Investment in roads has always played an important role in development. The theoretical literature argues that road expansion should stimulate development by promoting trade, reducing transaction costs and improving the flow of information.Footnote 1 There also is an empirical literature on roads, economic growth and rural incomes.Footnote 2 Several papers measure the linkages between the improvement of road access and economic growth.Footnote 3 In many of these papers, economic growth is shown to be linked to better opportunities for farmers to sell their goods at higher prices, which leads to income growth.Footnote 4
More recently researchers have tried to identify more nuanced ways that roads affect income. In particular, there are papers that show that roads can help rural individuals gain access to off-farm employment activities.Footnote 5 For example, a case study in Brazil shows that access to roads raises the returns to off-farm activities – even more than they raise the returns to on-farm activities.Footnote 6 Studies in other countries, such as VietnamFootnote 7 and Mexico,Footnote 8 also demonstrate the effect of roads on rural employment and have found a significant relationship between road expansion and off-farm job opportunities.
Mirroring the international literature, linkages between road expansion and its impact in China are also empirically estimated. Some studies show that improvements to transportation infrastructure have contributed to national economic growth,Footnote 9 farm income and poverty alleviation.Footnote 10 Other studies analyse the relationship between road expansion and off-farm activities. Using provincial-level data from the 1980s and 1990s, Shenggen Fan and Connie Chan-Kang demonstrate that road expansion has a positive and significant impact on off-farm GDP.Footnote 11 Likewise, Fan, Zhang and Zhang find that road density has positively affected the participation of farmers in off-farm employment activities.Footnote 12
While this new thread in the literature is interesting to both researchers and policymakers, a disaggregated analysis would be even more helpful since some types of off-farm work may be affected by road expansion differently. Off-farm work in the migrant sector is mostly located in large cities which are typically far from the migrants' homes. On the other hand, local off-farm job opportunities, which largely depend on the nature of local economic growth, allow farmers to live in their own homes while working off the farm. Owing to these inherent differences and because of the way road expansion reduces transaction costs for different types of off-farm employment, road expansion may be expected to have different impacts on local off-farm work and migration.Footnote 13
To our knowledge (with a couple of exceptions), there is little disaggregated empirical analysis on the impact of road expansion on off-farm employment in China. Using household-level data from the mid-1990s collected in Sichuan province, Yaohui Zhao finds that the presence of paved roads between villages and the areas outside of the village has a negative impact on migration,Footnote 14 and has an insignificant impact on local off-farm employment. Unfortunately, there is no way to judge the validity of the findings, since the author does not explain her modelling approach. Other studies on transportation do spend time discussing their empirical modelling approaches.Footnote 15 However, these papers are largely focused on the ability of households to access markets (in general) and do not focus directly on the impact of road expansion on off-farm employment. To date, we believe that no empirical work examines the impact of road expansion on total working time or the off-farm income of rural individuals.
The overall goal of this article is to understand the impact of road expansion on the off-farm work activities of farmers in rural China. In particular, we seek to track empirically how road expansion affects different types of off-farm employment opportunities (i.e. local off-farm and migrant employment opportunities) in terms of participation, total working time and off-farm earnings.Footnote 16 To meet this overall goal, we have three specific objectives. First, using survey data we provide a profile of China's rural road expansion, document the rise of off-farm employment and descriptively chart the way that roads and off-farm employment appear to move together. Second, we seek to understand if road expansion, all other things being equal, is associated with the level of participation in migrant and local off-farm labour markets. Finally, we empirically identify the different sources of the effects of road expansion on off-farm employment: is there an impact on total working time or income, or both?
The rest of this article is organized as follows. In the next section, we discuss the data collection. We then present a brief discussion of road expansion in China and the correlation between road expansion and off-farm employment. The econometric approach is presented, and the empirical results are discussed, in the fourth section. The last section concludes.
