https://orcid.org/0000-0002-5610-0479
![Sample our Economics, Finance,Business & Industry journals, sign in here to start your access, latest two full volumes FREE to you for 14 days]({"type":"image" , "src" : "/sda/5931/TOC-Subject-Sample-2021-Economics-Finance-Business-Industry.jpg"})
ABSTRACT
Previous studies suggest that ‘city–county consolidation’ has significant positive impacts on urban economic growth. Yet, the specific spillover effects on surrounding rural areas and the overall net impact on larger regions have not been fully investigated. We propose a difference-in-difference framework and find that consolidations have led to a significant loss of per capita gross domestic product (GDP) in nearby rural areas. The backwash mainly comes from the reduction of rural industrial output. We also find that this policy has an insignificant impact at the prefecture level, suggesting such consolidations do not promote aggregate growth and that any gains in urban areas are offset by losses in surrounding rural regions.
ACKNOWLEDGEMENTS
The authors thank the editor and three anonymous referees for comments and suggestions. The authors are also grateful to the seminar participants at the RUSE workshop, SRSA Conference, NARSC Conference, CES China Conference and Ohio State University graduate student workshop.
DISCLOSURE STATEMENT
No potential conflict of interest was reported by the authors.
Notes
1. Also known as ‘creating county-level city’ or Che Xian She Shi.
2. Also known as a ‘city–county merger’, ‘city–county amalgamation’ and Che Xian She Qu.
3. According to Chinese law, land is state owned. So-called ‘land selling’ in China is essentially equivalent to long-term land leasing.
4. There are many city types including province-level municipality, deputy-provincial city, prefecture-level city and county-level city. Prefecture-level cities are the most commonly used spatial unit for those interested in Chinese regions. Their data are easier to obtain, and in our case prefectures have both urban and rural areas that facilitate our analysis. There are nearly 300 prefecture-level cities. A typical prefecture-level city includes an urban core and several counties. Counties often include large rural areas whose economies are not necessarily highly integrated with the prefecture-level city’s urban core (Dingel et al., Citation2021).
5. There used to be a big difference between a ‘prefecture-level city’ and a ‘prefecture’. But this difference is smaller today. See Bo (Citation2020) for details.
6. The Hukou policy remains active, but it is now easier for rural residents to obtain ‘urban Hukou’ in less-populated cities.
7. County-level cities are a special form of county. They typically have better economic conditions than regular counties.
8. This policy is named ‘consolidation’ because (1) it increases the prefecture city’s land area, population and fiscal resources which are directly under the control of the prefecture city’s government; (2) there is an integration process during which rural land, residents and fiscal resources are ‘consolidated’ into the urban core; and (3) ‘consolidation’ is the literal translation from Chinese He Bing of Shi Xian He Bing.
9. This total number includes those that happened in four provincial-level municipalities of Beijing, Shanghai, Tianjin and Chongqing, and prefecture autonomous regions. However, given that the provincial-level cities enjoy various special fiscal and financial support from the central government, and prefecture autonomous regions are a minority entity which has more legislative rights and higher shares of a particular minority ethnic group, those cases are excluded from the empirical analysis.
10. For example, Gaoling county in Xi’An city experienced consolidation in 2015. The main advantage of being designated as an urban district is the increase of upper level government payment for urban construction. Before Gaoling was consolidated into the urban centre, the Xi’An government provided up to about 1 billion RMB urban construction investment to Yanta district (one of Xi’An urban districts). However, it only offered less than 1 million RMB to Gaoling. Thus, after the consolidation, the infrastructure construction in Gaoling could be directly guided by Xi’An’s government. Another example is Xindu county in Chengdu city. Xindu county is between two urban districts of Chengdu city. Due to its geographical location, Xindu county was consolidated into Chengdu’s urban centre in 2001. After consolidation, Chengdu’s government spent significant transportation expenditures in Xindu district to promote its economic growth.
