Table 1.
Estimated impact of the Grassland Ecological Compensation Program (GECP) on grassland quality: county and household levels.
| (1) | (2) | (3) | |
|---|---|---|---|
| County level | Household level | ||
| Models | DID | RE | FE |
| P × T | 0.032*** | 0.024*** | |
| (0.000) | (0.000) | ||
| Log (payment intensity) (yuan/ha) | 0.011* | ||
| (0.058) | |||
| Number of observations | 3,425 | 3,425 | 3110 |
| Number of counties/households | 574 | 574 | 821 |
| R2 | 0.989 | 0.806 | 0.958 |
Note. The dependent variable is NDVI in log form. NDVI stands for the Normalized Difference Vegetation Index, which is a measure for grassland quality. It was constructed based on infrared and near-infrared channel remote sensing images and has been widely used as an indicator of vegetation coverage. Column (1) provides the results from the DID (difference-in-differences) approach using county-level data (Eq. (1)). The treatment group (P = 1) includes the counties in five North and Northwestern program provinces that were covered in GECP-I, i.e., Xinjiang, Qinghai, Gansu, Ningxia, and Inner Mongolia. The control group (P = 0) includes the counties in five North and Northeastern provinces that were not covered by GECP-I (i.e., the first five-year period of GECP) until 2016, i.e., Shanxi, Hebei, Liaoning, Jilin, and Heilongjiang. The pre-program period (T = 0) is 2008–2010. The post-program period (T = 1) is 2011–2013. Column (2) presents the results from the county-level random-effect (RE) model. In Columns (1) and (2), year and province fixed effects are controlled for. Climate controls include monthly rainfall, temperature, and the Palmer Drought Severity Index (PDSI) for May–October. Socioeconomic control includes per-capita county fiscal income. Robust standard errors are clustered at the county level. Column (3) reports the results from the household-level fixed-effect (FE) model (Eq. (3)). Both household and year-fixed effects are controlled for. Household-, village-, and township-level time-variant variables also are controlled for. Household-level controls include the quantity of labor used in raising livestock, operated farm size, share of the joint operated area, the total number of different plots, a dummy variable for grassland harvesting, and a dummy variable for planting crop/fodder. Village-level controls include an indicator of whether a village has local grassroots measures in place to limit grazing intensity, an indicator of whether a village has a formal government-run monitoring system, and climate variables (cumulative rainfall and mean temperature from May to October in each year). Township-level controls include farm-gate livestock prices, hay prices, wages for non-pastoral employment, and grassland rental prices. Standard errors are clustered at the village-by-year level. A two-sided t test is performed for each coefficient. The exact p-values are in parentheses. *p < 0.10, ***p < 0.01.