Somewhere in between Towns, Markets and Jobs – Agricultural Intensification in the Rural–Urban Interface

Abstract

We propose a flexible conceptual and methodological framework to model the dynamics of agricultural intensification in the rural–urban interfaces of large cities. We focus particularly on the effects of polycentric urbanisation patterns and trade-offs between agricultural intensification and off-farm employment. In our conceptual framework – modelling household decision-making based on utility maximisation – we show that agricultural intensification in the rural–urban interface is likely to exhibit non-linear and complex spatial patterns due to location-dependent variation in output prices and wage rates. This is confirmed by our empirical analysis of a primary data set of 638 smallholder farms in the rural–urban interface of Bangalore. Applying Structured Additive Regression (STAR) techniques, we model two-dimensional urbanisation effects using household and village coordinates. Results imply that proximity to secondary towns and road infrastructure is the primary channel of urbanisation effects on the uptake of modern agricultural inputs. Furthermore, proximity to the large urban centre of Bangalore appears to increase the opportunity costs of agricultural intensification through improved access to off-farm labour opportunities. Overall, we show that patterns of agricultural intensification around urban centres are not necessarily radially symmetric.

Acknowledgements

The data were compiled within the Deutsche Forschungsgemeinschaft (DFG - German Research Foundation), Research Unit 2432 “Ecological and Social Systems at the Indian Rural-Urban Interface: Functions, Scales and Dynamics of Transition.”

Disclosure statement

No potential conflict of interest was reported by the author(s).

Notes

1. $T_z$ also includes leisure and, thus, free time that can be understood as a consumption good as well.

2. We assume standard second-order conditions for both equilibria. Thus, the indifference curve, $MR{S}_{C,T_Z}$, is convex to the origin of $C$ and $T_z$ (right origin in Figures 1 and 2). Furthermore, the second derivative of $MP{P}_L$ is nonpositive.

3. In fact, Appendix 1 shows that these crop classes are relatively more often marketed than other classes in our data set.

4. Other agricultural inputs such as irrigation or machinery might also be good proxies for agricultural intensification and modernisation. However, we only have cross-sectional data, and irrigation technology or machinery are normally long-term investments (Abdulai & Huffman, 2005; Dadi, Burton, & Ozanne, 2004; Irwin & Bockstael, 2004). Therefore, we do not use machinery or irrigation as dependent variables in our analysis, but we do consider them to be potential covariates.

5. Note that most modern inputs are subsidised in India and widely available so that we can generally assume that farmers have access if they choose to use them. Along with Bangalore and the secondary towns, normally every hobli – a small administrative unit of a few villages – has a distributor. No household reported difficulties obtaining the necessary input.

6. For a detailed introduction to P-splines see (Fahrmeir et al., 2013) and (Kneib & Fahrmeir, 2006).

7. The estimation of the model was conducted in R using the package ‘R2BayesX’ (Umlauf, Adler, Kneib, Lang, & Zeileis, 2013), which provides an interface to the free Software ‘BayesX’ for Bayesian inference. For more information on the estimation techniques and inference see Fahrmeir et al. (2013), Kneib and Fahrmeir (2006) and Umlauf, Adler, Kneib, Lang, and Zeileis (2015).

8. Estimation results are available on request.

9. We are grateful to an anonymous referee for suggesting this robustness test.

Additional information

Funding

This research was funded by the German Research Foundation (DFG), [grant number 279374797], and the Indian Department of Biotechnology (DBT) and was supported by the DFG Research Training Group 1644 “Scaling Problems in Statistics”, [grant number 152112243].