Abstract
The estimation of the effect of local human capital on wages only might not identify properly human capital spillovers. Appropriate identification requires considering the joint effect of local human capital on both wages and rents. Empirically, we study the effects of local human capital on household-level rents and individual-level wages for a sample of Italian local labour markets. Our results show a positive and robust effect of local human capital on rents, supporting the idea that human capital generates positive externalities at the local level. Our results also suggest that consumption and production externalities have a similar impact on wages.
Similar content being viewed by others
Notes
In this paper “schooling,” “human capital” and “education” are used interchangeably.
In what follows, we will often refer to externalities that affect utility and the quality of life as “consumption externalities”.
The main issue of our paper is closely related to Shapiro (2005). Shapiro develops a dynamic version of Roback’s model to analyze the effects of human capital on city growth. His evidence suggests that, while most of the literature has emphasized the impact of human capital on productive externalities, consumption externalities can be important as well. Shapiro’s strategy is further developed by Glaeser and Saiz (2004), who find that human capital increases consumption amenities within metropolitan areas.
The production function implies that workers of different education are perfectly substitutable. In this case, the “composition” problem emphasised by Ciccone and Peri (2002) does not arise.
LLMs are thus characterized by “self-contained” labour markets, in the sense that both the share of resident LLM employees in total LLM employees and the share of resident LLM employees in total LLM residents must be at least 75%. Labour mobility within local labour markets is by construction very high, whereas mobility from and to other local labour markets is low.
Our coefficient estimates, however, are not sensitive to weighting or not weighting the regressions.
The interviewees can be either owners or tenants. In the first case, the SHIW collects the rent the owner charges (or, if the dwelling is not rented or is the family residence, the best estimate for the rent that could be charged). In the second case, the tenant reports the actual rent. The fraction of hypothetical rents in our sample is 77% (see Table 1). On the one hand, hypothetical rents might reflect the interviewees’ inaccurate knowledge of the housing market. On the other hand, actual rents reflect contractual arrangements that have often been agreed years before. This explains why imputed rents are generally higher than actual rents (controlling for house characteristics).
We also analyzed whether the effect of local human capital is different in differently congested areas. By splitting the sample at the median of the log LLM population, we find that the coefficient for local human capital is equal to 11% for the relatively more congested areas (8% for the relatively less congested LLMs).
As suggested by Ciccone (2002), the introduction of increasing detailed spatial fixed effects allows us to control for spatially correlated omitted variables. For our sample of 238 LLMs, 1 LLM extend across region boundaries and 11 LLMs extend over province boundaries. The exclusion of interregion and interprovince LLMs does not alter the results.
We also tried house prices instead of rents as dependent variable. Results were very similar.
We also estimate a model in which private returns to education are non-linear in the years of schooling. For this purpose, we replace individual human capital with dummies for the highest educational attainment obtained by the individual. This has negligible effects on the estimates of local human capital returns. Padula and Pistaferri (2001) criticize the standard Mincerian approach by proposing a measure of implicit returns to education, which accounts for unemployment risk.
The results reported in the text have also been checked by using seemingly unrelated regression (SUR) techniques. In the theoretical model of Section 2, wages and rents are simultaneously determined. This implies that there might be correlation between unobservable shocks to wages and rents. In this case, SUR estimates are more efficient, whereas OLS are still consistent and unbiased. SUR estimates (not reported) confirm previous findings. For instance, in the benchmark cases of (2.1) and (3.1), the estimated coefficient for local human capital rises to 9.9% in the rent equation and 3.6% in the wage equation.
The inclusion of these additional controls reduces the rent and wage samples to 27,904 and 23,252 observations, respectively. As we checked, these reductions are not relevant for the results obtained before.
For the sake of brevity, point estimates for the control variables are not reported in the tables.
The Cannari–Signorini dataset is derived from a variety of sources (Census; Company Account Data Service; ISTAT’s Surveys on Export, Value added, Labour Force, Capital Stock): see Cannari and Signorini (2000) for details.
We report here only a subset of robustness checks that have been performed. Following de Blasio and Nuzzo (2003), we also controlled for the local endowments of social capital. In the tradition of Glaeser et al. (1992), we controlled for local competition, as measured by the ratio between average firm size in the LLM and the average size at national level. Moreover, we controlled for indexes of the LLM sector composition of economic activity. Results were only marginally different from those of the baseline case.
We thank a referee for this suggestion.
