Minnesota Connect

Click here to get the PDF file of my little project in Spatial Pattern Analysis Workshop, UC, Santa Barbara. The tile is:

“China’s Provincial Fiscal Inequality: A Cost-standardized Approach”

A common way to measure local fiscal capacity in China is by per capita local expenditure (PCEXP). The measure, however, overlooks the fact that the average cost of public service delivery is drastically different across regions. Even after adjusting for residential living cost, I find significant and negative elasticity between provincial population density and PCEXP. This leads to the surprising observations that some of the highest PCEXP in China are observed in Tibet, Xinjiang, and Qinghai, etc (see another post in the blog). Thus it is misleading to interpret the dispersion of PCEXP as representing fiscal inequality across provinces.

Here is the PDF file for the paper on China’s Provincial-level
Fiscal Disparities that we submitted for the 2008 International Conference on Public Administration.

The abstract:

Since China’s 1994 fiscal reforms, there have been increasing concerns about the distribution of fiscal capacities across the country. In this study we trace the evolution of intergovernmental relations in China since 1978, examine the fluctuation of provincial-level fiscal distribution during the period 1978-2006, and apply two approaches of factor decomposition to analyze the underlying factors of fiscal inequality. Gini indices over time show that that provincial-level fiscal dispersion in China remains high during the post-1994 period, as the equalization effects of central transfers became diminished. Caution should be noted in interpreting these results, however, because the fiscal dispersion is in part driven by high per capita local expenditures in remote provinces such as Tibet and Qinghai, probably due to higher cost of public service delivery. With regression-based inequality decomposition, we find that the fiscal dispersion can be primarily explained and attributed to the Prime Ratio, population density, and per capita GDP. The underlying reasons may include economic activity, cost of service delivery, and political considerations of equalization and national unity.

Package ineq in R can plot Lorenz curves. With a little twist, we can even have multiple Lorenz curves in one plot. The left-side plot above shows the Lorenz curves for per capita GDP, per capital local revenue, and per capital local expenditure for China provinces in 2006. The right-side plots show the corresponding Lorenz curves with fiscal measures that are adjusted by spatial price deflators. The first plot was generated by the codes below:

# ————————————————-
library(ineq)
Lc.1 <- Lc(PCGDP);Lc.2 <- Lc(PCRV); Lc.3 <- Lc(PCEP)
# ————————————————-
pdf(file=“LC.pdf”, width=6,heigh=6)

# Plot ———————————————
plot(Lc.1,lty=“dotted”);
lines(Lc.2$p, Lc.2$L,lty=“dashed”, lwd=1.2, col=4)
lines(Lc.3$p, Lc.3$L,lty=“dotdash”, lwd=1.5, col=2)

# Legends ———————————————
lines(c(.05,.15),c(.9,.9),lty=“dotted”)
lines(c(.05,.15),c(.85,.85),lty=“dashed”, lwd=1.2, col=4)
lines(c(.05,.15),c(.8,.8),lty=“dotdash”, lwd=1.5, col=2)

text(.15,.90,“PC Local GDP“, pos=4)
text(.15,.85,“PC Local Revenue”, pos=4)
text(.15,.8,“PC Local Expenditure”, pos=4)
# ————————————————-
dev.off()
# ————————————————-

To have multiple plots in one graph, I used:

Showing above are quintile-maps of per capita GDP (left) and per capita Expenditure (right) of Chinese provinces in 2005. Darker colors refer to higher data values. In terms of economic development, it is clearly that provinces along the east coast are the most developed while those to the west are most underdeveloped. Fiscally, however, the highest per capita expenditures occur at both ends of the continuum. The worse-offed provinces are the central ones, where the issue of social unrest may be the most pressing.

See the related post: Fiscal Issues & Geeky Stuff - The surprising “fiscal disparity” in China.

I had difficulty for quite a while today trying to open a data table to be joined with a Shapefile. Through VMware I usually use ArcView 3.2 rather than ArcGIS 9.2 to save some memory. However, I didn’t expect to come across these obstacles. Now I write them down to avoid getting frustrated again next time.

1. I clicked on “Tables” in the sidebar of ArcView but it would only take data in dBASE format, which is not supported by Office 2007 anymore (for whatever weird reason that I don’t know);

Solution: Click on “Add” on the topbar instead. It then accepts dBASE, INFO, and Delimited Text.

2. I saved my data file as Tab-Delimited text and then turned back to ArcView. But this time I got an error message: “STOP: The file ’XXXXX’ is not valid.”