cancel each other out, it i a more accurate meaure of the average forecat error than the ME. = 1 =1,, for the current year = 1 1 =1 +1,, for the year ahead 1 = 1 =1 +2,, for two year ahead. The root mean quared error (RMSE), which i a meaure of the relative ize of the forecat error, give greater weight to large than to mall error. = 1 2 =1,, for the current year = 1 2 =1 +1,, for the year ahead. = 1 2 =1 +2,, for two year ahead. 3.2 Programme` Forecat Accuracy Table 1 how the ummary tatitic for the inflation forecat made in the EAP of Greece, Portugal, and Ireland. A in the figure hown above, all available programme forecat (initial and of all review) for each country were conidered. Regarding current year forecat, the mean error for Greece i negative, indicating that forecat, on average, underetimated inflation in that country. The poitive ME in Portugal, which i cloe to the MAE, indicate that inflation forecat for Portugal tended to overetimate actual inflation. For Ireland, the ME i much maller than the MAE, which indicate that poitive error were in large part cancelled out by negative one. The MAE i higher for Greece than for the other countrie, although the difference to Ireland i not very big. But, the RMSE for Greece i twice a big a that for Ireland (the econd larget), which i due to the large error made in the initial inflation forecat for Greece. 14

17 Francico Joé Veiga Forecat Error in Price and Wage: The Experience with Three Programme Countrie Table 1: Programme Inflation forecat tatitic Country Current Year 1-year-ahead 2-year-ahead ME MAE RMSE ME MAE RMSE ME MAE RMSE Greece -0,42 0,58 1,00-0,31 0,80 1,23 0,69 0,87 1,09 Portugal 0,24 0,32 0,44 0,61 0,89 0,94 1,04 1,04 1,11 Ireland 0,08 0,44 0,52 0,16 0,91 0,95 1,01 1,01 1,07 Source: European Commiion, European Economy, Occaional Paper (everal iue) and AMECO. Although the 1-year-ahead forecat have a omewhat imilar pattern (epecially for the ME and RMSE), thoe of 2-year-ahead are quite different, with Portugal exhibiting the larget ME, MAE and RMSE. Thu, it i intereting to note that forecat accuracy for Portugal wa the highet when conidering the current year, but it wa the lowet for 2-year-ahead. For 1-year-ahead, Greece ha the lowet accuracy according to the RMSE (which penalize the large error of the initial programme forecat), and the highet according to the MAE. Table 2 preent the ummary tatitic for the forecat of change in unit labour cot (ULC). Regarding current year forecat, the lowet accuracy wa achieved for Ireland, according to any of the meaure. The ame applie to 1-year-ahead forecat, according to the MAE and RMSE. A for 2-year-ahead forecat, the lowet accuracy wa for Greece, according to all tatitic. Additionally, forecat for Greece were alway le accurate than thoe for Portugal. Table 2: Programme ULC forecat tatitic Country Current Year 1-year-ahead 2-year-ahead ME MAE RMSE ME MAE RMSE ME MAE RMSE Greece -1,44 1,76 2,42 1,28 1,54 2,32 2,64 2,91 3,78 Portugal -0,02 1,89 2,01-0,11 1,80 2,15-1,07 1,19 1,51 Ireland -2,18 2,72 3,23-0,57 3,04 3,34-1,13 2,90 3,29 Source: European Commiion, European Economy, Occaional Paper (everal iue) and AMECO. In um, there i ome indication that forecat for Greece were le accurate than for Portugal and Ireland regarding inflation (for current year and 1-year-ahead forecat). Regarding ULC forecat, thoe for Greece were till le accurate than for Portugal, but more accurate than thoe for Ireland. 4. Comparion Of EC Forecat Error With Thoe Of The IMF And OECD In order to check if the European Commiion forecat were more or le accurate than thoe of the IMF and OECD, we how in thi ection the ummary tatitic of thoe 15