The ISTAT DEMOS dataset provides an array of demographic and socio-economic variables for areas of Italy. Since information from this source is not available at the LLM level, the ISTAT DEMOS indicators that we use in the paper refer to the province level. The ISTAT DEMOS dataset does not provide figures for the provinces created after 1995 (Biella, Verbania-Cusio-Ossola, Lodi, Lecco, Rimini, Prato, Crotone and Vibo Valentia). For this reason, the rent and wage samples reduce to 27,413 and 22,977 observations, respectively.
According to Downes and Zabel (2002), local school characteristics explain a good deal of the variation in US house prices: houses in better school districts are more expensive. Differently from the USA, where the education system is mostly financed at the local level, the Italian schooling system is very centralized and egalitarian, with low variability in the quality of education across areas. This, however, does not apply to nurseries for infants of 0–3 of age, which are funded by local authorities or are private. We use the number of public nurseries for children aged 0–5 over local population of the corresponding age group from the ISTAT DEMOS source. This index never enters significantly in our equations. The effect of average human capital remains unchanged for wages and goes up for rents.
Again, only a subset of robustness checks accomplished has been reported. We also controlled for natural amenities, such as climate variables (average temperature and average days of rain). Moreover, we controlled for cancer and cardiovascular mortality rates, which proxy for environmental quality. Compared with the baseline of line 9, results were only marginally affected.
A potential source of endogeneity is represented by “selective migration” of talented workers across local markets. In particular, it might happen that workers with high (unobserved) ability tend to move to areas that are characterized by high average levels of schooling. In this case, the correlation between wages and local human capital may partially reflect unobserved ability, rather than true schooling externalities. We find, however, that selective migration is not an issue for our results. We exploit the confidential SHIW data on the birthplace of workers. This information is at the level of the 103 Italian provinces that cover the country. While this is certainly not ideal, we should still be able to detect selective migration through the different outcomes for those who work where they were born (the “stayers”) and the others (the “movers”). By interacting our explanatory variables with a dummy variable equal to 1 for the movers, we find that the interaction between workers’ characteristics and local human capital, on the one hand, and the dummy for movers, on the other hand, is never significant.
Measurement error problems might be present as well (see Krueger and Lindahl 2001).
Some implications of this identification assumption can be tested. Finding that the 1981 demographic structure predicts housing and labour market outcomes other than local education would cast some doubts on the exogeneity of the instruments. In this vein, we checked whether past demographics is correlated with labour force participation (conditioning out the control variables that appear in the wage equation) and found no support for such a claim.
The p value of the F statistic is always zero at the first four decimals.
Still, it is very difficult to take these estimates as conclusive. As emphasised in Angrist and Krueger (2001), a problem with interpreting IV estimates is that, in general, instruments do not affect observations in the same way. In case of heterogeneous responses, this technique provides an estimate that is mostly related to the specific group of people whose behaviour is sensitive to the instrument itself (see also Imbens and Angrist 1994). In our context, the demographic instrument is particularly relevant to those who are likely to quit school early, with little or no effect on those who are likely to go on to college. To provide additional robustness, we also used the lagged youth unemployment rate as instrument: the results are reported in Dalmazzo and de Blasio (2005). As suggested by Arkes (2003), this is a reasonable instrument for those who decide whether to go beyond compulsory schooling. To lessen the concern that the lagged youth unemployment rate can be correlated with current unemployment, and hence with wages and rents, we also controlled for the current unemployment rate in the wage and rent equations (see Cameron and Taber 2004). We found that the estimated coefficient for local average schooling remained significant only in the rent equation.
We thank a referee for this example.
