Exploring Disparities in Saving Determinants between Urban and Rural Households: Insights from Ethiopia

1. Introduction

The sustainability of long-term economic growth requires capital investment - infrastructure, technology, trained human capital and business expansion. Household saving is one of the most important sources of domestic capital for investment, especially at an early stage of development. The positive relationship between private saving and economic growth has long been recognized in development economics (Bisat, Mohammad, and El-Erain 1997; Loayza, Schmidt-Hebbel, and Servén 2000; Schmidt-Hebbel and Serven 2002). However, due to cultural, socioeconomic and institutional differences, savings rates and savings behavior vary widely across economies.

The fastest growing economies in East Asia save on average more than 30 percent of their gross national income, the advanced economies save about 25 percent, the Caribbean and Latin America save less than 20 percent and Sub-Saharan Africa save less than 15 percent of their GDP. Over the last thirty years, the cross-country dynamics of saving have also highlighted the widening regional imbalances. Savings rates have almost doubled in East Asian countries, declined in sub-Saharan Africa, stagnated in Latin America and the Caribbean and declined in some industrialized countries, particularly the United States and Japan(El-Seoud 2014; Grigoli, Herman, and Schmidt-Hebbel 2014)

Apart from the role of private saving for investment in developing countries, the risk associated with climatic conditions, underdeveloped financial services, and limited coverage by the social welfare system force private saving to be a primary source of income to smooth family consumption and overcome insecurity. The percentage of households covered by the social welfare system in Ethiopia is less than 18% (NBE 2022). The growing gap in investment savings and the problems in allocating savings represent a major challenge for developing countries. In addition, formal financial institutions can only mobilize the savings of low-income families to a limited extent. The local credit rotation and fixed fund association are customary in channeling credit and saving among low and middle-income households(Akaah, Dadzie, and Dunson 1987; Dadzie, Akaah, and Dunson 1989). In this context, investigating country-specific factors that influence the savings behavior of private households is very important to examine the effectiveness of government interventions in mobilizing domestic savings and to understand the complex development process.

Because of the role of private savings in economic growth, the Ethiopian government has implemented various packages of measures that promote private savings in the last two development plan periods (2010-2020). These packages of measures include the partial liberalization of the financial sector, enabling the free working of small financial cooperatives, and the expansion of microfinance and bank branches in urban and rural areas . As a result, the savings rate reached 24.2 percent of GDP and deposits in state and private banks increased from 78.4 billion to 1.6 trillion ETB between 2008 and 2022 (NBE 2022). The savings performance of MFIs also increased significantly from 122.7 million to 30.2 billion ETB, with voluntary savings accounting for 74 percent of the total on average (NBE 2022).

Despite the Ethiopian government’s efforts to promote private savings, a significant portion of the population remains financially excluded, and the development finance gap is widening. The number of people with savings accounts in state and private banks is limited to 9.1 percent of the total population; this figure is tiny even by African standards: Ghana (35%), Nigeria (44%) and Algeria (50%) (WDI 2021). Furthermore, the average savings in kind in urban and rural areas of Ethiopia is higher than in cash. Savings instruments have remained essentially unchanged over the last sixty years of Ethiopian history. A majority of the Ethiopian population practices traditional savings and the rotating insurance system known as Iqub^[1] and Idir^[2] respectively (Amha and Alemu 2014). Therefore, by examining the unique factors that influence saving decisions in urban and rural areas, we can better tailor financial strategies and policies to meet the specific needs of each group.

Macroeconomic performance shows that the gap between saving and investment is growing. Domestic saving, measured as a percentage of GDP, has increased from 10.32% in 2010 to 24.3% in 2020; however, due to the simultaneous increase in investment spending from 24.3% to 39.01%, the gap between saving and investment has increased from 12.3% of GDP to 17.5% over the period (NBE 2022). The low domestic savings compared to what is needed to finance development projects, the tendency of a considerable number of households to save in kind and the limited number of people with bank accounts indicate that household savings in Ethiopia are not optimally mobilized. Furthermore, it shows the importance of examining the factors that influence household saving behavior.

Most empirical evidence in developing countries, particularly in Ethiopia, is based on aggregate-level models where macroeconomic work ignores consumer heterogeneity by assuming representative household actors (Akaah, Dadzie, and Dunson 1987; Amha and Alemu 2014). This paper used microeconomic data to analyze household saving behavior in urban and rural areas compared to previous studies. Moreover, additional predictors such as the financial literacy index, perception of saving (self-control and foresight), access to credit, social capital, were included in the analysis. Against this background, the study answers the following research questions: First, do the factors that influence saving and motives for saving differ between urban and rural areas? Which determinants are more important in explaining the differences in household saving?

2. Literature Review

This section presents both theoretical and empirical theories of saving. Saving theories can be roughly divided into three large groups: Conventional economic theories, sociologically oriented theories and institutional perspectives (Chowa, Masa, and Ansong 2018; Deaton 1997).