Data
The data used in this article are from a nearly nationally representative survey in rural China conducted by the Center for Chinese Agricultural Policy, Chinese Academy of Sciences, in 2005 and 2008. In this survey, 100 villages were randomly selected from 50 townships in 25 counties located in 5 provinces. The sample villages were selected as follows. First, five provinces were each randomly selected to represent five of China's major agro-ecological zones: Jiangsu represents the eastern coastal areas (Jiangsu, Shandong, Shanghai, Zhejiang, Fujian and Guangdong); Sichuan represents the south-western provinces (Sichuan, Guizhou and Yunnan) plus Guangxi; Shaanxi represents the provinces on the Loess Plateau (Shaanxi and Shanxi), Inner Mongolia and the rest of the provinces in the north-west (Gansu, Ningxia, Qinghai and Xinjiang); Hebei represents the north and central provinces (Hebei, Henan, Anhui, Hubei, Jiangxi and Hunan); and Jilin represents the north-eastern provinces (Jilin, Liaoning and Heilongjiang).
After the provinces were selected, the second step of the sample selection involved choosing the counties, townships and villages. Five counties were selected from each province, one from each quintile from a list of counties arranged in descending order of per capita gross value of industrial output (GVIO). GVIO was used because Rozelle shows that it is one of the best predictors of standard of living and development potential and is often more reliable than net rural per capita income.Footnote 17 Within each county, the survey team chose two townships, one from each half of a list of townships also arranged in descending order of per capita GVIO. Finally, within each township, two villages were chosen following the same procedure as the township selection.
We then chose the sample households. In each village eight households were randomly selected in 2005. After excluding the households in which no individual was in the labour market (e.g. a household of two elderly people who did not work on-farm or off-farm), 773 households were included in this study. In 2008, we were able to re-interview 625 households (again, those households with at least one person in the labour market). In this manner, we were able to revisit 81 per cent of the households. All of the selected households answered similar questionnaires in both years.
It is possible to show that attrition between the 2005 and 2008 surveys has not compromised the nature of the sample. According to our data (from 2005), when we compare the households that were lost from the sample during the period between the two survey waves to those households that remained in the sample, there is no significant impact on the nature of our sample. The comparison was made in terms of a number of different characteristics of the household head (such as gender), the household (such as the number of elderly dependents and the household's per capita land holdings), and the township in which each sample household lived (such as per capita GDP at the township level). Likewise, and perhaps more importantly, households' access to roads did not differ between those that remained in the sample and those that dropped out of the sample. Hence, we do not consider that attrition between the first and second wave of the survey is a serious problem in the subsequent analysis.Footnote 18
The dataset collected for this study includes basic information about townships, villages and households in the study areas for both 2004 (collected in 2005) and 2007 (collected in 2008). Enumerators interviewed village leaders using a survey form designed to collect basic socio-economic information, such as total population and the number of individuals in the labour force. For highways outside the village and paved roads within the village, detailed questions (about information such as location and the type of road) were asked and recorded. In order to have a better understanding of road expansion over time, information on the nature of the road system in 1997 was also collected from records and from the recollection of respondents. Finally, basic social-economic indicators such as per capita GDP, per capita income and township-level GVIO were also collected.
At the same time, the team collected a great deal of information about each sample household during both waves of the survey. In addition to survey blocks enumerating the basic characteristics of households (for example, each household's land and labour endowments, production assets and housing), there were two sections that collected information that forms the core of this article's analysis. First, there is a section on the demographic information of each individual in the household, detailing facts such as gender, age, education and marital status. Second, the survey has a long section that records the working experience of each household member who was part of the labour force. Enumerators asked questions about whether each member of the household's labour force had an off-farm job, and if so, what type of off-farm job, how much time was spent on that job and how much income was earned from it, both monetary and non-monetary.