References
Acemoglu D, Angrist JD (2000) How large are human-capital externalities? evidence from compulsory schooling laws. NBER Macroecon Annu: 9–59
Adamson D, Clark D, Partridge M (2004) Do urban agglomeration effects and household amenities have a skill bias? J Reg Sci 44:201–223
Angrist JD, Krueger AB (2001) Instrumental variables and the search for identification: from supply and demand to natural experiments. J Econ Perspect 15:69–85
Arkes J (2003) Using state unemployment rates during teenage years as instruments to estimate the returns to schooling, mimeo. RAND, Santa Monica, CA
Berger MC, Blomquist GC, Peter KS (2003) Compensating differentials in emerging labor and housing markets: estimates of quality of life in Russian cities, IZA Discussion Paper Series No. 900
Bils M, Klenow P (2000) Does schooling cause growth? Am Econ Rev 90:1160–1183
Brandolini A, Cannari L (1994) Methodological appendix: the Bank of Italy survey of household income and wealth. In: Ando A, Guiso L, Visco I (eds) Saving and the accumulation of wealth. Essays on Italian household and government saving behavior. Cambridge University Press, Cambridge, pp 369–386
Cameron S, Taber C (2004) Estimation of educational borrowing constraints using returns to schooling. J Polit Econ 112:132–182
Cannari L, D’Alessio G (1995) Il rendimento dell’istruzione: alcuni problemi di stima. Bank of Italy, Temi di Discussione No. 253
Cannari L, Signorini (2000) Nuovi strumenti per la classificazione dei sistemi locali. In: LF Signorini (ed) Lo sviluppo locale. Meridiana Libri, Rome, pp 123–151
Charlot S, Duranton G (2004) Communication externalities in cities. J Urban Econ 56:581–613
Ciccone A (2002) Agglomeration effects in Europe. Eur Econ Rev 46:213–227
Ciccone A, Hall R (1996) Productivity and density of economic activity. Am Econ Rev 86:54–70
Ciccone A, Peri G (2002) Identifying human capital externalities: theory with an application to US cities. IZA Discussion Paper No. 488. Rev Econ Stud (in press)
Colussi A (1997) Il tasso di rendimento dell’istruzione in Italia: un’analisi cross-section. Polit Econ 13:273–294
Combes P (2000) Economic structure and local growth: France, 1984–1993. J Urban Econ 47:329–355
Dalmazzo A, de Blasio G (2005) Where do human capital externalities end up? Temi di Discussione No. 554, Bank of Italy, Rome
de Blasio G, Nuzzo G (2003) Putnam’s social capital and the Italian regions: an empirical investigation. Mimeo, Bank of Italy, Rome
Downes T, Zabel J (2002) The impact of school characteristics on house prices: Chicago 1987–1991. J Urban Econ 52:1–25
Duranton G (2004) Human capital externalities in cities: identification and policy issues. London School of Economics, London. In: Arnott R, McMillen D (eds) The Blackwell Companion to Urban Economics, Blackwell, Oxford (in press)
Duranton G, Monastiriotis V (2002) Mind the gaps: the evolution of regional earnings inequalities in the UK 1982–1997. J Reg Sci 42:219–256
Glaeser E, Maré D (2001) Cities and skills. J Labor Econ 19:316–342
Glaeser E, Saiz A (2004) The rise of the skilled city. Brookings-Wharton Pap Urban Aff 5:47–94
Glaeser E, Kallal HD, Scheinkman JA, Shleifer A (1992) Growth in cities. J Polit Econ 100:1125–1152
Glaeser E, Kolko J, Saiz A (2001) Consumer city. J Econ Geogr 1:27–50
Goldin C, Katz LF (1998) The origins of technology–skill complementarity. Q J Econ 113:693–732
Gyourko J, Kahn M, Tracy J (1999) Quality of life and environmental comparisons. In: Cheshire P, Mills ES (eds) Handbook of regional and urban economics, vol 3. Applied Urban Economics. Elsevier North-Holland, Amsterdam, pp 1413–1454
Haveman R, Wolfe B (1984) Schooling and economic well-being: the role of nonmarket effects. J Hum Resour 19:377–407
Henderson JV (2003) Marshall’s scale economies. J Urban Econ 53:1–28
Imbens G, Angrist J (1994) Identification and estimation of local average treatment effects. Econometrica 62:467–475
ISTAT (1997) I sistemi locali del lavoro 1991. ISTAT, Rome
Korenman S, Neumark D (1991) Does marriage really make men more productive? J Hum Resour 26:282–307
Krueger A, Lindahl M (2001) Education for growth: why and for whom? J Econ Lit 39:1101–1136
Lochner L (2004) Education, work, and crime: a human capital approach. Int Econ Rev 45:811–843
Lochner L, Moretti E (2004) The effect of education on crime: evidence from prison inmates, arrests, and self-reports. Am Econ Rev 94:155–189
Loh ES (1996) Productivity differences and the marriage wage premium for white males. J Hum Resour 31:566–589
Lucas RE (1988) On the mechanics of economic development. J Monet Econ 22:3–42
Mankiw G, Romer D, Weil D (1992) A contribution to the empirics of economic growth. Q J Econ 107:407–437
Moretti E (2003) Human capital externalities in cities. In: Handbook of Regional and Urban Economics, Elsevier North-Holland, Amsterdam (in press)
Moretti E (2004) Estimating the social return to higher education: evidence from longitudinal and repeated cross-sectional data. J Econom 121:175–212
OECD (2002) Redefining territories. The functional regions. OECD, Paris
Ottaviano GIP, Peri G (2004) The value of cultural diversity: evidence from US cities. CEPR Discussion Paper No. 4233. J Econ Geogr (in press)
Padula M, Pistaferri L (2001) Education, employment and wage risk. CSEF W.P. No. 67, Università di Salerno, Italy
Rauch JE (1993) Productivity gains from geographic concentration of human capital: evidence from the cities. J Urban Econ 34:380–400
Roback J (1982) Wages, rents and the quality of life. J Polit Econ 90:1257–1278
Shapiro J (2005) Smart cities: quality of life, productivity, and the growth effects of human capital. Rev Econ Stat (in press)
Weisbrod BA (1962) Education and investment in human capital. J Polit Econ 70:106–123
Acknowledgements
We thank Gilles Duranton for stimulating discussions and Luigi Cannari and Giorgio Gobbi, participants in the Third Labour Economics Workshop “Brucchi Luchino” Florence 2004, SOLE/EALE San Francisco 2005, ERSA Amsterdam 2005 and, in particular, two anonymous referees for comments and suggestions. We are grateful to Diego Caprara for editorial assistance. Some of the work for this paper was initiated while de Blasio was visiting the Centre of Economic Performance of the London School of Economics. He thanks the centre for their hospitality. The views expressed in the paper are those of the authors and do not necessarily correspond to those of the respective organizations.
