Financing Road Projects and Its Impact on Off-Farm Work

doi:10.21203/rs.3.rs-4838536/v1

Download PDF

Research Article

Financing Road Projects and Its Impact on Off-Farm Work

https://doi.org/10.21203/rs.3.rs-4838536/v1

This work is licensed under a CC BY 4.0 License

You are reading this latest preprint version

Since 2003, China has revamped its rural road construction policy, transitioning from "workers self-organised construction (Min Gong Jian Qin)" to a "one project, one discussion (Yi Shi Yi Yi)" approach and significantly increasing government investment. This study evaluates the impact of various financing sources—externally funded, self-financed, and mix financed—on rural off-farm income and migration intentions using the Chinese Household Income Project survey data, employing difference-in-differences and propensity score matching methodologies to analyze the socioeconomic impacts of rural road projects financed through different sources on off-farm work income and migration intentions of the rural Chinese population. Findings reveal that externally funded road projects significantly enhance off-farm income and encourage migration intentions for work. Conversely, self-financed projects that rely on local villagers' contributions and voluntary labor, show no positive impact on off-farm income and unexpectedly reduce migration intentions. Importantly, projects with mixed financing models were found to have no significant impact on either off-farm income or migration patterns, highlighting challenges in the coordination and efficacy of these hybrid financing strategies. These findings underscore the importance of continued investment in rural transport infrastructure, with an emphasis on poverty reduction and rural out-migration through diverse financing models for road construction. This research provides insights for policymakers and stakeholders aiming to enhance rural infrastructure and development through informed transport policy initiatives.

JEL codes: O18, J61, R42

rural road

financing sources

off-farm income

Chinese Household Income Project

PSM-DID

Transport infrastructure is a necessary means of mobility to access various services, opportunities, and social activities. The impacts of transport infrastructure are closely related to socio-economic and environmental issues such as social exclusion (Lucas, 2012), mobility of resources (Yu et al., 2013), productivity gains (Deng, 2013), territorial accessibility (López et al., 2008), regional trade (Lakshmanan, 2011), employment opportunities (McArthur et al., 2014), emission reduction (Creutzig et al., 2015), and decarbonization (de Blas et al., 2020).

One way to estimate the impact of transport infrastructure more explicitly is by investigating the impacts of transport infrastructure at various spatial levels, i.e., regional, national, and international (Rietveld & Bruinsma, 2012). Discussions on the impact of transport infrastructure at the regional level can be further divided into urban and rural transport infrastructure or inter-and intra-regional transport infrastructure. Most prior studies focused on the effects of large-scale infrastructure systems (e.g., inter-regional highways, railroads that connect distant markets across regions) on the spatial distribution of economic activity and aggregate macro outcomes such as GDP growth and income inequality (Banerjee et al., 2020; Démurger, 2001). In contrast to large-scale transport infrastructure, Aggarwal (2018) provided evidence on the impact of roads on a wide variety of economic outcomes in rural India. Jacoby and Minten (2009) indicated that there may be potentially important complementarities between rural road construction and urban economic development in Madagascar. The effects of small-scale transport infrastructure systems such as rural or local intra-regional roads, which are essential for spurring local growth and development (i.e., rural or local intra-regional roads are closely related to the mobility of resources, productivity gains, and territorial accessibility) are not as well understood. Moreover, billions of people around the world still do not have access to or cannot afford transportation services (Mikou et al., 2019). In many developing countries, the vast majority of rural areas remain disconnected from paved road networks (Transport & ICT, 2016).

One of the representatives of developing countries, China, has achieved remarkable economic growth in the past three decades, although this has been accompanied by urban-rural disparity. With the economic development, the intention of the central government to alleviate the disparity between urban and rural areas has started to drive many policies to focus on rural China. For instance, under the Eleventh Five-Year Plan (2006–2010), the State Council approved a Rural Road Construction Plan which represents the first national infrastructure construction plan for rural areas. This period marks the most investment in the history of China‘s rural road development. The Chinese government invested 197.8 billion yuan in the construction of rural roads to encourage the growth and development of rural areas, which resulted in a 30% annual growth rate for the network of rural roads between 2006 and 2010 (Researchinchina, 2011). Then, since 2010, the government has invested more than $100 billion annually in village road projects, which have essentially reached every rural household in the nation. By clarifying the role of rural roads in regional economy, we can gain more insight into the economic consequences of rural road infrastructure as well as implications for developing countries to narrow the urban-rural inequality and escape from the middle-income trap.

Rural road construction is an important link to road construction in China to connect people who cannot afford the cost of mobility, such as those with the lowest education level and income and previously did not benefit from highway construction. Existing research is largely supportive of policymakers’ claims: rural road construction is associated with increases in farm and off-farm economic growth as well as poverty reduction. Qin and Zhang (2016) found that better road access in natural villages led to agricultural specialization, increased investment in fertilizers and labor, and ultimately improved household farm income. Qiao et al. (2014) estimated the impact of road expansion on off-farm activities in rural China. Using individual data, their estimation results show that road improvement increases the income of local off-farm workers. However, there is limited empirical work on how best to finance and manage village road projects. Although Wong et al. (2013) discovered that when higher level governments finance a significant portion of road improvements rather than when the villages finance more, this leds to higher quality roads, they mainly focused on project quality and cost, overlooking other direct welfare impacts such as employment opportunities. Therefore, additional empirical research is necessary to identify the most effective financing and management strategies for village road projects, while considering the direct welfare impacts they can provide, such as income growth and employment opportunities.

To further explore the impact of rural road projects on off-farm work in China, our study evaluates the different impacts of rural road projects by their funding sources on off-farm work in rural China. From 2006 to 2020, the China’s central government aggressively promoted road improvement projects in rural areas, but the funding sources varied from self-finance to external sources. To analyze the impact, we used the difference-in-differences (DID) approach and propensity score matching (PSM) to estimate rural residents’ monthly off-farm income and intention to migrate for work.

This study adds to the body of knowledge on estimating the impact of investments in transportation infrastructure. Studies on major large-scale transportation infrastructure have limited external validity for the rural roads that we study, which connect poor rural areas to regional markets. Additionally, the development of rural roads can potentially accelerate structural transformations in less developed regions, contributing to sustainable growth. Therefore, understanding the impact of investments in rural roads is essential. Our study finds that different types of rural road project financing affect rural residents differently, further influencing the spatial and sectoral distribution of labor.

The rest of the paper is organized as follows: Section 2 provides an overview of the rural road program and off-farm work in China. Section 3 describes our data construction and empirical strategy. Section 4 presents the results, while Section 5 discusses potential mechanisms. Finally, Section 6 concludes the paper.

2.1 Rural road development in China

The Ministry of Transport of China generally defines rural roads as including county roads, township roads and village roads (MOTC, 2006a). County, township, and village roads¹ in China respectively serve as main connectors between the county seat and towns, production centers, and inter-county routes; within and between townships for economic and administrative services; and between villages and external networks (MOTC, 2006b).