Road Expansion and Off-Farm Employment in Rural China
Prior to the economic reforms, rural-to-urban migration was highly restricted in China. This urban–rural segregation was instituted following the devastating famine that occurred between 1959 and 1961.Footnote 19 Consequently, the income of rural households became tied to the participation of household members in on-farm work. In cases where rural individuals were able to move to cities to work, they faced high costs of living since they did not have access to necessities such as housing and food, and were unable to access many jobs which were reserved for urbanites. Such restrictions succeeded in controlling the size of the urban population for decades. As a result, the proportion of the population that lived in urban areas was less than 20 per cent by the late 1970s.Footnote 20
There are at least three reasons why the government loosened restrictions and began to allow farmers to move from rural areas to China's cities after the beginning of the economic reforms in the 1980s. First, following the decollectivization of farmland, many were eager to move out of farming since the plots allocated to farmers were small, which lead to a surplus of labour in many families. Second, as economic growth led to a reduction in the shortages of food and other goods, markets emerged and the cost of moving to the city fell. Rural individuals were able to find rudimentary housing and affordable food and clothing in the city. Finally, rapid economic growth in the cities and adjacent industrial zones resulted in a huge demand for low-paid labour. With little opportunity to increase on-farm income and with the availability of off-farm employment, tens of millions of farmers left their hometowns and migrated to the cities in the late 1980s and 1990s. Rural–urban migration not only increased the income of rural households, it also became an important force driving China's economic growth during the 1980s and 1990s.Footnote 21
The volume of labour that migrated from rural to urban areas in China during this period was unprecedented. Even though there was almost no rural–urban migration before the reforms began in the late 1970s, according to China's own statistical sources, the number of migrants was around 90 million in 2001.Footnote 22 This number increased to more than 130 million in 2006, which is about a quarter of the total labour force in rural China.Footnote 23 The number continues to increase rapidly.Footnote 24
There was also considerable movement of labour off farm to nearby factories and to local towns. According to the National Bureau of Statistics of China (NBSC), the average number of rural individuals with local off-farm employment was about 90 million during the late 2000s.Footnote 25 The survey data collected by NBSC, which included information on all provinces in China (with the exception of Tibet), showed that rural households in eastern coastal regions were most likely to work in local off-farm jobs (when compared to those in western and central regions).
Our survey data are consistent with national statistics. In our sample, as in the national-level statistics, members of rural households have two off-farm employment options. First, there is off-farm work in or near their hometowns where the individuals live in their own houses in their villages and commute to work. Hereafter, we call this type of work local off-farm employment. Alternatively, individuals can work off the farm outside of their hometowns. In these cases, the individuals do not live in their own homes, but instead they live in rented accommodation in the destination city or in a dormitory on the job site. This is what we term as “migration,” and those that engage in migration are called “migrants.”
According to our data, there are substantial levels of local off-farm employment and migration among the sample households. The share of individuals in the labour force in our sample that has found local off-farm employment is 28 per cent (see Table 1). At the same time, 24 per cent of the individuals in our sample are migrants. The share of the sample individuals who only farm (that is, they have no local off-farm employment and are not engaged in migration) is less than 50 per cent.
Characteristics of Labourers with and without Off-Farm Work a
Notes:
a Samples include those aged 16–60 years old and who are not in school. bTotal off-farm income, including non-cash income, and 2007 income is adjusted using CPI.
Source:
Authors' survey.
Table 1 also shows the characteristics of those working off the farm. Specifically, our data show that migrants are younger, more educated and less likely to be married than individuals with local off-farm employment. Table 1 also shows that the per capita GVIO of the towns which have higher shares of individuals who work in local off-farm employment is more than the per capita GVIO of towns which have higher shares of migrants. Presumably, individuals in China's poorer regions are more likely to be migrants since there are fewer local off-farm job opportunities. In contrast, individuals in non-poor regions have relatively more opportunities to work locally. Our data also show that the average working time of migrants is longer than those working in local off-farm employment; the income from migration is also higher.
In contrast, individuals who only farm are more likely to be female, older and less educated than individuals with either type of off-farm employment. As shown in row 8, per capita land holdings of individuals who only farm are higher than those of individuals with off-farm work (1.96 mu versus 1.55 mu/1.39 mu).Footnote 26 Table 1 also shows that individuals who only farm are more likely to be in poor regions than those with local off-farm employment.