18 Nota Económica Julho '17 (7-24) intitution forecat. Following Cabanilla and Terzi (2012), current-year forecat are taken from the Commiion pring forecat, wherea 1-year-ahead forecat come from the Commiion autumn forecat. The 2-year-ahead forecat, not ued in that tudy, are alo taken from the autumn forecat. The ame procedure i ued regarding the IMF forecat, taken from it World Economic Outlook, and the OECD forecat, taken from it Economic Outlook. 6 The firt forecat conidered for each intitution i that of autumn 2009 and the lat i that of autumn We tart by comparing forecat for Greece, Portugal and Ireland, and then we how Ordinary leat Square (OLS) regreion reult for etimation including other euro area countrie. 4.1 Forecat Statitic For Greece, Portugal And Ireland The forecat ummary tatitic for the inflation forecat of the EC, IMF and OECD, for Greece, Portugal and Ireland, are hown in Table 3. IMF forecat were generally the leat accurate, according to the MAE and RMSE, regardle of the country and time horizon conidered. 7 Regarding current year forecat, the OECD performed omewhat better than the EC for Greece and Ireland, while their accuracy wa practically the ame for Portugal. The EC 1-year-ahead forecat were generally more accurate than thoe of the OECD for Greece, lightly better for Ireland, and about the ame for Portugal (the EC ha a lower MAE, but a higher RMSE). Regarding 2-year-ahead forecat, the EC performed wore than the OECD for Greece and Ireland, but better for Portugal. Table 4 how the ummary tatitic for EC and OECD forecat of annual percentage change in unit labour cot. 8 The EC forecat for Greece were more accurate than the OECD when conidering the current year and the year ahead, but the oppoite happened for 2-year-ahead. Regarding forecat of change in ULC for Ireland, the OECD wa alway more accurate than the EC. A comparion i not poible for Portugal, a there were no ULC forecat available for thi country in the OECD Economic Outlook. In um, it i not poible to tate that the European Commiion forecat were alway the mot accurate. But, it i afe to argue that the EC inflation forecat for the 3 programme countrie conidered were alway more accurate than thoe of the IMF. 9 Additionally, the 6 The Spring forecat of the EC were generally releaed in May, thoe of the IMF in April, and thoe of the OECD in June. The Fall forecat of the EC were generally releaed in November, thoe of the IMF in October, and thoe of the OECD in December. Although the releae date were not exactly the ame, they were cloe enough to make EC, IMF and OECD forecat comparable. Neverthele, the fact that the IMF i the firt and the OECD i the lat intitution to preent forecat, may be a diadvantage for the former and an advantage for the latter. 7 A mentioned in the previou footnote, the IMF i alway the firt of thee three intitution to publih it forecat. Uing lightly more recent information and knowing the IMF forecat may help the EC and the OECD produce better forecat. 8 There are no forecat for unit labour cot or compenation of employee in the IMF WEO. 9 Thi may eem trange, a both the EC and the IMF were part of the Troika adminitering the Greek EAP, and we would expect their forecat to be aligned. In practice, however, Programme forecat eem to be more aligned with EC forecat than with thoe of the IMF. 16

19 Francico Joé Veiga Forecat Error in Price and Wage: The Experience with Three Programme Countrie Table 3: Inflation forecat tatitic Current Year (Spring Forecat) 1-year-ahead (Fall Forecat) 2-year-ahead (Fall Forecat) ME MAE RMSE ME MAE RMSE ME MAE RMSE European Commiion Greece -0,60 0,86 1,05 0,16 0,55 0,73 0,75 0,93 1,15 Portugal 0,10 0,32 0,35 0,21 0,63 0,79 0,46 1,01 1,19 Ireland 0,19 0,35 0,41-0,18 0,69 0,76 0,30 0,75 0,89 International Monetary Fund (World Economic Outlook) Greece -1,02 1,22 1,59-0,88 1,02 1,46-0,06 1,05 1,26 Portugal -0,09 0,61 0,69-0,44 0,93 1,19-0,28 1,29 1,46 Ireland -0,11 0,43 0,52-0,13 1,09 1,18 0,07 0,82 0,84 OECD (Economic Outlook) Greece -0,41 0,46 0,77-0,63 0,92 1,31-0,25 0,75 0,89 Portugal -0,19 0,33 0,35-0,21 0,66 0,75-0,40 1,21 1,45 Ireland 0,17 0,17 0,24 0,15 0,71 0,77-0,05 0,56 0,58 Source: European Commiion, European Economy, Economic Forecat and AMECO; International Monetary Fund, World Economic Outlook; and OECD, Economic Outlook. Table 4: Unit labour cot forecat tatitic Current Year (Spring Forecat) 1-year-ahead (Fall Forecat) 2-year-ahead (Fall Forecat) ME MAE RMSE ME MAE RMSE ME MAE RMSE European Commiion Greece -0,96 1,30 2,20 0,60 0,76 1,07 1,71 1,73 2,55 Portugal 0,47 1,25 1,51 0,67 1,01 1,22 0,22 1,03 1,41 Ireland 0,20 3,08 3,49-1,68 2,36 2,96-0,71 1,82 2,49 OECD (Economic Outlook) Greece 2,45 2,45 3,54 1,31 2,50 3,35 2,74 3,66 4,42 Ireland -1,76 1,76 2,24-0,60 1,90 2,47-1,10 1,70 2,01 Source: European Commiion, European Economy, Economic Forecat and AMECO; and OECD, Economic Outlook. 17