Author information
Authors and Affiliations
Corresponding author
Additional information
Responsible editor: Christian Dustmann
Appendix
Appendix
1.1 Derivation of the model solution
The maximum problem for an individual i who lives and works in area c can be written as:
where p c =1. The first-order conditions are given by:
Rearranging Eqs. A2 and A3 and dividing side by side, one obtains:
From Eq. A4, one can find an expression for L ic to be substituted into the worker’s budget constraint. This yields the optimal value of Y ic as a function of worker’s income:
Again, one can exploit the budget constraint together with Eq. A5 to get the optimal value of L ic :
By substituting Eqs. A5 and A6 into the utility function (1), one obtains worker i’s indirect utility function:
Dividing Eq. A7 by h i and defining v ic ≡V ic /h i , one obtains Eq. 2 in the text, which is the utility obtained per unit of efficiency by worker i. “Free mobility” of workers will imply that each worker must receive the same utility per unit of individual education across all areas. Thus, condition (3) in the text holds true.
Each firm j maximizes profits by choosing the inputs {L jc , h j N jc }, land and efficiency units of labour, respectively. Here, h j can be interpreted as the average level of efficiency possessed by the N jc workers in the firm. (More formally, the labour input as measured in efficiency units is given by \({\sum\limits_{s = 1}^{N_{{jc}} } {h_{s} } }\). Since workers of different ability are perfect substitutes in this model, this expression can be rewritten as h j N jc not only when all workers in firm j have the same level of efficiency, but also when h j is the average value of efficiency of the workers in the firm.)
The problem of firm j in area c is then:
The first-order conditions for problem (A8) are thus given by:
From Eqs. A9 and A10, together with Eq. 4 in the text, one obtains that:
and
One can exploit Eqs. A11 and A12 to obtain expressions for {L jc , h j N jc }, which can be substituted into the production function (4) in the text. Thus, one obtains the “free-entry” condition (5). (By using Eqs. A11 and A12 to substitute {L jc , h j N jc } away from the profit expression, one can verify that profit is zero in equilibrium.)
Finally, the “free-entry” condition (5) can be coupled together with the “free-mobility” condition (expressed in labour efficiency units), given by:
where \(\eta = {\left( {1 - \mu } \right)}^{{1 - \mu }} \cdot \mu ^{\mu } \).
By taking the logs of expressions (5) and (A13), one obtains two log-linear equations in (log w c , log r c ), i.e.
By finding log w c from Eq. A14 and substituting it into Eq. A15, one obtains Eq. 6 in the text. Then, expression (6) can be exploited to get rid of log r c in Eq. A14 to obtain expression (7).
