Productivity and Infrastructure in the Italian Regions

Picci, Lucio Working Paper Productivity and Infrastructure in the Italian Regions Quaderni - Working Paper DSE, No. 230 Provided in Cooperation with: University of Bologna, Department of Economics Suggested Citation: Picci, Lucio (1995) : Productivity and Infrastructure in the Italian Regions, Quaderni - Working Paper DSE, No. 230, Alma Mater Studiorum - Università di Bologna, Dipartimento di Scienze Economiche (DSE), Bologna, https://doi.org/10.6092/unibo/amsacta/5102 This Version is available at: https://hdl.handle.net/10419/159073 Standard-Nutzungsbedingungen: Die Dokumente auf EconStor dürfen zu eigenen wissenschaftlichen Zwecken und zum Privatgebrauch gespeichert und kopiert werden. Sie dürfen die Dokumente nicht für öffentliche oder kommerzielle Zwecke vervielfältigen, öffentlich ausstellen, öffentlich zugänglich machen, vertreiben oder anderweitig nutzen. Sofern die Verfasser die Dokumente unter Open-Content-Lizenzen (insbesondere CC-Lizenzen) zur Verfügung gestellt haben sollten, gelten abweichend von diesen Nutzungsbedingungen die in der dort genannten Lizenz gewährten Nutzungsrechte. Terms of use: Documents in EconStor may be saved and copied for your personal and scholarly purposes. You are not to copy documents for public or commercial purposes, to exhibit the documents publicly, to make them publicly available on the internet, or to distribute or otherwise use the documents in public. If the documents have been made available under an Open Content Licence (especially Creative Commons Licences), you may exercise further usage rights as specified in the indicated licence. https://creativecommons.org/licenses/by-nc/3.0/ Productivity and Infrastructure in the Italian Regions. Lucio Picci1 University of Bologna August 1995 J.E.L.: H54, C80 Abstract. Weaddresstheissueofwhetherpublicinfrastructureplayanimportantroleindetermining factor productivity in Italy, and we show that the evidence is mixed. Publiccapitalissignificantinexplainingoutputinmostcases.However,whentheattention is drawn on the long-run properties of the data, or when care is taken to rule out contemporaneous short-run effects, then public capital results to be either non-significant, orsignificantbutofnegligibleimportance.Weconcludethattheinfluenceofinfrastructure on output is probably due, to a great extent, to short-run demand-side phenomena. 1 Dipartimento di Scienze Economiche, Strada Maggiore 45, I-40125 Bologna, Italy. Tel. +39-51-6402608, Fax. +39-51-6402664. E-mail: [email protected] IwouldliketothankforthevaluablecommentsDouglasHoltz-EakinandSergioPastorello. 1 Introduction. The huge public debt affecting the Italian economy is an obvious reason of concern for economistsandpolicy makersalike. Jappelliand RipadiMeana(1990)arguethatpolicies whosegoalistoreducethedebt/outputratiocannotabstractfromtherolethatinfrastructure playin theeconomy. If infrastructuredo influence outputsignificantly, thencutting public capital investment, by reducing future potential output, could even worsen the problem. On the other hand, if that is not the case, then policy makers, in a sense, could do without caring too much on what kind of budget cuts they make, as long as they make them. The interest in the relationship between infrastructure and output or, more to the point, factorproductivity,however,isnotlimitedtothisissue.IntheUnitedStates,muchattention has been dedicated to the analysis of the causes of the productivity slowdown of the last two decades.A whole thread of the literature, starting from the seminal paper of Aschauer (1989), has tried to impute the decline of factor productivity to the decline in public investment2. The roleplayed byinfrastructure inthe economy has beenanalyzed by acountless number of works. At a simple and intuitive level, we note that infrustructure affect factor productivity by "assisting" public capital: a given truck is much more productive on a freeway, than on a country road. Moreover, new infrastructure embodies technological progress: to build a brand-new mass-transit system is tantamount to the introduction of a new technology into a local economy. Infrastructure, also, allows for technological progress: weren’t roads available at all, we