20 Nota Económica Julho '17 (7-24) EC forecat for inflation and ULC do not eem to be, on average, le accurate than thoe of the OECD. 4.2 Comparion Of Forecat Error For Programme Countrie With Thoe For Oher Euro Area Countrie In order to check if the forecat accuracy for the three programme countrie analyed above wa different from that for other 14 euro area countrie, 10 we etimated individual OLS cro-country regreion for each variable (inflation and ULC), intitution (EC, IMF and OECD), and ummary tatitic (ME, MAE and RMSE). In each regreion, the dependent variable i a ummary tatitic, and the explanatory variable are the dummy variable for the three programme countrie and a contant. The etimated model i the following: = , where FS i i a forecat tatitic (ME, MAE or RMSE) for country i; Greece i, Portugal i and Ireland i are dummy variable that take the value of one for the repective countrie and zero otherwie; ε i i a white noie error term; and a, b, g, and δ are parameter to be etimated. The reult of the regreion for inflation forecat tatitic are preented in Table Regarding current year forecat, the dummy variable for Greece i poitive and tatitically ignificant for the MAE and RMSE, which indicate that EC inflation forecat for Greece were le accurate than for the remaining euro area countrie. The ame applie to the IMF and the OECD. There i alo ome indication that IMF and OECD forecat for Ireland were alo le accurate than for the other euro area countrie (but more accurate than thoe for Greece). Finally, the regreion for the MAE, indicate that OECD inflation forecat for Portugal may have been more accurate than for the other euro area countrie. A different picture emerge for the 1-year-ahead forecat, a EC forecat for the three programme countrie eem to have been more accurate than for the remaining euro area countrie. Thi i not true for IMF forecat, ince there i a poitive and tatitically ignificant coefficient for Greece in the RMSE regreion, and the dummy variable for Portugal and Ireland are not tatitically ignificant. A for OECD year ahead forecat, there i weak evidence of higher error for Greece (for the RMSE) and lower one for Portugal and Ireland, than for the remaining euro area countrie. Regarding the 2-year-ahead forecat of the EC, there i no robut evidence of different forecat accuracy for the programme countrie. For the IMF, there i indication of higher error for Ireland and lower one for Portugal. Finally, the OECD 2-year-ahead forecat eem to have been more accurate for Greece and Portugal. 10 Only the firt 17 member of the euro area were conidered. We did not include Latvia and Lithuania, a they were not euro area member during the entire period under analyi. 11 A the ME i not a good meaure of accuracy, we will only refer to it when it clearly indicate biaed forecat. Thu, mot of the comment in thi ection refer to the reult for the MAE and the RMSE. The R-Squared of each etimation i alo reported in Table 5. 18

21 Francico Joé Veiga Forecat Error in Price and Wage: The Experience with Three Programme Countrie Table 6 preent the reult of regreion for the forecat of change in unit labour cot (ULC) by the European Commiion and the OECD. Current year ULC forecat for the programme countrie eem to have been le accurate than for the other euro area countrie. In fact, all dummy variable for programme countrie are poitive and tatitically ignificant in the regreion for the MAE and the RMSE. The ame applie to 1-year-ahead forecat of the EC for Portugal and of the OECD for Greece and Ireland. Regarding 2-year-ahead forecat, they eem to have been le accurate for Greece and Portugal than for the remaining euro area countrie. Thu, overall, ULC forecat for the 3 programme countrie tended to be le accurate than for the other euro area member. Table 5: Regreion for inflation forecat Current Year (Spring Forecat) European Commiion 1-year-ahead (Fall Forecat) 2-year-ahead (Fall Forecat) ME MAE RMSE ME MAE RMSE ME MAE RMSE Greece *** 0.471*** 0.610*** *** *** 0.338*** (-11.71) (8.023) (8.883) (0.283) (-5.976) (-4.406) (3.838) (0.910) (0.762) Portugal 0.198*** *** ** *** * (3.900) (-0.628) (-0.422) (-5.794) (-2.810) (-3.898) (-1.181) (-1.416) (-2.083) Ireland 0.108* *** *** * (2.122) (-1.224) (-1.397) (1.036) (-4.196) (-3.387) (0.531) (2.029) (1.141) R-quared