Portal de Periódicos da CAPES

Estimation of Local Employment Growth: Do Sectoral Aggregation and Industry Definition Matter?

2013; Routledge; Volume: 48; Issue: 11 Linguagem: Inglês

10.1080/00343404.2012.756578

1360-0591

Francesca Mameli, Alessandra Faggian, Philip McCann,

Regional Development and Policy

AbstractMameli F., Faggian A. and McCann P. Estimation of local employment growth: do sectoral aggregation and industry definition matter?, Regional Studies. Over the last two decades, numerous attempts have been made to explain the determinants of local growth, with as yet little overall consensus. The aim of this paper is to reveal a potential problem of parameter heterogeneity in growth regressions associated with the use of data at different levels of aggregation. Using Italian data and focusing on both manufacturing and service industries, it is shown how different empirical outcomes can be generated from exactly the same spatial units simply by changing the levels of sectoral aggregation. Moreover, the results point to some advantages associated with using more disaggregated data.Mameli F., Faggian A. and McCann P. 地方就业成长评估:部门聚集和产业定义是否具有影响,区域研究。过去二十多年来,已有众多努力尝试致力于解释地方成长的决定因素,但却显少有普遍的共识。本文目的在于揭露成长回归中,有关在不同聚集程度的资料使用的参数异质性之潜在问题。本研究运用意大利的资料,并同时聚焦制造业和服务业,展现如何仅透过部门聚集程度的改变,便可使完全一致的空间单元产生不同的经验结果。研究结果亦指出多加运用分类资料的若干优点。Mameli F., Faggian A. et McCann P. Une estimation de la croissance locale de l'emploi: est-ce que l'agrégation sectorielle et la définition de l'industrie importent?, Regional Studies. Pendant les deux dernières décennies, de nombreuses tentatives ont été faites pour expliquer les déterminants de la croissance locale et, jusqu'ici, peu de consensus général s'est dégagé. Ce présent article cherche à mettre à jour un problème d'hétérogénéité paramétrique dans les régressions de croissance lié à l'emploi des données à différents niveaux d'agrégation. À partir des données italiennes, et portant à la fois sur le secteur industriel et les industries des services, on montre comment on peut obtenir des résultats empiriques différents des mêmes unités géographiques simplement en changeant les niveaux d'agrégation sectorielle. Qui plus est, les résultats indiquent quelques avantages relatifs à l'emploi des données plus détaillées.Mameli F., Faggian A. und McCann P. Schätzung des lokalen Beschäftigungswachstums: Kommt es auf sektorale Aggregation und Branchendefinition an?, Regional Studies. In den letzten zwei Jahrzehnten wurde oft versucht, die Determinanten des lokalen Wachstums zu erklären, wobei jedoch bisher nur ein geringer Gesamtkonsens erzielt werden konnte. Mit diesem Beitrag soll ein potenzielles Problem der Heterogenität von Parametern in Wachstumsregressionen bei der Verwendung von Daten auf unterschiedlichen Aggregationsebenen aufgezeigt werden. Mit Hilfe von italienischen Daten und unter Berücksichtigung der produzierenden Industrie und der Dienstleistungsbranche wird gezeigt, wie aus genau denselben räumlichen Einheiten durch eine einfache Änderung der Ebenen der sektoralen Aggregation unterschiedliche empirische Ergebnisse erzeugt werden können. Darüber hinaus lassen die Ergebnisse darauf schließen, dass die Verwendung von stärker disaggregierten Daten mit einigen Vorteilen verbunden ist.Mameli F., Faggian A. y McCann P. Estimación del crecimiento de empleo local: ¿es importante la agregación sectorial y la definición de industria?, Regional Studies. En las dos últimas décadas se ha intentado en numerosas ocasiones explicar los determinantes del crecimiento local, aunque hasta ahora con poco consenso general. El objetivo de este artículo es mostrar un posible problema de la heterogeneidad de los parámetros en las regresiones de crecimiento asociada al uso de datos en diferentes niveles de agregación. Usando datos italianos y centrándonos en las industrias de manufactura y de servicios, mostramos cómo pueden generarse diferentes resultados empíricos desde exactamente las mismas unidades espaciales con tan solo cambiar los niveles de la agregación sectorial. Además, los resultados indican que el uso de datos más disgregados aporta algunas ventajas.KeywordsIndustrialSectorsAggregationLocalEmploymentGrowthKeywords产业的部门聚集地方就业成长KeywordsIndustrielSecteursAgrégationLocalEmploiCroissanceKeywordsIndustrieSektorenAggregationLokalBeschäftigungWachstumKeywordsIndustrialSectoresAgregaciónLocalEmpleoCrecimientoJEL classifications: R12R13O18 AcknowledgementsThis research was partly funded by the Regione Autonoma della Sardegna with the programme 'Borse di Ricerca per Giovani Ricercatori'.Notes1. In particular, while specialization is always negatively related to employment growth, diversity plays a positive role for services and has a negative influence on