In 1978, there were only 586,000 km of rural roads in China, and many towns and villages were not accessible by road. By the end of 2002, China's rural roads reached 1.337 million kilometers. In 2003, following the central government's plan to build a “new socialist countryside,” the Ministry of Transport of China put forward the slogan of "repairing rural roads, serving urbanization, and putting farmers on cement roads", and began to implement the project of "connecting villages in the eastern region, townships in the central region, and counties in the western region". In 2005, the General Office of the State Council issued the "Rural Road Management and Maintenance System Reform Program"; in the same year, the State Council adopted the "The Rural Road Construction Plan (RRCP)", putting forward medium- and long-term rural road development goals. The medium-term goal was threefold: All townships (towns) were to have access to asphalt (cement) roads by the end of 2010; incorporated villages in the eastern and central regions would have access to asphalt (cement) roads; and incorporated villages in the western region would have general road access. By 2010, the country's rural road mileage reached 3.1 million kilometers. The long-term goals were for the incorporated villages to have access to asphalt (cement) roads and for the country's rural road mileage to reach 3.7 million kilometers by 2020, comprehensively improving the density and condition of rural roads and, as a result, service levels.

From 2006, China began the development phase of the Eleventh Five-Year Plan, and the Ministry of Transport established he "five-year 100 billion yuan" project to be implemented by regional governments. Thus, China's rural road construction entered a new period of rapid development of the largest scale in history. The Eleventh Five-Year Plan was a period that marked the most investment in the history of China‘s rural road development. In the five years, the central government’s investment in rural road construction funds amounted to 197.8 billion yuan, with an average annual increase of 30% (Researchinchina, 2011). Central government investment has greatly driven local investment in rural road construction, and a total of 950 billion yuan was invested during the five-year period toward the reconstruction of 1.868 million kilometers of rural roads, including 527,000 kilometers of new rural roads. The total mileage of rural roads reached 3.45 million kilometers.

2.2 Finance of rural roads in China’s village

China has seen significant rural road development over the past two decades, as the government implements policies to improve connectivity and reduce poverty in rural areas. The main sources of rural road provision in China can be divided into external funding and farmer self-financing, as shown in Fig. 1.

Higher levels of government play an important role in the external financing of rural infrastructure projects (Yan et al., 2004). Since 2003, China's investment policy has been adjusted and the "workers self-organised construction of rural roads (Min Gong Jian Qin)" has been cancelled in favour of the "one project, one discussion (Yi Shi Yi Yi)" system, while central and local governments have increased their investment in rural road construction through vehicle purchase taxes, national debt funds, road maintenance fees and other general financial resources (Wang, 2011).

On the other hand, villages in China are self-governed by village committees, which are staffed entirely by local villagers. These village committees are collectively responsible for managing village affairs (Wong et al., 2013). In 1987, local elections were introduced in villages of China (Oi & Rozelle, 2000). Therefore, village leaders and committees play an important role in the self-financing of rural infrastructure projects by farmers (Liu et al., 2009). This self-financing often involves funds raised directly from the farmers, their voluntary labor contributions, and investments from local enterprises. The combination of these resources allows rural communities to finance and develop infrastructure projects on their own terms, reducing dependency on external sources and creating a sense of collective ownership and responsibility among the residents.

2.3 Off-farm work

Depending on the type of off-farm employment opportunities, off-farm work can be divided into local off-farm work and migrant work (Qiao et al., 2014). Off-farm work is a significant source of household income in many countries, helping to achieve economic equality between rural and urban populations, reduce poverty, and ensure food security in rural regions (Chang & Mishra, 2008; De Janvry & Sadoulet, 2001; Zereyesus et al., 2017). In the context of China, the significant role of off-farm work in rural household welfare and regional development has also been amply documented (Ma et al., 2018). The development of off-farm labor markets can be seen as one of a number of possible mechanisms to help address this urban-rural gap by providing rural households with opportunities to improve their incomes.

Despite the significant contributions of off-farm work, there has less attempts thus far to examine the nexus between rural road construction and off-farm work participation in China. Fan et al. (2004) found that road density had a positive impact on farmers' participation in off-farm work. Qiao et al. (2014) suggested that the road expansion encouraged farmers to participate in local off-farm work, rather than migrate.

3.1 Data

As the Rural Road Construction Plan (RRCP) was implemented during 2006–2020, a data set for this period is approperiate for our research. To analyze the impact of rural road projects financed by different sources on off-farm work, we must research the dynamic process of the residents' income, which means that we should utilize a panel data set. For this study, we chose the rural sample of the Chinese Household Income Project (CHIP), which are provided by the China Institute for Income Distribution. The CHIP survey has been conducted in 1989, 1996, 2003, 2008, 2014, and 2019 respectively, and these basic survey data formed a database of 1988, 1995, 2002, 2007/2008, 2013, and 2018.

We use individual-level and village-level data from the CHIP2007 and CHIP2008 rural household survey databases. This data set forms a representative sample of rural households for each year, using 10 families each from 800 rural communities across the nine provinces of Hebei, Jiangsu, Zhejiang, Anhui, Henan, Hubei, Guangdong, Chongqing, and Sichuan. We can compose a panel data set of these two years of sample data from CHIP2007 and CHIP2008.

In CHIP datasets, information on employment is only available for those over the age of 16. Therefore, we selected all individuals with off-farm work information as our sample. After combining the village-level data with the individual data and eliminating invalid data, our database contained a total of 20,791 observations (10,196 individuals from CHIP2007 and 10,595 individuals from CHIP2008). In this study, we constructed our dependent variables based on the information from the CHIP individual data. There are two dependent variables in this study. First, income from off-farm work was calculated based on the repondents’ answer to the questionnaire item "What is the average monthly total income from all paid off-farm work (including self-employment)?”. Second, intention to migrate out for work was measured based on the repondents’ answer to the questionnaire item "Are you planning to migrate out for work or business in the near future?”.

In the village-level data from CHIP datasets, they reported whether each village had road improvement projects in 2005, 2006, 2007, and 2008. In addition, they documented the specific source of funding for road improvement projects in each village in 2007 and 2008. Moreover, they divided the source of project funds into self-financing by the village and external funds (higher level governments’ grants and other funds). Based on the above information, in our study we classified whether the village had road improvement projects into the following four categories to construct our treatment variables: 1) received no fund of road improvement projects; 2) had road improvement projects with project funding from external funding only; 3) had road improvement projects with project funding from external funding and village self-financing; and 4) had road improvement projects with project funding from village self-financing only.

Intuitively, several factors affect the rural residents’ off-farm work; thus, the features of the individual and village need to be controlled. The variables for individual-level control include gender, level of education, age, ethnicity, health status, marital status, and household size. In rural areas, residents over 16 years old are the primary decision makers, so their demographics should greatly affect their off-farm work decisions. This is the theoretical basis behind our controlling for these factors in our empirical model.

We also control several variables for the village, which include the population, percentage share of the labor force working in the agriculture sector in the village, average income level, the money earned in a day as a temporary worker in the village, whether a natural disaster occurred in the respective year and the percentage of losses caused by the disasters compared to the previous year's agricultural production, and the distance from the village to the nearest township government. As a society that relies heavily on social capital, the village population can reflect the likelihood that a person will have access to acquaintances to help with off-farm work. The percentage of the village labor force working in the agricultural sector reflects the structure of the rural economy and whether residents are dependent on agricultural work. The distance from the village to the nearest township government reflects the feasibility and difficulty of going to the township for off-farm work. The money earned in a day as a temporary worker in the village provides an indication of the degree of market development in the village. The average income level of the village reflects the state of the economic development of households in the community, while the occurrence of natural disasters temproralily affects the income.