International Monetary Fund (World Economic Outlook) Greece *** 0.737*** 1.003*** *** *** (-13.22) (13.22) (11.63) (-7.392) (0.864) (4.673) (-0.254) (0.369) (1.024) Portugal ** *** (-1.171) (-0.940) (-0.854) (1.042) (1.552) (1.428) (0.671) (-2.989) (-5.002) Ireland ** ** * 0.272*** 0.265*** (-0.927) (2.386) (1.190) (-2.469) (-0.175) (1.484) (-1.836) (3.993) (3.748) R-quared OECD (Economic Outlook) Greece a-0.410*** 0.218*** 0.488*** ** * ** ** (-8.017) (8.797) (16.15) (-2.637) (-0.251) (2.049) (0.0413) (-2.803) (-2.852) Portugal 0.178*** ** *** ** ** *** *** (3.478) (-2.797) (-1.408) (4.279) (-2.617) (-2.914) (1.556) (-4.150) (-4.606) Ireland *** *** ** *** ** (-3.606) (3.585) (2.299) (1.078) (-3.160) (-3.067) (-1.062) (0.411) (0.424) R-quared Source: European Commiion, European Economy, Economic Forecat and AMECO; IMF, World Economic Outlook; and OECD, Economic Outlook. Note: OLS model etimated for a ample of 17 euro area countrie, from autumn 2009 to autumn The etimated coefficient for the contant i not hown. T-tatitic are in parenthee. Significance level at which the null hypothei i rejected: ***, 1%; **, 5%, and *, 10%. 19

22 Nota Económica Julho '17 (7-24) Table 6: Regreion for forecat of annual change in Unit Labour Cot (ULC) Current Year (Spring Forecat) 1-year-ahead (Fall Forecat) 2-year-ahead (Fall Forecat) European Commiion ME MAE RMSE ME MAE RMSE ME MAE RMSE Greece *** 0.488** 1.186*** 0.688** *** 0.708*** 1.150*** (-6.392) (2.838) (5.605) (2.983) (-0.995) (-0.922) (5.864) (3.985) (4.888) Portugal 0.359** 2.276*** 2.470*** *** 1.424*** 1.673*** ** 0.807*** 1.084*** (2.870) (13.25) (11.68) (-6.909) (8.382) (7.001) (-2.628) (4.540) (4.608) Ireland 0.633*** 0.437** 0.489** 0.756*** (5.063) (2.546) (2.313) (3.276) (0.469) (-0.293) (0.654) (0.0971) (0.0316) R-quared OECD (Economic Outlook) Greece 2.751*** 1.577*** 2.473*** 1.824*** 1.365*** 1.895*** 3.645*** 2.111*** 2.591*** (14.44) (15.81) (17.58) (6.143) (7.218) (7.952) (9.720) (9.642) (10.34) Ireland *** 0.879*** 1.177*** *** 1.016*** (-7.640) (8.819) (8.365) (-0.296) (4.060) (4.264) (-0.530) (0.672) (0.707) R-quared Source: European Commiion, European Economy, Economic Forecat and AMECO; and OECD, Economic Outlook. Note: OLS model etimated for a ample of 17 euro area countrie (EC) or 14 euro area countrie (OECD), from autumn 2009 to autumn The etimated coefficient for the contant i not hown. T-tatitic are in parenthee. Significance level at which the null hypothei i rejected: ***, 1%; **, 5%, and *, 10%. 4.3 Comparion Of Forecat Error Combining The Three Intitution A final econometric analyi conited on etimating regreion for a ample which combine the forecat tatitic of the 3 intitution. In order to check if the EC forecat were in general more or le accurate than thoe of the IMF and OECD, dummy variable for thoe intitution were added to the lit of explanatory variable. The reult are preented in Table 7. In the regreion for current year inflation forecat, the dummy variable for the IMF i poitive, while that for the OECD i negative. Since both are tatitically ignificant, thi indicate that EC inflation forecat were, on average, more accurate than thoe of the IMF, but le accurate than thoe of the OECD. A for ULC, the OECD dummy variable i not tatitically ignificant, indicating that there wa no difference in accuracy relative to EC forecat. MAE regreion indicate that EC inflation forecat 1 and 2-year-ahead may have been more accurate than thoe of the IMF and OECD. For ULC, there i no difference in 1-year-ahead forecat, but the EC eem to have been more accurate than the OECD in 2-year-ahead forecat (in the MAE etimation). 20

