Table D.3.
Robustness to Corrections for Spatial Autocorrelation
| (1) | (2) | (3) | (4) | (5) | (6) | |
|---|---|---|---|---|---|---|
|
| ||||||
| Corrected OLS | Spatial GMM | Corrected OLS | Spatial GMM | Corrected OLS | Spatial GMM | |
|
| ||||||
| Full Sample | Restricted Sample | Full Sample | Restricted Sample | Full Sample | Restricted Sample | |
|
| ||||||
| Dependent Variable is Log Population Density in: | ||||||
| 1500 CE | 1000 CE | 1 CE | ||||
|
|
|
|
||||
| Log Years since Neolithic Transition | 1.087*** [0.184] | 2.038*** [0.387] | 1.480*** [0.213] | 2.713*** [0.498] | 1.930*** [0.316] | 3.322*** [0.404] |
| Log Land Productivity | 0.576*** [0.053] | 0.583*** [0.092] | 0.497*** [0.066] | 0.575*** [0.095] | 0.394*** [0.076] | 0.448*** [0.093] |
| Log Absolute Latitude | −0.314*** [0.108] | −0.257* [0.141] | −0.229* [0.123] | −0.117 [0.138] | 0.057 [0.101] | 0.124 [0.115] |
| Mean Distance to Nearest Coast or River | −0.392** [0.195] | 0.318 [0.351] | −0.528** [0.207] | 0.373 [0.370] | −0.685*** [0.168] | −0.423 [0.294] |
| Percentage of Land within 100 km of Coast or River | 0.899*** [0.319] | 1.395*** [0.417] | 0.716** [0.351] | 1.550*** [0.409] | 0.857** [0.371] | 1.143** [0.461] |
|
| ||||||
| Continent Dummies | Yes | Yes | Yes | Yes | Yes | Yes |
| Observations | 147 | 96 | 142 | 94 | 128 | 83 |
| R-squared | 0.73 | 0.70 | 0.67 | 0.62 | 0.69 | 0.72 |
Notes – (i) log land productivity is the first principal component of the log of the percentage of arable land and the log of an agricultural suitability index; (ii) the spatial GMM regressions employ the numbers of prehistoric domesticable species of plants and animals as instruments for log transition timing; (iii) a single continent dummy is used to represent the Americas, which is natural given the historical period examined; (iv) the spatial GMM regressions do not employ the Oceania dummy due to a single observation for this continent in the corresponding regression samples; (v) standard errors corrected for spatial autocorrelation are reported in square brackets; (vi) the spatial distribution of countries in ℜ2 is specified using aerial distances between geodesic centroids; (vii) the spatial autocorrelation in error terms is modelled as declining linearly along a 4,000 km radius from each observation; (viii) *** denotes statistical significance at the 1 percent level, ** at the 5 percent level, and * at the 10 percent level, all for two-sided hypothesis tests.