关于夜灯数据的三个事实

ThreeFacts About Night Lights Data

作者：

约翰·吉布森（John Gibson）

吉阿·波·吉布森（Geua Boe-Gibson）

The DMSP night lights data used in economics are old and not very accurate. Newer VIIRS night lights data have 60 percent higher predictive power for state-level GDP in the United States. Predictive accuracy is far higher in the cross section than for time series changes, either annually or quarterly. Night lights predict more weakly for agriculture than for manufacturing and other industries. These three facts suggest a need for caution in using night lights data, which may be unsuitable for many economics research purposes in many places.
经济学中使用的DMSP夜灯数据过时且不太准确。较新的VIIRS夜灯数据对美国州级GDP的预测能力高60％。与每年或每季度的时间序列变化相比，横截面的预测精度要高得多。夜灯对农业的预测比对制造业和其他行业的预测要弱。这三个事实表明在使用夜灯数据时需要谨慎，这可能不适合许多地方的许多经济学研究目的。