23 Francico Joé Veiga Forecat Error in Price and Wage: The Experience with Three Programme Countrie The reult for the country dummy variable indicate that inflation current year forecat were, on average (and conidering the 3 intitution), le accurate for Greece than for the remaining euro area countrie. A for ULC forecat (by the EC and OECD), they were le accurate for the 3 programme countrie, epecially for Greece and Portugal, than for the other. For 1-year-ahead forecat, the MAE regreion indicate that greater than average accuracy may have been achieved in inflation forecat for Ireland, while there wa lower than average accuracy in the 1-year-ahead ULC forecat for Portugal. Finally, concerning 2-year-ahead, inflation forecat for Portugal were more accurate and thoe for Ireland were le accurate than average. ULC forecat were le accurate for Greece than for the other countrie. Table 7: Regreion forecat combining the three intitution Current Year (Spring Forecat) 1-year-ahead (Fall Forecat) 2-year-ahead (Fall Forecat) Inflation Unit Labour Cot ME MAE RMSE ME MAE RMSE ME MAE RMSE Greece *** 0.476*** 0.701*** * (-4.640) (3.835) (5.332) (-1.853) (-0.744) (0.750) (0.875) (-0.807) (-0.741) Portugal * ** *** (1.278) (-1.816) (-1.157) (0.389) (-0.736) (-1.291) (0.624) (-2.450) (-2.955) Ireland ** ** 0.150* (-0.630) (0.886) (0.432) (-0.172) (-2.293) (-1.204) (-1.223) (2.478) (1.977) IMF * 0.179** *** 0.182** *** 0.164* (-0.806) (1.824) (2.047) (-4.380) (2.249) (1.568) (-3.559) (2.014) (1.198) OECD *** ** *** 0.136* *** 0.229* ( ) (-2.966) (-2.677) (-4.290) (1.800) (1.062) (-5.510) (1.826) (1.467) R-quared Greece ** 1.830*** 1.249*** *** 1.408** 1.869*** (0.731) (2.353) (3.572) (3.030) (1.023) (1.047) (3.617) (2.609) (3.296) Portugal *** 1.823*** *** 1.341*** * (-0.734) (2.924) (3.549) (-1.558) (3.622) (4.018) (-1.777) (1.555) (1.574) Ireland 0.686*** 0.428** 0.489** 0.838*** (4.656) (2.679) (2.514) (3.946) (0.677) (-0.113) (1.014) (0.245) (0.111) OECD * 0.582** ( ) (0.247) (0.210) (-0.773) (1.126) (0.865) (-1.904) (2.347) (1.548) R-quared Note: OLS model etimated for a ample of 17 euro area countrie, from autumn 2009 to autumn There are 49 obervation for inflation and 31 for ULC. The etimated coefficient for the contant i not hown. T-tatitic are in parenthei. Significance level at which the null hypothei i rejected: ***, 1%; **, 5%, and *, 10%. Source: European Commiion, European Economy, Economic Forecat and AMECO; IMF, World Economic Outlook; and OECD, Economic Outlook. 21

24 Nota Económica Julho '17 (7-24) 5. Concluion The analyi preented in thi tudy lead u to conclude that programme forecat for Greece were le accurate than for Portugal and Ireland regarding inflation. Initial inflation forecat (in May 2010) for Greece, for the current year and the year ahead, were far below from the actual value. The ame, although to a maller degree, happened with the initial forecat of the econd adjutment programme for Greece. Forecat for unit labour cot (ULC) were le accurate for Ireland regarding the current year and year ahead, and le accurate for Greece for 2-year-ahead. Comparing to ULC programme forecat for Portugal, forecat for Greece were alway le accurate. Although programme forecat for Greece were, overall, the leat accurate of the 3 programme forecat, thoe made in more recent review of the Greek EAP eem to have gotten cloer to actual value. The forecat error in programme countrie may in part have been due to greater difficulty to forecat in more volatile environment. In fact, IMF and OECD forecat for thee three programme countrie alo had relatively large error, with IMF inflation forecat being le accurate than thoe of the European Commiion for any of the three countrie. 12 Additionally, current year inflation and ULC forecat for Greece were generally le accurate than thoe for the other euro area countrie, regardle if they were made by the EC, IMF, or OECD. Actually, ULC forecat for Greece, Ireland and Portugal were alway le accurate than for the other euro area member. Overall, EC forecat do not eem to be le accurate than thoe of the IMF and OECD. In fact, regarding inflation, they are on average more accurate than thoe of the IMF. REFERENCES Arti, M. (1996) How accurate are the IMF hort-term economic forecat? Another examination of the World Economic Outlook, IMF Working Paper, WP/96/89, Augut. Cabanilla, L. G.; Terzi, A. (2012) The acuracy of the European Commiion forecat re-examined. Economic Paper No. 476, European Commiion, Directorate-General for Economic and Financial Affair, Bruel, Belgium. European Commiion ( ) European Economy, Economic Forecat, European Commiion, Directorate-General for Economic and Financial Affair, Bruel, Belgium. Forecat conulted: Spring 2009, Autumn 2009, Spring 2010, Autumn 2010, Spring 2011, Autumn 2011, Spring 2012, Autumn 2012, Winter 2013, Spring 2013, Autumn 2013, Winter 2014, Spring 2014, Autumn European Commiion ( ) European Economy, Occaional Paper, European Commiion, Directorate-General for Economic and Financial Affair, Bruel, Belgium. Number conulted: Greek EAP 61, 68, 72, 77, 82, 87, 94, 123, 148, 159, 192; Irih EAP 76, 78, 84, 88, 93, 96, 115, 127, 131, 154, 162, 167, 195, 210; Portuguee EAP 79, 83, 89, 95, 111, 117, 124, 153, 164, 171, 191, 202, Again, it i worth noting that the EC and, epecially, the OECD could have taken the IMF forecat into account when computing their own. Furthermore, they may ue lightly more recent data, which alo help producing more reliable forecat. 22

