Syndicate content

Presumed poorer until proven net-seller: measuring who wins and who loses from high food prices

Gero Carletto's picture

(with contributions from Will Martin)

The most recent food price crisis has caught, once again, the statistical systems of most countries with their “pants down”! The low level of preparedness of the system was, yet, another stark reminder of the poor state of affairs still found in national statistical systems of most developing countries. The inability of domestic institutions and their development partners to produce timely and relevant data hinders country policy-makers and their development partners in their efforts to devise informed responses to the pressing issues of the day.

Anecdotally, one would be tempted to infer the existence of a strong positive relationship between higher food prices and poverty. After all, it is the poor who spend a higher share of their food on basic staples and have the least means to buy food with their meager income. And several studies using the available, imperfect data tend to confirm that relationship. This is despite the fact that three quarters of poor people live in rural areas and the majority of them earn their living from farming. Some poor farmers produce more food than they consume and hence benefit from higher prices, but many others are net buyers of food and hence lose out when food prices rise. But identifying which households gain and which lose, and hence the overall impact on poverty, requires knowledge of this relationship for all vulnerable households. A major problem is that we still lack the data for accurately gauging who, for a given level of production and pattern of food consumption and purchases, is more likely to be negatively impacted by higher food prices.

Household surveys, even ones collecting full consumption data, as well as information on production and utilization of food items, such as the Living Standards Measurement Study (LSMS), may not provide an adequate solution in part due to its delimited period of recall. This presses analysts to make simplistic assumptions about the dynamics of household food consumption, purchases and sales in the course of the year.

In regards to food consumption and purchases, one often has to assume that quantity and value of food consumption and purchases are distributed uniformly across the 52 weeks or 26 fortnights of the year. This is because annualization is based on the extrapolation of reported consumption and purchases over much shorter reference periods, most frequently one or two weeks. In some cases, annualization can be somewhat refined if information is collected on the number of months the food items is typically consumed in the course of the year. An imperfect solution that has been tried in some surveys is to ask respondents to recall consumption and purchase patterns for the main staple crops of interest for longer periods, say a quarter or the whole year. This too is likely to result in unacceptably large measurement errors due to poor recall. In regards to agricultural production and utilization, the information is generally collected for a reference agricultural season, usually the last completed season depending on the interview date. However, for accurately determining the amount of a given commodity that a household may have sold, households should ideally be interviewed multiple times between two harvests. On the whole, a survey design that features multiple visits to households as part of a national effort and that could reliably solicit the necessary information on the dynamics of consumption, purchases, and sales throughout a 12-month period is seldom feasible due to its prohibitive costs and logistic complexity. As a result, the available data and statistical techniques can lead to gross misclassification of who is a net-buyer and who is a net-seller of that product in the course of a 12-month reference period, and thus of the impact of a price change.

An alternative and more cost-effective solution would be to complement the data collected through a household survey with more frequent data collection efforts based on “nested” systems of local resident enumerators in each or in a subsample of survey clusters. Whenever possible, the use of cellular phones could also help with the collection, transmission and monitoring of this nested, high-frequency data collection. In light of high demand for all sorts of high frequency data, local resident enumerators could serve multiple purposes, thus lowering the unit costs of the information collected. For instance, agricultural production information, particularly for continuously-harvested crops like cassava or banana, labor inputs or environmental data could be collected at set frequencies and transmitted to headquarters for fast processing. It would also be important for such system to collect food price information to better capture seasonal fluctuations. As shown in various contexts, compared to better-off households, the poor are more likely to sell when prices are low (i.e. immediately after harvest) and buy when prices are high (i.e., in the lean months preceding the harvest), with obvious consequences on their ability to meet their food requirements and our capacity to estimate the implication of food price changes.

An example of such a nested system is being piloted by the Bureau of Statistics in Uganda where more than 200 crop card monitors are collecting ancillary information at a higher frequency on crop production in each rural enumeration area of the Uganda National Panel Survey. Comparison of the method with recall information collected through the household survey supports the claim. The analytical advantages of “nesting” such resident system of high-frequency data collection with a nationally representative household surveys are obvious, and include the possibility of using the rich information from the household survey to understand and cross-validate the findings from the high-frequency data, and vice-versa. Building on such system to collect additional information on purchases and sales of main staple crops, as well as prices, is being explored.

The experience from Uganda also highlights the importance of proper selection, training and supervision of the resident enumerators. In terms of enumerators’ selection, in order to ensure stability and sustainability of the system, the preferable option would be to train local residents that meet some basic criteria i.e. literate and numerate individuals who have deep roots in the community and are unlikely to move out in the near future. However, low levels of literacy as well as rent seeking behavior and local nepotism may be constraints to an efficient recruitment. Because of the innate difficulty of supervision of such highly decentralized system, the importance of putting in place the right incentives for these enumerators must be emphasized.

For the time being, and in absence of better data, it is probably safer to err on the conservative side and presume a poor household worse-off when prices of staple food go up. However, the ability to unambiguously identify who is a net-buyer or a net-seller would greatly help in quantifying the true impact of food price fluctuations and assist policy-makers in properly targeting interventions.