As shown in Table 1, we divided the villagers into a treatment group and a control group depending on whether they belonged to the villages with specific financed road projects in 2007. For example, those belonging to the villages with externally funded road projects were placed in the treatment group, and those without were placed in the control group. The average monthly off-farm income of rural residents in the treatment group in 2007 was 1,363 yuan, while that of the control group was 1,397 yuan. In the 2007 group, 660 rural residents belonged to the treatment group and 7,686 belonged to the control group, and the treatment group accounted for 7.9% of the total. In the 2008 group, the rural residents were also divided into treatment and control groups on the basis of whether they belonged to a villages with externally funded road projects in 2007. The average monthly off-farm income of rural residents in the treatment group in 2008 increased to an average of 1,499 yuan, while that of the control group increased to an average of 1,487 yuan. The treatment group contained 631 rural residents, and the control group contained 8,035, with the treatment group accounting for 7.3%.

Table 1
Descriptive statistics for monthly off-farm income and intention to migrate for work.
Variables	2007		2008
	Treated	Control	Treated	Control
Externally funded road projects
Off-farm income	1363	1397	1499	1487
Off-farm income (logarithm)	7.040	7.044	7.136	7.085
Observation	660	7,686	631	8,035
Intention to migrate for work	0.257	0.254	0.227	0.187
Observation	447	5,293	401	5,019
Self financed road projects
Off-farm income	1456	1382	1576	1471
Off-farm income (logarithm)	7.029	7.047	7.147	7.077
Observation	1,326	7,020	1,427	7,239
Intention to migrate for work	0.170	0.272	0.124	0.203
Observation	1,010	4,730	917	4,503
Mix financed road Project
Off-farm income	1366	1407	1407	1526
Off-farm income (logarithm)	7.040	7.046	7.068	7.098
Observation	2,676	5,670	2,786	5,880
Intention to migrate for work	0.246	0.258	0.204	0.183
Observation	1,838	3,902	1700	3,720
Source: CHIP2007/2008.
Note: The observations are different for the two years because the data are not balanced panel data and there are some missing observations and new observations.

Table 1 presents the descriptive statistics for the dependent variables in the treatment and control groups. The descriptive statistics for the other control variables are presented in Table 2. Two of the control variables, namely the average household income in the village and the distance from the village to the nearest township government, are ordinal variables. The survey question used to collect data on the average household income in the village was "In 2007, which of the following ranges did the per capita annual net income of farmers in this village fall into?" Respondents were presented with 19 response options, ranging from "Below 500 yuan" to "Above 20,000 yuan". The survey question used to collect data on the distance from the village to the nearest township government was "Based on the usual mode of travel, approximately how long does it take to travel from your village to the nearest seat of township government?" Respondents were provided with 5 response options, ranging from "Within 15 minutes" to "More than 90 minutes".

Table 2
Descriptive statistics of control variables.
Variables	Observation	Mean	SD	Minimum	Maximum
Individual level
Gender (male=1)	20787	.628	.483	0	1
Education	20791	4.282	4.626	0	31
Age	20775	36.044	12.254	17	97
Ethnicity (minorities=1)	20791	.008	.09	0	1
Health status (level)	20783	1.849	.703	1	5
Marriage states	20771	2.358	2.189	1	6
Household size	20791	4.489	1.454	1	18
Experience of migrated for work (yes=1)	20791	.618	.486	0	1
Village level
Average household income in the village (level)	20782	11.648	3.453	1	18
Population (logarithm)	20791	7.668	.642	5.209	10.535
Labor force working in agriculture sector (%)	20791	45.679	24.648	0	99
The money earned in a day as a casual worker in the village	20791	43.869	13.375	0	120
Distance from the village to the nearest township government (level)	20791	1.825	.81	1	5
Interaction term of natural disaster occurrence in the respective year and agricultural production lose (%)	20664	7.784	11.92	0	90
Source: CHIP2007/2008.

3.2 Econometric model specification

We specified the empirical model below:

$$\:{\:\:\:\:\:\:\:\:\:\varvec{Y}}_{\varvec{i},\varvec{c},\varvec{t}}={\varvec{\beta\:}}_{0}+{\varvec{\beta\:}}_{1}{\varvec{A}}_{\varvec{t}}+{\varvec{\beta\:}}_{2}{\varvec{D}}_{\varvec{r},\varvec{c}}+{\varvec{\beta\:}}_{3}\left({\varvec{A}}_{\varvec{t}}\cdot\:{\varvec{D}}_{\varvec{r},\varvec{c}}\right)+{\varvec{\beta\:}}_{4}^{{\prime\:}}{\varvec{z}}_{\varvec{i},\varvec{c},\varvec{t}}+{\varvec{\beta\:}}_{5}^{{\prime\:}}{\varvec{I}}_{\varvec{c},\varvec{t}}+{\varvec{v}}_{\varvec{p}}+\:{\varvec{e}}_{\varvec{i},\varvec{c},\varvec{t}}$$

for i = 1,..., N; t = 0,1; c = 0,…,V;and r = 1,2,3 (1)

The dependent variable Y_ict in Eq. (1) represents the monthly off-farm income or intention to migrate out for work for the ith rural resident in the cth village at time t. D_r,c is a dummy variable for the treatment and control groups, indicating whether rural resident i in village c received the rural road improvement project financed by specific fund r. When the sample belongs to the treatment group, this variable takes the value of 1. According to the CHIP database, we can classify D_r,c into three categories: D_1,c, the villages with only externally funded road improvement projects (EFRIP) were assigned a value of 1, and those without were assigned a value of 0. D_2,c, the villages with externally funds and self finance (mix financed) road improvement projects (MFRIP) were assigned a value of 1, and those without were assigned a value of 0. D_3,c, the villages with only self finance road improvement projects (SFRIP) were assigned a value of 1, and those without were assigned a value of 0. We chose to treat a particular funded road project as the treatment group and to place other categories in the control group, rather than including no road projects as a control group. The reason is to avoid selection bias.

Out of the 652 villages included in the CHIP2007², 52 received a rural road improvement project financed only by external funds in 2007, which is about 8%. This percentage was 3.5% and 3.9% in 2005 and 2006, respectively. 101 villages received a only self financed rural road improvement project in 2007, which is about 15.5%. This percentage was 28.1% and 24.0% in 2005 and 2006, respectively. 207 villages received a mix financed (self financed and externally funded) rural road improvement project in 2007, which is about 31.8%. This percentage was 20.9% and 22.5% in 2005 and 2006, respectively. 292 villages³ received no rural road improvement project or no funding for a rural road improvement project in 2007, which is about 44.8%. This percentage was 47.4% and 49.6% in 2005 and 2006, respectively. From 2005 to 2007, the percentage of villages with road improvement projects first decreased and then increased. Among them, the external funded road improvement projects got doubled, while the rate of villages with MFRIP had a significant increase. However, the percentage of villages with SFRIP decreased significantly. The increase of EFRIP and MFRIP in the two years was likely due to the implementation of the RRCP.

At is the year dummy variable. When the sample belongs to the follow-up period (2008), this variable takes the value of 1. In the final empirical analysis, we primarily focus on the coefficient of the interaction term Dr,c × At. This coefficient represents the effect of the implementation of the rural road improvement project. As our data contains only two time periods, under the model established in Eq. (1), we use ordinary least squares (OLS), fixed-effects (FE) and logit regressions models to obtain unbiased and consistent regression results. Table 2. shows the descriptive statistics for the important control variables. Z_ict and I_ct are the individual-level and village-level control variables mentioned previously.

In addition to the above control variables, we include a dummy variable vp for each observation's province in our model; this controls for the unobservable effects related to different provinces and administrative regions. In addition, e_ict is the error term in the model.

In our DID analysis, because we only have a two-year sample, we cannot test the parallel hypothesis using a general method, such as plotting the dependent variable before the treatment. Thus, we need another way to ensure the comparability between the treatment and control groups in the base period. We found a significant difference in the features of households and villages between the treatment group and control group. This may be due to specific differences between individuals and villages. It may also be due self-selected bias; after the implementation of the rural road improvement project, rural households could choose to remain in or leave their village.