25 Francico Joé Veiga Forecat Error in Price and Wage: The Experience with Three Programme Countrie International Monetary Fund ( ) World Economic Outlook Databae, IMF, Wahington, DC. Iue: October 2009, April 2010, October 2010, April 2011, September 2011, April 2012, October 2012, April 2013, October 2013, April 2014, October Keereman, F. (1999) The track record of the Commiion forecat, Economic Paper No. 137, European Commiion, Directorate-General for Economic and Financial Affair, Bruel, Belgium. Melander, A.; Simanidi, G.; Grenouilleau, D. (2007) The track record of the Commiion forecat: An update, Economic Paper No. 291, European Commiion, Directorate-General for Economic and Financial Affair, Bruel, Belgium. OECD ( ) Economic Outlook Databae. OECD, Pari. Iue: December 2009, June 2010, December 2010, June 2011, December 2011, June 2012, December 2012, June 2013, December 2013, May 2014, November 2014 Öller; L.E.; Barot, B. (2000) The accuracy of European growth and inflation forecat, International Journal of Forecating, 16 (3), Piany-Ferry, J. Sapir, A.; Wolff, G. (2013) EU-IMF Aitance to Euro-Area Countrie: An Early Aement, Bruegel Blueprint 19, Bruel. 23

26 (Página deixada propoitadamente em branco)

27 Manuel Correia de Pinho Maria Manuel Pinho Faculdade de Economia da Univeridade do Porto e CEF.UP/ mpinho@fep.up.p Eforço Fical em Portugal: Uma Avaliação no Período Manuel Correia de Pinho Maria Manuel Pinho RESUMO O nível de ficalidade praticada por um paí tem implicaçõe no plano económico e ocial, podendo er aferido por recuro ao conceito de receita fical, de carga fical e de eforço fical. O eforço fical é expreo por um índice que relaciona a carga fical (rácio entre a receita fical e o PIB) com uma de trê medida alternativa: o PIB per capita, uma carga fical padrão e a capacidade fical da economia. Com bae em dado para o período e para o 28 Etado-membro da União Europeia, procede-e à etimação do índice de eforço fical atravé de metodologia diferente. Em trê do quatro índice calculado, o reultado ugerem que o eforço fical em Portugal eguiu uma tendência de redução entre 1996 e 2010 e de aumento no período de implementação do programa de ajutamento económico e financeiro de , em que, no entanto, tenham ido atingido o valore obervado em No quarto índice calculado, o reultado apontam no entido de uma tendência de aumento do eforço fical em Portugal dede A evidência empírica ugere ainda que o eforço fical em Portugal tem ido inferior à média do paíe da União Europeia. Palavra-chave: carga fical, capacidade fical, fronteira fical etocática, eforço fical. ABSTRACT The tax level in a country ha an impact in both the economic and the ocial phere and can be aeed uing the concept of tax revenue, tax burden and tax effort. The tax effort i expreed by an index relating the tax burden (ratio between the tax revenue and GDP) with one of three alternative meaure: GDP per capita, benchmark tax burden and tax capacity. Uing data for the period and the 28 European Union Member State, we etimate tax effort indexe on the bai of different methodologie. For three of the four computed indice, the reult ugget that the tax effort in Portugal followed a downward trend between 1996 and 2010 and an upward one in the period of the implementation of the economic and financial adjutment program, without, however, having reached the figure oberved in For the fourth index, the reult point toward a trend of increae in the tax effort in Portugal ince The empirical evidence alo ugget that the tax effort in Portugal ha remained below the EU Member State average. Keyword: tax burden, tax capacity, tochatic tax frontier, tax effort JEL Claification: C51, E62, H21 Received for publication: November 14, 2016 Reviion accepted for publication: February 3, 2017

