Alternative methods to produce poverty estimates: When household consumption data are not available (Part II)

In Part I of this blog, we discuss the question of how to monitor progress in reducing poverty when the poorest countries of the world are more likely to be missing data for poverty comparisons over time. Drawing from the paper, in this second part of the blog we offer a typology of missing data for poverty comparisons and describe the corresponding candidate imputation approaches for addressing the missing data problem.
 
For presentation purposes, we group all situations of missing data into three broad categories as follows (Table 1):

1. cross-sectional household consumption data are completely missing
2. cross-sectional household consumption data are partially missing, and
3. although cross-sectional consumption data are available, panel household consumption data are (completely or partially) missing

These categories can also be thought of as being ranked, in a roughly decreasing order, according to their severity of data scarcity. A typical example of the surveys that belong to Group A are non-consumption surveys where no consumption data are collected by design, such as the popular Demographic and Health Surveys (DHS), (most) Labor Force Surveys (LFS), and other surveys such as school-based surveys. In fact, implementing a household consumption module requires considerable resources and logistic arrangements – thus, most specialized surveys like the DHS, LFS, as well as almost all small-scale surveys (e.g., project-level monitoring and evaluation (M&E) surveys) would normally fall into this category.

Examples for Group B include (i) rounds of the same household consumption survey that are not comparable over time, (ii) where we have a more recent LFS that has no consumption data but has a similar design to an older household consumption survey, or (iii) where the (cross-sectional) consumption data are unavailable at more disaggregated administrative levels than those in the current survey.  
 
Given the significantly higher costs of implementing longitudinal surveys, it is quite rare for a country to have panel data for monitoring both progress in reducing poverty over time as well as factors leading to change in poverty status at the household level. This is especially true for developing countries.
For example, two major household consumption surveys that are commonly employed to provide poverty estimates in China and India—the China Household Income Project (CHIP) survey and the National Sample Survey (NSS)—are both cross-sectional surveys. For countries where panel surveys exist, few such surveys are likely to be representative of the population over a long period of time and without a great deal of effort. One major reason is that the surveyed household unit can change (e.g., household members can die, split off to form a new household, or simply migrate to another place) and it is very costly to track all household members over time. But as global living standards are rising, this data situation can be improved as greater demand for this type of data emerges and more resources are being invested in fielding panel surveys.
 
In Table 1, we list the imputation methods that can be used to provide poverty estimates in the absence of consumption data. These imputation techniques vary depending on whether the consumption data needed are cross-sectional or panel. For data situations in Group A, the most commonly used method is to generate a wealth index from household assets and the physical characteristics of the house (e.g., the material of the floor or the wall, or which type of toilet is available). For data situations in Group B, techniques have been developed to offer survey-to-survey imputation (i.e., imputation from one survey to another) for sub-cases i and ii, and survey-to-census imputation (i.e., imputation from a survey into a census) for sub-case iii. Finally, data situations in Group C can be addressed with recently developed methods that construct synthetic panel data from cross-sectional data, which can substitute for actual panel data to some extent.
 
We provide more technical details about these techniques and also offer examples with data from Vietnam in the paper. Some of the methods we reviewed are more established, but some are rather recent. While micro survey data are becoming more widely available and collected more frequently in developing countries, we expect these methods to be useful for the foreseeable future, including for back-casting consumption from a more recent survey for better comparison with older surveys.
 
In addition, these imputation methods may also be appropriate in contexts where survey costs and/or survey implementation pose a challenge. For example, perhaps most national statistical agencies are keen on producing annual poverty statistics. But given the costly expenses and demanding logistics of fielding a household consumption survey every year, they often implement household consumption surveys every few years at best, particularly for developing countries. In such contexts, poverty rates can be imputed for the intervening years between the surveys, or for project zones of influence, at just a fraction of the cost of fielding a full-fledged consumption survey by, say, using other non-consumption data or implementing a lighter (non-consumption) version of the survey.
 
Seen in this light, imputation techniques can offer a low-cost and reasonably feasible approach to poverty estimation. While we should be mindful of the various assumptions underlying imputation methods, we would earnestly call for more attention to further developing these methods, particularly validation studies that can provide richer evidence in various contexts.

Table 1: Categories of Missing Household Consumption Data and Commonly Employed Imputation Methods