Risk Sharing and Transaction Costs: Evidence from Kenya s Mobile Money Revolution. William Jack and Tavneet Suri

Risk Sharing and Transaction Costs: Evidence from Kenya s Mobile Money Revolution William Jack and Tavneet Suri

Research Questions What is the role of the financial sector in development? How important is access to financial intermediation? Cell phone innovations are leap-frogging : allow financial transactions across areas where banks are thin Here we think about dramatically reducing transactions costs in an economy where networks are important

Research Questions In developing economies, govt. safety nets (health, u/e insurance) do not exist; neither does formal insurance A large fraction of Sub-Saharan Africa is agricultural (low and extremely variable incomes) Households often resort to informal mechanisms to smooth risk between them (e.g. Suri, 2011)

Research Questions Lots of evidence that HHs help each other smooth risk But this insurance is not perfect or efficient Why not? Reasons in the literature: Moral hazard Asymmetric information Commitment problems A more boring reason: transaction costs

Summary of Findings There are gains to smoothing consumption from lowering transaction costs The consumption of households who use mobile money is about 7% - 10% less sensitive to income shocks Transactions costs pose a significant barrier to optimal risksharing Simple technologies like mobile money can alleviate such inefficiencies

Outline of Presentation Simple theoretical implications Background on mobile money in Kenya Testing the theory Survey data Results on consumption smoothing Results on remittances Falsification test and other robustness checks Conclusions

Theoretical Implications Set up a simple three person risk sharing environment Use a simplex to understand the role of transaction costs As transaction costs are lowered, households are able to smooth risks more completely Consumption of individuals with access to the technology will be less responsive to shocks Households will share risk with more members of their network

Outline of Presentation Simple theoretical implications Background on mobile money in Kenya Testing the theory Survey data Results on consumption smoothing Results on remittances Falsification test and other robustness checks Conclusions

Background on Mobile Money Called M-PESA Remote account storage accessed by simple SMS Cash-in/out services provided by M-PESA agents Limits on transaction sizes ($500) and on money stored on account ($750) Fees charged on all transactions except deposits

Growth of Mobile Phones in Kenya 14 12 Millions of subscribers 10 8 6 4 2 0 1999 2001 2003 2005 2007 2009 Fixed lines Mobile lines

Adoption of M-PESA 14000000 12000000 Number of Registrations 10000000 8000000 6000000 4000000 2000000 0 Apr-07 Oct-07 Apr-08 Oct-08 Apr-09 Oct-09 Apr-10 Oct-10

What do People Use M-PESA For? 100% 80% 60% 40% 20% 0%

Frequency of M-PESA Use 40% 30% 20% 10% 0% Daily Weekly Every 2 Weeks Monthly Every 3 Months Every 6 Months Less Often 2008 2009

M-PESA Agents Individuals trade e-float for cash with M-PESA agents Stand-alone agents, shop-keepers, supermarkets, gas stations, etc. Agents must have a bank account and internet access Face a non-trivial inventory management problem, predicting customer demand for both e-float and cash

Growth of Agent Network 25000 20000 Number of Agents 15000 10000 5000 0 Apr-07 Oct-07 Apr-08 Oct-08 Apr-09 Oct-09 Apr-10 Oct-10

Agent Network: June 2007

Agent Network: Dec 2007

Agent Network: June 2008

Agent Network: Dec 2008

Agent Network: June 2009

Agent Network: Dec 2009

Agent Network: April 2010

4 Distance to the closest agent (km) 3.5 Improving Agent Access 3 2.5 2 1.5 1 22% Change 40% Change 28% Change 14% Change 33% Change Round 1 Round 2 0.5 0 Mean Distance (km) 5th Percentile 25th Percentile 50th Percentile 75th Percentile Percentiles of Distance to Closest Agent