To deal with this difference in features, we conduct a PSM-DID analysis, the results from which are as follows:

PSM - DID = E[ $\:{\varvec{Y}}_{1\varvec{i}\varvec{c}}^{\varvec{T}}-{\varvec{Y}}_{0\varvec{i}\varvec{c}}^{\varvec{T}}\left|\:\varvec{p}\right({\varvec{I}}_{0\varvec{c}}$ ), D _r,c =1]- E[$\:{\varvec{Y}}_{1\varvec{i}\varvec{c}}^{\varvec{C}}-{\varvec{Y}}_{0\varvec{i}\varvec{c}}^{\varvec{C}}\left|\:\varvec{p}\right({\varvec{I}}_{0\varvec{c}}$), D_r,c =0] (2)

Here, $\:{Y}_{1ic}^{T}\:$and $\:{Y}_{0ic}^{T}$ are the dependent variables of the treatment group in the base and follow-up periods, respectively; $\:{Y}_{1ic}^{C}$ and $\:{Y}_{0ic}^{C}$ are the dependent variables of the control group in base and follow-up periods, respectively. P($\:{I}_{0c}$) is the propensity score in the base period. The initial observed covariates from village level $\:{I}_{0c}$ in 2007 are used to predict the probability of whether the samples belong to the treatment group. We use logit models to conduct the regression and use the regression results to calculate the conditional probability of samples entering the treatment group; this conditional probability is the propensity score P($\:{I}_{0c}$). Once the propensity scores are estimated, we employ radius matching, to create a matched sample of treated and control units with similar propensity scores. With the matched sample in hand, we apply a DID analysis to estimate the treatment effect, incorporating the propensity score weights obtained from the matching.

Since CHIP2007 and CHIP2008 are unbalanced panel data sets⁴, we also use fixed effects DID estimates to exploit the panel data structure. we apply an within transformation as follows:

$$\:{\varvec{Y}}_{\varvec{i},\varvec{c},\varvec{t}}-{\stackrel{-}{\:\varvec{Y}}}_{\varvec{i},\varvec{c}}={\varvec{\beta\:}\varvec{{\prime\:}}}_{1}\left({\varvec{A}}_{\varvec{t}}-{\stackrel{-}{\varvec{A}}}_{\varvec{t}}\right)+{{\varvec{\beta\:}\varvec{{\prime\:}}}_{2}\left({\varvec{D}}_{\varvec{r},\varvec{c}}-{\stackrel{-}{\varvec{D}}}_{\varvec{r},\varvec{c}}\right)+\:\varvec{\beta\:}\varvec{{\prime\:}}}_{3}\left({\varvec{A}}_{\varvec{t}}\cdot\:{\varvec{D}}_{\varvec{r},\varvec{c}}-{\stackrel{-}{\varvec{A}}}_{\varvec{t}}\cdot\:{\stackrel{-}{\varvec{D}}}_{\varvec{r},\varvec{c}}\right)+{\varvec{\beta\:}}_{4}^{\varvec{{\prime\:}}\varvec{{\prime\:}}}\left({\varvec{z}\varvec{{\prime\:}}}_{\varvec{i},\varvec{c},\varvec{t}}-{\stackrel{-}{\varvec{z}\varvec{{\prime\:}}}}_{\varvec{i},\varvec{c}}\right)+{\varvec{\beta\:}}_{5}^{\varvec{{\prime\:}}\varvec{{\prime\:}}}\left({\varvec{I}\varvec{{\prime\:}}}_{\varvec{c},\varvec{t}}-{\stackrel{-}{\varvec{I}\varvec{{\prime\:}}}}_{\varvec{c}}\right)+\:{(\varvec{e}}_{\varvec{i},\varvec{c},\varvec{t}}-{\stackrel{-}{\varvec{e}}}_{\varvec{i},\varvec{c}})$$

Where Z$\:\varvec{{\prime\:}}$_ict and I$\:\varvec{{\prime\:}}$_ct are the time-varying individual-level and village-level control variables, respectively.$\:\:{\stackrel{-}{\:Y}}_{i,c}$, $\:{\stackrel{-}{A}}_{t}$, $\:{\stackrel{-}{D}}_{r,c}$, $\:{\stackrel{-}{z{\prime\:}}}_{i,c}$, $\:{\stackrel{-}{I{\prime\:}}}_{c}$ and $\:{\stackrel{-}{e}}_{i,c}$ are the time means of the respective variables for individual i in village c. Eq. (3) could control for unobserved time-invariant individual-specific characteristics, and common shocks that are assumed to be derived mainly from macroeconomic policies and institutional reforms.

4.1 Characteristics of rural road improvement projects

Based on the data available in the CHIP2007 database, we divided the sample villages into four different categories based on their rural road improvement projects. Table 3 presents an overview of the project and village characteristics for each of the four subgroups, as summarized by various descriptive statistics. In terms of the size of rural road improvement project funding in 2007, the table indicate that villages receiving only EFRIP had the highest average funding of 275,936.5 yuan. Villages that received a MFRIP had an average funding of 388,568.6 yuan, whereas villages that relied exclusively on SFRIP had the lowest average funding of 186,818.9 yuan. The average population in the villages varies slightly across the four groups, with MFRIP villages having the highest average population (2718) and villages with no road projects having the lowest (2211). The villages with SFRIP have the highest average household income level of 11.62, while villages with EFRIP have the lowest average household income level of 10.44⁵. The percentage of labor force working in agriculture sector is highest in villages with no road projects (52.8%) and lowest in EFRIP villages (44.6%). The table also shows that the average distance from the village to the nearest township government is slightly longer for SFRIP villages (2.05) compared to the other three groups. Finally, the percentage of agricultural production lost due to natural disasters is highest in EFRIP villages (16.6%) and lowest in villages with no road projects (9.4%).

Table 3
Characteristics of three type of rural road improvement projects and the features of the villages in which they are implemented.
Rural road improvement project	EFRIP	MFRIP	SFRIP	No projects
No. of villages	52	101	207	292
Average funding (Yuan)	275,936.5	388,568.6	186,818.9	N/A
Average population (People) in the village	2428	2718	2372	2211
Average household income in the village (level)	10.44	10.97	11.62	10.89
Average labor force working in agriculture sector (%) in the village	44.6	49.8	49.3	52.8
Average money earned in a day as a casual worker in the village (Yuan)	43.4	43.3	39.8	43.2
Average distance from the village to the nearest township government (level)	1.93	1.91	2.05	1.81
The percentage of agricultural production lost due to the occurrence of natural disasters in the village compared to a normal year (%)	16.6	12	10.4	9.4
Source: CHIP2007.

4.2 Appropriatenss test for PSM

As mentioned before, we use PSM-DID to ensure comparability between the treatment and control groups. However, two conditions must be satisfied before we use PSM. First, the propensity scores of the treatment and control groups should have sufficient common support. Most of the observations in our analysis are in common support, which minimizes the loss of observations after matching. Second, there should not be an obvious difference between the control variables in the treatment and control groups. Table 4 shows the results for the difference of control variables after matching. There were no obvious differences for most control variables, which means that we achieved our goal for PSM. Our estimates on SFRIP and MFRIP also satisfy the above two conditions.

