Analysing firm-level impacts of high-speed railways on reducing business costs: evidence from China

Figures & data

Figure 1. Bar chart of propensity scores.

Note: The propensity score represents the probability that each city individual will be affected by the HSR, i.e., the probability of being selected to the treatment group. The vertical axis represents the nuclear density. On support means that samples are successfully matched after the PSM method, i.e., propensity scores of both treatment and control groups are in the range of common support, and off support means that samples are not successfully matched after the PSM method, indicating data of these samples ought to be omitted (Abadie & Imbens, 2016).

A bar chart showing results of propensity scores for both treated and untreated samples after PSM. The distribution of propensity scores of control variables in the treatment group approximates an ideal normal distribution in a symmetric form.

Table 1. Variable definitions of PSM-DID analysis.

Variable	Definition	Data sources
FXCit	Fixed costs of each firm	China Industrial Business Database
VBCit	Variables costs of each firm
BSCit	Total business costs of each firm
Cit	HSR dummy (city)	China Railway Yearbook
Yit	HSR dummy (year)^4
GDPit	GDP per capita of the city where the firm is located	China Statistical Yearbook^5
INDit	The proportion of the secondary industry of the city	China City Statistical Yearbook^6
FDIit	The proportion of output created by FDI in the total output of the city
INTit	The Internet penetration rate of the city
PRit	Population density
Controlit	All control variables
Oit	Ownership dummy

Table 2. Balance test results of PSM.

	Pre-matching	Post-matching
GDP	4.21***	−0.36
Industry	0.17	−0.97
FDI	5.67***	0.27
Internet	4.98***	−0.44
Pop	10.27***	−0.21
Pseudo R^2	0.110	0.002
Mean-bias	32.6	3.6
Med-bias	31.7	2.7

Notes: The values outside the parentheses represent t-statistics and values in the parentheses represent p-values corresponding to t-statistics. */**/*** indicates the difference in means between the two groups is statistically significant at the 0.1/0.05/0.01 level.

Table 3. Impacts of HSR on fixed and variable costs.

Dependent variable: Business costs fixed (columns 1–3) and variable costs (columns 4–6) for each firm
	Fixed costs			Variable costs
	(1)	(2)	(3)	(4)	(5)	(6)
HSR coefficient	−0.044** (1.268)	−0.025** (1.239)	−0.026*** (0.383)	−0.136** (−2.334)	−0.128*** (−0.537)	−0.163*** (−2.893)
GDP	−0.046 (−0.293)	−0.032 (−0.213)	−0.033 (−0.577)	−0.011 (−1.192)	−0.025*** (0.653)	−0.038*** (5.040)
Industry	0.032 (1.002)	−0.004 (2.456)	0.003 (1.489)	−0.001 (0.023)	−0.002 (1.989)	−0.003 (−0.053)
FDI	−0.034 (4.032)	−0.045*** (−4.045)	−0.048*** (2.127)	−0.037 (1.074)	−0.088*** (1.565)	−0.057*** (0.311)
Internet	0.007 (3.208)	0.013 (3.721)	0.009 (3.265)	−0.002 (−0.991)	−0.006 (−1.293)	−0.004 (−0.932)
Population	−0.036 (−0.077)	−0.046** (−1.219)	−0.099*** (2.221)	−0.005 (2.345)	−1.295** (−3.883)	−1.911*** (0.012)
Ownership	0.003** (0.079)	0.998*** (2.344)	0.233*** (−0.679)	0.007 (4.327)	0.009 (2.431)	0.183** (0.008)
Constant	0.171***	0.178***	0.177***	0.177***	0.173***	0.174***
	(0.990)	(−0.203)	(0.566)	(2.242)	(0.404)	(0.558)
R^2	.833	.776	.993	.835	.977	.997
Observations	5,520	12,816	18,368	5,520	12,816	18,368
Industry controls	Yes	Yes	Yes	Yes	Yes	Yes
Region FE	Yes	Yes	Yes	Yes	Yes	Yes
Time FE	Yes	Yes	Yes	Yes	Yes	Yes

Notes: The table represents results from difference-in-difference analysis estimated by OLS. The t-statistics clustered by firms are reported in parentheses. */**/*** indicates the difference in means between the two groups is statistically significant at the 0.1/0.05/0.01 level. Columns 1, 2 and 3 represent primary, secondary, and tertiary industries, so are columns 4, 5 and 6, respectively. The first three columns show impacts of HSR on fixed costs and the final three columns show impacts of HSR on variable costs.

Table 4. Test of spatial heterogeneity.

Dependent Variable: Business costs (fixed and variable costs) for each firm
	Eastern China						Central and western China
	Fixed costs			Variable costs			Fixed costs			Variable costs
	(1)	(2)	(3)	(4)	(5)	(6)	(7)	(8)	(9)	(10)	(11)	(12)
HSR ($C\times Y$)	−0.027** (1.256)	−0.046*** (2.532)	−0.044*** (1.285)	−0.086*** (−1.723)	−0.143*** (3.426)	−0.168*** (1.008)	−0.008 (2.003)	−0.047 (1.009)	−0.078** (−0.957)	−0.089 (3.259)	−0.057** (1.676)	−0.126*** (2.229)
Constant	0.098***	0.096***	0.097***	0.087***	0.093***	0.096***	0.038	0.065	0.083	0.036	0.067	0.088
	(1.235)	(−0.998)	(0.034)	(1.113)	(2.987)	(−1.145)	(1.331)	(0.558)	(1.295)	(−1.213)	(0.783)	(1.019)
R^2	.736	.853	.777	.935	.965	.878	.632	.867	.968	.732	.968	.969
Observations	3,537	7,203	11,352	3,537	7,203	11,352	1,983	5,613	7,016	1,983	5,613	7,016
Industry controls	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes
Region FE	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes
Time FE	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes

Notes: The table represents results from difference-in-difference analysis estimated by OLS. The t-statistics clustered by firms are reported in parentheses. */**/*** indicates the difference in means between the two groups is statistically significant at the 0.1/0.05/0.01 level.

Abadie, A., & Imbens, G. W. (2016). Matching on the estimated propensity score. Econometrica, 84(2), 781–807. https://doi.org/10.3982/ECTA11293

 Web of Science ®Google Scholar