Financial Intermediation in Kenya

Transaction Costs I 1,400 1,200 1,000 Tariff 800 600 400 200 0 0 5,000 10,000 15,000 20,000 25,000 30,000 35,000 40,000 Amount deposited and sent Postapay M-PESA: Reg to reg Western Union

Transaction Costs II Frequency Round 1 Cost (KShs) Hand Delivery by Self 13.5% 1.6 Bus Delivery Through Driver 3% 158.7 Western Union 0.4% 108 Postal Bank 2.9% 173.1 Direct Deposit 6.7% 85 M-PESA 60.8% 49.8 Other 3.3% 78.0 Average distance travelled is 200km which is a 400 KShs ride

Outline of Presentation Simple theoretical implications Background on mobile money in Kenya Testing the theory Survey data Results on consumption smoothing Results on remittances Falsification test and other robustness checks Conclusions

Household Survey 3,000 households Rural and a lot of urban and semi urban 3 rounds (so far) 2008, 2009, 2010 Non-negligible attrition, but not bad for a survey that has urban and semi urban areas included

Households Surveyed Uganda Somalia Tanzania Indian Ocean

Network Coverage Nairobi

Economic Activity POPULATION DENSITY (number of people per sq. km) > 600 300-600 100-300 50-100 20-50 <= 20 No data OTHER FEATURES District boundaries Selected national parks and reserves Water bodies

Summary Statistics I Round 1 Round 2 Mean SD Mean SD M-PESA User 0.432 0.496 0.698 0.459 Own Cell Phone 0.692 0.462 0.758 0.428 Per Capita Consumption 73137 131229 64025 87078 Per Capita Food Consumption 31825 31123 30092 25612 Total Wealth 129447 422649 136954 700517 HH Size 4.285 2.224 4.398 2.324 Education of Head (Years) 6.974 5.670 7.546 5.008

Summary Statistics: Economic Round 1 Round 2 Mean SD Mean SD Financial Access Dummies Bank account 0.504 0.500 0.515 0.500 Mattress 0.759 0.428 0.750 0.433 Savings and Credit Cooperative 0.189 0.391 0.176 0.381 Merry Go Round/ ROSCA 0.405 0.491 0.460 0.498 Household Head Occupation Dummies Farmer 0.289 0.453 0.273 0.445 Professional Occupation 0.232 0.422 0.195 0.397 Househelp 0.093 0.290 0.103 0.304 Run a Business 0.145 0.353 0.162 0.369 Unemployed 0.063 0.244 0.077 0.267

Summary Statistics: By Adoption Early Adopters Late Adopters Non- Adopters Own Cell Phone 0.940 0.885 0.368 Per Capita Consumption 87728 57380 38371 Education of Head (Years) 8.683 7.701 5.611 Negative Shock 0.604 0.527 0.578 Agricultural Shock 0.126 0.115 0.144 Illness Shock 0.441 0.357 0.410 Send Remittances 0.660 0.506 0.167 Receive Remittances 0.556 0.484 0.175 Bank account 0.733 0.522 0.184 Mattress 0.679 0.745 0.857

Summary Statistics: Remittances I Round 1 Round 2 Sent Received Sent Received Overall Remittances No of Remittances per Month 2.860 2.211 2.375 1.929 Total Value 10065.8 13006.9 7059.5 5093.7 Total Value (% of Consumption) 0.036 0.050 0.033 0.029 Average Distance 234.3 288.6 214.3 235.0 Net Value Received 2354.2-882.3 Demand for overall remittances fell from round 1 to round 2

Summary Statistics: Remittances II Round 1 Round 2 Sent Received Sent Received M-PESA Remittances Number of Remittances 0.931 0.805 1.616 0.847 Total Value 7,965 9,924 7,879 4,790 Average Distance 344 335 239 237 Non M-PESA Remittances Number of Remittances 1.930 1.406 0.759 1.080 Total Value 9,709 13,674 4,615 5,058 Average Distance 195 274 172 231 Shift into M-PESA; and M-PESA sent over longer distances

