The Cantor and Land model of unemployment and crime separates the effects of long- and short-run unemployment. In the long run, increases in unemployment are expected to increase crime, whereas the same increases are expected to decrease crime in the short run. This model has been tested for decades, generally supporting these predictions. In this article, we investigate spatial variations in these relationships using geographically weighted regression. Using crime data from Vancouver, Canada (commercial burglary, residential burglary, mischief, theft, theft from vehicle, theft of vehicle, and aggregate property), we find global models do not exhibit statistically significant unemployment–crime relationships, but they do emerge in local (geographically weighted) regression. These results have important implications for theoretical development, policy formation, and policy evaluation.

Cantor and Land模型的失业和犯罪将长期和短期失业的影响分开。从长远来看，失业率的上升预计将增加犯罪，而从短期来看，预计失业率的上升将减少犯罪。该模型已经测试了数十年，总体上支持了这些预测。在本文中，我们使用地理加权回归调查这些关系中的空间变化。使用加拿大温哥华的犯罪数据（商业入室盗窃，住宅入室盗窃，恶作剧，盗窃，盗窃车辆，盗窃车辆和财产总计），我们发现全球模型并未显示出统计上显着的失业与犯罪之间的关系，但它们确实在局部（按地理加权）回归。这些结果对理论发展具有重要意义，