Interestingly (and importantly for reasons of external validity), these figures are consistent with those from national statistical sources and other case studies.Footnote 27 This consistency not only reinforces the idea that there are different types of workers in different types of jobs in rural China, but also suggests that our sample is fairly representative of households in rural China.
Road expansion in China
China was poorly endowed with transportation infrastructure when the economic reforms began in 1978. Under the centrally planned system, the government was the sole investor in all sectors of the economy, and priority was given to heavy industry. Road building fell far below what it should have been, given China's ambitions and growth targets. As a result, with the exception of an emerging transportation network in the north and north-eastern regions of China – where the nation's heavy industry was located – there was little development of roads in other regions until the late 1970s and early 1980s.Footnote 28 The total length of the highway network was only 883,000 km in 1978; the road density was less than 100 km per thousand square km, a level that indeed put China among the world's poorest nations.Footnote 29
Recognizing that the nation's infrastructure was not keeping pace with the needs of its developing economy, the government began to focus on road development in the mid-1980s. Between 1990 and 2010, road expansion was one of the highest priority items in the annual “Report on the work of the government” document.Footnote 30 The government also expended a great deal of effort on advertising its road expansion policy in order to gain support for the huge fiscal expenditures that were being invested. In the 1990s, one of the most common slogans in China's countryside was: “Want to get rich? Build a road first!”Footnote 31
China's leadership has backed up its plans and propaganda campaigns with serious investment commitments. Particularly since the early 1990s, China's government has invested heavily in all dimensions of the nation's road network,Footnote 32 including towards the expansion of the road network in counties and towns,Footnote 33 improving the quality of roads,Footnote 34 and increasing the number and length of expressways in both eastern coastal regions and western inland regions.Footnote 35 The total length of China's highway system increased from less than 1 million km in 1978 to more than 4 million km at the end of 2010.Footnote 36 By the end of 2010, the total length of the nation's network of expressways was 74,100 km, making China second only to the United States in overall expressway density.Footnote 37
Researchers have now begun to document and analyse the unfolding road-building frenzy in China. Previous studies have shown that highway expansion leads to the rapid development of the local economy along the new roads.Footnote 38 Using provincial and county-level data, Fan and Zhang also show that road expansion contributes to local economic development.Footnote 39 Road expansion also has had a significant positive impact on farmer income and poverty alleviation – especially in poor areas.Footnote 40
Despite the progress in road building during the 1980s and 1990s, it was widely acknowledged in the early 2000s that more road networks were needed – especially in rural areas. Living conditions in rural areas were still poor after two decades of reform.Footnote 41 Studies show that rural infrastructure was also still poor and was partially to blame for the wide and widening gap between rural and urban areas. For these reasons, in the early 2000s, the government launched the “building a new socialist countryside” campaign.Footnote 42 As part of its commitment to improve livelihoods in rural areas, the government dramatically increased its investment in rural areas in order to improve rural infrastructure, and road expansion was at the top of the list of priorities.Footnote 43 Between 2006 and 2010, the government invested 197.8 billion yuan in the construction of rural roads, leading to an annual growth rate of 30 per cent for the rural road network.Footnote 44 By 2011, the length of the rural highway network had reached 3.5 million km,Footnote 45 accounting for 87 per cent of China's total road network.Footnote 46 During this time, the quality of rural roads was also improved.Footnote 47
The extent of road expansion from our survey data is consistent with these macro-statistics that come from government reports. In Figure 1, we use two variables to measure road expansion: a) the distance to the nearest highway from a village;Footnote 48 and b) the length of paved road within a village's boundaries. As shown in Panel A, the average distance from a village to the nearest highway was 3.35 km in 1997; it decreased to 1.66 km by 2007, a reduction of about 50 per cent. The rise of paved roads within the boundaries of villages is also significant. Our data show that between 1997 and 2007, the total length of paved road within our sample villages increased from 0.22 km to 3.84 km (Panel B).
