Segmenting the markets for savings among the poor across countries potx

Segmenting the markets for savings among the poor across countries Report prepared for the Bill and Melinda Gates Foundation By Bankable Frontier Associates1, Somerville MA Executive

Segmenting the markets for savings among the poor across countries

Report prepared for the Bill and Melinda Gates Foundation

By Bankable Frontier Associates1, Somerville MA

Executive summary

Findings:

• We analyze macro-level FinScope datasets from seven African countries and a micro-level household panel from South Africa and find quantitative evidence supporting savings patterns observed elsewhere: the poor do save, using a variety of formal and informal savings instruments, and a substantial percentage save proportionately more than higher income neighbors within the same community

• The strong take-up in South Africa of a new category of basic bank accounts demonstrates the strong desire for appropriate formal products among those who have never banked before Attitudes towards savings differentiate those among the poor who adopted the product as their first bank account ever from those who did not However, the ‘dropout’ rate, at 23% of adopters, is high

• Merely opening or having a savings account is not the same as using it regularly Savings may also be measured by its intensity of saving (how much of income is saved) as well

as its duration (the period over which it accumulates before being accessed) The portfolio of savings in financial instruments is categorized into four clusters by duration and formality; and the flows into each cluster are quantified

Implications for a strategy to scale up savings of the poor:

• The objective must be carefully specified: merely counting new savings accounts opened does not capture underlying savings activity; and given the already high intensity of savings, poor people may be unable to save much more as a proportion of income However, households may choose to rebalance their portfolio of financial assets towards safer and longer duration instruments which match their timing needs This re-balancing effect should also be measured following new product introduction

• In order to test a business case, and/or justify subsidy, the potential size of the savings market among the poor needs to be measured This could be done by segmenting the likely market using the combination of a basic standard national survey, building on the methodology developed by Fin-Scope, and a more detailed look at household flows provided by the micro-level study

• We quantify here the possible effect if poor households were to rebalance their portfolios of financial assets using actual numbers from South Africa: more than $159 million may flow into formal savings instruments within a year

Recommendations

• We recommend the collection of baseline data in each country which enables a baseline to be drawn; and further research into adoption patterns of successful savings patterns in order to inform segmentation of the market for savings In particular, there

is value in analyzing cases where additional savings options have been added to basic

bank accounts so that households can easily diversify their savings portfolio

1 This report was written by a project team comprising David Porteous, Daryl Collins, Jeff Abrams and David Toniatti Marguerite Robinson has provided useful comments throughout

1 Introduction


The microcredit movement has demonstrated that poor people can and do repay loans We have known for a while that poor people also can and do save (Rutherford 2000, Robinson 2004) However, discussions about savings instruments and behavior often remain undifferentiated: Central banks publish average national household savings rates which cover a multitude of household types and circumstances; and providers offer ‘one size fits all’ savings products But who actually saves among the poor? And how?

One of the lessons from rapid growth of consumer credit is the need to differentiate among potential customers—not only in terms of their risk worthiness but also their propensity to behave in certain ways Better segmentation and understanding of the potential market for savings instruments is helpful in at least two respects: First, by enabling better market sizing,

it can help financial institutions assess the business case for building ‘big pipes’, that is, basic bank accounts offered en masse which can be used for saving Such accounts are also usually the first step for unbanked people on the ladder of formal financial instruments We accept that such ‘pipe-laying’ is indeed necessary as an initial approach to connect large numbers to the formal financial system However, it may not be sufficient to sustain regular usage of the new accounts or to serve the different savings needs of many of the newly-connected A next phase of product development, based on finer segmentation, may offer more tailored and wider-ranging product features Second, then, finer segmentation supports the development of

‘add-on’ savings products which service diverse needs

This paper examines how different groups among the poor are saving based on evidence arising from two data sources which are further described in immediately following sub-sections:

• for cross-country analysis and single country adoption analysis, we draw on cross-country

