Urban Wage Behaviour and Food Price Inflation in Ethiopia

Abstract

Theoretically, increases in food prices could benefit the poor by increasing the demand for unskilled labour, and hence their wages. This paper tests this hypothesis in urban Ethiopia. We exploit a unique panel of monthly price and wage data from 111 urban markets to first construct welfare-relevant measures of real wages, before employing various panel estimators to formally test wage-food price integration. We find moderate rates of long-run adjustment to increases in food prices, but that adjustment is very slow. This implies highly adverse short-run welfare impacts of higher food prices on the urban poor.

Acknowledgements

This paper would not have been possible without the data collection efforts of Ethiopia’s Central Statistical Agency (CSA), to whom we are very grateful. Alemayehu Seyoum Taffesse was instrumental in pointing us to the CSA data, and provided invaluable inputs on earlier drafts. Ibrahim Worku and Mekdim Dereje also provided excellent research assistance. Will Martin, Martin Ravallion, John Hoddinott, Bart Minten and David Stifel provided very useful comments and ideas. Any errors are our own. The statistical code used in this study is available upon request: please email [email protected]

Disclosure statement

No potential conflict of interest was reported by the authors.

Notes

1. In rural areas lack of income data in the 2004/05 national household survey prevent the identification of net food consumption, although Ticci (2011) makes an effort using 1999/2000 income data. Under some assumptions she finds that the rural poor are better off, but under other assumptions they are worse off.

2. Other research in the Ethiopian context has focused on determinants of inflation and price transmission, though we do not review this separate literature here. See Ulimwengu, Workneh, and Paulos (2009) and Haji, Gelaw, Bekele, and Tesfay (2011).

3. At one extreme, applications of Deaton’s (1989) net benefit ratio approach typically focus on the short-run partial equilibrium impacts of higher food prices. In this approach poor urban wage earners are very hard hit by higher food prices since they are all net food consumers. At the other extreme general equilibrium models – such as Jacoby’s (2016) relatively simple model for rural India, and Van Campenhout, Pauw, and Minot’s (2013) computable general equilibrium model for Uganda – model sizeable wage adjustments to higher food prices over the long-run. Ivanic and Martin (2014) experiment with different short-run and long-run assumptions, and clearly show that these assumptions matter a great deal.

4. On the other hand, wage data do not give a complete welfare picture insofar as they say nothing about the importance of potentially detrimental coping mechanisms, such as working longer hours, or cutting back on other expenditures, such as micronutrient-rich foods or education expenditures or enrolment.

5. The data used in Bangladesh, as well as a Philippines study by Lasco et al., (2008), cover a very limited number of markets and are largely measured on an annual basis.

6. Though this estimate of informal labour shares would rise substantially if domestic workers (maids, guards, cooks) were included in these calculations.

7. In Table 1 we have reported unemployment statistics for the sake of completeness, yet Ethiopian data show that the bulk of the urban unemployed are actually relatively well educated.

8. The number of markets in each region is also approximately proportional to the region’s share of the total urban population, to ensure a sufficient degree of national representativeness. The full list comprises 32 markets in Southern Nations Nationalities and Peoples (SNNP) region, 24 in Oromia, and 20 in Amhara (the three biggest regions), while 12 markets are surveyed in Addis Ababa (by far the largest urban centre with around three million people). The smaller regions include only a handful of markets.

9. While we might have used the urban poverty line, we note that since urban poverty was just under 40 per cent anyway this choice would have been immaterial.

10. Calculating these regionally disaggregated indices is quite important because of significant heterogeneity in consumption patterns across the regions. For example, in the SNNP region enset (false banana) is a major staple, but in the rest of the country maize, teff, wheat and some coarse grains are the principles staples (again with variation).

11. These five regions together account for 90 per cent of the country’s total population and the 11,056 market level monthly observations in these regions constitute 82 per cent of the total (CSA 2014).

12. Discussion with local experts revealed that many of the urban poor – such as slum dwellers – pay very low rents, suggesting that the vast majority of their income is indeed spent on food. This share would be expected to rise during periods of high real food price changes.

13. As we shall discuss below, the appropriate lag length and whether the relationship given by Equation (4) holds will be jointly tested.

14. We provide summary results of market-level unit root tests and other results mentioned below in the Online Appendix.

15. We thank an anonymous referee for indicating this point.

16. Given that the DFE and PVECM models assume cross-sectional independence, which is strongly rejected by Pesaran’s test, we only report results that do not make that assumption. However, results obtained from DFE and PVECM models can be obtained upon request.

17. These structural rigidities between rural and urban labour markets are not modelled in Ivanic and Martin (2014).