1.2 Description of the variables
Variable | Description | Source |
---|---|---|
Rents | Log of the annual rent. For each household, the interviewee can be either the property owner or the tenant. In the first case, the SHIW collects the rent the owner charges (or, if the dwelling is not rented or it is the family residence, the best estimate for the rent that could be charged). In the second case, the tenant reports the actual rent paid | SHIW |
Wages | Log of hourly wages. Hourly wages are calculated by dividing the annual earnings (from any activity as employee, including fringe benefits, net of taxes and social security contributions) by the total amount of hours worked in a year (average hours worked per week×months worked×4.3333). The sample is trimmed at the 1st and 99th and percentile of the distribution of earnings | SHIW |
Local human capital | Average years of schooling (1991) in the LLM where the dwelling is located or the individual resides | ISTAT |
Surface area | Area in m2 | SHIW |
Age of the house | Calculated as the difference between the year of the survey and the year the house was built, which is a datum available from the SHIW | SHIW |
Bathrooms | Indicator variable equal to 1 if two or more bathrooms are available in the dwelling | SHIW |
Heating system | Indicator variable equal to 1 if a heating system is available in the dwelling | SHIW |
Imputed rents | Indicator variable equal to 1 if the rent recorded is imputed by the interviewee | SHIW |
House’s location | Series of dummies for the location of the dwelling (isolated area, countryside; town outskirts; between outskirts and town centre; town centre; other; hamlet) | SHIW |
LLM population | Log of the LLM population | ISTAT |
South | Indicator variable equal to one for the following Italian regions: Abruzzi, Molise, Campania, Puglia, Basilicata, Calabria, Sicilia, and Sardegna | SHIW |
Individual human capital | Number of years of study required to achieve the highest qualification earned by the individual. We derived the length of education by assigning: 2 years to no qualification; 5 years to elementary school; 8 years to middle school; 11 years to professional secondary school diploma; 13 years to high school; 16 years to a diploma or other short course university degree; 18 years to a bachelor’s degree; and 20 years to a postgraduate qualification | SHIW |
Experience | Calculated as the difference between worker’s age at the survey date and the age at first job held, which is a datum available from the SHIW | SHIW |
Subjective house rating | Series of dummies for the subjective (the survey asks “How do you rate this dwelling?”) rating of the dwelling (luxury; highly desirable; mid-range; modest; low income; very low income; rural; other) | SHIW |
Subjective location rating | Series of dummies for the subjective (the survey asks “How would you rate the location where the dwelling is located?”) rating of the dwelling’s location (highly desirable; run-down; neither highly desirable nor run-down) | SHIW |
Job qualification | Series of dummies for the employment work status (blue-collar worker or similar; office worker or school teacher; junior manager, clerk; manager, senior official) | SHIW |
Industries | Series of dummies for the sector of activity of the firm in which the individual works (agriculture; manufacturing; building and construction; wholesale and retail trade, lodging and catering services; transport and communications; services of credit and insurance institutions; real estate and renting services, other professional, business activities; general government and other private and public services) | SHIW |
Firm size | Series of dummies for the size of the firm in which the individual works (up to 4; from 5 to 19; from 20 to 49; from 50 to 99; from 100 to 499; 500 or more; not applicable, public-sector employee) | SHIW |
Per capita GDP | Per capita net disposable income in the province in thousand lira | ISTAT |
LLM unemployment rate | LLM 1993 unemployment rate | ISTAT |
LLM physical capital | Ratio between stock of capital (valued at the replacement price) and value added in each LLM | Cannari–Signorini |
LLM infrastructures | Ratio between kilometers of roads and LLM’s surface in km2 | Cannari–Signorini |
LLM population | Log of the LLM population | ISTAT |
LLM plant intensity | Ratio between the number of plants and the LLM’s surface in km2 | Cannari–Signorini |
LLM share of population under 35 | Ratio of inhabitants less than 35 over the entire population of the LLM | ISTAT |
Theatre | Theatre shows over the population residing in the province | ISTAT DEMOS |
Cinema | Cinema halls over the population residing in the province | ISTAT DEMOS |
Crime | First-degree murders, robberies and blackmail divided by the population residing in the province | ISTAT DEMOS |
Nurseries | Public nurseries for children aged 0–5 over the corresponding age group for the population residing in the province | ISTAT DEMOS |
Doctors | Doctors in public hospitals over the population residing in the province | ISTAT DEMOS |
Hospital beds | Beds in public hospitals over the population residing in the province | ISTAT DEMOS |
1981 share of population 0–5 | Share of the LLM population between the age of 0 and 5 in 1981 | ISTAT |
1981 share of population 5–10 | Share of the LLM population between the age of 5 and 10 in 1981 | ISTAT |
Rights and permissions
About this article
Cite this article
Dalmazzo, A., de Blasio, G. Production and consumption externalities of human capital: an empirical study for Italy. J Popul Econ 20, 359–382 (2007). https://doi.org/10.1007/s00148-005-0038-7
Received:
Accepted:
Published:
Issue Date:
DOI: https://doi.org/10.1007/s00148-005-0038-7