9 Tips to Improve Your PPI Use >

Frank Ballard

1. Average poverty likelihoods, not PPI scores.

To calculate a poverty rate for a group of individuals, first convert each individual’s PPI score to a poverty likelihood using the Look-up Table. Second (and finally) average all of the poverty likelihoods to calculate the poverty rate. This simple method is the only way to accurately calculate a poverty rate estimate using the PPI.

2. When sampling, ensure that samples are random and representative.

Most organizations collect PPI data to better understand particular groups—for example, new clients or customers of a particular product. Sometimes the only feasible method to collect data on these groups is through sampling. Sampling can be efficient and save time; however, the sample of the groups you wish to research should be randomly chosen and representative of the groups in question. If a sample is not random and representative, it may severely bias your results. For example, if an organization works in many regions within a country, data collected from only one location should not be used to draw a conclusion about the organizations’ total customer base.

3. Forget sampling! Integrate data collection into existing processes.

If your organization comes into regular contact with its clients or customers, take advantage of this time by collecting data from them (provided it is appropriate to do so). Many organizations operationalize PPI data collection by including it into client on-boarding or signup. Doing so often leads to collecting PPI data on all clients, called census PPI data collection. This avoids the issue of sampling altogether and can improve opportunities for data analysis.

4. Stay current, use the most recent PPI for your country.

It is very attractive to continue using an outdated PPI when a new PPI becomes available. Staff won’t have to be trained on the new indicators and systems won’t have to be updated. However, waiting to switch can lead to data of poor quality. As PPIs age, their prediction accuracy becomes more uncertain. Due to changes in an economy and fluctuating asset prices, previous PPI indicators may no longer be associated with poverty. To ensure your data is the most accurate possible, update to the new PPI when it becomes available. In most cases, legacy poverty lines are developed so that data collected with two separate PPIs can be compared

To ensure you are notified, you can sign up to receive our newsletter.

5. Review the Interpretation of Indicators for each PPI.

A PPI’s Interpretation of Indicators is an invaluable source for information on how to correctly administer the PPI. It includes definitions of terms used by the government’s statistics office that help avoid uncertainty during an interview. Should a migrant worker be considered a member of a household? A household has a certain asset, but it’s broken. Should it be counted? These types of questions are typically answered in the Interpretation of Indicators.

At the very least, the manager in charge of training the PPI enumerators should review this document in detail. An even better scenario is one in which all enumerators have reviewed the document.

The Interpretation of Indicators can sometimes be lengthy. We try to simplify the document by creating a short Interview Guide that includes helpful information for the most common issues that enumerators can carry with them during a survey.

6. Keep track of clients and the PPI you are using.

While it may not seem important now, use and track a unique identifier for each interviewee and include space in your database to track the PPI version used (e.g., 2012 India PPI). You’ll thank yourself later when you analyze data over time. Having a unique identifier for clients allows different PPI data collections to be analyzed together. Also, as new PPIs become available for your country, it can be difficult, if not impossible, to compare data over time if you are unsure which PPI version was used. (Also, to compare data over time, you'll need to have the date that your data was collected.)

7. Track more than the PPI.

The PPI enables you to estimate the financial poverty of your clients. While this is an important piece of information to have when working with people living in poverty, tracking more data will enable you to gain a better understanding of your clients. You can track demographic data, transactional data, customer satisfaction surveys, and any other client-level data that may be of interest to your organization. A health clinic recently analyzed acceptance rates with the PPI and discovered that even when it offered free cataract surgery, people that were most likely to be living in poverty tended not take advantage of their services. The clinic used this information to change its strategy for reaching their target population.

While collecting more data allows for further analysis, carefully consider what data you not only would find valuable to know but also would use. Collecting data can be costly. The more data that is collected, the longer that data collection will take and the more likely a respondent will suffer from survey fatigue.

8. Audit your data.

To be sure that the data you collect on customers is of good quality—whether PPI data or any other type—validate it. This involves re-surveying a random and representative sample of respondents and checking for response discrepancies. Consistent discrepancies should be investigated and corrected.

Organizations should also clean their data. This involves reviewing a dataset to:

  1. Remove any records with missing data from data analysis;
  2. Identify unusual responses that indicate poor quality data; and
  3. Find and correct misspellings or variations of the same response (e.g, “first child” and “1st child” mean the same thing, but would be treated as separate during data analysis).

9. Use consistent terminology.

Communicate effectively within your organization by using common terminology about the PPI. Below we list commonly used terms.

census: Census data collection involves collecting data from every member of a population in a defined period. For example, the PPI is often applied as an annual census to all incoming clients. 

Design Documentation Memo: This document is a detailed technical description of the PPI design and construction. The appendices include tables that can be used for targeting.

enumerator: An enumerator is an individual that collects the PPI from households or other individuals.

Interpetation of Inidicators: This document contains the official instructions (where available) for each indicator in the PPI scorecard. The instructions are taken from an enumerator manual for the national household income or expenditure survey. They provide definitions of terminology used for PPI indicators.

Interview Guide: This is a short guide based on the Interpretation of Indicators that includes the most important definitions, examples and suggested ways of asking the PPI questionnaire. It is meant to be an aid for enumerators as they administer the PPI.

(PPI) poverty likelihood: A poverty likelihood is associated with a poverty score. It indicates the probability that a household lives under a given poverty line. It is used to refer to a single household or individual.

poverty line: A poverty line is the level of expenditure deemed necessary to achieve a given standard of living in a particular place.

(PPI) poverty probability: This is the same as the poverty likelihood.

(PPI) poverty rate: A poverty rate is the percentage of people or households in a given population whose expenditure falls below a given poverty line. It is calculated by averaging households' poverty likelihoods.

PPI indicator: A PPI indicator is a question whose responses are correlated with poverty. Responses to indicators are linked with greater or lesser likelihood that a household has expenditure below a poverty line.

PPI look-up table: A PPI look-up table is a chart used to convert PPI scores into PPI likelihoods for a number of poverty lines.

PPI score: A PPI score is the sum of points for all chosen responses to indicators in a PPI scorecard. PPI scores range from 0 (most likely poor) to 100 (least likely poor). PPI scores should be converted into poverty likelihoods using the PPI look-up tables.

PPI scorecard: The PPI scorecard is the PPI survey, along with point values for responses to PPI indicators.

sample: A sample is a subset of a population used for surveying. The purpose of a sample is to learn about the characteristics of the population without conducting a full census.