wouldn’t have trucks. At the same time, all theseeffectsprobablytaketimeto manifestthemselves, sincetheyinvolvecreationofnew business, relocation of existing activity, and learning3. The point of interest, obviously, is notwhether public capital is significant in determining productivity,athesisthatveryfew people,ifanybody,wouldquestion.Whatisinteresting tounderstandistowhatextentinfrastructureaffectproductivitybeyonditsdirectprovision of amenities, on one side, and what are the mechanisms that make them influence production.Both questions have important policy implications. Ananswer tothe first one 2 See Eberts (1990), Munnell (1990), Holtz-Eakin (1994), and Holtz-Eakin and Schwartz (1994). 3 For a revue essay, also illuminating other points touched by this work, see Gramlich (1994). 1 would helppolicy makersin choosingthe optimalamount of publicinvestment; ananswer to the second one could provide useful guidelines on how to make public investment succesful. Using Italian regional data, this work tries to address the first of these two questions. We try to determine the importance of infrastructure in determining production in the Italian regions4over the period 1970-1991. Only very recently the availability of the necessary data has made this type of study possible for the Italian economy. We conclude that infrastructure are significant in explaining regional output, but that this is probably due to the presence of short-run demand-side, as opposed to supply-side, effects. Section 2, with reference to a few representive works, examines the existing empirical evidence on the issue, both for the U.S. and for Italy. Section 3 deals with the analysis of the empirical results. Section 4 concludes. 2 The Existing Evidence. Aschauer (1989) shows the results of various regressions of different concepts of factor productivity onfactor inputs, including public capital, for the U.S. as a whole. The results are quite astounding: public capital seems to be highly significant in influencing productivity,withpositiveelasticitiesofaround0.35.SimilarresultsarefoundbyMunnell (1990a) who, again for the U.S. as a whole, finds elasticities of output with respect to public capital of comparable magnitudes. Munnell(1990b)analysestheimpactofpubliccapitalonproductivityusinginsteadapanel of U.S. state data. The results of her pooled OLS regressions indicate a coefficient for public capital lower (0.15) than in thetime series studies mentioned above, but still highly significant. 4 Italyisdividedinto20regions, sometimes,fordescriptivepurposesonly,lumped into five macro-regions. The regions (and the macro-regions) are: Piedmont, Val d’Aosta, Liguria and Lumbardy (North-West); Trentino-Alto Adige, Veneto, Friuli Venezia Giulia and Emilia Romagna (North-East); Tuscany, Marche, Umbria and Lazio (Center); Abruzzo, Molise, Apulia, Campania,BasilicataandCalabria(South);SicilyandSardinia(Islands).SouthandIslands together are called the Mezzogiorno of Italy. 2 The firstwave of works onthe topic left the impression thatpublic capital was indeed very importantindeterminingproductivity,andthatitsdecreasedgrowthrateswereaprominent candidate to explain the productivity slowdown that the U.S. economy has experienced in the last decades. After a first phase of enthusiasm, however, some caution was asked for by other less-optimistic results. Holtz-Eakin (1994) criticizes the econometric analysis carried out byMunnell(1990b),andarguesthat,inacross-stateanalysisofproductivity,statespecific effects are potentially important. Holtz-Eakin then rejects the hypothesis that individual state effects are not relevant and finds evidence against the hypothesis that public capital plays a role in determining output. The same result also emerges when different econometric techniques, such as instrumental variables panel estimation and estimation using long differences of the data, are used. Holtz-Eakin and Schwartz (1994) take a different approach, and develop a neoclassical growthmodel"à laSolow"that incorporatesinfrastructureas oneof theproductioninputs. They check whether the data conform to the predictions of their model, and again find essentially no role for public capital. Evidence for Italy is scant, due at least in part to lack of the