Does the Urban Population Pay More for Food? Implications in Terms of Poverty

Abstract

The relation between urban agglomeration and final food consumer prices is controversial. Pressure over the land in large cities results in higher prices in general and for feeding products in particular. On the other hand, large cities provide greater competition among firm, which might drop prices down. Previous literature studying this issue was mainly focused on developing countries, finding empirical evidence of higher food costs in large urban concentrations. Such evidence is missing, however, for developed countries. In this paper, we are interested in measuring the differences in the cost of food products among several city sizes for the case of Spain. A comparison that applies a standard price index would not be appropriate because it would ignore consumer substitution capacity. To make a proper comparison, a “true” food products costs index should be obtained. We have estimated a demand system for food products consumed by Spanish households to measure their costs in cities of different sizes across Spain and over the recent period 2008–2015. The data come from the Spanish Household Budget Survey (HBS). We found that the cost of attaining a given level of utility in food consumption is greater in the largest cities. Additionally, as an example of the political implications of this analysis, we analyze the effect over the quality of life by adjusting the poverty lines with our index and observe that the poverty rates of the largest urban areas in a developed country, such as Spain, might be substantially underestimated if differences in cost of living are not taken into account.

Appendices

Appendix 1

Table 5 AIDS estimates of the model defined in Eq. (14)

Appendix 2. A fixed basket approach

The methodology for calculating indicators of cost of life is not the standard procedure for statistical agencies and policy makers. Instead, an homothetic indicator that assumes a fixed basket of products purchased for all the consumers is usually calculated. This approach is simpler in coumputation and less data intensive but ignores the potential heterogeneity in the preferences of consumers, which might result in a bias (see Paredes and Iturra 2013). In order to highlight the differences between the type of the indicator porposed here and one based on a common basket of commodities imposed to all the consumers, we have calculated a Laspeyres index that sets as basket based on the average shares in the municipalities with more than 100,000 inhabitants in 2008. With these shares (w_r) and the median prices (unit values) in each size of municipalities in each period, the following index has been computed:

$$ {\mathrm{L}}_{hr}=\frac{\sum_j{\overline{p}}_{jh}{w}_{jr}}{\sum_j{\overline{p}}_{jr}{w}_{jr}} $$

(16)

Being ${\overline{p}}_h$ and ${\overline{p}}_r$ the vectors with the median prices paid by the households in areas h and r for the j = 1, …10 food products, respectively, and w_r the vector that contains the average budget shares of the municipalities with more than 100,000 inhabitants in 2008. Table 6 shows the results for the period under study 2008–2015.

Table 6 Laspeyres index and ajusted poverty lines by type of municipality, 2008–2015

Figures in Table 6 reveal that the homothetic index is considerably lower than the Cost of Living Index calculated in Table 2 in the biggest municipalities: if a common basket is fixed, a comparison between 2008 and 2015 on these largest municipalites would result in a very minor variation on the cost of food. Something similar occurs with the smallest municipalities, those with less than 10,000 inhabitants: on average, the Food Cost Index of Table 2 is 3% higher than the Laspeyres index of Table 6. These results indicate that the setting of a common consumption pattern might result in an underestimation of the “true” cost of living. The differences between the two approaches compared here are the consequence of neglecting the potential variations along time and across space in consumption preferences.

Additionally, the poverty line taken as reference in the fourth section has been adjusted now by applying this Laspeyres index, in order to compare the results with those adjusted by applying the non-homothetic index calculated from the AIDS estimates. The lower rows in Table 6 show how these poverty rates can be remarkably different to those reported in Table 3, being the rates shown in Table 3 larger by approximately 2% on average.