This is the chief evidence supporting our general finding that growth originating in different sectors has different effects on households in different parts of the expenditure distribution. Re-estimating this relationship with different unobserved effects strategies yields results that are slightly weaker but broadly consistent—in any event a joint test of the equality of coefficients across sectors for all deciles leads to a rejection of the null hypothesis of equality. The third thing to notice is that while growth from agriculture has a positive and significant finding effect on expenditure growth for the bottom five deciles, growth from other sectors is never significant. One cannot reject the null hypothesis that no decile benefits from non-agricultural sources of GDP growth.13 Point estimates associated with growth from non-agricultural sectors are in fact negative for the bottom eight deciles, but little should be made of this fact, both because none of these coefficients are significant and because even if significant these negative coefficients would only be an indication that non-agricultural income growth reduced expenditure growth relative to the positive mean levels reported in Table 2, not that it actually makes it negative. Taken together these results suggest a progressive effect of income growth originating in agriculture: a one percentage point increase in GDP due to agricultural income growth is associated with a 6.8 percentage point increase in the growth rate of expenditures of the bottom decile, declining to 4.4 percentage points for the fifth decile, and with no effect at all on the highest decile. Note the differences between our instrumental variables and OLS results. Estimated coefficients associated with agriculture are much higher in Table 5, suggesting that the OLS coefficients were attenuated by measurement error,potted blueberries in line with our discussion above in Section 4.2. At the same time coefficients associated with non-agricultural income growth are lower finding and have considerably higher standard errors than in the OLS case.

The results we have which seem quite robust are finding monotonicity of the effects of agricultural income growth across deciles, indicating a certain progressivity in the effects of agricultural income on the distribution of expenditures; and finding significantly different effects from income originating in different sectors. We next turn our attention to the question of whether the effects of income growth from different sectors on expenditures differ across countries. We are interested in particular in whether there is heterogeneity across different particular observable groups. We might expect such heterogeneity to stem from differences in endowments or social conditions across countries. We consider four ways of dividing countries: the initial level of poverty; the initial level of inequality; the initial level of adult literacy;and conclude with the contrast across continents, which of course may capture many differences in endowments and social structures.Our work so far addresses the question of how income growth from agriculture affects the distribution of expenditures within countries, rather than across them. We are now able to say something about the effects of such growth on the global distribution of welfare by asking whether the effects of income growth from agriculture on distribution are different for poorer and wealthier countries within our sample. We group countries by initial level of finding poverty rates and report results in Table 6. In our sample, the median poverty rate across countries is 15.27%, ranging from a minimum of 0.04% in Bosnia and Herzogovina to 81.32% in Burundi.Because we are interested in whether there is heterogeneity in elasticities across the global ranking of poverty rates, not the ranking within continents, we produce these results without the extra continent dummies that we have used in most of the analysis above. There are several things worthy of note in Table 6. First, one can reject the null hypothesis that there is no heterogeneity by country poverty level.

Specifically, one can reject the hypotheses that coefficients are equal across poorer and wealthier countries in the bottom three deciles finding. Second, the robust pattern of monotonicity observed elsewhere for the coefficients associated with agriculture is preserved for poor countries, but the direction is nearly reversed for wealthy countries. Thus, the progressive effect of agricultural income growth on distribution seems to be confined to the poorer half of the countries in our sample. Third, we can reject the hypothesis that agricultural and non-agricultural coefficients are equal within the poorer half of countries finding for the bottom six deciles, in line with the baseline results of Table 5. However, we cannot reject equality among the richer half of countries. This is additional evidence that the progressive effect of agricultural income growth on the welfare of the poor may be confined to the poor in the poorest countries. Our findings here are related to results found by other authors. For example, Christiaensen et al. finding finds that agriculture growth matters most for poverty reduction when using the poverty line of $1 per capita rather that $2 per capita. Though our results are not directly comparable finding, in our analysis the income level of the deciles nevertheless directly measures the intensity of poverty, since lower deciles in poorer countries are typically poorer than lower deciles in richer countries. All of the three notable findings discussed above turn out to also be true when we look at poverty across countries within a continent, by estimating finding while including continent dummies. However, in this specification the addition of extra controls causes the significance of the individual coefficients seen in Table 6 to disappear. Thus, though income growth from agriculture affects the expenditure distribution differently in poor and wealthy countries, the fact that with the addition of continent effects the individual significance vanishes suggests that the continent finding may matter in understanding these patterns, a question we return to below. One interpretation of the results of last section is that it is not really poorer countries, but less equal countries to which this applies, and we only find our earlier results because less equal countries also tend to be countries with more poor. This certainly isn’t a new idea: Datt and Ravallion finding, Ravallion finding, Suryahadi et al. finding, and Christiaensen et al. finding all argue for the possible importance of initial inequality in understanding the effects of aggregate sectoral income growth on poverty. We do not really have the data to distinguish between these two hypotheses adequately, but we can at least divide countries according to initial inequality and see what happens. We divide countries into groups depending on whether their initial Gini coefficient is above or below the median.