Road Development, 1997–2007
Road expansion and off-farm employment
To study the correlation between road expansion and the changes in off-farm employment, we need to create a set of variables to measure the nature of roads and the nature of off-farm employment. Since road quality and type may have different impacts on the off-farm activities of rural individuals,Footnote 49 we use two road variables in this study: the distance to the nearest highway from a village; and the per capita length of paved road within a village's boundaries.Footnote 50 We then divide farmers into different groups according to their access to roads. We first divide the sample into three subgroups according to the distance to the nearest highway from a village: a “near” group is defined as such if the distance to the nearest highway from a village is less than 0.50 km; a “middle” group if the distance is less than 3 km but greater than 0.5 km; and a “far away” group if the distance is greater than 3 km.Footnote 51
Using these cut-off definitions, we can see a correlation between the type of road and the nature of off-farm employment. As shown in Table 2, 41 per cent of the individuals in our sample belong to the near group, with the average distance to the nearest highway being 0.08 km. In addition, 39 per cent of the sample individuals belong to the middle group. The average distance of the middle group to the nearest highway is 1.66 km. Finally, 21 per cent of the sample individuals belong to the far away group, with an average distance to the nearest highway of 9.80 km.
Road Expansion and Off-Farm Work
Notes:
aDistance (km) to the nearest highway from a village. bPaved road within the boundaries of a sample village. cWe defined a “near” group as such if the distance to the nearest highway from a village was less than 0.5 km; a “middle” group if the distance was less than 3 km but greater than 0.5 km, and a “far away” group if the distance was greater than 3 km. dA “yes paved roads” group was defined as such if there was paved road within the boundaries of a sample village, and a “no paved roads” group if there was no paved road within the boundaries of a sample village.
Source:
Authors' survey.
We also use our other road variables, paved roads within the village, to create analytical groupings. Since about two thirds of the sample villages contain no paved roads inside their boundaries, we divide the sample into two parts. Specifically, as seen from Table 2, 68 per cent of individuals live in villages without paved roads (henceforth the “no paved roads” group). In contrast, 32 per cent of individuals live in villages with paved roads (henceforth the “yes paved roads” group).
Table 2 shows that the expansion of roads appears to have a positive impact on the off-farm employment of our sample individuals. According to our data, in the case of villages that are far away from the highway, the share of migrants is 60 per cent, which is higher than in middle or near villages (49 per cent and 38 per cent, respectively). The opposite is true in the case of local off-farm employment. As shown in column 4 of Table 2, in near villages, more than 60 per cent of the sample individuals have local off-farm employment, but only 40 per cent of those in far away villages work in local off-farm employment.
Table 2 also shows that road expansion is correlated with total working hours and income. This is especially true for those working in local off-farm employment. As shown in column 6, the total working time of sample individuals in near villages is 1,444 hours. In contrast, the totals of working hours of sample individuals in middle and far away villages are lower, ranging between 771 and 1,046 hours. Similarly, the off-farm income of sample individuals in near villages is 8,000 yuan, which is about 50 per cent higher than that of individuals in far away villages. Hence, according to our data, in villages with access to roads, local off-farm workers work longer hours and their off-farm income is higher. The story is slightly different with migrants. Even though migrants from far away villages work longer hours than migrants from near and middle villages, their income is not significantly higher.
A similar pattern of correlations appears if we examine the relationship between the presence of paved roads within the boundaries of the village and different types of off-farm employment (Table 2). According to the data, 49 per cent of sample individuals work as migrants if they are from the no paved road villages. In contrast, only 41 per cent of those from paved road villages work as migrants. At the same time, for local off-farm workers, the presence of paved roads within the village has positive correlations with total working hours and off-farm income. Sample individuals in villages with paved roads work longer hours (1,399 versus 1,046) and earn more off-farm income (9,006 yuan versus 5,973 yuan) than those in villages with no paved roads.
Multivariate Approach and Empirical Findings
In this section, we seek to isolate the impact of road expansion on off-farm employment by estimating a series of econometric models. We do so because it is possible that the descriptive results in the previous section are misleading since the impact of other factors which affect off-farm employment outcomes are not excluded. In order to isolate the effect of road expansion on off-farm employment, we define an econometric model and describe the findings below.
