The goals of this study are: i) to determine the relative importance of spatial factors in explaining household wealth; ii) to identify the spatial characteristics of the chronically poorest, the consistently well off and households escaping from poverty as well as descending into poverty; iii) to determine effects of compound disadvantages on the likelihood of chronic poverty; and iv) to assess the evidence of spatial poverty traps.

Quantitative analysis is conducted using panel data collected from 1275 households, each surveyed four times with a structured questionnaire over an 11-year period from 1997 to 2007. We identified four distinct groups. The chronically poor are defined as households remaining consistently in the bottom third (tercile) of households ranked by wealth in each of the four survey years. Roughly 12.9% of the nationwide sample were found to be ‘chronically poor’. The consistently non-poor are defined as households consistently in the upper tercile of households ranked by wealth, and this group composed 16.2% of the total sample. The third and fourth groups were those households found to have risen from poverty (starting in the bottom tercile and ending in the top tercile, the ‘ascending’) and those who were in the top asset tercile in 1997 and fell to the bottom tercile by 2007 (the ‘declining’). Relatively few households in the sample were in either the upwardly mobile category (3.8%) or the downwardly mobile category (3.6%).

Findings show that spatial factors indeed are a substantial determinant of wealth, explaining a relatively similar share of the total variation in wealth as household-specific factors. The chronically poor and the consistently non-poor households tended to cluster into areas with particular spatial characteristics. Bi-variate analyses show a pattern of correlation between spatial characteristics and chronic poverty. By contrast, there were very few spatial features associated with the location of households rising from and falling into poverty.

With respect to general isolation and remoteness, we find that the chronically poor are disproportionately likely to be far from a motorable road, and more likely to live in areas with relatively little access to education. This is particularly true in terms of higher education. The overwhelming majority (70%) of the chronically poorest households reside in divisions where fewer than one in four household heads have more than eight years of education. This is true of only 21% of the consistently wealthy. Households rising from and descending into poverty are equally likely to come from well-connected or isolated areas.

There is strong evidence that areas with land constraints and with relatively low agricultural potential are more likely to contain chronically impoverished households. Nearly four in five households consistently in the bottom wealth tercile are found in an agriculture zone considered to be of mid-low to lowest potential. Perhaps the most striking determining factor is the prevalence of poverty in areas of land constraints. Nearly 75% of the chronically poor households are found in divisions where median farm size is smaller than two acres. By contrast, fewer than 7% of the chronically poor are in divisions where median farm size is greater than four acres. Statistical correlations indicate that land availability decreases with population density.

The strong correlation between poverty and rising land constraints has been fuelling both poverty and conflict throughout Africa for decades, and there is no reason to expect Kenya to be immune.

Much literature on spatial poverty traps suggests that the likelihood of poverty increases when spatial disadvantages overlap. Results of Probit estimation confirm this, and highlight some specific relationships. For example, low average rainfall, market isolation and land constraints increase the probability of chronic poverty above and beyond their individual effects. We refer to this as ‘compounded effects’ – certain features in combination increase the likelihood of a household being poor more so than the sum of their individual effects.

Although there is strong correlation between spatial factors and static welfare, there are four other important conclusions from the study. First, not all households in areas characterised by ‘spatial poverty traps’ are chronically poor. Although there is some clustering of poor households, they are often surrounded by others that manage to remain above the bottom tercile, or even rise out of poverty in some cases, indicating that spatial factors are not wholly determinant of poverty.

Second, not all chronically poor are in ‘spatial poverty traps’. We see a number of households that are consistently in the bottom third of the sample in terms of wealth, who do not reside in areas of low or variable rainfall, market isolation, severe land constraints or other spatial features found in this analysis, to be correlated with poverty.

Third, there is little or no evidence of spatial factors playing a defining role in the ability to rise from poverty. In fact, the proportion of households that have climbed out of poverty is not greatly different between areas of low and high mean wealth.

Fourth, household-specific factors are also shown to be of considerable importance in explaining the variation in household wealth across this nationwide sample. The degree of variation in wealth within communities is as large as the degree of variation across communities. In fact, results show that the relative explanatory power of spatial factors, though substantial, is slightly less than that of household-specific factors.

Together, these points call into question the appropriateness of defining areas as poverty ‘traps’. While evidence suggests that spatial disadvantages have an increasing and compounding effect on the likelihood of chronic poverty, one’s poverty status and especially one’s ability to escape from poverty are not clearly defined by location. These conclusions, if they are found to hold elsewhere in rural Africa, may warrant a reassessment of whether spatial ‘traps’ or perhaps ‘spatial disadvantage’ may be a more accurate way of describing the spatial dimensions of poverty in this region. Just as there are many composite facets to an area being spatially disadvantaged, there are also many factors driving chronic poverty and poverty dynamics. This includes spatial factors, but also household-specific factors. The considerable heterogeneity of smallholder households typically found even within a given community underscores the limits of conceptualising poverty primarily in spatial terms, and highlights the need for policy to also address the important household-level factors leading to high levels of variation in wealth with communities.