Table 4
Balanced test for unmatched and matched samples after radius matching (caliper=0.006). (externally funded road projects)
Variables	Unmatched (Mean)		Matched(Mean)		t-Test
	Treated	Control	Treated	Control	t-value	P-value
Average household income in the village (level)	11.442	11.891	11.619	11.549	0.51	0.612
Population (logarithm)	7.604	7.644	7.596	7.593	0.12	0.902
Labor force working in agriculture sector (%)	43.232	45.194	42.974	46.427	-3.63	0.000
The money earned in a day as a casual worker in the village	41.895	44.206	41.952	42.368	-0.78	0.433
Distance from the village to the nearest township government (level)	1.911	1.787	1.872	1.871	0.00	0.998
Interaction term of Natural disaster occurrence in the respective year and agricultural production lose (%)	11.182	7.294	10.42	10.442	-0.04	0.976

4.3 Main results for the EFRIP

The results obtained from the DID estimation method are displayed in Table 5. It should be noted that all standard errors have been corrected for heteroskedasticity, and provincial dummy variables were included in all models. Columns (2), (4), (6), and (8) exclude all villages with road improvement projects in 2006. Controlling for the previous year's road projects helps to minimize the impact of ongoing or uncompleted projects that could have confounded our treatment effects. We strategically adopted this approach to mitigate potential distortions that could arise from the influence of prior-year road projects. Given that the CHIP 2007/2008 dataset does not provide specific completion dates for each road improvement project, we opted to exclude villages with 2006 road projects from this subset of our analysis. This approach complements our baseline analysis which includes all villages, thereby providing a more comprehensive understanding of the effects under investigation.

Table 5
Impact of externally funded road improvement project on off-farm work.
	DID estimations				PSM-DID estimations (radius matching)
	(1)	(2)	(3)	(4)	(5)	(6)	(7)	(8)
	OLS		Logit		OLS		Logit
Dependent variable	Monthly off-farm income (logarithm)		Intention to migrate for work (dummy variable)		Monthly off-farm income (logarithm)		Intention to migrate for work (dummy variable)
Year	.176***	.166***	-.26**	-.259	.049	.158*	-.081	-.333
	(.022)	(.032)	(.125)	(.188)	(.041)	(.084)	(.234)	(.329)
Road7E	.009	-.033	-.341**	-.448*	-.023	-.040	-.315*	-.677**
	(.033)	(.069)	(.146)	(.243)	(.037)	(.085)	(.165)	(.287)
Year_Road7E	.048	.183**	-.015	.445	.076*	.202**	-.037	.899**
	(.039)	(.07)	(.195)	(.368)	(.042)	(.079)	(.234)	(.423)
Individual	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes
Village	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes
Province	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes
_cons	6.75***	6.67***	-1.57***	-2.67 ***	7.03***	6.31***	-.866	-7.52***
	(.095)	(.143)	(.551)	(.842)	(.169)	(.262)	(1.06)	(1.39)
Observations	16876	7404	11055	4624	16407	7182	10722	4461
R-squared	.114	.105			.110	.098
Pseudo R2			.409	.412			.403	.431
Standard errors are in parentheses
* p<.01, p<.05, * p<.1

The OLS results in column (1) suggest that EFRIP had no significant effect on rural residents' monthly off-farm income. However, after accounting for the previous year's road improvement projects in column (2), a significant effect of the EFRIP on the monthly off-farm income of rural residents can be observed at a 5% significance level. Specifically, the monthly off-farm income was 18% higher than it would have been without the EFRIP. The logit regression results in columns (3) and (4) indicate that the EFRIP did not significantly influence rural residents' intention to migrate for work. Table 5 also presents the results of the PSM-DID estimations, which align with the DID estimations. Column (6) shows a positive and significant effect of the EFRIP on rural residents' monthly off-farm income after controlling for the previous year's road improvement projects. However, column (8) reveals that the EFRIP has a positive impact on rural residents' intention to migrate for work at a 5% significance level.

Table 6
Fixed effect DID estimates of the effects of rural road improvement projects.
Dependent variables	(1)	(2)	(3)	(4)
Dependent variables	Monthly off-farm income (logarithm)		Intention to migrate for work (dummy variable)
Externally funded road improvement projects
Year_Road7E	.004	.125**	.006	.783
	(.028)	(.052)	(.378)	(.524)
Observations	16876	7404	868	430
Groups	10367	4663	434	215
R-squared	.008	.016
Pseudo R²			.274	.303
Self-financed road improvement projects
Year_Road7S	.077***	.117	-.17	-1.026
	(.028)	(.081)	(.326)	(.716)
Observations	16876	7404	868	430
Groups	10367	4663	434	215
R-squared	.01	.016
Pseudo R²			.274	.303
Mix-financed road improvement projects
Year_Road7M	-.021	.001	.222	.382
	(.019)	(.030)	(.251)	(.399)
Observations	16876	7404	868	430
Groups	10367	4663	434	215
R-squared	.009	.015
Pseudo R²			.275	.298
Standard errors are in parentheses
* p<.01, p<.05, * p<.1
Note: Column (2) to column (4) excluded all observations from villages with road improvement projects in 2006. R squared of FE within regression is derived from within-group R-square.

Table 6 displays the results of the fixed effects DID estimates. The coefficient of the treatment variable on monthly off-farm income has slightly decreased, while the coefficient of the treatment variable on the intention to migrate for work has also slightly decreased, and the degree of significance has changed. In the case of unbalanced panels, the difference between the OLS and fixed effects estimates should be considered as evidence that non-response is not ignorable for the DID estimation (Lechner et al., 2016). Based on the results, we find that the EFRIP will increase farmers' off-farm income without influencing their intention to migrate out for work. There's a possibility that certain individuals might not have participated in the CHIP2008 survey as they had already migrated for work. Furthermore, the increased off-farm income among the local populace could potentially attract individuals from less prosperous areas to the village in subsequent years, thereby decreasing their need to migrate further.

4.4 Main results for the SFRIP

Table 7 presents the estimated results of the SFRIP impact assessments. All standard errors were corrected for heteroskedasticity, and provincial dummy variables were included in all models. Columns (2), (4), (6), and (8) exclude all villages with road projects in 2006. The OLS results in column (1) indicate that the SFRIP had a positive and significant effect on rural residents' monthly off-farm income at a 1% significance level. However, after controlling for road improvement projects in the previous year in column (2), there is no significant effect of the SFRIP on rural residents' monthly off-farm income. The logit regression results in columns (3) and (4) reveal that the SFRIP did not significantly impact rural residents' intention to migrate for work. However, in the PSM-DID estimation of the SFRIP in column (8), the SFRIP had a negative and significant effect on rural residents' intention to migrate for work at a 5% significance level. The fixed effects DID estimates, which are consistent with the DID estimates, are reported in Table 6.