Econometric model
The econometric model we use can be specified as:
$$\eqalign{Off\hbox{-}farm_{\rm i\comma t}&={\rm a}_{0}+{\rm a}_{1}^{\ast}Road_{\rm i\comma t-1}+{\rm a}_{2}^{\ast}Individual_{\rm i\comma t} +{\rm a}_{3}^{\ast}Household_{\rm i\comma t}\cr & \quad+{\rm a}_{4}^{\ast}Social\hbox{-}economics_{\rm i\comma t} + {\rm a}_{5}^{\ast}Year_{\rm i\comma t}+e_{\rm i\comma t}}$$
where off-farm is a binary variable, which is equal to 1 if individual i has an off-farm job, and is equal to 0 otherwise. The variable, Road, is specified in one of two ways (and is used in separate models): 1) the distance to the nearest highway from a village; or 2) per capita length of paved road within a village's boundaries. In this model, we use lagged measures of road (that is, the measure of road length from the previous time period of our data). It is possible that the impact of road expansion may be overestimated if there is any degree of reverse causality (i.e. if off-farm participation might be associated with road expansion) and this source of endogeneity is ignored. For this reason, lagged road expansion is used in the analysis as is common practice in other similar studies.Footnote 52 The second reason for using lagged road expansion is that it might take some time for the impact of road expansion to appear.
The model in the above equation (1) also controls for other factors. Specifically, Individual is a vector of the characteristics of the individuals in our sample. We include several variables to measure the characteristics which may account for some of the heterogeneity in off-farm employment across individuals. For example, years of education are included as a measure of human capital; and age and age-squared are included to control for the life-cycle effects that may influence the decision to participate in migrant and local off-farm labour markets. In addition, as in other studies,Footnote 53 married (a dummy variable which is equal to 1 if the individual is married, and is equal to 0 otherwise) and gender (a dummy variable which is equal to 1 if the individual is male, and is equal to 0 if female) are also included.
The Household variable is a vector of household characteristics. Per capita Land is used to control for factors affecting farming income at the household level and, by implication, differences in the potential of individuals to increase income from working off-farm.Footnote 54 We also include the number of children (aged 0–7) and the number of elderly (aged 65 and over) in the household to measure the impact of the number of dependents on off-farm work participation, as do other studies.Footnote 55
Finally, we include two variables to measure the impact of the socio-economic factors of our individual's town/villages. The first variable is ratio of migration in the village's total labour force during the previous time period. This variable, among other things, is used to measure the impact of social networks on an individual's migration decision, which is common in previous studies.Footnote 56 The second variable is the per capita GVIO of the township. This variable, as discussed above, is included to control for the nature of the local economy. A year dummy (Year) is added to control the impact of time. The symbol e is the error term.
Since off-farm work participation is a discrete variable, OLS may produce biased and inconsistent parameter estimates.Footnote 57 Therefore, Logit models are used in this study. Because an individual's off-farm employment decision may be affected by the off-farm employment status of other individuals in the family (for example, if one household member is farming, the other may be more likely to work off-farm), all of the models correct for clustering at the household level.
Estimation results
The results of the econometric estimation of equation (1) are shown in Table 3, columns 1–6. In general, most of the regression results are consistent with the descriptive analysis above. Most coefficients on the control variables are as expected and statistically significant. For example, the results show that educated male labourers are more likely to engage in off-farm work; land endowment on the other hand is negatively related to off-farm work participation. These findings are consistent with most previous studies.Footnote 58 We also find that social networks have a positive and significant impact on participation in the migrant labour force.Footnote 59
Logistic Models of Off-Farm Work Determination
Note:
z-statistics are in parentheses. *, **, *** statistically significant at the 10%, 5%, 1% levels, respectively.
Source:
Authors' survey.
