Subnational politics and foreign direct investment in Mexico

ABSTRACT

Focusing on Mexico, this article makes two departures from existing studies of the determinants of foreign direct investment (FDI): (1) it disaggregates investment into three types (resource-, market- and efficiency-seeking); and (2) it models variation in investment subnationally, across the 32 Mexican states. Using panel data for foreign investment between 2000 and 2009, we find that the predictors of subnational variation in investment go beyond simple geographic and economic conditions and include factors such as local political party control, social stability and the perceived effectiveness of state authorities. Moreover, the three types of investment are shaped by distinct social, political and economic dynamics. Insofar as the location and type of foreign investment can affect economic development and inequality within – and not just between – countries, the subnational distribution of investment is of consequence for both academics and policy makers.

KEYWORDS:

• Foreign direct investment
• Mexico
• political parties
• corruption

Notes

¹Most of the existing subnational studies focus on East Asian economies, especially China and Vietnam (e.g. Zhou, Delios and Yang, 2002; Chadee, Qiu and Rose, 2003; Meyer and Nguyen, 2005; Malesky, 2008).

²Dunning and Lundan (2008) note that there is a fourth type of investment: strategic investment, which seeks to obtain strategic assets or capabilities, such as technologies. However, Máttar, Moreno-Brid and Peres (2003: 134) note specifically that this kind of strategic investment is not a motivator of investment in Mexico.

³The Pierson's correlation between the three types of investment on a national level are −.16, .05 and −.09, none of which are statistically significant at p < .05.

⁴Some work has already been done in this area (Dussel Peters, 2009), but many issues still remain to be addressed.

⁵An additional consideration is that data for some of the independent variables (particularly the indicators of administrative competence) are unavailable for earlier years.

⁶Numerous other parties exist. The PRI is the long-dominant party and PAN and PRD are the two parties that have consistently challenged the PRI's domination in state and national politics.

⁷These incentives are difficult to standardize because they differ greatly from state to state, are often ad hoc, and may represent very different values or incentives even to investors in the similar sectors.

⁸An alternative view of corruption has been advanced by Egger and Winner (2005): corruption may be positively associated with investment because it speeds the potentially cumbersome legal requirements for investment and large investors are often willing to take advantage of it.

⁹In spite of well-known concerns with indices of this sort (e.g. that in aggregate form what they measure is not transparent), we feel that the use of this wide-ranging index is justifiable because it is likely that it captures the reputational quality of the state government.

¹⁰Wooldridge (2002) tests for serial autocorrelation of each of the three series did not reject the null hypothesis of no serial autocorrelation in the panels at p < .05.

¹¹Given log-transformed dependent variables, for log-transformed predictors (GDP, GDP per capita, mining production) coefficients are interpreted as percentage change in y associated with 1 per cent change in x (i.e. elasticity); for dichotomous predictors (location dummies, party politics variables) 100*(coefficient) is interpreted as the effect of the discrete change in x from 0 to 1; for non-transformed predictors, 100*(coefficient) percentage change is interpreted as the effect of one unit change in x.

¹²To check the robustness of the findings, the three combined models were also estimated with the inclusion of the lagged dependent variable and using panel-corrected standard errors. The results reported here remain stable across these alternative models, although there is some difference in the size of the mean effect of independent variables, particularly with the inclusion of the lagged dependent variable.

¹³To counter this problem, Mejía (2002) recommends incorporating indicators that take into account different criteria of the regulatory framework – regulatory reform, business opinion about the state procedures, quality improvement programs and the average days for procedures for opening a business – much of which we gauge with the Corruption and Efficiency and Effectiveness indices.