Table 7
Impact of self-financed road improvement project on off-farm work.
	DID estimations				PSM-DID estimations (radius matching)
	(1)	(2)	(3)	(4)	(5)	(6)	(7)	(8)
	OLS		Logit		OLS		Logit
Dependent variable	Monthly off-farm income (logarithm)		Intention to migrate for work (dummy variable)		Monthly off-farm income (logarithm)		Intention to migrate for work (dummy variable)
Year	.166***	.167***	-.284**	-.169	.178***	.176***	-.147	-.194
	(.023)	(.032)	(.127)	(.185)	(.035)	(.049)	(.178)	(.252)
Road7S	-.013	-.101*	-.385***	-.45*	-.017	-.085	-.214*	-.414*
	(.026)	(.052)	(.119)	(.248)	(.028)	(.056)	(.125)	(.237)
Year_Road7S	.082***	.074	.02	-.626	.078***	.060	-.145	-.831**
	(.031)	(.075)	(.18)	(.381)	(.034)	(.080)	(.194)	(.376)
Individual	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes
Village	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes
Province	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes
_cons	6.74***	6.66***	-1.36**	-2.64***	6.82***	6.75***	-.771	-1.75
	(.095)	(.142)	(.551)	(.842	(.131)	(.225)	(.791)	(1.11)
Observations	16876	7404	11055	4624	16361	7140	10629	4425
R-squared	.114	.105			.112	.093
Pseudo R2			.410	.414			.438	.435
Standard errors are in parentheses
* p<.01, p<.05, * p<.1

4.5 Main results for the MFRIP

The Table 8 and Table 6 display the results of the impact estimates for the MFRIP. It is important to note that all standard errors were corrected for heteroskedasticity, and provincial dummy variables were included in all models in Table 8. The DID and PSM-DID estimations demonstrate consistent results. The OLS regression results in columns (1), (2), (5), and (6) suggest that the MFRIP had no significant effect on rural residents' monthly off-farm income. The logit regression results in columns (3) and (7) reveal that the MFRIP did not have a positive and significant effect on rural residents' intention to migrate for work. After controlling for the road improvement projects in the previous year, no significant effect of the MFRIP on rural residents' intention to migrate for work can be observed in columns (4) and (8). Furthermore, the results of the fixed effects DID estimates are consistent with the DID estimates in Table 6.

Table 8
Impact of Mix-financed road improvement project on off-farm work.
	DID estimations				PSM-DID estimations (radius matching)
	(1)	(2)	(3)	(4)	(5)	(6)	(7)	(8)
	OLS		Logit		OLS		Logit
Dependent variable	Monthly off-farm income (logarithm)		Intention to migrate for work (dummy variable)		Monthly off-farm income (logarithm)		Intention to migrate for work (dummy variable)
Year	.187***	.166***	-.32**	-.311	.188***	.159***	-.336**	-.353
	(.023)	(.033)	(.131)	(.196)	(.025)	(.036)	(.145)	(.216)
Road7M	-.015	-.009	-.154*	-.166	-.021	-.017	-.175*	-.218
	(.018)	(.030)	(.087)	(.133)	(.018)	(.030)	(.090)	(.138)
Year_Road7M	-.023	.037	.212	.323	-.010	.049	.155	.254
	(.023)	(.037)	(.13)	(.212)	(.023)	(.037)	(.135)	(.222)
Individual	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes
Village	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes
Province	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes
_cons	6.73***	6.65***	-1.51***	-2.59***	6.70***	6.37***	-1.01	-2.22**
	(.094)	(.141)	(.55)	(.842)	(.111)	(.182)	(.645)	(1.01)
Observations	16876	7404	11055	4624	16587	7278	10841	4522
R-squared	.114	.104			.117	.107
Pseudo R2			.401	.412			.409	.438
Standard errors are in parentheses
* p<.01, p<.05, * p<.1

Our empirical results from the PSM-DID estimation, controlling for the previous year’s road construction, reveal that, first, externally funded road projects contribute to an increase in rural residents' off-farm income and their intention to migrate out for work. Second, self-financed projects do not have a significant impact on rural residents' off-farm income, but they do have a negative effect on rural residents' intention to migrate out for work. Third, the combination of self-financing and external funds for road improvement projects appears to have no impact on rural residents' off-farm income or their intention to migrate out for work. In this session, we aim to explore the reasons behind these observations as shown in Fig. 2.

The improvement of roads enhances the connectivity of villages. According to the type of rural roads, the funding sources for county and township roads include external funds and self-financing by farmers, while village roads are solely financed through self-financing by farmers (MEEC, 2004). County roads and township roads are generally built according to the construction standards of graded roads, while the construction standards of village roads are determined according to the actual local needs and economic conditions (MOTC, 2006a). As a result, externally funded and mix-financed roads are more likely to be connected to county and town governments, linking villages to more developed areas. Therefore, based on the different financing sources of road improvement projects, we can further categorize the improved connectivity into two types: first, connectivity between villages and within villages; and second, connectivity from villages to towns or counties.

Moreover, we believe that the different types of improved connectivity affect off-farm work in different ways. First, the improved accessibility of rural roads reduces the mobility costs for rural workers between areas with off-farm job opportunities and rural regions (Huang et al., 2022). This reduction in labor migration resistance leads to a situation where off-farm work in other locations is referred to as "migrate out for off-farm work". Second, the improved accessibility of rural roads reduces transportation and transaction costs, futher attracting industrial consentration (Wen, 2004) and development (Redding & Turner, 2015). The growth of rural non-farm industries increases the likelihood that rural residents will participate in local off-farm work, a situation we refer to as "engaging in local off-farm work". Finally, the improved accessibility of rural roads promotes agricultural growth by increasing agricultural modernization. This increase in agricultural modernization leads to an increase in agricultural labor productivity (Qin & Zhang, 2016). In some cases, as agricultural labor productivity rises, surplus labor is displaced, causing the surplus labor to engage in off-farm work, either locally or by migrating out of the county. On the other hand, the increase in agricultural productivity may also increase the demand for agricultural labor, thus reducing the surplus labor.

Based on the above background, we put forth two potential explanations for our results: firstly, it is hypothesized that externally funded road projects possess superior quality and design compared to other rural road projects, which subsequently enhances connectivity and accessibility to improved economic opportunities. Secondly, we postulate that rural residents who benefit from externally funded road projects are more likely to migrate out for work, consequently increasing their off-farm income, as employment opportunities outside their villages tend to offer higher remuneration.

5.1 Potential superior quality and design

The positive impact of externally funded roads may be attributed to the potential superior quality and design of such projects. This can be understood within the institutional context of rural China's infrastructure provision. Higher levels of government in China, particularly at the county level, may prioritize the quality of infrastructure more than local village committees (Wong et al., 2013). This could result in externally funded projects having a better overall quality. Furthermore, disparities in access to technology and information between village committees and higher levels of government could contribute to these findings, as limited knowledge and experience in infrastructure design among village committees could hinder their ability to create effective projects.

5.2 Different off farm workplace

To investigate the mechanisms underlying the varying effects of differently financed rural road projects on off-farm work, we conducted additional analyses. These analyses employed the same DID framework that was utilized in the preceding section. All observations are from villages with no road improvement projects in 2006. We constructed three dependent variables: average weekly work hours, intention of seek better paying off-farm jobs (yes=1), and off-farm work location (dummy variable, outside county=1) to examine potential mechanisms. It's important to note that the sample for these additional analyses was only include individuals who are engaged in off-farm work. Similarly, the average weekly work hours recorded encapsulate only the time dedicated to off-farm work. The data concerning off-farm work location provides information only about respondents' current job location. However, individuals who commute to work outside their home county and those who may be seasonal migrants are unclear in CHIP 2007/2008.