Outline of Presentation Simple theoretical implications Background on mobile money in Kenya Testing the theory Survey data Results on consumption smoothing Results on remittances Falsification test and other robustness checks Conclusions

Empirical Strategy I Use an extension of Gertler and Gruber (2002) specification: where c ijt is consumption of HH i in location j at time t Shock is a measure of the income shock β is the coefficient of interest X ijt are covariates (demographics, economic)

Empirical Strategy II Consumption c = Shock + User + User * Shock + controls Shocks don t hurt users so much ( ) User Users are richer ( ) Non-user ( ) Shocks hurt ( ) Shock No shock Shock status

Basic Results OLS A Panel A Panel Without Nairobi M-PESA User 0.553*** -0.090** -0.016-0.008 [0.037] [0.036] [0.047] [0.049] Negative Shock -0.207*** 0.241** 0.232 0.120 [0.038] [0.116] [0.169] [0.141] User*Negative Shock 0.101** 0.176*** 0.156** 0.150** [0.050] [0.050] [0.062] [0.065] Shock, Users -0.105*** 0.052* 0.055 0.050 [0.033] [0.028] [0.035] [0.037] Shock, Non-Users -0.207*** -0.069** -0.068-0.056 [0.038] [0.032] [0.043] [0.045]

Different Shock Measures Total Consumption Non-Health Consumption Weather Shock Illness Shock Illness Shock M-PESA User -0.0260-0.0446-0.0279 [0.0358] [0.0420] [0.0407] Negative Shock -0.0603-0.0704-0.2052 [0.3352] [0.1640] [0.1686] User*Shock 0.3329** 0.1547** 0.1595** [0.1511] [0.0738] [0.0692] Shock, Users -0.0878 0.0545 0.0101 [0.0903] [0.0418] [0.0404] Shock, Non-Users -0.2084*** -0.0623-0.1275** [0.0959] [0.0500] [0.0483]

Using Agent Roll Out Agents w/in 1km Agents w/in 2km Agents w/in 5km Agents w/in 20km Distance to Agent Negative Shock 0.152 0.122 0.148-0.176 0.619*** [0.152] [0.153] [0.160] [0.140] [0.203] Agents -0.022-0.003 0.018-0.002 0.051 [0.039] [0.031] [0.024] [0.006] [0.054] Agents*Shock 0.055*** 0.050*** 0.021** -0.002-0.058*** [0.019] [0.015] [0.010] [0.005] [0.019]

Agent Roll Out: Correlates I Agents w/in 2km Agents w/in 5km Dist to Agent Coefficient Coefficient Coefficient Log Wealth 0.0225* 0.0091 0.0022 Cellphone Ownership -0.0318-0.0272-0.0016 HH Head Can Read -0.0393 0.0141 0.0382* HH Head Can Write -0.0252 0.0273 0.0256 HH Head Education -0.0024 0.0011-0.0010 HH Has a Bank account 0.0285 0.0252-0.0011 HH has a SACCO -0.0180 0.0159-0.0128 HH has a ROSCA 0.0376-0.0128 0.0076 Negative Shock 0.0240 0.0002 0.0023 Rainfall Shock -0.0071 0.0028-0.0009 Illness Shock -0.0014-0.0056-0.0037

Agent Roll Out: Correlates II Distance to Nairobi Agents w/in 1km Agents w/in 2km Agents w/in 5km Dist to Agent Period 1 Changes Period 1 Changes Period 1 Changes Period 1 Changes -0.0026-0.0016-0.0100* -0.0042 0.0151-0.0037 0.0001-0.0020 [0.0028] [0.0013] [0.0055] [0.0028] [0.0091] [0.0045] [0.0056] [0.0012]