FinScope datasets from seven African countries, which have not been used for this purpose before;

• in South Africa, we use a 2004 micro-level household panel called the Financial Diaries, which track all the income, expenditure and financial flows of a small sample of poor and relatively poor households over a ten month period

These sources allow us to explore potentially useful means of segmentation of the savings markets in these countries, as the basis for strategies by financial providers, government and donors to scale-up savings Specifically, the paper addresses the following questions:

• In section 2, what does evidence from the take-up of a new basic bank account by first time users among the poor in one of the countries (South Africa) tell us about adoption patterns for basic formal instruments?

• In section 3, how to define and measure savings, and who are savers by these definitions?

In addition to considering evidence on individual instrument usage, we specifically construct a portfolio of savings instruments used by the poor based on flows and balances held in different classes of financial instruments

• Finally, in section 4, how large is the potential market for savings instruments among the poor? We illustrate the implications of one approach, based on using actual numbers of

portfolio distribution combined with assumptions on adoption, to yield an initial estimate

1.1 FinScope household survey data

FinScope surveys, developed by FinMark Trust, ask in detail about the usage of and attitudes toward financial instruments by the adult population as a whole in a country FinScope surveys have been completed in seven countries in southern and east Africa, Botswana, Namibia, South Africa, Kenya, Tanzania, Uganda and Zambia, with the field work mostly in 2005 and 2006 The average sample size was around 3000 respondents in each country, together representing some

86 million individuals These will be known here as the ‘FinScope countries’ The surveys are designed to be nationally representative of adult individuals, and in most countries, the true number likely (with 95% confidence) to fall with a range of 5% above or below the weighted

survey number Questionnaires differ somewhat among the countries, affecting to some extent the ability to undertake cross country econometric analysis.2

FinScope data enables savers to be defined based on their declared usage of savings instruments, from a pre-coded list of options which differ somewhat by country and are listed

in Annex A These options include informal options, such as savings clubs, alongside formal options such as bank accounts, but focuses on financial instruments other than personal ones:

in other words, cash savings in the home, with a money guard or savings in the form of livestock are not included here, although these options are offered in certain FinScope surveys

1.2 The Financial Diaries

The Financial Diaries (“Diaries”) dataset seeks to understand the usage of financial instruments

by poor households at a detailed level The Diaries continuously tracked a full set of cash flows across 152 households (“the Diaries households”) from February through November 2004

The Diaries methodology is distinct from FinScope in at least two relevant ways First, the Diaries use the household as the unit of analysis, which is helpful because money is fungible through the household For instance, one member may be saving while the other is borrowing (or otherwise dissaving), and presumably this is related to the household’s overall cash management strategy Second, one-off surveys can tell us whether a respondent has a certain

instrument and even if he/she uses that instrument, but it falls short of telling us the intensity

with which it is used The Diaries data measures such intensity (e.g., how often and to what extent a particular instrument is used) and therefore allows us to analyze usage more deeply

Table 1 below shows the range of the Diaries’ household incomes while Figure 1 shows the LSM profile of the Diaries population compared to the total population

Table 1: Financial Diaries Sample by US$ per day income 3 (% of households)

Urban Rural Overall sample

2

We are grateful to FinMark Trust for allowing access to FinScope SA data which is owned by a consortium of mainly

private funding organizations for the purposes of this research

3 Dollar per day calculations are done by taking average daily income per capita in South African rand, deflating by a

factor of 1.98 to convert from 2004 to 1993 prices, then dividing by a PPP exchange rate factor of 1.67 to arrive at a

dollar per person per day figure for each household Note that had average daily per capita been adjusted using 2004

market exchange rates rather than 1993 PPP exchange rates, 32% of the sample would have been considered below $2

per day, rather than the 10% shown Where ZAR is converted to US$, the average exchange rate for the Diaries period

of 6.50/US$ is used

Box A: Definitions of poverty

The standard measure, $2 per head per day Purchasing Power Parity (PPP)-adjusted, can be used only for the Financial Diaries sample, but not for the cross-country FinScope surveys, since household income is not collected in all countries, and when it is, it is within bands which do not conform to the cutoff thresholds