28 Nota Económica Julho '17 (25-46) 1. Introdução O nível de ficalidade praticado por um paí tem implicaçõe no plano económico e ocial. Em primeira intância, o nível de ficalidade oferece uma perceção da medida em que a ociedade apreciam a maior ou menor provião pública de ben. Porém, o nível de ficalidade é também um fator determinante da competitividade externa da economia e pode ter um efeito de crowding out obre o recuro afeto à atividade económica da família e da emprea. Importa, ainda, ublinhar que o nível de ficalidade é uma variável que condiciona a definição da política económica. O nível de ficalidade é, aqui, aferido por recuro ao conceito de receita fical, carga fical 1 e eforço fical. A receita fical correponde ao montante arrecadado com impoto e contribuiçõe ociai efetiva e é exprea em unidade monetária. A carga fical correponde ao quociente entre a receita fical e o Produto Interno Bruto (PIB) a preço de mercado e é exprea em percentagem. O eforço fical correponde à relativização da carga fical pelo PIB per capita ou a uma comparação entre a carga fical e uma carga fical virtual reultante de um padrão vigente num conjunto de paíe e/ou num conjunto de período (carga fical padrão) ou, ainda, a uma comparação entre a carga fical e o máximo de receita fical cobrável por um paí em função da ua caracterítica económica, ociai e intitucionai (capacidade fical) (Peino e Fenochietto, 2010: 65). Em qualquer do cao, o eforço fical é expreo por um índice que é determinado de acordo com um do trê modo equematizado na Figura 1. Figura 1: Modo de determinação do eforço fical Impoto (u.m.) PIB per capita (u.m.) Contribuiçõe ociai efetiva (u.m.) Receita fical (u.m.) Carga fical (%) Índice de eforço fical Carga fical padrão (%) PIB (u.m.) Capacidade fical (%) A carga fical padrão e a capacidade fical ão etimada por modelo econométrico que produzem valore dependente quer da etrutura do modelo quer da variávei explicativa utilizada determinando, em conequência, índice de eforço fical com valore diferente. A literatura exibe divergência, por veze acentuada, no reultado obtido pela divera invetigaçõe. A procura de robutez para a análie efetuada com bae nee índice 1 Em Pereira et al. (2012), Santo (2013) e Pereira (2014), ete conceito é deignado por nível de ficalidade. 26

29 Manuel Correia de Pinho Maria Manuel Pinho Eforço Fical em Portugal: Uma Avaliação no Período etimula, por io, o aprofundamento do etudo com diferente quadro metodológico, bae de dado e epecificaçõe. Cabe anotar que o índice de eforço fical não têm um ignificado concreto, ma ão indicadore de ordenação no epaço e de evolução no tempo. No entanto, ete indicadore devem er uado com ponderação em comparaçõe entre paíe ou período de tempo porque não refletem a qualidade do ben de provião pública que o impoto proporcionam. Acrece, ainda, que a medida da grandeza ubjacente pode incorporar alteraçõe concetuai e metodológica no epaço e no tempo. Ete etudo pretende analiar a evolução do eforço fical em Portugal entre 1995 e Em particular, procura-e avaliar e o aumento da carga fical decorrente da aplicação do plano de ajutamento económico e financeiro relativo ao período conduziu a um aprofundamento do eforço fical. Paralelamente, comparar-e-á o eforço fical obervado em Portugal com o eforço fical regitado no paíe da União Europeia. Com ete propóito, o trabalho etá etruturado em trê ecçõe. Na ecção 2, ão reviitada a abordagen e aplicaçõe encontrada na literatura e alientado o contributo relativo ao fatore determinante do eforço fical, a medida de eforço fical e o reultado obtido em divero etudo que abrangem a economia portuguea. Na ecção 3, deenvolve-e um exercício de aplicação empírica viando ituar a evolução do eforço fical em Portugal no quadro do paíe da União Europeia. Na ecção 4, itematizam-e a concluõe. 2. Literatura Sobre O Eforço Fical: Abordagen E Aplicaçõe A literatura obre o eforço fical detaca o fatore que o determinam e identifica o conceito uado para o medir. Neta ecção, ão reviitado o principai contributo que a literatura regita a ete propóito e ão apreentada a poiçõe de Portugal em ordenaçõe do eforço fical extraída da literatura. 