Table 2 Kernel propensity score matching model.

From: Do open data impact citizens’ behavior? Assessing face mask panic buying behaviors during the Covid-19 pandemic

Treatment threshold

All stores

Higher college graduation rate areas

Lower college graduation areas

1%

3%

1%

3%

1%

Coef. of treatment dummy

−  0.522

−  0.758**

−  1.184***

−  0.720*

1.485

Std.err.

(0.572)

(0.272)

(0.329)

(0.271)

(2.099)

Adjusted \(R^{2}\)

0.11

0.31

0.44

0.36

0.07

Number of obs.

29,592

15,760

12,202

10,562

12,718

  1. This approach matches treated areas with similar non-treated areas based on observed characteristics, then applies the DiD to the matched areas. Each column in the table represents separate DiD regressions. All stores are included in the analysis in Columns 1–2. The stores only in higher college graduate rate areas are included in Columns 3–4. The stores only in lower college graduate rate areas are included in Column 5. The outcome variable is the daily number of sold mask numbers per household. All models present standard errors clustered at the area level. Treatment dummy thresholds have been changed to 1% and 3% for robust tests (see “Methods” section). \(^{***}\) \(p<0.01\), \(^{**}\) \(p<0.05\), \(^{*}\) \(p<0.1\).
Back to article page