Working Paper: NBER ID: w29569
Authors: Arman Khachiyan; Anthony Thomas; Huye Zhou; Gordon H. Hanson; Alex Cloninger; Tajana Rosing; Amit Khandelwal
Abstract: We apply deep learning to daytime satellite imagery to predict changes in income and population at high spatial resolution in US data. For grid cells with lateral dimensions of 1.2km and 2.4km (where the average US county has dimension of 55.6km), our model predictions achieve R2 values of 0.85 to 0.91 in levels, which far exceed the accuracy of existing models, and 0.32 to 0.46 in decadal changes, which have no counterpart in the literature and are 3-4 times larger than for commonly used nighttime lights. Our network has wide application for analyzing localized shocks.
Keywords: No keywords provided
JEL Codes: R0
Edges that are evidenced by causal inference methods are in orange, and the rest are in light blue.
| Cause | Effect |
|---|---|
| CNN model predicts changes in local economic activity (R15) | changes in income and population (J11) |
| daytime satellite imagery features (Y91) | changes in income and population (J11) |
| features learned from satellite imagery (Y91) | economic outcomes (F61) |
| CNN model outperforms nighttime light data (C45) | more accurate reflection of local economic conditions (R11) |
| localized economic shocks (F69) | changes in income and population (J11) |