2.1 Determinante Do Eforço Fical A literatura afirma que o eforço fical de um paí depende da ua capacidade fical e, por eta via, da ua caracterítica económica, ociai e intitucionai (Peino e Fenochietto, 2010). Aim, a identificação do fatore determinante da capacidade fical é uma quetão central na invetigação obre o eforço fical. O nível de deenvolvimento económico de um paí é amplamente acolhido como determinante do eforço fical e a variável que, habitualmente, o repreenta é o PIB per capita (Lotz e Mor, 1967; Gupta, 2007; Peino e Fenochietto, 2010; Fenochietto e Peino, 2013; Santo, 2013; Pereira, 2014) 2. À medida que o PIB per capita e eleva, o cidadão médio dipõe de uma proporção maior do eu rendimento para pagar 2 Lotz e Mor (1967) recorrem ao PNB per capita para repreentar o nível de deenvolvimento económico, enquanto Gupta (2007), Peino e Fenochietto (2010), Fenochietto e Peino (2013), Santo (2013) e Pereira (2014) uam o PIB per capita expreo em paridade de poder de compra. 27

30 Nota Económica Julho '17 (25-46) impoto e conforma-e mai com a norma ficai. Por outro lado, de acordo com a lei de Wagner, a procura de erviço público é elática em relação ao rendimento, endo de preupor a exitência de uma relação poitiva entre o PIB per capita e a capacidade fical. O grau de abertura da economia ao exterior é, também, identificado pela literatura como determinante da capacidade fical (Lotz e Mor, 1967; Gupta, 2007; Peino e Fenochietto, 2010; Fenochietto e Peino, 2013; Santo, 2013; Pereira, 2014). A literatura alienta a preunção de que exite uma relação poitiva entre o grau de abertura da economia ao exterior e a capacidade fical, o que é explicado, principalmente, pela circuntância de a tranaçõe com o exterior implicarem formalidade que facilitam a identificação da bae tributária e dificultam a evaão fical. Contudo, Gupta (2007) concluiu que o efeito pode er ambíguo na medida em que a liberalização do comércio externo pode er coneguida atravé de redução da tributação. O grau de deigualdade na ditribuição do rendimento é outro fator relevado pela literatura como endo determinante da capacidade fical, no entido em que uma ditribuição do rendimento mai equitativa facilita a arrecadação de receita fical, na medida em que tal ituação favorece a conformidade do cidadão com a norma ficai (Peino e Fenochietto, 2010; Fenochietto e Peino, 2013). De acordo com a literatura, a corrupção tem um impacto negativo ignificativo obre a capacidade fical (Tanzi e Davodi, 2000; Bird e Martinez-Vazquez, 2008; Peino e Fenochietto, 2010; Fenochietto e Peino, 2013). Ete impacto reulta de a corrupção tender a informalizar a economia e, conequentemente, a reduzir a bae de incidência fical. Em particular, no paíe meno deenvolvido, ete efeito pode er potenciado pelo devio de impoto arrecadado em benefício do agente tributário. A literatura ugere que o grau de intervenção do Etado na economia determina poitivamente a capacidade fical. Como alientam Varano et al. (1998), epera-e que a provião pelo Etado de educação, aúde e proteção ocial liberte a ociedade de aplicar o eu recuro na obtenção dete ben no mercado e, por ea via, aumente a ua capacidade para pagar impoto. A inflação é referida pela literatura como um fator que determina negativamente a capacidade fical (Varano et al., 1998; Peino e Fenochietto, 2010; Fenochietto e Peino, 2013). O mecanimo atravé do qual e concretiza o efeito da inflação obre a capacidade fical é conhecido por efeito Tanzi e ignifica que no cao do impoto uualmente epecificado em termo nominai e arrecadado defaadamente, um aumento da inflação tipicamente reduz a receita fical real (Romer, 2006: 542). A compoição etorial do produto é, de acordo com Gupta (2007), um fator determinante da capacidade fical vito que há etore onde é mai difícil controlar a evaão fical (por exemplo, a agricultura de ubitência). No memo entido, inclinam-e Peino e Fenochietto (2010), Fenochietto e Peino (2013) e Pereira (2014). O nível de educação é apontado em Peino e Fenochietto (2010) e em Fenochietto e Peino (2013) como um fator determinante da capacidade fical dado que populaçõe com maior nível de ecolaridade compreendem melhor a neceidade de pagar impoto, aceitam melhor a função reditributiva do Etado e revelam maior conformidade com a norma ficai. 28