necessary data. Bracalente and Di Palma (1982) compute a series of indexes to measure infrastructure for the Italian regions for the year 1977. Usingboth OLS analysis and rotated-factorregression analysis, they find that infrastructure are significant in explaining regional development. The determination of the direction of causality, however, remains an open question in that work. Jappelliand Ripa di Meana (1990) estimate aseries of reduced-form equations for output, where both private consumption and public investment are used as regressors. The estimated coefficients of public capital are significant and bigger than the coefficients of privateconsumption; the authorsinterpret thisresultas evidence infavor ofthe hypothesis that the effect of public capital is important in determining output. Picci (1994) uses a data set developed by Rossi et al. (1993) and by Picci (1995a) on aggregate public capital for the post WWII period and for the between-war period, to estimate a number of regressions similar to the ones considered by Aschauer (1989) and Munnell (1990). The results generally indicate a significant role for public infrastructure, with very high output elasticities. This type of analysis permits, to some extent, to address 3 the question of the directions of the casual relation between infrastructure and output. Granger-causation analysis between multifactor productivity (or "Solow residuals") and thegrowthrateofpubliccapitalgiveshoweverambiguousresults,dependingonthesample period considered. 3 The Analysis of Regional Data. In what follows, we analyze the incidence of infrastructure on output using a recently developed dataset on regional infrastructure covering the period from 1970 to 1991, for a total of 22 annual observations. Regional public capital stocks have been computed using the perpetual inventory technique. The necessary regional public investment time series have been obtained by apportioning the national aggregate in Rossi et al. (1993) using the yearlydataon"publicworks"collectedbyIstat(Istat,variousyears). Thewholeprocedure is fairly involved; full details are in Picci (1995a). Private capital also has been taken from Picci (1995a), where it has been computed using a benchmark for the regional capital stock for the census year 1981 and regional gross private investments for the remaining years. Output is regional gross product, and labor is regional units of labor (source: SVIMEZ (1993); ISTAT (1990a, 1990b, 1992). Table1 showsthe averagegrowthrate forper capita output,laborunits, private and public capital, together with their beginningand end-of-sample levels, for Italy as a whole and for its five macro-regions. The Mezzogiorno (comprising the Southern and the Insular regions) has a per-capita income much lower than the national average. Convergence has not occurred over the two decades considered, and the two low income macro-regions have both scored below-average per-capita income growth rates. Labor units increased by little and rather evenly across Italy5. Private capital has increased dramatically, mostly in the Mezzogiorno. In comparison, publiccapital has increased byless, whithgrowth rates of around3.5% yearly acrossItaly. A more detailed break-up of the data would show that most of the increase of the capital 5 However,whileintheNorthofItaly thisgrowthisreflectedbyasensibleincrease in the ratio of labor units over population, with a more or less constant population, in the Mezzogiorno population has also grown. 4 stock - both private and public - in the Mezzogiorno occurred during the first part of the sample period, while the huge effort to industrialize the less-developed Southern and Insular regions was still ongoing. During the ’80’s, with the dismission of the "Cassa del Mezzogiorno"(a state agencyaimed at the development of the Mezzogiorno), the massive investments in that part of Italy that had characterized the previous decades came to an end. Notethat, according tothe data, and contraryto intuition, the less-developedMezzogiorno has a stock of infrastructure bigger than the two industrialized Northern macro-regions. Picci (1995b) compares measures of regional infrastructure endowment computed using perpetual inventory data - let’s call them, "PI" (perpetual