Measuring microenterprise profits: Must we ask how the sausage is made?

https://doi.org/10.1016/j.jdeveco.2008.01.007Get rights and content

Abstract

A large share of the World's poor is self-employed. Accurate measurement of profits from microenterprises is therefore critical for studying poverty and inequality, measuring the returns to education, and evaluating the success of microfinance programs. However, a myriad of problems plague the measurement of profits. This paper reports on a variety of different experiments conducted to better understand the importance of some of these problems, and to draw recommendations for collecting profit data. In particular, we (i) examine how far we can reconcile self-reported profits and reports of revenue minus expenses through more detailed questions; (ii) examine recall errors in sales, and report on the results of experiments which randomly allocated account books to firms; and (iii) asked firms how much firms like theirs underreport sales in surveys like ours, and had research assistants observe the firms at random times 15–16 times during a month to provide measures for comparison. We conclude that firms underreport revenues by about 30%, that account diaries have significant impacts on both revenues and expenses, but not on profits, and that simply asking profits provides a more accurate measure of firm profits than detailed questions on revenues and expenses.

Introduction

Otto von Bismarck famously remarked that people who like sausages and laws should not see how either of them is made. Economists may wish to say the same about profit data from microenterprises in developing countries. Self-employment and household enterprises are major sources of employment in developing countries, with Vijverberg and Mead (2000) reporting that about one-half of the households sampled in the World Bank's Living Standards Measurement Surveys operate one or more non-farm enterprises. Self-employment is particularly important among the poor: as an example, Banerjee and Duflo (2007) note that 69% of the urban poor in Peru operate a non-agricultural business. Accurate measurement of profits from microenterprises is therefore critical for studying poverty and inequality, measuring the returns to education, evaluating the success of microfinance interventions, and many other important questions of interest. However, a myriad of potential problems plague the measurement of profits. The majority of microenterprises in developing countries do not keep financial records, making data collection generally reliant on recall. Money and goods are fungible between the business and the household. Inputs may be purchased in one period and sold in another, and production can be highly seasonal. And as with other income, individuals may be sensitive about revealing how much they earn, and concerned about the information being used for tax purposes.

In this paper we go inside the sausage factory of profit reporting and conduct experiments to measure the importance of these various problems, and to draw recommendations on how to collect profit data. We use data from two panel surveys of microenterprises conducted in Sri Lanka between 2005 and 2007. In the baseline surveys firm owners were asked directly for their profits in the last month, and also to report revenue and expenses. The level of reported profits is substantially higher than the level of reported revenues minus expenses at both the mean and the median. Moreover, the correlation between reported profits and reported revenue minus expenses is only 0.2–0.3, and 30% of firms have negative revenue minus expenses. Vijverberg (1991) and Daniels (2001) report similar correlations in other countries. They conclude that net revenue, the sum of money from the business used by the household and money left after meeting business expenses and using some money in the household, may be the best single measure, but note that they have no rigid standard to compare this to.

We examine how far one can go towards reconciling the difference between profits and revenue–expenses through asking more detailed survey questions and through better matching of revenues with the expenses incurred to meet these revenues. Responses to questions related to use by the household of enterprise resources reduce the gap in the level of profits compared with revenue minus expenses. But even with these adjustments, the correlations between profits and revenues minus expenses remains in the 0.3 range. We use reported markups of sales over input costs to adjust for the mis-match in the timing of the purchase of inputs and sales resulting from those inputs. We find these corrections bring the levels of the two measures much closer to one another, and lead to a marked improvement in the correlations — to 0.61–0.73. The importance of matching revenues and expenses varies with characteristics of the enterprise, in a manner we discuss further in the body of the paper.

We next turn to the issue of recall error. We ask firms for sales data with different amounts of recall, and find firms understate revenues by about 10 to 15% with recall over four months compared to one month. However, there is little recall error in asking for annual sales compared to asking monthly sales at quarterly intervals. To correct for recall error we randomly allocated some firms ledger books, to keep diary records of firm revenue and expenses. Firms complied well with this over a one month period, but compliance fell over longer periods. The use of diaries led to significantly higher expenses and to higher revenues of similar magnitudes, suggesting that recall leads firms to underreport both revenue and expenses. However, the use of books did not have any sizeable or significant effect on reported profits, suggesting that profits are less affected by recall errors. The use of books does not improve the correlation between reported revenue minus expenses and reported profits.

Finally we examine whether firms deliberately underreport revenues. As in the corruption literature we ask firms about firms like theirs, with the expectation that firms will answer in large part based on their own behavior. The majority of firms think that revenues are underreported, with a median level of underreporting of 20% in one sample and 30% in the other. We had research assistants observe firms in the second sample 15 to 16 times during a month and record transactions, and use this to estimate actual revenues for these firms. The reported revenues of firms are 31% lower than we estimate, confirming the level of underreporting suggested in the self-reports.

