Does democracy affect environmental quality in developing countries?

Abstract

This article examines the impact of democracy on environmental conditions in a large sample of developing countries for the period 1976–2003. This relationship is explored empirically using three indicators of environmental quality: carbon dioxide emissions, water pollution and deforestation damage. We find evidence that democracy is conducive to environmental improvement but that this result depends on the measure of the environmental quality that is used. We also find remarkable differences in results across our different sub-samples. The conclusion therefore is that there is no uniform relationship between democracy and the state of the environment.

Acknowledgements

We are grateful to Marisa Scigliano and an anonymous referee for helpful suggestions.

Notes

¹ See also Stern (2008). According to Conceicao (2003), if the concentration of CO₂ in the atmosphere reaches twice the pre-industrial era, the effect will be catastrophic, with developing countries losing 2 to 9% of their Gross Domestic Product (GDP).

² Of possible significance is also trade. For instance, Frankel and Rose (2002) find that increased trade may have a beneficial impact on certain forms of pollution. In the same vein, Shen (2008) finds that trade liberalizations in China increase air pollutants but reduce water pollutants.

³ The impact of democracy has been discussed in the past, but the focus has been on whether more democratic governments have better prospects of extricating their nations from economic poverty or promoting their economic development. See, for example, Persson and Tabellini (2006) for a recent discussion.

⁴ On theoretical grounds, democracy may also have a negative impact on the environment. Early writers such as Hardin (1968) and Heilbroner (1974) made the argument that freedom needs to be constrained to prevent environmental ruin–given an increasing resource scarcity and individuals' incentive to harvest the common good.

⁵ Larger corporations have considerable economic and political powers. Democratic governments are in a better position to ensure that environmental interests of their citizens are not overridden by political pressures exerted by corporations.

⁶ Analogously, one may expect a positive correlation between corruption and environmental degradation. Recent discussions of a possible relationship include Lopez and Mitra (2000), Welsch (2004) and Damania (2005).

⁷ Compared with earlier studies, Fredriksson and Wollscheid (2007) use more recent data, but their panel consists of data from the late 1990s only. Pellegrini and Gerlagh (2006) utilize a longer time span, but use only a few years from the 1990s.

⁸ Conjointly, the article extends the literature by looking at more countries over a longer time period.

⁹ Our data is an unbalanced panel. That is, some countries have more observations than others. The list of countries and the number of observations per country is available upon request.

¹⁰ Hsiao (1986) and Baltagi (1995) inter alia discuss the advantages and shortcomings of using panel data analysis.

¹¹ Even though freedom and democracy are not synonymous, many of the questions that are used to generate the ratings address the notion of democracy directly. As the result, a number of recent studies (e.g. Barro, 1996; Arvin and Barillas, 2002) use the Freedom House ratings as their measure of democracy.

¹² Thus, the environmental Kuznets curve is an inverted U-shaped relationship between income (as the measure of the level of economic development) and environmental degradation (as defined by pollution, deforestation, etc.). The name originates from Kuznets' proposed inverted U-shaped relationship between income growth and income inequality. See, for example, Selden and Song (1994), Grossman and Krueger (1995), Holtz-Eakin and Selden (1995), Hilton and Levinson (1998), Lekakis and Kousis (2001) and Sobhee (2004) for a discussion.

¹³ A GLS regression is superior to an Ordinary Least Squares (OLS) procedure in this case. In a GLS model, the fixed effects are estimated by weighted least squares. Under such procedure, the weight for each observation is different and is the reciprocal of the normalized SD of the disturbance for the observation obtained from an OLS estimation. By contrast, OLS gives an equal weight to each observation. GLS corrects for this by giving less weight to the outliers.

¹⁴ By controlling for country fixed effects, the problem of omitted variables bias, which seriously afflicts cross-country regressions, is reduced.

¹⁵ Bhattarai and Hammig (2001) also find a diverse pattern of results in their study of the relationship between deforestation and regime type. However, their results with respect to different regions are different compared to ours. They find that in Africa and Latin America more democratic countries have less deforestation, with the opposite result being true for Asia.