11. In the empirical analysis, per capita GDP is adjusted by the annual national GDP deflator to account for inflationary effects. However, we do not have a county GDP deflator to address local price-level differentials, nor do we account for other welfare effects such as congestion or pollution, meaning the results should be interpreted in this context. Yet, Chinese governments almost entirely stressed GDP as their preferred welfare outcome during the peak-consolidation period, meaning we evaluate counties using the measure stressed by various governmental levels during the sample period. To be sure, this problem is mitigated by our use of propensity score matching to identify similar rural control counties that also have a similar cost of living, so that our comparisons are more of a ‘apples to apples’ cost-of-living comparison than at first glance.
12. Consolidations between 2000 and 2006 are selected for three reasons. (1) After 2008, the central government introduced another policy called ‘province-managed county’ (PMC). The PMC policy shifted rural county management away from the prefecture to the provincial government. Thus, prefecture governments lost control of those counties and any associated fiscal revenue. To avoid losing fiscal revenue, prefecture governments could use city–county consolidations to convert some rural counties into urban districts, or the converted rural counties will not become PMC. Therefore, the main rationales for post-2008 city–county consolidations have changed, which is why we omit post-2008 consolidations. (2) Post-2008 consolidations mainly occurred in 2013 and 2014, while our available data are from 1997 to 2016. Thus, the data period is not lengthy enough to assess the longer term effects of those later consolidations. (3) No consolidations occurred in 2007 or 2008, which further supports our emphasis on consolidations between 2000 and 2006. In the case where a prefecture experienced more than one consolidation during our study period, its treatment starts the first time a consolidation occurred.
13. We omit four provincial-level municipalities: Beijing, Shanghai, Tianjin and Chongqing (they are also called ‘cities’, but are administratively equivalent to provinces), prefecture autonomous regions and regions that experienced consolidation after 2008. We also use a linear interpolation to replace missing values.
14. The total area of each county is not strictly time invariant because of occasional village-jurisdiction adjustments.
15. Because of county-level data availability, other measures of public service quality are unavailable. However, these two variables are often used as explanatory variables in Chinese studies of subnational growth.
16. To assess whether there are changing patterns more than 10 years after consolidation that obscure our findings, we next added indicators for: exactly 10 years after treatment, exactly 11 years after treatment and so on up to 15 years after treatment. Note that each additional year-indicator beyond 10 years has fewer and fewer observations given the sample period ends in 2016, meaning those coefficients are increasingly estimated less precisely. Nonetheless, these results indicate that the individual indicator coefficients are not statistically different. The only indicators that are insignificantly different from zero are the 12th- and 13th-year indicators. There is no clear trend in the coefficients after 10 years, though the 15th-year indicator’s coefficient was less than the 14th-year coefficient and virtually identical to the 10th-year coefficient – indicating no clear ‘U’-shape pattern in which the estimated effects eventually trend to zero in some year after 10 years. The results are available from the authors upon request.
17. The reasons to include ‘deputy-prefecture cities’ are (1) in China’s local governance structure, namely province–prefecture–district/county system, ‘deputy-provincial cities’ are administrated by provincial governments, the same as regular ‘prefecture-level cities’; (2) within a province, ‘deputy-provincial cities’ are often treated identically as regular ‘prefecture-level cities’, but with a higher ranking among all sub-province administrative units; and (3) geographically, ‘deputy-provincial cities’ are the same as regular ‘prefecture-level cities’, which all take on many of the same governance functions as US counties.
18. The industrial sector includes manufacturing, mining and utilities.
19. The specification with leads and lags were also estimated for these alternative models. All lead coefficients are insignificant, which indicates the pre-trend assumption holds. The results are available from the authors upon request.
20. The main reason for insignificance might be the population data used, because in China the residential population at the county level is not provided annually. To fit our dataset structure, we use the household registration population, which takes longer to reflect population change. That is why we obtain the negative but insignificant result.
21. We cannot fully replicate Tang and Hewing’s results because of data availability constraints.
22. Because prefecture-level sample sizes are too small, LASSO does not converge. Thus, we do not use LASSO for prefecture-level analysis.
23. A graph for the Goodman–Bacon decomposition without controls is shown in Figure A6 in the supplemental data online. The current Stata command to illustrate the Goodman–Bacon decomposition results only works when there are no weight or adjustment of control variables. Thus, we discuss the corresponding DiD specification without controls.