inventory) indexes -, with the indexes in Biehl, Bracalente, Di Palma and Mazziotta (1990), based on the physical consistency of the capital stock - let’s call them "BBDPM" indexes -. Both indexes are obtained by deflating regional capital stocks either by regional population - in the case of "population-serving" infrastructure, such as schools, hospitals, etc. - or by regional area - in the case of "space-serving" infrasructure, such as roads and railroads. A comparison of the indexes gives some indication on the relative efficiency of the different regions in producing infrastructure: a PI index higher than a BBDPM index, would show relative inefficiency in building infrastructure, indicating the presence of little infrastructure in relation to the amount of resources spent to produce it, and vice-versa. Picci (1995b), while warning about the risk of drawing hasty conclusions, shows that, in this respect, there is a very sharp difference between the Mezzogiorno and therest ofItaly: all the Mezzogiorno regions have PI indexes generally much higher than the BBDPM indexes.Thisisparticularly trueforCampania, Calabria andSicilia, thethree big Southern regions plagued by organized crime. On the other hand, only Liguria, among the Northern regions, shows the same carachteristic. This should come as no surprise: Liguria, a mountainous and densely populated region, is certainly characterized by higher costs in buildinginfrastructure.ForalltheotherNorthernandCentralregions,PIindexesarelower then BBDPM indexes. In other words, there is a case for overestimation of the public capital stock computed using the permanent inventory technique for Southern Italy in general, even more so for Campania, Calabria, Sicily, and also for the Northern region of Liguria.Figure1showsbothindexesforallregions.The45degreeslinedivides"efficient" regions (below) from "inefficient" ones (above). 5 The empirical analysis is carried out by estimating a production function where public capital is used in conjunction with private capital and labor input to assess its importance in determining output. The estimated equation is: where is output, is private capital, is labor, is public capital, and denotes logs. The and subscript indicate, respectively, region and time. The error term has the following structure: is aregion specific component; is a time specific component, and is aidiosincratic i.i.d error. Differentspecificationof themodels -and, aswe haveseen, often differentresultsfollow from different assumptions about the error term. Not considering gives pooled OLS, the technique used by Munnell (1990). Holtz-Eakin (1994), in his analysis on U.S. state productivity, argues that state effects are potentially important. Once he considers them, he overturns Munnell’s results. Itis importantto stressthat detecting asignificant publiccapitalcoefficient does notimply thatpubliccapitalcausesoutput.First,therelationcouldbeexplainedbyreverse-causation, with policy makers responding with increased public investment to better economic conditions. Moreover, even if infrustructure do influence output, a distinction between "supply" and "demand" effects should be drawn. It could be that the significant infrastructure coefficient in the estimated equation is due to its effect on the underlying determinants of productivity. On the other hand, such a coefficient could result from the effects of the increased public expenditure on demand. This type of demand effect could beparticularly important in Southern Italy, where the sizeof public investment, especially during the first part of the ’70’s, was a considerable fraction of output (see figure 2). Is there a way to discriminate between these two type of effects? It seems reasonable to lyrt = α0 + α1lkrt + α2llabrt + α3lkpubrt + εrt ,(1) y k lab kpub l rt εit = fr + δt + ηrt . frδtηrt f 6 expect that short-lived demand effects should be detected by analysis that focus on the short-run time series properties of the data, such as fixed-effects OLS regressions. On the other hand, long-run supply effects should be detected by focusing on the long run cross-sectional dimension of the data, for example by estimating long-differences of the data. The first and second columns of table 2 