Other measures can be applied to get similar results from FinScope In South Africa alone, in which more detailed analysis is undertaken, we used:

• Living Standard Measures (LSMs), which are segmentation tools used in consumer

marketing in South Africa The LSM is a wealth proxy, calculated entirely on observable

goods, which runs from 1 (very poor and rural) to 10 (wealthy and urban) (see Annex B) LSM1-3 constitute 33% of the SA population, roughly equivalent to the number (30%) living

on under $2 per capita per day in Bannerjee & Duflo (2007) In local terms, LSM1-5 are considered financially underserved, and are targeted in the Financial Services Charter, designed to increase financial access to the previously ‘unbanked’

• Questions asked by FinScope about hunger and basic services, such as those who report that their household has experienced some shortage of food or a lack of clean drinking water, which can be combined into a poverty proxy

Since LSMs are not measured in the other FinScope countries and the questions about hunger and services are not consistently asked across all, in cross-country econometrics, we use a simple quality of housing indicator as a poverty proxy

Note that only 10% of Diaries households qualified as poor under the application of the $2 per day measure (see Table 1); whereas 19% were in LSM 3 (see Figure 1) All Diaries households live in what are regarded locally as poor communities and indeed all are at or below LSM6, and most below LSM5; but when speaking of the poor here, we focus on households in LSM3

Figure 1: Diaries and LSM distribution

Source: Financial Diaries and FinScope SA 2006 for LSM

2 “If
you
build
it,
they
will
come…”

Who
comes,
and
when?


The supply of appropriate formal savings instruments is so suppressed in most developing countries that when a suitable instrument is offered, the take-up is often overwhelming This has been the experience of leading banks like Indonesia’s BRI and Kenya’s Equity But who comes, and how quickly? These parameters are sometimes little understood but are vital for making the business case for ‘laying large pipes’, that is, for a new savings product roll out South Africa’s so-called “Mzansi” bank account offers a case study to analyze adoption patterns using FinScope data across time Mzansi is a brand name of a category of basic bank accounts with similar features which was launched in late 2004 by a consortium of four large commercial banks and the state-owned Postbank, as a coordinated effort to increase financial access

Features of the Mzansi account include inter alia: (i) low or no minimum balance, (ii) no

monthly service charge, (iii) at least one free monthly deposit, (iv) nominal interest of up to 3.25%, and (v) various other transactions (deposit, withdrawal, bank transfers, payments, etc.) via multiple channels (e.g., branch, ATM, P.O.S and some internet and/or mobile banking) at fees set by each institution Mzansi therefore embodies many basic elements of good design for basic bank accounts Thus, at a minimum, Mzansi allows holders to save money in a regulated institution via free monthly deposits, without having savings eroded by minimum fees.4

The take-up of Mzansi among the previously ‘unbanked’ has been impressive After less than 2

years from product launch, almost two million individuals had opened and kept Mzansi

accounts5 and of these, 1.2 million (60%) had never before had a bank account (“Mzansi 1st

timers”).6 76% of all Mzansi 1st-timers said the purpose of opening an account was to save The client base of Mzansi in 2006 is also quite evenly distributed across income terciles, as shown in Table 2: in this, Mzansi compares favorably with other large savings programs highlighted in a recent WSBI study (2008), in particular Bansefi of Mexico However, whereas the savings banks shown below are government institutions, Mzansi is a consortium of private and public banks The private banks launched Mzansi in terms of their commitments to development under the Financial Sector Charter; all feared cannibalizing their existing account holder base and some have subsequently complained that the revenues on the new accounts are not sufficient for