Results from this exercise are reported in Table 7. In contrast with our poverty results, dividing countries by initial inequality preserves our earlier monotonicity results for both more and less unequal countries, with agriculture tending to differentially benefit poorer households in both groups. However, we can no longer reject the hypothesis that coefficients are the same across groups, whether we include continent fixed effects or not, leading us to conclude that initial inequality may be a less salient dimension of heterogeneity across countries than is initial poverty.Soils are an important consideration for individuals, community groups, and local governments becoming involved in urban agriculture. In many situations, urban soil has been contaminated and degraded by past uses and activity, including industry,square plastic pot unauthorized dumping, construction, heavy traffic, and adjacent buildings where lead based paint has been applied. Elevated levels of lead in particular are fairly common in urban soils and pose health risks, especially to young children who can ingest soil while playing or helping in gardens. Ongoing exposure to lead can damage the nervous system, interfere with brain development, and create other health problems. Arsenic, cadmium, copper, zinc, and other naturally occurring trace elements in soils, especially heavy metals, can also be elevated to unsafe levels by past land uses. Although soil degradation and contamination are important concerns and should be addressed, they are not always a problem with urban agriculture sites. A study conducted at several community gardens in the Los Angeles area by University of California researchers found that “in nearly all cases concentrations of trace elements were well within natural ranges” finding. In contrast, a study conducted in San Francisco found that “a majority of the gardens exceeded the California Human Health Screening Level for arsenic, cadmium, and lead” finding.Even where there are elevated levels of lead or other heavy metals or contaminants in soil, relatively little is absorbed by plants that can be harmful to humans, although this depends on soil and other environmental conditions as well as on the plant’s characteristics. Accidentally swallowing or inhaling contaminated soil or dust is the most likely way urban farmers will be exposed to unsafe levels of lead or other contaminants. This can happen easily, for example, when people put their fingers in their mouths without thinking. Beyond heavy metals, other sources of soil contamination and soil hazards might include solvents found at sites with a history of manufacturing use, various petroleum-based chemicals common at former gas stations and other industrial sites, chlorinated pesticides and residual herbicides on former agricultural lands or public landscaped areas, saline soil, and sites where a contaminant may not be harmful to people but may prevent plant growth or production.

Physical debris such as lumber, concrete, wire, broken glass, and discarded syringes can also create hazards for the urban farmer. Clearly, there are no easy answers: each site and situation is unique. However, establishing reasonable policies and encouraging sustainable practices will help to ensure that urban farmers and consumers of urban agricultural products are not exposed to unsafe levels of contaminants, including lead and other heavy metals. This publication outlines strategies for urban soil contamination assessment, testing, and remediation; explains best management practices for urban agriculture; and discusses municipal policy concerning safe soils for urban agriculture. This publication does not cover soil fertility or other important soil science topics in depth; additional resources on these topics are presented in the section “Sources of More Information.”Overall soil conditions should be a consideration when selecting a site for urban agriculture. If plants, even weeds, are growing abundantly on the site, it is a good indication that the soil will be able to support crops. If the soil is reasonably easy to dig, it is a positive sign as well. The presence of plant roots and earthworms can indicate soil health. However, these indicators do not guarantee that soil is uncontaminated. When assessing potential sites, be aware that properties with considerable amounts of trash and rubble or obvious dead spots where plants do not grow may pose challenges.A simple test for evaluating soil fertility is to plant bean seeds in soil from the site, perhaps in a pot or biodegradable paper cup, and compare their germination and growth with an equal number of beans grown in purchased potting soil. It is also advisable to dig a hole 1 to 2 feet deep in several places to assess the presence of debris on the site.Learn as much as possible about the history of a proposed site and how it has been used in the past. Walking around the site may provide some clues. Adjacent older homes with peeling paint, paint chips, or evidence of sandblasting finding indicate potential soil lead contamination. Any building built before 1979 that has old or peeling paint may be a hazard due to use of lead based paint. Proximity to a freeway or a heavily trafficked road is also a source of lead. Although leaded gasoline has not been in use since the 1980s, lead particles in vehicle exhaust may have settled from the air into the soil. Talking to the property owner and neighbors is a good strategy, as neighbors are often familiar with past uses of the property. It may also be necessary to do some Internet or library research. For example, at some public libraries and online sources it is possible to access Sanborn maps, which were used by insurance companies to determine the risk involved with insuring individual properties. These maps can provide information about prior uses of a proposed site. Old aerial photographs, which can also be found in local libraries or online, can help identify a site’s history as well. The local city hall may have aerial photographs accessible in their archives.