show, respectively, results for pooled regressions andfixed-effects estimates of regional production functions for the Italian regions6. In the pooledOLSregressionresults,theestimatedcoefficientforpubliccapitalisembarassingly negative and significant. By estimating OLS regressions separetely for the 20 regions (results not reported here but available from the author), we obtain generally positive and significant estimates of the public capital coefficients, as in Picci (1994), for the Italian economy, and in Aschauer (1989), for theUS economy. The(unweighted) average ofthese 20 estimated coefficients, that tend to be bigger for the Southern regions, is equal to 0.504. Thefixedeffectsestimates,unlikethepooledOLSestimates,showsignificantandpositive public capital elasticity. The F-test on the null hypothesis that all fixed effects are equal (that is, that there are no fixed effects) is strongly rejected. As in Holtz-Eakin (1994) we are ledto concludethat OLSpooled regressions arenotconsistent. Unlike inHoltz-Eakin, assuming fixed-effectsdoes not change the results obtained with theaggregate time series approach. The estimated public capital ouptut elasticities is very high, and comparable to the results obtained by running separate OLS regressions. Note also that the estimated labor output elasticity is implausibly bigger than one. Random effects estimates, shown in column c of table 2, provide similar results. The Hausmantest,however,rejectsthenullhypothesisthattheregionaleffectsareuncorrelated with the right-hand side variables of the regression. In this case, random effects estimates are biased, while fixed effects estimates are still consistent. Fixed effects estimation accounts only for variation in the time series dimension of the data,andleavesnoroom forcross-sectionalvariation.Attheopposite endof thespectrum, OLS regressions on long-differences of the data consider only cross sectional variation. 6 All panel data regressions include time effects. 7 Table 1 Summary statistics. Output per capita N. West N. East Center South Islands Italy Avg. growth rate: 3.36% 4.04% 3.57% 3.49% 3.16% 3.52% 1970 12.20 10.63 10.36 6.75 7.06 9.66 1991 20.81 19.66 18.12 11.71 11.74 16.80 Per capita output: million of ’85 lire. Labor Units N. West N. East Center South Islands Italy Avg. growth rate .38% 1.15% 1.12% .89% .89% .83% 1970 6243.0 3942.0 3829.3 4048.0 1887.0 19949.4 1991 6786.5 4891.5 4726.7 4806.5 2238.6 23449.8 Private Capital N. West N. East Center South Islands Italy Avg. growth rate 6.35% 5.38% 8.19% 11.78% 9.90% 7.34% 1970 164674.5 125344.0 76512.4 56237.5 35740.2 458508.7 1991 384359.3 267076.8 208139.1 195326.8 110079.0 1164981 Public Capital N. West N. East Center South Islands Italy Avg. growth rate 3.87% 3.86% 3.06% 3.44% 3.69% 3.56% 1970 72718.0 67924.9 72717.1 103456.1 52944.7 369760.9 1991 131819.2 122936.9 119383.7 178190.3 93955.5 646285.6 Private and public capital: billion of 1985 lire. 14 Table 2 Dependent variable: Gross Regional Product. All variables are in logs. Italy. Variable: a) OLS b) F.E. c) R.E. d) L.D. Constant 2.445 - .200 .453 (21.63) (1.28) (3.98) Private Capital .248 .097 .145 .011 (20.76) (7.81) (12.59) (.293) Labor .863 1.080 .737 .807 (49.41) (18.79) (39.13) (3.56) Public Capital -.063 .430 .355 -.012 (-4.15) (15.51) (14.58) (-.08) .993 .945 .948 .459 Test: : no FE. P-Value: .000 Hausman Test: FE vs. RE. P-Value: .000 OLS: Pooled OLS; F.E.: Fixed Effects; R.E.: Random Effects; L.D.: Long Differences. t-statistics are between parentheses. R2 H0 15 Table 3 Dependent variable: Gross Regional Product. All variables are in logs. Italy excl. Cal, Camp, Sic, Lig. Variable: a) OLS b) F.E. c) R.E. d) L.D. Constant 2.184 - .134 .512 (17.21) (.81) (2.95) Private Capital .224 .088 .126 .012 (18.46) (6.47) (9.98) (1.27) Labor .849 1.22 .758 .789 (45.21) (18.58) (32.16) (2.39) Public Capital -.005 .413 .374 -.090 (-.26) (13.68) (14.04) (-.48) .994 .945 .951 .438 Test: : no FE. P-Value: .000 Hausman Test: FE vs. RE. P-Value: .000 OLS : Pooled OLS; F.E.: Fixed Effects; R.E.: Random Effects; L.D.: Long Differences. t-statistics are between parentheses. R2 H0 16 Table 4 Dependent variable: Gross Regional Product. All variables are in logs. North and Center excl. Lig. Variable: a) OLS