manufacturing.2. For a recent treatise on the scale-dependency of agglomeration externalities and the MAUP, see Burger et al. (2007).3. The two issues can never be entirely separated because using a large number of spatial units, as done in this paper, implies that, even by holding these spatial units constant, there will still be some spatial autocorrelation in the data. However, as noted, this problem holds for all papers in this tradition which do not employ spatial econometric techniques while using large numbers of spatial units. On the other hand, while there are some by holding these spatial units constant, there will still be some spatial autocorrelation in the obviously commonalities between the MAUP and the sectoral aggregation issue discussed here; mathematically these are also rather different issues, in that the MAUP can be related to nested, non-nested and overlapping sets, whereas the problem discussed here only relates to nested subsets.4. The methodology and criteria used to identify the LLSs is described by Sforzi et al. Citation(1997).5. In response to changes in society and the economy, the number of Italian LLSs in 2001 has decreased to 686 areas. In order to standardize the data, 1991 LLS definitions were adopted for both years of the analysis.6. According to the theory developed by Marshall Citation(1920), and later formalized by Arrow Citation(1962) and Romer Citation(1986), many types of knowledge spillovers (abbreviated as MAR externalities) are intra-industry phenomena.7. Lacking plant-level data, the Herfindahl index was approximated here by using the employment distribution over plants (the industrial plant size distribution was computed by exploiting the census information on the number of employees and local units in the fourteen employment dimensional classes identified by ISTAT), as suggested in Schmalensee Citation(1977) (see also Lafourcade and Mion, Citation2003; and Mameli et al. Citation(2008).8. The productivity gains of urban scale economies are generally found to be positive by empirical research. On the other hand, a recent meta-analysis undertaken by Melo et al. Citation(2009) over thirty-four studies dealing with agglomeration showed how the magnitude of the estimates is affected by the single-study characteristics. Differences in the reviewed results depend upon the specific country effects, the industrial coverage, the specification of agglomeration economies, the presence of controls for both unobserved heterogeneity and labour quality.9. In equation (2) the interpretation of the results for the variables den (population/area), infraden (infrastructure/area) and inf itself is slightly difficult when these variables are all included together. This is because the specification in logs means that the group of these variables can be written as:which itself can be rearranged to give:As such, the interpretation of equation (2) is somewhat less straightforward than that of equation (1). However, as explained in this paper, the literature suggests that each of these variable specifications may have an impact on the response variable, and it is simply the need to capture any such effects that provides the justification for the inclusion of these variables.10. Since heteroskedasticity has been detected using a Breush–Pagan/Cook–Weisberg test, White robust standard errors are used to partially correct for this problem.11. This has allowed employment density to be excluded as a measure of the size of local economy and instead population density to be used. The highest value of the correlation matrix is 0.71, while all other values are mostly below this threshold.12. As a rule of thumb, a variable merits further investigation if it has a VIF value greater 10 or a tolerance lower than 0.1.13. For innovative sectors.14. Frenken et al. Citation(2007) specifically addressed this issue by distinguishing between related and unrelated varieties at different levels of sectoral aggregation. Unrelated variety (used as a proxy of portfolio effects that protect a region against shocks in unemployment) is measured at the two-digit sector level, while related variety (an indicator of Jacobs externalities) is measured at the five-digit level within two-digit classes. The authors find that only related variety enhances employment growth.15. A likelihood ratio (LR) test is used to compare the most comprehensive model with that resulting from a restricted specification. The LR statistics are 167.23 at the two-digit level and 1964.52 at the three-digit level (Chi-squared distributed with seven degrees of freedom) and the associated p-values are very low (p < 0.001). Note that the same model cannot be compared at different digit levels because data are not nested.