Self employment rates

Measuring Microenterprise Measuring Microenterprise ProfitsProfits

Christopher Woodruff, UC San Christopher Woodruff, UC San DiegoDiego

(Based on joint work with Suresh de Mel and David (Based on joint work with Suresh de Mel and David McKenzie)McKenzie)

Presentation for conference on Innovations in Presentation for conference on Innovations in Development Theory and Survey Data: Implications Development Theory and Survey Data: Implications

for Policyfor Policy

UTCC, Bangkok, August 4-6, 2008UTCC, Bangkok, August 4-6, 2008

Self employment ratesSelf employment rates

Source: Gollin 2002

Measuring microenterprisesMeasuring microenterprises

Important…Important…– Around a quarter of the urban labor Around a quarter of the urban labor

force in a typical low-income countryforce in a typical low-income country– Owners represent an important part Owners represent an important part

of the urban poorof the urban poor But challengingBut challenging

– Lack of administrative dataLack of administrative data– Only about 20% of microenterprises Only about 20% of microenterprises

keep any recordskeep any records

DataData

Data from two surveys in Sri Data from two surveys in Sri LankaLanka– 618 non-ag enterprises in southern 618 non-ag enterprises in southern

Sri Lanka surveyed multiple times, Sri Lanka surveyed multiple times, retail, manufacturing, servicesretail, manufacturing, services

– Data from 174 retail shops in KandyData from 174 retail shops in Kandy 1/3 surveyed monthly1/3 surveyed monthly 1/3 surveyed weekly1/3 surveyed weekly 1/31/3rdrd surveyed every other day surveyed every other day

OutlineOutline

Cash flow vs. profitsCash flow vs. profits Recall issuesRecall issues Deliberate mis-reportingDeliberate mis-reporting

– GeneralGeneral– Specific to interventionsSpecific to interventions

Prospects for obtaining harder Prospects for obtaining harder data?data?

What is the best measure of What is the best measure of performance?performance?

Profits vs. cash flowProfits vs. cash flow– Possible that perhaps owners don’t Possible that perhaps owners don’t

understand well the concept of understand well the concept of profitsprofits

– Very low correlations between Very low correlations between profits and R-C: around 0.2, in our profits and R-C: around 0.2, in our data and in other datadata and in other data

– Levels differ markedly as well, with Levels differ markedly as well, with R-C much lowerR-C much lower

Profits and R-CProfits and R-C

Profits:Profits:““What was the total income the business earning What was the total income the business earning

during the month of March after paying all during the month of March after paying all expenses including the wages of employees, but expenses including the wages of employees, but not including any income you paid yourself. That is, not including any income you paid yourself. That is, what were the profits of your business during what were the profits of your business during March?”March?”

Revenue / CostRevenue / Cost– Costs in eleven categoriesCosts in eleven categories

Profits v. (R-C)Profits v. (R-C)

Table 2: Reported Profits Vs Reported Revenue Minus Expenses

round 1 retail manufacturingMean 3357 5257 3598S.D. 2534 4095 3212Median 3000 4000 2650% negative 0 0 0Mean 1174 3017 1271S.D. 7368 7631 5548Median 1000 2623 1425% negative 29 27 26

Correlations: Pearson 0.286 0.468 0.202

SLMS

Reported Revenue Minus Expenses

Reported Profits

Why is the cash flow/profit Why is the cash flow/profit correlation so low?correlation so low?

retail manufacturingPercentage of firms reporting that:Goods used for home consumption 63.5 14.7Firm inputs given as gifts but reported as a business expense 2.3 2.3Firm owners pay themselves a salarybut fail to include this as profits: 0.7 1.7Business revenue is used to payhousehold expenses, but is notcounted as profit 13.3 14.1Adjusting Expenses and Profits

Mean 5752 4199S.D. 4471 3929Median 4810 3000% negative 0 0Mean 4976 1883S.D. 9290 7470Median 3825 1615% negative 21 25

Correlations: Pearson 0.457 0.174

Adjusted Revenue Minus Expenses

SLMS

Adjusted Profits


retail manufacturingPercentage of firms reporting that:Goods used for home consumption 63.5 14.7Firm inputs given as gifts but reported as a business expense 2.3 2.3Firm owners pay themselves a salarybut fail to include this as profits: 0.7 1.7Business revenue is used to payhousehold expenses, but is notcounted as profit 13.3 14.1Adjusting Expenses and Profits

Mean 5752 4199S.D. 4471 3929Median 4810 3000% negative 0 0Mean 4976 1883S.D. 9290 7470Median 3825 1615% negative 21 25

Correlations: Pearson 0.457 0.174


SLMS

Adjusted Profits

Timing mismatchesTiming mismatches

Inputs purchased one month may be sold Inputs purchased one month may be sold in a later monthin a later month– SeasonalitySeasonality

We asked firms for markup between input We asked firms for markup between input and sales prices (up to three products)and sales prices (up to three products)

““Consider the most important item which you Consider the most important item which you manufacture. If you buy Rs. 1000 worth of raw materials manufacture. If you buy Rs. 1000 worth of raw materials how much of revenue will you receive from the final how much of revenue will you receive from the final products that you manufacture with these raw materials on products that you manufacture with these raw materials on

average?”average?”