Table 3 also provides results that show that road expansion does have a positive and significant impact on off-farm employment. As shown in the first column, the estimated coefficient on the variable measuring distance to the nearest highway from a village is negative and statistically significant in the total (or migration + local) off-farm work participation equation. In other words, the results show that the smaller the distance is, or the better the transportation infrastructure is, the more likely it is that rural individuals will have off-farm employment. The positive impact of road expansion on off-farm work is consistent with the descriptive findings above and the findings of previous studies.Footnote 60
Our results, however, go beyond the impact of road expansion on off-farm employment in general. According to our findings, road expansion has different effects on different types of off-farm employment (that is, migration and local off-farm employment). As shown in the first row of Table 3, the estimated coefficient on the distance to the nearest highway from a village variable is positive and significant in the migration equation (column 2) and is negative and significant in the local off-farm employment equation (column 3). In other words, the estimation results show that road expansion encourages local off-farm employment, but, all other things being equal, discourages migration.
A similar story is found in our econometric analysis when we examine the impact of the alternative measure of road expansion, per capita paved roads in our sample villages, on different types of off-farm employment. The estimated coefficient on the per capita length of paved roads within a village variable is positive and statistically significant in the local off-farm employment equation (column 6). In other words, the estimation results show that when paved road per capita is higher, rural households send more of their members to work in local off-farm employment. In contrast, the estimated coefficient on the paved road per capita variable is insignificant in the case of migration (column 5).
To control the impact of all time-invariant factors that may be affecting the estimated relationship between road expansion and off-farm employment (and the endogeneity problems that such unobserved heterogeneity can create), we take advantage of the panel nature of our data and estimate a fixed effect (FE) model (controlling for household fixed effects). We follow the lead of Warr.Footnote 61 Specifically, the FE model that we estimate is:
$$\eqalign{\Delta Off\hbox{-}farm_{\rm i} = & \beta_{0} + \beta_{1}^{\ast} \Delta Road_{\rm i} + \beta_{2}^{\ast}\Delta Married_{\rm i} + \beta_{3}^{\ast} \Delta Household_{\rm i} \cr & + \beta_{4}^{\ast} \Delta Social\hbox{-}economics_{\rm i} + \varepsilon _{\rm i}}$$
Here, ΔOff-farm, ΔRoad, ΔHousehold and ΔSocial-economics are changes between time periods 1 and 2 of the variables defined above (and used in the estimation of equation (1)). However, in this version of the equation, we only include a subset of the variables (that is, only those variables that vary over time – since all others are differenced away). In fact, except for marriage, all the other individual variables, such as age, gender and education, are excluded since they do not vary over time.
The results of the FE Logit models are similar to those of the pooled Logit models (from equation (1), Table 3, columns 1–6). As shown in Table 3, the estimated coefficients of interest (that is, those on the variable measuring the distance to the nearest highway from a village) are positive and significant in the migration equation, and negative and significant in the local off-farm employment equation. In other words, the results from both the pooled and fixed effect Logit models are consistent and robust. The expansion of roads has an impact on whether the rural households decide to migrate or find employment in the local off-farm labour market. However, the effect is different: it is negative or neutral in the case of migration and positive in the case of local off-farm employment.Footnote 62
Because of the nature of the labour allocation problem in rural China today, it might be argued that accounting for the three different options (farming, out migration and local employment) at the same time is a more appropriate way to approach the empirical estimation of the link between roads and off-farm employment. Therefore, we also define a multinomial Logit model and simultaneously model the labour allocation problem (i.e. whether an individual works on the farm, in the local off-farm employment market or in the migrant labour market). Using this approach, we find that the results are almost identical (Table 4). Road expansion is associated with rising interest in local off-farm employment and less (or relatively less) interest in migration.
Multinomial Logit Models of Employment Choice
Notes:
Marginal effects are reported; z-statistics are in parentheses. *, **, *** statistically significant at the 10%, 5%, 1% levels, respectively.
Source:
Authors' survey.
