Rural Roads and Local Market Development in Vietnam

Abstract

We assess impacts of rural road rehabilitation on market development at the commune level in rural Vietnam and examine the geographic, community, and household covariates of impact. Double difference and matching methods are used to address sources of selection bias in identifying impacts. The results point to significant average impacts on the development of local markets. There is also evidence of considerable impact heterogeneity, with a tendency for poorer communes to have higher impacts due to lower levels of initial market development. Yet, some poor areas are also saddled with other attributes that reduce those impacts.

Acknowledgements

For helpful comments, we thank Kathleen Beegle, Hai Anh Dang, Emanuela Galasso, Shahidur Khandker, John Strauss, Martin Ravallion, and the Journal of Development Studies' editor and (anonymous) referees. We gratefully acknowledge funding support from CIDA (TF034859), the World Bank's East Asia and Pacific Transport Unit and the Poverty and Social Unit in the Vietnam Resident Mission Office, and the DFID trust fund for Poverty Analysis and Policy Advice. The findings, interpretations, and conclusions expressed in this paper do not necessarily represent the views of the World Bank, its Executive Directors, or the countries they represent.

Notes

1. See, for example, Gannon and Liu (1997), Escobal and Ponce (2004), Lokshin and Yemtsov (2005), Dercon et al. (2006) and Khandker et al. (2009).

2. The now classic contribution is Krugman's (1991) ‘new economic geography’ model in which one sector of the economy is subject to increasing returns to scale and is (hence) non-competitive. Yet the new economic geography literature does not pay much attention to rural transport improvements and there has been little effort to link these literatures.

3. Wanmali (1992) makes similar arguments about the effects of roads on spatial patterns of local economic activity. The agricultural growth linkages literature as summarised in Haggblade et al. (2007) is also relevant. Here the impacts are first felt in agriculture with eventual externalities to the non-farm sector.

4. The Vietnam Rural Transport Project I, see World Bank (1996) for details.

5. In one of the few cases in which impacts of a poor area development project (including rural roads) were tracked over time, the impacts declined rather than showing cumulative gains (Chen et al., 2009).

6. While impact heterogeneity has received surprisingly little attention in the context of rural roads, it has been emphasised in social sector programmes; see, for example, Galasso and Ravallion (2005).

7. Least cost techniques refer here to the minimum cost engineering solution that ensures a certain level of motorised passability.

8. We looked at education, healthcare, family planning, child nutrition, reforestation, opening up new land, anti-opium, job creation, TV and radios distribution programmes, as well as various components of the Hunger Elimination and Poverty Reduction Programme, including credit loans, school and healthcare fee exemptions, free land and new infrastructure.

9. Lao Cai, Thai Nguyen or Nghe An are located in what was previously known as North Vietnam or the Democratic Republic of Vietnam.

10. Early on during data collection, we mapped many of the non-project communes and judged them to be sufficiently far from our road links to be confident that contamination is unlikely.

11. A welfare ranking implemented by commune authorities was used to divide households into the poorest, middle and richest thirds of each commune's households. Five were then randomly selected from each of these equal sized groupings. The household ranking is undoubtedly subjective, but stratified sampling on this basis should ensure a sample that is reasonably representative of each commune's socioeconomic groups.

12. Full details on the consumption model are available from the authors.

13. Distance to the closest market town is defined as the distance to the closest large town, as identified by the commune informant.

14. These are non-parametric regressions, using locally weighted smoothed scatter plots, in which the unit of observation is the commune.

15. The variables are defined in the notes to Table 1, when they are not self-explanatory.

16. The implementation of PSM is carried out in Stata with the program psmatch2.

17. For the theory of propensity score matching and propensity score weighting, see Rosenbaum and Rubin (1983) and Hirano et al. (2003), respectively. For an empirical application in the same setting, see van de Walle and Mu (2007).

18. The detailed results of the logit model and density graphs of the propensity scores are given in van de Walle and Mu (2007). To save space we only summarise them here.

19. Following Rosenbaum and Rubin (1985), we carried out a balancing test using the standardised mean difference - the difference in covariate means in project and non-project communes as a percentage of the standard deviation in the full sample. This drops significantly from 14 per cent before to 9 per cent after matching.

20. With the exception of market frequency, the employment and schooling variables, other variables are dichotomous so that the numbers are interpretable as probabilities that communes have the outcome.

21. The 2001survey was fielded about four months after all projects were completed. Approximately 11 per cent of project communes had finished their road project less than one year before.

22. Khandker et al. (2009) also find larger impacts for the poor although their analysis is carried out at the household level.

23. We leave out transportation access and road density as they are highly correlated with other explanatory variables that we judge more important and we are limited in degrees of freedom. We also exclude measures of social services as there is little variance across communes.

24. Note that the fact that the dependent variables are estimated does not invalidate the parameter estimates or their standard errors. The estimated impact is the true impact plus an error term that ends up in the composite regression error. The overall predictive power falls but the estimates are still valid.

25. Given the number of commune attributes and outcomes, the discussion here focuses on the estimated coefficient signs rather than their magnitudes. The coefficients indicate how each attribute affects the road project's impact on the probability of having a market or a related indicator (in the case of dichotomous variables), or its impact on the percentage change in employment or enrolments. For example, in the markets regression (Table 4), the coefficient on the initial value is −0.27 - meaning that the probability across communes that better roads lead to new local markets is reduced by 27 per cent by having a market in 1997.