b) F.E. c) R.E. Constant 1.151 - .197 (7.78) (1.52) Private Capital .200 .390 .418 (11.80) (16.67) (20.91) Labor .700 .846 .588 (31.53) (13.48) (26.56) Public Capital .236 .171 .178 (9.37) (6.61) (7.27) .997 .977 .991 Test: : no FE. P-Value: .000 Hausman Test: FE vs. RE. P-Value: .019 OLS : Pooled OLS; F.E.: Fixed Effects; R.E.: Random Effects; t-statistics are between parentheses. R2 H0 17 Table 5 Dependent variable: Gross Regional Product. All variables are in logs. Public capital is lagged one period. Italy Italy excl. Cal, North and Camp, Sic, Lig. Center excl. Lig. Variable: a) F.E. b) R.E. c) F.E. d) R.E. e) F.E. f) R.E. Constant - 1.623 - .117 - -.018 (13.42) (10.41) (-.143) Private Capital .246 .279 .102 .154 .418 .447 (17.62) (21.42) (6.25) (10.41) (16.32) (20.70) Labor 1.342 .832 1.196 .738 .810 .553 (20.38) (38.83) (18.40) (31.69) (12.54) (24.29) Lagged Public Capital .013 .013 .410 .359 .179 .192 (3.26) (3.12) (12.77) (13.02) (6.66) (7.48) .919 .944 .945 .953 .977 .981 Test: : no FE. P-Value: .000 .000 .000 Hausman Test: FE vs. RE. .000 .000 .038 P-Value: F.E.: Fixed Effects; R.E.: Random Effects; t-statistics are between parentheses. R2 H0 18 Table 6 Dependent variable: Gross Regional Product. All variables are first differences of logs. Italy Italy excl. Cal., South and Center Camp., Sic, Lig. exc. Lig. Variable: a) b) c) d) e) f) g) h) i) OLS F.E. R.E. OLS F.E. R.E. OLS F.E. R.E. .014 .007 - .008 Constant .014 .158 .014 .015 - (4.17) (1.45) (1.60) (4.34) (2.00) (4.16) (4.26) Private Capital .011 .013 .013 .017 .021 .017 .060 -.179 .011 (.55) (.486) (.558) (.80) (.77) (.81) (.74) (-1.46 (.12) Labor .542 .524 .534 .529 .506 .526 .546 .487 .527 (9.79) (9.16) (9.97) (8.47) (7.90) (8.42) (6.47) (5.52) (6.20) Public Capital .120 .158 .136 .083 .129 .089 .073 .211 .108 (1.78) (2.00) (1.98) (1.20) (1.64) (1.27) (1.07) (2.56) (1.51) .622 .623 .626 .659 .670 .661 .784 .797 .459 Test: : no FE. P-Value: .568 .279 .072 H. Test: FE vs. RE. P-V: .999 .999 .999 OLS : Pooled OLS; F.E.: Fixed Effects; R.E.: Random Effects; H. Test: Hausman Test. t-statistics are between parentheses. R2 H0 19 Table 7 Dependent variable: Gross Regional Product. All variables are in logs. Constant returns to scale in the private inputs . Italy Italy excl. Cal, North and Camp, Sic, Lig. Center excl. Lig. Variable: a) F.E. b) R.E. c) F.E. d) R.E. e) F.E. f) R.E. Constant - -.068 - .102 - .198 (-.404) (.584) (1.48) Private Capital ------ Labor .908 .844 .922 .853 .594 .585 (72.69) (76.92) (66.45) (70.84) (24.71) (33.21) Public Capital .472 .300 .484 .291 .203 .182 (19.37) (16.31) (17.86) (15.15) (8.09) (11.73) .960 .947 .958 .944 .887 .893 Test: : no FE. P-Value: .000 .000 .000 Hausman Test: FE vs. RE. .000 .000 .905 P-Value: F.E.: Fixed Effects; R.E.: Random Effects; t-statistics are between parentheses. α1+α 2=1 R2 H0 20 Table 8 Dependent variable: Gross Regional Product. All variables are in logs. Constant returns to scale in the private and public inputs . Italy Italy excl. Cal, North and Camp, Sic, Lig. Center excl. Lig. Variable: a) F.E. b) R.E. c) F.E. d) R.E. e) F.E. f) R.E. Constant - 2.07 - 2.190 - 1.425 (27.36) (25.76) (18.96) Private Capital ------ Labor .439 .593 .457 .623 .384 .420 (17.22) (27.52) (15.55) (25.57) (18.21) (20.02) Public Capital .436 .216 .433 .197 .138 .117 (13.47) (8.45) (11.73) (6.95) (4.86) (4.26) .947 .930 .945 .953 .867 .849 Test: : no FE. P-Value: .000 .000 .000 Hausman Test: FE vs. RE. .000 .000 .002 P-Value: F.E.: Fixed Effects; R.E.: Random Effects; t-statistics are between parentheses. α1+α 2+α 3=1 R2 H0 21 Figure 1 High Investment / Low Capital Stock High Investment / High Capital Stock Low Investment / Low Capital Stock Low Investment / High Capital Stock Public Investment and Infrastructure. The regions are: North-West: Piedmont (PIE), Val d’Aosta (VAA), Liguria (LIG) and Lumbardy (LOM); Noth-East:Trentino-Alto Adige(TAA), Veneto (VEN), Friuli-Venezia Giulia(FVG) and Emilia-Romagna (ER); Center: Tuscany (TOS), Marche (MAR), Umbria (UMB) and Lazio (LAZ); South: Abruzzo (ABR), Molise (MOL), Apulia (PUG), Campania (CAM), Basilicata (BAS) and Calabria (CAL); Islands: Sicily (SIC) and Sardinia (SAR). 22 Figure 2 NW: North-West; NE: North-East; C: Center; S: South; I: Islands. 23