Table 4A: Adjustments Using Markups

retail manufacturingMean 8508 6238S.D. 6044 5663Median 7000 5000% negative 0 0Mean 7431 3401S.D. 20967 9110Median 5775 2100% negative 21 20Mean 7227 3938S.D. 6859 5039Median 5540 2188% negative 8 9

Correlations: Adjusted Profits Pearson 0.609 0.733


Adjusted Profits

Markup-Adjusted Revenue Minus Expenses

SLMS


Table 4A: Adjustments Using Markups

retail manufacturingMean 8508 6238S.D. 6044 5663Median 7000 5000% negative 0 0Mean 7431 3401S.D. 20967 9110Median 5775 2100% negative 21 20Mean 7227 3938S.D. 6859 5039Median 5540 2188% negative 8 9

Correlations: Adjusted Profits Pearson 0.609 0.733


Adjusted Profits

Markup-Adjusted Revenue Minus Expenses

SLMS


Table 4B: Markups and Inventory Turns

Pearson 0.698 0.138Spearman 0.767 0.297p-value 0.000 0.000Pearson 0.701 0.659Spearman 0.742 0.698p-value 0.000 0.000

Correlations: Adjusted Profits vs. Markup Adjusted

≥ 75% of inventories

sold in same month

< 60% of inventories

sold in same month

Correlations: Adjusted Profits vs. Adjusted Rev-Exp


This is the same timing mismatch that This is the same timing mismatch that Samphantharak and Townsend (2008) Samphantharak and Townsend (2008) resolve using monthly data on inventoriesresolve using monthly data on inventories– Clearly, better to have monthly dataClearly, better to have monthly data– We had no success getting monthly data in We had no success getting monthly data in

quarterly surveys quarterly surveys (beginning of last month, end of last (beginning of last month, end of last month, expected end of this month) month, expected end of this month)

– Markup allows this correction to be made Markup allows this correction to be made even in cross-sectional dataeven in cross-sectional data

– We interpret the high correlations as We interpret the high correlations as suggesting that owners do understand what suggesting that owners do understand what ‘profits’ are‘profits’ are

Profits v R-CProfits v R-C

Improvements in the correlations Improvements in the correlations of the measures are reassuring of the measures are reassuring

We interpret the data as We interpret the data as suggesting that profits are likely a suggesting that profits are likely a better reflection of reality than R-better reflection of reality than R-C, based on a comparison of levelsC, based on a comparison of levels– In baseline month, unskilled workers earn 6-7000 In baseline month, unskilled workers earn 6-7000

LKR/monthLKR/month– Owners say they would need 8000 LKR in wages to Owners say they would need 8000 LKR in wages to

shut downshut down

Recall issuesRecall issues

Compare March 2005 sales reported Compare March 2005 sales reported in April with March 2005 sales in April with March 2005 sales reported in July. Later report is 10% reported in July. Later report is 10% lower on average, 16% at medianlower on average, 16% at median

But, compare annual sales reported But, compare annual sales reported in April 2006 with sum of monthly in April 2006 with sum of monthly sales reported quarterly. Annual sales reported quarterly. Annual recall only 3% lowerrecall only 3% lower

Recall issues: 1 month vs. 4Recall issues: 1 month vs. 4

Recall issues: AnnualRecall issues: Annual

Recall issues: the effect of Recall issues: the effect of booksbooks Provided simple ledgers for half of Provided simple ledgers for half of

the enterprises (daily/weekly the enterprises (daily/weekly recording)recording)– Expenses on inputsExpenses on inputs– Other expensesOther expenses– Goods taken from businessGoods taken from business– Total business revenueTotal business revenue– Business income used by householdBusiness income used by household

Recall issues: the effect of Recall issues: the effect of booksbooks Find:Find:

– Increased reported revenue and Increased reported revenue and expensesexpenses

– Much higher level of goods taken for Much higher level of goods taken for home usehome use

– Little effect on reported profitsLittle effect on reported profits

Compliance with booksCompliance with books

Compliance fairly high for first few Compliance fairly high for first few monthsmonths– 68% in first month, 53% in third, 60% in 68% in first month, 53% in third, 60% in

fourth, 43% in ninth monthfourth, 43% in ninth month– In April 2008, only 17% still keeping books, In April 2008, only 17% still keeping books,

only 8% as detailed as the ledgersonly 8% as detailed as the ledgers In Kandy, with weekly interviews, In Kandy, with weekly interviews,

higher compliancehigher compliance– One month later, 52% still keeping books; One month later, 52% still keeping books;

a year later, only 20% keeping books in a year later, only 20% keeping books in anything like this formatanything like this format

Deliberate misreportingDeliberate misreporting

We might suspect firms would We might suspect firms would underreport profitsunderreport profits– 2/3rds say ‘firms like theirs’ under 2/3rds say ‘firms like theirs’ under

report profitsreport profits Fear of taxesFear of taxes Trying to show difficulty of running Trying to show difficulty of running

businessbusiness Don’t want to reveal true state of Don’t want to reveal true state of

businessbusiness

Deliberate misreportingDeliberate misreportingTable 9: If the truth is 10,000, how much would owners report?

