Rural roads and local development in India

As of 2020, over one billion people in the world live more than two kilometres away from road infrastructure, with one-third living in India. Intuitively this problem appeals as a substantial issue preventing people in disconnected regions to realize their economic opportunities and improve living standards. To tackle this infrastructure connectivity problem in India, the Government of India started Pradhan Mantri Gram Sadak Yojana (PMGSY), literally Prime Minister's Village Road Program, in 2000. In 15 years, the PMGSY program had financed the construction of all-weather roads to about 200,000 villages, incurring nearly 40 billion U.S. dollars investment expenditure.

Usually the problem of understanding the role of rural roads on local development hinges upon both theoretical and empirical research gaps. Theoretically, it could be that road connectivity supports the development of economic coordination, only if other missing necessary conditions for development are in place. This means, for instance, a village already has enough economic capacity in terms of agricultural or firm production and transportation infrastructure could enhance the economic activity right away. In this regards, improved road connectivity can support rural development by improving employment chances and reducing poverty in disconnected regions.

Yet, empirical estimating of the aforementioned theoretical arguments is a convoluted task due to the difficulty in excluding confounding factors or holding them fixed (i.e. ceteris paribus). Perhaps a quasi-experimental design could help to somewhat test the impact of rural roads on local development. Asher and Novosad (2020) using a dataset from PMGSY project for six Indian states, namely Rajasthan, Maharasthra, Madhya Pradesh, Gujarat, Chattisgarh, Odissa. They try to reveal the causal impact of road pavement in India on firms, occupation, agriculture, consumption and transportation. Yet, Asher and Novosad (2020) find inconclusive results concerning the hypothesis if the PMGSY project improved the social welfare in India. I was interested in why the results do not hold, therefore, I have replicated the results. Contrary to the original approach of authors, I replicate for individual Indian states since there is a multitude of differences in terms of culture, economic and political institutions across states in India. In particular, I test whether rural road construction is connected with increases in farm and non-farm economic growth.

Figure 1: Political map of India

According to the PMGSY project, only villages that meet specific population level are eligible to receive a road pavement. Consequently, some villages with 999 people living should be ineligible to receive a road pavement but one village with 1,000 people living should be eligible. Thus, 1,000 village population is an example of the cutoff value for eligibility to receive a road pavement and this population value is used as a running variable (in other words as an instrument). Theoretically, there should be not much difference between villages with 999 and 1,000 people living. This theoretical familiarity between villages to the left and right of the cutoff level (which is 1,000 in our example) creates a quasi-experimental design utilizing the discontinuity in receiving new road by 2012. Here this discontinuity becomes a decisive variable to test the impact of road pavement on development. This method is called fuzzy regression discontinuity design(RDD) and uses discontinuities in the probability or expected value of road treatment conditional on village population. With a fuzzy RDD method one analyses, the differences between left and right to the cutoff point (for example 1,000 people) and compares the differences to calculate an average treatment effect around a selected bandwidth, which in our case is 60 and 80. So for a bandwidth of 60, we compare villages that have 970-999 people living to 1,000-1,029 people living to derive a treatment effect value (these values are depicted in table 1 below). Shortly, the fuzzy RDD estimator calculates the local average treatment effect (LATE) of receiving a new road for a village with a population around the cutoff level [1].

Figure 2: Effect of road prioritization on road treatment in India, by state

Findings do not indicate a substantial discontinuous increase in the probability of receiving road pavement according to the PMGSY rule based on the village population level. Only two out of six states in our sample demonstrate a fuzzy discontinuous increase at the cutoff level, which was around 20-30 percent. Therefore, this rule was not fully followed in India and high fuzziness, perhaps because of corruption, political pressure and many other confounders. Thus, I can conclude that PMGSY rule is not a very valid and informative instrument for a two-staged least-squares analysis (2SLS). There are further problems connected with using PMGSY data and rule to test the hypothesis: i) smaller villages could be connected if they lay in the path of connecting a prioritized village; ii) groups of villages within 500 meters of each other could combine their populations; iii) members of parliament and state legislative assemblies were allowed to make suggestions; iv) the presence of a weekly market could also influence allocation.

I present the second stage results from two states where figure 2 indicates a fuzzy discontinuity, namely Rajasthan and Madhya Pradesh.

Table 1: Impact of New Road in Madhya Pradesh and Rajasthan villages

Contrary to what Asher and Novosad (2020), I find a somewhat significant impact of rural roads on villages in Madhya Pradesh and Rajasthan. Rural road pavements increase firm-level activity in disconnected regions of India by 26 to 32 percent depending on the bandwidth choice (60, 80 or 100) and type of clustering (triangular or rectangular). Findings for aggregate production, on the other hand, do not indicate a reliable impact. This indeterminant impact of rural roads on aggregate production in Indian villages could be mostly due to the time dimension of our analysis since we only have data for the short-term impact of rural roads on Indian state economies. In simple words, this indicates that transportation infrastructure supports economic coordination and impersonal exchange at least on firm-level dynamics.

However, the problems behind this empirical testing are deeper. Firstly, the data utilized to test the impact of PMGSY on Indian village economies are at maximum four years after road completion. It means that we can only test the short-to-medium-run impact of roads. But how valid is it to make a conclusion based on only the short- or medium-run impact of road pavement in villages with low-education, high poverty and many more socio-economic problems? What about the long-run impact of road pavement on economic activity in Indian villages? Findings in this writing are only a short-run impact not even medium because, for example, after a decision has been made concerning road pavement to connect a village, it takes around one to two years to finalize that infrastructure project. If a firm or investor will decide to make an investment, this will require more than one or two years additional. In short, the four-yearly data span does not provide legitimately valid data to test our hypothesis.

Development is a long-run process and short-run analysis of roads on local development or poverty reduction can only reveal inconclusive results. It is because structural changes cannot be captured and traced during a short-run analysis.

At the end, when one tries to reveal any causal links in social sciences, perhaps caution is the best. Most of the causal links revealed in some countries under investigation only turn out to be local contextual realities. Perhaps economic infrastructure has a crucial impact on local development of disconnected regions but the mission should be to explain why this causal link holds to help for the empirical identification process as well. As of now, I have more questions to understand than valid answers ahead of me.

Notes and references:

[1] For an introductory information on RDD, see Angrist and Pischke (2008) as an introduction to the topic. Angrist, J. D. and Pischke, J.-S. (2008). Mostly harmless econometrics: An empiricist’s companion. Princeton university press.

[2] Asher, S. and Novosad, P. (2020). Rural roads and local economic development. American Economic Review, 110(3):797-823.