While it is beyond the scope of this article to identify definitively what lies behind the empirical results at which we arrive, it is possible to speculate. Specifically, the different impacts of road expansion on local off-farm employment and migration may be down to at least two reasons. First, it could be that transaction costs fall when the road infrastructure is improved. For individuals with options for local off-farm employment, road expansion may reduce the cost of getting to/finding work. Since these rural workers have to commute on a daily basis, even a small reduction in travel time (for example) could lead to significant gains in welfare on an annual level. In contrast, in the case of migrants, who live and work far from home, the impact of the expansion of roads inside a village may be small (since they often only come home – and use the roads – once a year). Hence, such a dynamic may be seen as making local off-farm employment more attractive than migration, all else being equal.
A second reason is that road expansion may facilitate the expansion of the local economy, which in turn may raise the prospects of local off-farm employment over migration. As shown in previous studies, road expansion can have a positive and significant impact on the development of the local economy.Footnote 63 As a result, both job opportunities and the level of income from local off-farm work might be expected to increase. This is consistent with the findings in the regression analysis in this study. As shown in Tables 3 and 4, the development of the local economy (as measured by per capita GVIO) is associated with a rise in the probability that a rural individual will find off-farm work employment, especially at the local level.
Our results can also show that the effects of road expansion go beyond access to employment; road expansion can affect total working time and the earnings of those engaged in local off-farm employment. As shown in Table 5, if the distance to the nearest highway from a village is reduced by 1 km, local off-farm income will increase by 697 yuan (or 9.55 per cent of total local off-farm income). We find similar results when using the alternative variable (per capita paved roads). Table 5 also shows that an increase in local off-farm income is associated with an increase in the total number of hours that a rural individual can work off-farm owing to the improvement of the road network. In other words, this study seems to show that as roads expand, rural individuals working in local off-farm employment can work more since they spend less time commuting. The same result is found in the case of road expansion and a rise in off-farm earnings.
Determinant of the Working Time and Income of Local Off-Farm Work
Notes:
t-statistics are in parentheses. *, **, *** statistically significant at the 10%, 5%, 1% levels, respectively.
Source:
Authors' survey.
Conclusion
Infrastructure has been, and will continue to be, one of the most important parts of the public investment portfolio in rural China. Previous studies show that road expansion has had a positive impact on economic growth and rural income. In this study, we go beyond this aggregated result. The findings of the empirical analysis suggest that road expansion has different impacts on different types of off-farm employment. Specifically, road expansion leads to more local off-farm employment and relatively less migration.Footnote 64 As a result, with the improvement of local transportation, some migrant labourers might turn instead to off-farm work. We also show that road improvement increases both the total working time and income of local off-farm workers.
This study has important policy implications. First, although China has made great progress in building up its rural infrastructure since the 1980s, public services are still severely underfunded in most communities, especially in poor regions.Footnote 65 Previous studies show that road expansion has a positive impact on local economic development.Footnote 66 This study shows that road expansion plays an important role in increasing the participation of rural individuals in off-farm work (especially in the case of local off-farm employment) and well-being. In other words, it can be argued that increasing the level of investment in rural roads will contribute to both rising rural incomes and a reduction of the urban–rural inequality that plagues China today. On the other hand, the previous studies by Fan, Zhang and Zhang, and Fan and Chan-Kang show that the return to road investments in richer regions (for example, in China's coastal regions) is higher than that of poor regions (for example, in China's western regions).Footnote 67 Hence, all else being equal, private companies may have less of an incentive to invest in poor regions if roads are built in both places (and if the results from the 1990s are still true). In other words, if the objective is to stimulate local economic development, raise rural incomes and reduce inequality, it may be that even more public investment in roads is needed in poor rural regions.
Second, the results show that the impact of rural road expansion is contributing to the success of another set of development policies. China's central government is pursuing a strategy that seeks to channel rural people into small cities and towns rather than big cities.Footnote 68 In this study, we show that an improved road network encourages rural individuals to join the local off-farm workforce rather than migrate to large and more distant urban districts. Our findings would urge the government, assuming that they are really interested in developing inland cities, to continue to support road expansion in rural areas.
