Econometric Analysis of the Data

There is more than theory in existence to explain the relationship between poverty and environmental degradation. One common theory is that of the environmental Kuznets curve.

According to this theory, for very poor societies, increased income leads to increased environmental degradation. As people use more resources to become wealthier, and as improved technology allows for more efficient use of these resources, environmental degradation increases. However, at a certain point, a society is sufficiently wealthy that increased environmental quality becomes as important as increased material gains. As the care of the environment becomes a priority, people use some of their wealth, or sacrifice some potential wealth, to conserve it. Thus, this theory predicts that at a certain point, increased prosperity leads to decreased environmental degradation. We will keep this theory in mind during the following analysis.

Linear regression analysis, such as is done here, assumes the following relationship between the dependent variable y and the independent x variables:

Here, y_i is the level of deforestation in district i, is a vector of independent variables associated with that district, is a set of coefficients determining how these independent variables affect deforestation, and is an error term—that is, the difference between the true value of deforestation and the value predicted from independent variables.

At first, a linear relationship is assumed between deforestation and poverty. That is, the only dependent variable used was the proportion of the population not in poverty. I also make the assumption that each error termis normally distributed with a mean of zero and a constant variance, and that the error terms are independent of one another. This ignores the spatial data available.

The result is positive relationship between the amount of population not in poverty and deforestation, as this fitted line shows:

Each point represents one of Nepal’s 75 districts.

To examine the relationship in more complexity, a quadratic relationship was assumed. That is, the functional relationship assumed is

(deforestation index) = +(proportion of population not in poverty)+(proportion of population not in poverty)².

The fitted line and results are shown here.

Parameter	Estimated Value	95% Confidence Interval Bounds (OLS)
	-93.73	-155.02, -32.45
	315.57	93.64, 537.51
	-194.80	-385.88,-3.71

All coefficients are significant, and clearly the data is consistent with a Kuznets curve type relationship.

However, a more refined analysis takes account of spatial autocorrelation between the points. ArcGIS provides tools to determine if data are spatially autocorrelated. ArcGIS reports that both poverty and deforestation levels are strongly spatially autocorrelated.

This, for example, is the spatial autocorrelation report for deforestation levels.

To analyze the data taking this autocorrelation into account, a program called Geographically Weighted Regression was used. The coordinates of each of the data points were taken to be the centroids of the districts:

A Gaussian analysis was done using these coordinates and poverty and deforestation value. The coefficients obtained were the same as for the quadratic regression above. However, the program determined that none of the coefficients were significant at the 90% level. This means that deforestation levels are much more strongly linked to deforestation levels in neighboring districts than they are to poverty levels. It cannot be concluded from this analysis that poverty has an effect on deforestation.

But this is not the final word. Future refinements will be implemented in this study.