Remittances Overall Shock Illness Shock Prob [Receive] Number Received Total Received (Root) Prob [Receive] Total Received (Root) M-PESA User 0.160*** 0.253** 10.77*** 0.182*** 12.48*** [0.047] [0.127] [3.71] [0.041] [3.079] Shock -0.030 0.032 2.613-0.187-8.556 [0.143] [0.427] [11.70] [0.149] [11.13] User*Shock 0.135** 0.343* 8.067* 0.144** 8.385 [0.063] [0.177] [4.668] [0.070] [5.312] Shock, Users 0.066* 0.104 5.180 0.071* 6.470** [0.037] [0.112] [3.283] [0.042] [3.289] Shock, Non-Users -0.028-0.094-0.397-0.044-0.599 [0.041] [0.120] [2.652] [0.044] [3.061]

Network Size? Distance Travelled Network Size Fraction of Network Overall Illness Overall Illness Overall Illness M-PESA User 71.35-16.93 0.174*** 0.194*** 0.102*** 0.116*** [63.50] [53.52] [0.065] [0.053] [0.036] [0.031] Shock -111.7-111.3-0.264-0.478** -0.024-0.199 [130.6] [149.5] [0.211] [0.223] [0.131] [0.126] User*Shock -186.6** -9.33 0.203** 0.253*** 0.101** 0.110* [81.0] [90.86] [0.087] [0.097] [0.048] [0.060] Shock, Users -57.71* -10.03 0.112** 0.121** 0.046* 0.045* [31.31] [40.46] [0.056] [0.057] [0.024] [0.026] Shock, Non-Users 94.07-79.23-0.026-0.057-0.007-0.014 [63.49] [71.99] [0.058] [0.062] [0.038] [0.044]

Outline of Presentation Simple theoretical implications Background on mobile money in Kenya Testing the theory Survey data Results on consumption smoothing Results on remittances Falsification test and other robustness checks Conclusions

Falsification Test I Agents w/in 2km Maize Consumption Crop Consumption OLS Panel OLS Panel Shock*Agents -0.009-0.058 0.091 0.055 [0.083] [0.068] [0.085] [0.065] Shock Measure (Positive) 0.418*** 0.412*** 0.400*** 0.377*** [0.074] [0.068] [0.069] [0.062] Agents -15.181-13.537 [16.855] [16.796]

Falsification Test II Total Food Food Expenditure Dist to Agent Agents w/in 2km M-PESA User -0.0031-0.0353 [0.0740] [0.0763] Shock 0.1349 0.0446 0.6298** -0.0308 [0.1915] [0.1995] [0.2990] [0.1987] User*Shock 0.2140** 0.1756** [0.0866] [0.0872] Agent Variable -0.0001-0.0738 [0.1068] [0.1023] Agent*Shock -0.0781** 0.1070*** [0.0318] [0.0350]

Shocks Coefficient SE M-PESA User -0.0228 0.0287 Cellphone Ownership -0.0267 0.0319 Agents within 1km 0.0033 0.0263 Log Distance to Agent 0.0089 0.0490 HH Head Education 0.0034 0.0026 HH Has a Bank account 0.0033 0.0310 HH has a SACCO 0.0070 0.0247 Occupation - Business -0.0715** 0.0353 Occupation Farmer 0.0450 0.0352 Occupation - Professional -0.0130 0.0338 Occupation - Sales 0.0579 0.0461 Household size 0.0106 0.0105

Conclusions I Life in developing countries can be precarious Crop failure, health shocks, job loss, etc. Given M-PESA reduces transaction costs so much and seems to be used primary for P2P transfers, it may help reduce vulnerability When facing an income shock, households can easily receive (and spend) money from friends and family

Conclusions II Mobile banking expands access to financial services It improves the ability to weather serious shocks Negative shocks cause non-users of M-PESA to reduce their consumption by about 7% M-PESA users can smooth these shocks perfectly Some of this from users being wealthier, educated, etc. But, M-PESA itself a significant source of risk sharing

Thank You

How do People Send Money? Direct deposit 7% Other 6% Hand 32% M-PESA 46% Bus 9%

Without Transaction Costs

With Transaction Costs