Capital grants and regional public investment in Spain: fungibility of aid or crowding-in effect?

Abstract

The aim of capital grants to subcentral governments is to increase their investment. However, recipients may try to allocate additional resources to cut saving or deficit, generating a crowding-out effect on self-financed investment. Using data from the Spanish Autonomous Communities during the period 1984 to 1999, the effect of capital transfers on regional public investment, saving and deficit are estimated. Econometric results demonstrate that capital grants have no relevance when explaining the dynamics of saving. However, in deficit regressions they are shown to be both significant and negative. Both results lead us to a partial long-run crowding-out effect of grants on self-financed investment.

Acknowledgements

The author is grateful for helpful comments from an anonymous referee, Marc Baudry, Thorsten Baymdur-Upmann, Alex Esteller, Lars Feld, Lidia Gonçalvez, Jean-Michel Josselin, Cristophe Tavéra, and participants in the X Encuentro de Economia Publica (Tenerife, 2003), the European Public Choice Society Meeting (Aarhus, 2003), and the 59th IIPF Meeting (Prague, 2003). The usual disclaimer applies.

Notes

¹ For the specific purposes of this paper, see Zampelli (1986), Meyers (1987), Islam and Choudhury (1990), Zou (1994), Chernick (1998), Gamkhar (2002) and Knight (2002).

² Spanish regions are called Comunidades Autonomas (Autonomous Communities or ACs). There are 17 ACs. See Suárez-Pandiello (1999) and Castells (2001) for a survey on Spanish fiscal federalism.

³ See Correa and Manzanedo (2002) for a comprehensive analysis of regional policy in Spain during the period 1983–1999.

⁴ Source of data: Analistas Financieros Internacionales (http://www.sectorpublico.com). While resources from FCI and CI must be devoted to investment, a – minor – part of ESF covers current expenditures.

⁵ All figures have been deflated using the Spanish GDP deflator and divided by regional populations in 1996. The statistical source for both variables is the INE (http://www.ine.es).

⁶ Navarra and País Vasco amount to around 7% of the total Spanish population and 8% of Spanish GDP. Their contribution to financing public services in the poor regions is lower than corresponding regions with similar per capita GDP but under the Common Regime, e.g. Catalunya.

⁷ The three provincial governments in País Vasco collect the taxes and transfer the lion's share to regional government, because the latter retains the most important spending powers.

⁸ In 2002 the system was modified to lessen vertical fiscal imbalances, but it is irrelevant for the sample used in this paper.

⁹ As García-Milá and McGuire (1990) posit: ‘Because the level of ACs expenditures differs very little from the level of grants to the ACs, estimation of a traditional local government demand equation, in which expenditures per capita are regressed on a set of variables describing tastes and resources of the jurisdictions, with grants being a key variable, makes little sense in the Spanish context’.

¹⁰ For the whole period 1985–1999, mean per capita GDP were the following (Spain = 100): Baleares (143), Madrid (132), Cataluña (124), Rioja (114), Aragon (108), Comunidad Valenciana (102), Canarias (100), Cantabria (93), Asturias (90), Castilla y Leon, (89), Galicia (83), Murcia (82), Castilla La Mancha (79), Andalucia (70), and Extremadura (68). Source: FBBV (various years). These figures do not significantly change through this period.

¹¹ Investment comprises net direct spending by regional public administration and capital grants from it to other agents (mostly local governments and regional public firms) to be invested.

¹² Therefore, exogeneity of KG must be tested.

¹³  where s ² are estimated variances using OLS residuals, n is the number of individuals and T the number of periods. See Greene (1997).

¹⁴  where are squared correlations among FGLS residuals. See Breusch and Pagan (1980).

¹⁵ According to Nickell (1981), the bias is given by where matrices and represent the mean differenced X and y.

¹⁶ If β > 0 then due to the direction of the bias. But estimations show that .