4 However, worth noting is that the nominal interest rate paid (up to 3.25%) is lower than recent inflation.

5 By 2006, although 2,518,946 had adopted it, because 573,972 dropped out, only 1,944,474 still held an account.

6 At 2007, 3,925,804 people had adopted Mzansi, of whom 77% still had it, hence 23% of all adopters had dropped out

Of those who still had it, 65% were Mzansi 1 st -timers, similar to the proportion for all adopters

them to sustain the offering The question of profitability (or the need for subsidy) makes it all the more important that adequate market sizing is undertaken for such new products so that returns, whether for state or private institutions, can be undertaken

Table 2: Client base of Mzansi compared to that of large government savings banks

(South
Africa)
 (Mexico)
 (Tanzania)
 (India)
 (Thailand)


Mzansi
 Bansefi
 TPB
 NSI
 Thai
GSB


% of clients in:

Source: WSBI (2008)

To understand the pattern of Mzansi adoption, we compare Mzansi 1st-timers to those already banked (some of whom also opened a Mzansi account) and to the unbanked in the country; we also look at Mzansi ‘dropouts’ Table 3 below provides background data to highlight several relevant issues about who has taken up the Mzansi offering:

• Young people (ages 16-29) were much more prevalent among ‘Mzansi 1st-timers’ than they were among the ‘non-Mzansi banked’ segment: In 2006, young people represented 62% of Mzansi 1st-timers and only 29% of non-Mzansi banked Thus, relative ‘youth’ positively influenced Mzansi adoption

• There is a dramatic difference between the ‘banked’ and ‘unbanked’ with respect to expressed behavior such as working to a budget: 70% of the banked and only 23% of the unbanked claim to do so At 55%, Mzansi 1st-timers are closer to the banked than unbanked; in other words, there is a significant correlation between working to a budget and Mzansi use/adoption, although the available data alone cannot prove causality

• Also, there is a significant difference between the ‘banked’ and ‘unbanked’ with respect to attitudes towards savings: of the ‘banked’, 40% say they “sacrifice to save” and 69% say they “try to save regularly”; compared to 12% and 16%, respectively, for the unbanked Mzansi 1st-timers are much more like the banked in this respect too: 46% say they “sacrifice

to save” and 55% “try to save regularly”

• Mzansi has had a relatively high penetration in rural areas: Just as 40% of all South Africans are rural, 39-40% of all Mzansi users/adopters are rural, much higher than the proportion among those banked through other products

• A substantial proportion (23%) of Mzansi adopters had dropped it by year-end 2006; the dropout rate for Mzansi 1st-timers was essentially the same (22%).7

Table 3: Comparison of Mzansi to Non-Mzansi Banked and Unbanked (2006 Finscope data)


 
 All
Mzansi
 Adopters
 (including
 dropouts)



 Mzansi
 Current
 Users
 (excluding
 dropouts)


Mzansi
1 st ‐ timers
 (and
 currently
 using)
 dropouts
Mzansi


Banked,
not
 Mzansi
1st‐

timer
or
 dropout


Unbanked,
 not
Mzansi
 dropout
 Population
Total
 Number
 2,518,946 1,944,474 1,157,451 573,972 14,486,846 14,918,530 31,136,800

7 Finscope 2006 data did not allow further analysis of the breakdown of Mzansi dropouts between Mzansi-1 st -timers and non-1 st -timers; however, Finscope 2007 data does allow this, and can be analyzed for this in the future.