Putting the results of these various exercises together, we conclude that direct reports of profits, adjusted for household use of enterprise resources, are likely to be less noisy and at least as reliable as asking firms for all the details of the revenues and expenses. The reports one gets seem to give reasonable rankings across firms in terms of observed transactions, but are likely to understate the true profit levels.

The remainder of the paper is structured as follows. Section 2 describes the data and Section 3 calculates the correlations between profits and revenue minus expenses. 4 Adjusting for unreported categories, 5 Mismatching input purchases and sales adjust for unreported categories and for mismatching of revenue and expenses respectively. Section 6 examines recall errors, Section 7 describes our bookkeeping experiments. Section 8 considers deliberate misreporting and Section 9 asks how well reported profits reflect reality by comparison with wage data. Section 10 concludes.

Section snippets

Data

We use data from two panel surveys of Sri Lankan microenterprises, both designed by the authors and collected between 2005 and 2007. The two surveys use similar instruments; the firms in each sample differ in ways we describe, but all are very small scale enterprises. The first survey we use is the Sri Lanka Microenterprise Survey (SLMS), carried out in three Western and Southern districts of Sri Lanka: Kalutara, Galle and Matara.1

Simple measures of profits

We motivate our initial two measures of business profits on the basis of the way these questions are asked in the Mexican National Microenterprise Survey (ENAMIN), one of the more comprehensive and regular microenterprise surveys carried out in developing countries.4 The first measure of business profits is obtained by asking the

Adjusting for unreported categories

The first explanation we consider for the difference between reported revenue–expenses and reported profits is that there are categories of expenses or forms of profit which are not captured in the basic questions above. After the first round of the SLMS we re-interviewed a subset of firms with large differences between reported revenue minus expenses and reported profits, and asked them to explain the difference between the two measures. These interviews revealed four unreported categories:

  • i)

Mismatching input purchases and sales

A second major source of discrepancy between reported revenue–expenses and reported profits can lie in the timing of transactions. Firms report the amount of cash revenue received and cash expenses incurred during a month. However, inputs purchased in one month may not be sold until another month. As Samphantharak and Townsend (2006) note, this problem becomes more acute the higher the frequency at which data are collected or the shorter the recall period asked. However, the longer the recall

How important are recall errors?

The fact that few microenterprises keep formal business records means that surveys usually must rely on recall to collect business revenue and profit data. Relatively few studies have tried to measure the accuracy of this recall. However, the results of two small studies (summarized in Liedholm, 1991) carried out on 81 entrepreneurs in Honduras and 80 entrepreneurs in Jamaica from 1979–80 give cause for concern. Each study collected data twice-weekly from firms for one year, and then at the end

Bookkeeping experiments

One obvious solution to recall problems, fungibility issues, and mismatching of expenses and revenues is to try and get microenterprise owners to keep better records. This approach has long been a staple of household consumption surveys, with some surveys asking households to keep diaries of all expenditures during a set period. Deaton and Grosh (2000) summarize several studies of this nature, including experiments comparing diaries to recall. The use of diaries was found to increase food

Deliberate misreporting

Issues such as recall problems, fungibility with the household, and mismatching of purchases and sales can all be resolved to some degree through questionnaire design. However, a final cause for concern is the tendency of business owners to deliberately misreport revenue and expenses. As with corruption, it is difficult to directly ask firm owners whether they do this. We therefore followed a common approach in the corruption literature and asked firms to consider other enterprises similar in

How well do reported profit levels reflect reality?

Once we account for the overlap of household and enterprise activities and for the timing of input purchases and sales, we find that the two methods of calculating profits are quite highly correlated with one another. While these adjustments also close the gap the mean or median levels of estimated profits, profit levels reported in response to the direct question on profits average 6300 rupees over seven quarters, about a third higher than the 4900 rupee average of revenues–expenses. Can we

Conclusions and discussion

Few small firms keep business accounts in developing countries, yet data on the profits from such businesses is crucial for answering many important economic questions. We began this paper by showing that there is a very low correlation between what firms report as profits, and what they report as revenue–expenses. We show that a large part of these differences can be reconciled through adjusting revenue–expenses for business goods used for home consumption, and for mismatching of revenues with

References (11)

There are more references available in the full text version of this article.

Cited by (199)

  • Targeting in tax behavior: Evidence from Rwandan firms

    2022, Journal of Development Economics
View all citing articles on Scopus

University of Peradeniya, World Bank, and University of California, San Diego, respectively. The authors thank Susantha Kumara and Jayantha Wickramasiri for outstanding research assistance, and the editor, two anonymous referees, and seminar participants at the World Bank for helpful comments. AC Nielsen Lanka and the Kandy Consulting Group administered the surveys on which the data are based. Financial support from NSF grant # SES-0523167 and the World Bank's Knowledge for Change Trust Fund is gratefully acknowledged.

View full text