ProfitKMS-all SLMS KMS SLMS SLMS

Mean 6744 8204 12254 11417 780125th percentile 5000 7000 10000 10000 600050th percentile 7000 8000 12000 10000 800075th percentile 8000 10000 15000 13000 10000

ExpensesRevenue

Only weak evidence of reversion to the mean, typically found in wage data

Do the ‘firms like yours’ reflect own Do the ‘firms like yours’ reflect own firms?firms?

For 1/3For 1/3rdrd of the sample in Kandy, of the sample in Kandy, RAs visited every 2 days, RAs visited every 2 days, recorded all transactions for an recorded all transactions for an ~hour each visit.~hour each visit.

Compare reported revenue with Compare reported revenue with revenue estimated from these revenue estimated from these visits, and with the average % visits, and with the average % under reportingunder reporting

Do the ‘firms like yours’ reflect own Do the ‘firms like yours’ reflect own firms?firms?

Table 10: Comparison of Reported Sales to Directly Monitored SalesFor 58 KMS firms receiving visits every 2 days

Mean S.D.Hours of transactions observed in month 8.2 1.8

Reported Sales for September 37090 26205Estimated Sales based on observedtransactions and weekly hours reported 53630 39224

Implied % underreporting 30.8

Compare with “firms like yours” response: average under reporting of 32.5%

Impact of interventionsImpact of interventions

Deliberate mis-reporting is Deliberate mis-reporting is particularly problematic for particularly problematic for analyzing impact of interventionsanalyzing impact of interventions

Two models of behavior:Two models of behavior:– Warm glow (wears off over time)Warm glow (wears off over time)– Switch to “honest reporting” state Switch to “honest reporting” state

(might not wear off)(might not wear off) Ideally, would have ‘harder’ dataIdeally, would have ‘harder’ data

Prospects for harder data?Prospects for harder data?

In some contexts administrative In some contexts administrative data is an optiondata is an option– Wage workersWage workers– MFI interventions (repayment)MFI interventions (repayment)

But generally any administrative But generally any administrative data on incomes and revenues of data on incomes and revenues of microenterprises will be less microenterprises will be less rather than more reliablerather than more reliable

Harder measuresHarder measures

One possibility is independent One possibility is independent valuations of inventoriesvaluations of inventories

A. Rangana Stores- Doluwa Road, Hindagala. (July 14, 2008) 1. Enterpriser Owner’s answer about current inventory= around Rs. 150,000 2. RA (Susil) manual measurement = Rs. 214,000 (around 2 hours)

Research Assistant’s Visual Estimate: Time spent

RA1. M. Chanaka Rs. 170,000 20 minutes RA3. K.G.M. Kodagoda Rs. 260,000 20 minutes RA4. D.M.R.K. Disanayaka Rs. 204,000 20 minutes RA6. M.M.E.M Gnanarathna Rs. 185,000 20 minutes B. Kapila Stores, Hindagala Junction, Hindagala (July 14, 2008) 1. Enterpriser Owner’s answer about current inventory= No answer given 2. RA (Susil) manual measurement = Rs. 115,000 (around 2 hours)

Research Assistant’s Visual Estimate: Time spent

RA2. K.G.C.N.S. Kehelwatta Rs. 105,000 20 minutes RA5. Rasika Rajapaksha Rs. 106,600 20 minutes


A poor man’s version of A poor man’s version of Samphantharak and TownsendSamphantharak and Townsend

Low correlations between profit reflect Low correlations between profit reflect timing mismatches more than a timing mismatches more than a misunderstanding of profitsmisunderstanding of profits

We find little evidence of long-term We find little evidence of long-term recall bias, no effect of books on recall bias, no effect of books on reported profitsreported profits

Monitoring of sales levels suggests Monitoring of sales levels suggests firms under report sales by 30%, about firms under report sales by 30%, about what they say “firms like their” do.what they say “firms like their” do.


Measurement issues more Measurement issues more problematic when combined with problematic when combined with interventionsinterventions

Did not address a couple of Did not address a couple of relevant issues:relevant issues:– Allocation of assets used jointly by Allocation of assets used jointly by

business and householdbusiness and household– Depreciation of fixed assetsDepreciation of fixed assets

Self employment rates

Documents

cash flowprofit correlation

cash flowpossible

month markup

survey data

harder data

annual sales

sum of monthly sales

monthly datawe