Age
16‐29
 51% 49% 62% 57% 29% 43% 37%

As a big ‘pipe-building’ project, Mzansi appears successful: within three years of product launch, it connected 2.5 million (15%) of the previously-unconnected The material differences

in adoption rates across distinct segments could be helpful in designing future large-scale rollouts However, the substantial dropout rate (23%) also suggests the need to go beyond

measuring success merely in terms of accounts opened; and to look at underlying patterns of

usage and who is most likely to continue using While one-size-fits-all can make very beneficial strides as a ‘phase one’ approach, there is also a need for ‘phase two’ follow-up offerings, in order to increase meaningful usage and, in turn, increase customer retention rates

3 Moving
beyond
take­up,
to
measure
usage
of
savings
services.


3.1 Defining and measuring savings

A key issue in analyzing savings behavior is how to define savings At one level, savings constitutes all additions to household net worth, where the wealth is likely to be held in physical assets as well as financial assets Figure 2 below shows the breakdown of net worth between financial assets and physical assets at the beginning and end of the study Physical assets (including illiquid home values which are inherently hard to value) certainly make up the larger proportion of net worth; however, simply because someone holds more physical assets than financial assets should not imply a firm “preference” for saving in physical over financial

It may rather simply reflect an ongoing lack of viable financial alternatives in which to accumulate long-term savings Diaries households did not actively “save” in physical assets during the year – the value of physical assets barely changed at all Financial assets, on the other hand, were actively used and actively grew over time The median household grew financial assets at a rate of 14% in just under a year A key question emerging from Figure 2 is:

if households are able to mobilize relatively so much financial savings in this period, then why have they not accumulated financial assets over time so that they represent a larger share of net worth? A large part of the answer is that households are able to save a great deal in financial assets over the short term, but may be unable or unwilling to accumulate them over the long-term, an issue we return to in Section 3.1.3

Figure 2: Total asset profile of Financial Diaries households (US$)

We now consider three different definitions of financial savings and their measures in the available data:

• The usage of financial instruments;

• The intensity of savings (savings flows in financial instruments as a % of household income); and

• The duration of savings (the period over which savings balances accumulate, either by instrument or aggregated)

3.1.1 Reported savings instrument usage

FinScope and the Diaries ask respondents which instruments they use The available list of savings instruments from FinScope South Africa and the Diaries is compared below, along with their categorization as ‘regulated’ or not, based on the status of the provider

Table 4: Instrument definitions

x8

Endowment/Investment/Savings

8 The Diaries tracks “bank accounts”, but does not distinguish details within that general category Mzansi had not been launched at the time of the Diaries

Pension fund Regulated X

The instrument-based definition is therefore that a ‘saver’ uses at least one instrument from the list provided While this definition is perhaps easiest to measure, the difficulty with this approach is that the full list of possible instruments has to be quite long If there are missing categories on the pre-coded survey list or if respondents do not adequately understand the items on the list, the response rate may be lower than expected Also, the exact list of instruments offered varies in each FinScope country in order to capture the options considered most appropriate to that country (see Annex A)

Applying the instrument-based definition results in the cross-country profiles shown in Table 5 below Whereas typically around half of the population report having at least one formal or informal savings instrument in most of these countries, the proportion in Tanzania and Uganda

is much lower When we consider only the poor, the rates of instrument usage in most cases

do not drop significantly; however, among the poor, a significant proportion in countries like Kenya, Tanzania and Uganda use only informal (unregulated) savings instruments

Table 5: Overview of FinScope data

Botswana Kenya Namibia South Africa Tanzania Uganda Zambia

1 % of total

population using at

least 1 defined

savings product

2 Of the poor:

% using at least 1

de-fined savings product

3 Of poor who save:

(a) % using formal

and informal savings

products

(b) % using only

infor-mal saving products

16.8%

10.9%

19.1%

59.2%

6.7%

3.4%

20.5%

9.6%

6.7%

58.9%

3.2%

82.5%

9.1%

9.1%

While the use of informal instruments such as rotating or accumulating savings is common across the region, FinScope data provides a view on the risk associated with this, and attitudes towards informal groups of this kind For countries in which the question was asked, Table 6 shows personal experience of loss as a percentage of those who use an informal instrument — ranging from as high as 20% in Botswana to 3% in Uganda.11 Perhaps as a result of the losses, some ambivalence towards group-based mechanisms emerges from the data, with around a third of those using informal instruments expressing mistrust in them, although the framing of this question differs across countries and affects comparability with Uganda