Table 9
The estimations of the potential mechanisms.
	(1)	(2)	(3)	(4)	(5)	(6)	(7)	(8)	(9)
	OLS			Logit			Logit
Dependent variable	Average weekly working hours			Off-farm working location (outside county=1)			Looking for better paying off-farm jobs (Yes=1)
Year	-3.42***	-3.6***	-3.42***	.645***	.607***	.635***	-1.70***	-1.78***	-1.76***
	(.735)	(.758)	(.743)	(.154)	(.160)	(.153)	(.118)	(.122)	(.120)
Road7E	.554			.443**			.134
	(.954)			(.197)			(.198)
YearRoad7E	-.171			-.502*			-.153
	(1.23)			(.277)			(.244)
Road7M		-2.13***			-.259**			-.515***
		(.609)			(.108)			(.104)
YearRoad7M		.847			.066			.266**
		(.814)			(.152)			(.134)
Road7S			.265			-.168			-.249*
			(.932)			(.177)			(.143)
YearRoad7S			-.153			-.215			.480**
			(1.32)			(.259)			(.202)
Individual	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes
Village	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes
Province	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes
_cons	63.8***	64.2***	63.7***	-6.02***	-6.10***	-6.12***	1.85***	2.02***	1.86***
	(3.23)	(3.21)	(3.23)	(.652)	(.647)	(.650)	(.503)	(.507)	(.503)
Observations	7380	7380	7380	7335	7335	7335	6343	6343	6343
R-squared	.077	.079	.077				.136	.140	.137
Pseudo R2				.481	.481	.481
Standard errors are in parentheses
* p<.01, p<.05, * p<.1

The results in Table 9 show that, first, EFRIP had no significant effect on rural residents' average weekly work hours or their intention to seek better-paying off-farm jobs. However, EFRIP had a negative effect on rural residents' choice of off-farm work outside the county at the 10% significance level. Secondly, MFRIP had no significant effect on rural residents' average weekly work hours and their choice of off-farm work outside the county. However, MFRIP had a positive effect on their intention to seek better-paying off-farm jobs at the 5% significance level. Finally, SFRIP had no significant effect on rural residents' average weekly work hours, but a positive impact on their intention to seek better-paying off-farm jobs at the 5% significance level. SFRIP had no effect on rural residents' choice of off-farm work outside the county.

The results presented in Table 9 offer valuable insights into the diverse impacts of differently financed rural road projects on off-farm work. Our findings suggest that the source of financing for rural road projects may influence rural residents' off-farm work behavior differently. First, EFRIP does not significantly affect rural residents' average weekly off-farm work hours or their intention to seek better-paying off-farm jobs, but negatively influences their choice of off-farm work outside the county. This suggests that EFRIP facilitates shorter distance migration, allowing residents to pursue better jobs without necessitating long-distance relocation. Second, the positive impact of MFRIP and SFRIP on rural residents' intention to seek better-paying off-farm jobs, while not affecting their average weekly work hours or off-farm work location preferences outside the county, indicates that MFRIP and SFRIP may increase access to information on job opportunities, labor market conditions, and skill requirements. This enhanced access to information could lead rural residents to higher aspirations to explore better-paying jobs in both local and non-local labor markets. Notably, the negative impact of SFRIP on rural residents' migration intentions highlights its role in not only improving physical connectivity but also in fostering local employment opportunities and strengthening community bonds. This effect can be linked to the creation of better-paying local jobs, facilitated by enhanced access to markets and services through improved infrastructure. Consequently, SFRIP may contributes significantly to the development of social capital within rural communities, as it encourages both local employment and stronger social ties.

This study utilized microeconomic survey data and several empirical methods (DID analysis and PSM-DID) to analyze the impact of differently financed rural road projects on the off-farm work of rural residents in China. From our empirical results, we concluded as follows. First, the implementation of the EFRIP increases the off-farm income of the rural residents and their intention to migrate out for work. Second, the implementation of the SFRIP has no impact on the off-farm income of the rural residents, but it decreases rural residents intention to migrate out for work. Third, the implementation of the MFRIP has no impact on the off-farm income of the rural residents and their intention to migrate out for work.

There are three sets of possible explanations for our findings given the institutional environment regarding the infrastructure provision in rural China. First, higher levels of government in China (especially at the township and county level) may actually place more emphasis on the quality of infrastructure than local village committees (Wong et al., 2013). Differences in access to technology and information between village committees and higher levels of government may also have supported our findings. Compared to higher levels of government, village committees often have limited knowledge and experience in infrastructure design, so they may not know how to build a better infrastructure project that would enhance human resource allocation. In the context of mixed financed projects, inherent coordination and efficacy challenges arise from hybrid financing strategies. Second, differently financed rural roads may increase accessibility to various off-farm workplaces. Specifically, the EFRIP might improve off-farm employment opportunities outside the village but within the county, while the SFRIP might enhance local off-farm employment opportunities. Third, social capital plays a crucial role in the success of road projects. In rural China, where adherence to higher-level government instructions is common, social capital—comprising trust, networks, and norms among community members and between communities and government levels—significantly impacts infrastructure project effectiveness. Projects that effectively leverage social capital can boost community engagement and resource pooling. Conversely, a deficiency in social capital can amplify the challenges faced by self-financed and mixed financed projects, especially in aligning local expectations with the strategic objectives of higher-level government investments.

The results suggest that continued efforts to invest in transport infrastructure in rural areas are warranted. China's sustainable economic growth is directly linked to the development of rural areas. Road construction undoubtedly plays an important role in China's rural development, as our study shows. In order to reduce poverty and promote labor mobility in rural areas, it is important to use different types of financing models to build rural roads.

Our research has certain limitations that should be noted. Our study focuses only on the short-term impacts of rural road projects on off-farm income and intention to migrate, and thus the long-term effects of these projects remain uncertain. Given this limitation, future research should focus on both short-term and long-term effects of these projects will provide a more comprehensive understanding of their implications for rural development. In addition, as rural communities evolve and become increasingly aware of non-agricultural uses of available resources, identifying new markets and opportunities for off-farm work will be crucial for rural residents. Consequently, exploring reliable methods to estimate relationship between rural roads and emerging markets, and their potential impact on off-farm work should be essential next steps in this research area.

Author Contribution

L.B.: Conceptualization, Methodology, Data curation, Formal analysis, Resources, Writing – original draft, Writing –review & editing. M. K.: Writing – review & editing, Supervision. A. C.: Writing – review & editing, Supervision.

Data Availability

The data supporting the results of this study were obtained from the Chinese Household Income Program (CHIP) survey database.

This work was supported by the Support for Pioneering Research Initiated by the Next Generation of FLOuRISH Institute (Grant No. 3-7), Tokyo University of Agriculture and Technology (TUAT), granted by the Ministry of Education, Culture, Sports, Science and Technology (MEXT) Japan.

This work was original research that has not been published previously, and not under consideration for publication elsewhere. All the authors listed have approved the manuscript that is enclosed.