Table 6: Experience of informal savings mechanisms (FinScope)

Botswana Kenya Namibia Africa South Tanzania Uganda Zambia

9 The Diaries tracks “informal savings clubs”, but does not distinguish details within that general category

10 Note that the concept of ‘savings in the house’ was intentionally distinguished from cash on hand in Diaries

interviews, so that casual cash balances day to day were not confused with more intentional savings efforts

11 Note that this number for Uganda is lower than the finding by Wright and Mutesasira (2001) that some 26% of clients from focus group and individual interviews had lost savings in the informal sector

% of those using informal

instruments reporting

% of those using informal

instruments who agree with

statement “I don’t trust

informal groups ”

Note: * indicates where statement was expressed in opposite form ^only available for 735 of 4962 respondents

The relatively high rates of mistrust of informal groups among those continuing to use them (Table 6, bottom row) is striking; such behavior may stem from a lack of ‘safer’ alternatives

3.1.2 The intensity of savings

Counting the number of savings accounts, or even preferably the number of people with accounts, is not indicative of the significance or intensity of that usage: in the Diaries sample, 74% of households report using a formal savings instrument such as a bank account but there was wide variation in usage levels We therefore need another dimension when assessing

savings behavior: the intensity of savings, measured as the cash flow into defined financial

instruments over time as a percentage of household income

Figure 3 below shows the median monthly savings of a Diaries household over the 10 month period divided by average monthly income over the period Savings intensity varies month by month and seems to increase towards year end.12 The median intensity over the entire 10 months was 21% Since median monthly income was $290, this means that $60 per month was saved in a financial product of some sort during this period At such low income levels, this

rate is high; trying to encourage proportionally more savings may be unrealistic, although it is

possible that access to a safe savings instrument might give more reason to save what otherwise might have been spent

12 Our discussions with households indicate that many savings activities are intended to fund year-end activities, such as Christmas feasts, home improvement, school fees/uniforms Therefore, we suspect that much of the savings

accumulation that happens during the year is spent in the December/January period Unfortunately, because the

study was completed just before this period, we were not able to confirm our suspicions with actual cash flow data

Figure 3: Intensity of savings by month (Financial Diaries) 13

The intensity can also be analyzed by income group and by the instruments used Table 7 below shows that the poorest group does not save markedly less as a percentage of income than their

middle-income counterparts However, the poor do save in different financial instruments

Households earning less than $5 per day tend to accumulate savings almost entirely in informal instruments, such as savings clubs and hiding savings in the house Very little savings happens

in the bank, even though over half of these households report having bank accounts Households earning above $5 per day save substantially more in formal instruments, such as bank accounts, provident funds and retirement annuities; although, note that this higher-income segment continues to use informal instruments as well

Table 7: Calculating savings flows and cycles, based on $ per day income (Financial Diaries)

Income

per day Percent of sample accumulation as Mean

Saving in the house guarding Money Savings clubs accounts Bank Provident fund or savings annuities

Total

3.1.3 Duration of savings

In addition to instrument usage and intensity, savings may be measured by duration: the length

of the period over which households manage to accumulate savings using the instrument before withdrawing it to use for a variety of purposes, including investment in physical assets Duration can be calculated for each Diaries household and each instrument by counting the number of days households manage to accumulate money over the period before the balance falls (see Annex C for more detail) Note that longer duration is not necessarily good: as discussed more below, the important issue is that households can match the timing of their underlying reason for savings with an instrument which has an appropriate time profile

Figure 4 below shows the average duration across the sample for the five different types of savings instruments captured in the Diaries On average, both money guarding and savings in the house had the lowest durations, with households managing to hold on to their

13 Savings flows are defined here as monthly flows into: bank accounts (net), a savings place in the home (net), savings clubs (gross), provident funds (gross), retirement annuities (gross), and money guarding (gross).