Aggarwal, S. (2018). Do rural roads create pathways out of poverty? Evidence from India. Journal of Development Economics, 133, 375-395. https://doi.org/10.1016/j.jdeveco.2018.01.004
Banerjee, A., Duflo, E., & Qian, N. (2020). On the road: Access to transportation infrastructure and economic growth in China. Journal of Development Economics, 145, 102442. https://doi.org/10.1016/j.jdeveco.2020.102442
Chang, H.-H., & Mishra, A. (2008). Impact of off-farm labor supply on food expenditures of the farm household. Food Policy, 33(6), 657-664. https://doi.org/10.1016/j.foodpol.2008.02.002
Creutzig, F., Jochem, P., Edelenbosch, O. Y., Mattauch, L., Vuuren, D. P. v., McCollum, D., & Minx, J. (2015). Transport: A roadblock to climate change mitigation? Science, 350(6263), 911-912. https://doi.org/10.1126/science.aac8033
de Blas, I., Mediavilla, M., Capellán-Pérez, I., & Duce, C. (2020). The limits of transport decarbonization under the current growth paradigm. Energy Strategy Reviews, 32, 100543. https://doi.org/10.1016/j.esr.2020.100543
De Janvry, A., & Sadoulet, E. (2001). Income strategies among rural households in Mexico: The role of off-farm activities. World Development, 29(3), 467-480. https://doi.org/10.1016/S0305-750X(00)00113-3
Démurger, S. (2001). Infrastructure development and economic growth: an explanation for regional disparities in China? Journal of Comparative economics, 29(1), 95-117. https://doi.org/10.1006/jcec.2000.1693
Deng, T. (2013). Impacts of transport infrastructure on productivity and economic growth: Recent advances and research challenges. Transport Reviews, 33(6), 686-699. https://doi.org/10.1080/01441647.2013.851745
Fan, S., Zhang, L., & Zhang, X. (2004). Reforms, investment, and poverty in rural China. Economic Development and Cultural Change, 52(2), 395-421. https://doi.org/10.1086/380593
Huang, Q., Zheng, X., & Wang, R. (2022). The impact of the accessibility of transportation infrastructure on the Non-Farm employment choices of rural laborers: Empirical analysis based on China’s micro data. Land, 11(6), 896. https://doi.org/10.3390/land11060896
Jacoby, H. G., & Minten, B. (2009). On measuring the benefits of lower transport costs. Journal of Development Economics, 89(1), 28-38. https://doi.org/10.1016/j.jdeveco.2008.06.004
Lakshmanan, T. R. (2011). The broader economic consequences of transport infrastructure investments. Journal of Transport Geography, 19(1), 1-12. https://doi.org/10.1016/j.jtrangeo.2010.01.001
Lechner, M., Rodriguez-Planas, N., & Fernández Kranz, D. (2016). Difference-in-difference estimation by FE and OLS when there is panel non-response. Journal of Applied Statistics, 43(11), 2044-2052. https://doi.org/10.1080/02664763.2015.1126240
López, E., Gutiérrez, J., & Gómez, G. (2008). Measuring regional cohesion effects of large-scale transport infrastructure investments: an accessibility approach. European Planning Studies, 16(2), 277-301. https://doi.org/10.1080/09654310701814629
Lucas, K. (2012). Transport and social exclusion: Where are we now? Transport policy, 20, 105-113. https://doi.org/10.1016/j.tranpol.2012.01.013
Ma, W., Renwick, A., Nie, P., Tang, J., & Cai, R. (2018). Off-farm work, smartphone use and household income: Evidence from rural China. China Economic Review, 52, 80-94. https://doi.org/10.1016/j.chieco.2018.06.002
McArthur, D. P., Thorsen, I., & Ubøe, J. (2014). Employment, transport infrastructure, and rural depopulation: a new spatial equilibrium model. Environment and Planning A, 46(7), 1652-1665. https://doi.org/10.1068/a46120
MEEC. (2004). Report on Financing Rural Road Construction. Retrieved 20th December from https://english.mee.gov.cn/Events/Special_Topics/AGM_1/Pub04AGM/meetingdoc04/201605/t20160524_344913.shtml
Mikou, M., Rozenberg, J., Koks, E. E., Fox, C. J. E., & Peralta Quiros, T. (2019). Assessing rural accessibility and rural roads investment needs using open source data. World Bank Policy Research Working Paper(8746). https://doi.org/10.1596/1813-9450-8746
MOTC. (2006a). Regulations on Rural Road Construction and Management (Order No. 3 of the Ministry of Transport of the People's Republic of China, 2006). Retrieved 10th, December from https://xxgk.mot.gov.cn/2020/jigou/glj/202006/t20200623_3312897.html (in Chinese)
MOTC. (2006b). National rural road statistics standards. Retrieved 10th, December from https://www.mot.gov.cn (in Chinese)
Oi, J. C., & Rozelle, S. (2000). Elections and power: the locus of decision-making in Chinese villages. The China Quarterly, 162, 513-539. https://doi.org/10.1017/S0305741000008237
Qiao, F., Rozelle, S., Huang, J., Zhang, L., & Luo, R. (2014). Road expansion and off-farm work in rural China. The China Quarterly, 218, 428-451. https://doi.org/10.1017/S0305741014000629
Qin, Y., & Zhang, X. (2016). The road to specialization in agricultural production: Evidence from rural China. World Development, 77, 1-16. https://doi.org/10.1016/j.worlddev.2015.08.007
Redding, S. J., & Turner, M. A. (2015). Transportation costs and the spatial organization of economic activity. Handbook of regional and urban economics, 5, 1339-1398. https://doi.org/10.1016/B978-0-444-59531-7.00020-X
Researchinchina. (2011). China Road Construction Industry Report, 2010. http://www.researchinchina.com/Htmls/Report/2011/6053.html
Rietveld, P., & Bruinsma, F. (2012). Is transport infrastructure effective?: transport infrastructure and accessibility: impacts on the space economy. Springer Science & Business Media.
Transport, & ICT. (2016). Measuring rural access: using new technologies. World Bank.
Wang, J. M. X., Wei. (2011). The evolution rule of road investment and financing system and reform thinking. Transportation Finance and Accounting(09), 20-25 (in Chinese).
Wen, M. (2004). Relocation and agglomeration of Chinese industry. Journal of Development Economics, 73(1), 329-347. https://doi.org/10.1016/j.jdeveco.2003.04.001
Wong, H. L., Luo, R., Zhang, L., & Rozelle, S. (2013). Providing quality infrastructure in rural villages: The case of rural roads in China. Journal of Development Economics, 103, 262-274. https://doi.org/10.1016/j.jdeveco.2013.03.002
Yu, N., De Jong, M., Storm, S., & Mi, J. (2013). Spatial spillover effects of transport infrastructure: evidence from Chinese regions. Journal of Transport Geography, 28, 56-66. https://doi.org/10.1016/j.jtrangeo.2012.10.009
Zereyesus, Y. A., Embaye, W. T., Tsiboe, F., & Amanor-Boadu, V. (2017). Implications of non-farm work to vulnerability to food poverty-recent evidence from Northern Ghana. World Development, 91, 113-124. https://doi.org/10.1016/j.worlddev.2016.10.015

The three rural road types do not contain each other.
CHIP2007 has data for 800 villages, but there is a lack of information on road improvement projects. Of these, 652 villages have data on whether there were road improvement projects in the current year and the specific source of funding. We have only used the villages without missing information in our analysis.
Of the 292 villages, 284 villages reported no road improvement projects and 8 villages had road improvement projects but no funding.
10,196 individuals were surveyed from CHIP2007, of which 8076 were also surveyed in CHIP2008. In addition, 2,519 new respondents were surveyed in CHIP 2008.
According to CHIP2007, the average household income in the village (level) is equal to ten if the average household income in the village is in the income range of 3000–3500 Yuan. If it is equal to 11, the income range is 3500–4000 Yuan. If the income level is 12, the income range is 4000–5000 Yuan.

No competing interests reported.

Download PDF

Submission checks completed at journal
05 Aug, 2024
First submitted to journal
31 Jul, 2024

You are reading this latest preprint version

Financing Road Projects and Its Impact on Off-Farm Work

Status:

Version 1

Abstract

Figures

1. Introduction

2. Background

2.1 Rural road development in China

2.2 Finance of rural roads in China’s village

2.3 Off-farm work

3. Data and Methodology

3.1 Data

3.2 Econometric model specification

4. Empirical Results

4.1 Characteristics of rural road improvement projects

4.2 Appropriatenss test for PSM

5. Potential Mechanisms and Discussion

5.1 Potential superior quality and design

5.2 Different off farm workplace

6. Conclusion

Declarations

Author Contribution

Data Availability

References

Footnotes

Additional Declarations

Status:

Version 1