Conventional economic theories include the neoclassical, in particular the Permanent Income Hypothesis (PIH) proposed by Friedman (1957), and the Life Cycle Hypothesis (LCH) by Modigliani (1986). The interpretation of the LCH on household saving is based on the change in behavior with age. The age of individual respondents is the most important predictor of differences in household saving behavior. In any economy, young people will save too little because they earn too little when they are young and have to borrow to meet their growing consumption needs. As they reach middle age, their income is likely to rise, so they save more and more to repay their debts and accumulate assets for retirement. As they get older, people begin to deplete their savings to maintain average lifetime consumption. Thus, in a given economy, the savings rate varies between young, middle-aged and older people (Ando and Modigliani 1963; Deaton 1997) . Friedman’s PIH emphasizes the temporal dimension of the change in household income. Saving and the saving rate increase as income rises if the household perceives the change in income as a short-term/transitional phenomenon(Eisner 1958; Friedman 1957) .

The original conventional economic theories are based on simplifying assumptions that run counter to the specific nature of a developing economy. They ignore the inheritance motive and place little emphasis on factors such as the size of families, low life expectancy, the volatility of household income, the small number of old people and the lack of developed financial institutions. It is also over-ambitious about the degree of decoupling of consumption from income; few people engage in long-term saving and de-savings (Deaton 1997). However, several empirical analyzes have extended conventional economic theories by incorporating the concepts of prudence motives, saving habits and household behavior in the face of liquidity constraints (Bernheim, Skinner, and Weinberg 1997). The two common extended theories are the intertemporal consumption theory (based on the liquidity constraint) and the certainty– equivalence hypothesis (based on the precautionary motive).

In contrast to neoclassical consumer theory, behavioral economists and economic psychologists do not believe in the unrestricted rationality of economic actors and the absence of information asymmetries. The application of behavioral psychology to explain household saving behavior has been advanced by economists such as (Angner and Loewenstein 2007; Fisher et al. 1978; Jensen 1987; Sigot 2002). They emphasized the importance of considering not only income but also differences in personal attitudes as a predictor for studying variations in household saving and wealth accumulation(Fisher Jr et al., 1978; Jensen 1987)

Saving by economic agents depends on their willingness (psychological factor) and ability (objective factor) to save. Higher income is not a guarantee of saving; willingness to save is explained by the degree of economic optimism/pessimism (Hosseini 2011; Katona 1975); perceived locus of control (Lunt and Livingstone 1991; Perry and Morris 2005); perceived ability to save(Sherraden and McBride 2010); and future orientation (Webley and Nyhus 2006) are also helpful in explaining differences in household saving behavior. By linking economics and psychology, behavioral economics has introduced new behavioral assumptions that stand in contrast to unbounded rationality and selfishness.

The economic approach to class and social stratification suggests that an important explanation for class inequalities lies in access to and use of material resources and the institutions that regulate this access(Chowa, Masa, and Ansong 2018; Crompton 2008). Empirical evidence in developing countries indicates that poverty at the household level is a characteristic beyond the control of the individual, such as race (Oliver and Shapiro 2013); gendered cultural norms(Chowa, Masa, and Ansong 2018); cultural origins (Al-Awad and Elhiraika 2003); financial socialization in families, schools or communities and (Chiteji and Hamilton 2005). The development experiences of East Asian countries emphasize the importance of sociological and institutional factors for rapid wealth accumulation and high pro-capital saving.

Accordingly, in developing countries, ownership of factors of production (mainly land and house) and wealth accumulation are highly linked to social class, community membership and political affiliation. The communal nature of wealth ownership and particular lifestyles significantly impact saving and wealth accumulation. Individuals from the upper economic classes with socioeconomic and political affiliations have financial information, income and an educational advantage that enables them to save more and build life-changing wealth. On the other hand, a person who comes from the lower economic class or stratum has a high probability of being dragged down by a vicious cycle of poverty, which in turn harms wealth accumulation and saving (Caskey 1997; Chiteji and Hamilton 2005).

There is various empirical evidence on the determinants of household saving behavior. Although the statistical significance of the covariates varies. The empirical determinants of household saving can be grouped into four categories: conventional factors (income and age), sociological factors (demographics, class, education), behavioral factors (self-control and foresight), and institutional factors (ease of access, immediate availability, access to credit). Empirical evidence on Beveheverioul and institutional determinants of household saving is limited in sub-Saharan African countries (Amha and Alemu 2014).

According to Davis and Schumm (1987), the effect of income on household saving depends on whether the family is low or high income; for high income families, saving is a positive function of income, but the relationship is not significant for low income families. However, recent empirical data from middle- and higher-income sub-Saharan African countries suggest that income positively influences saving. Evidence from Morocco (Arestoff et al. 2009), Kenya (Kibet et al. 2009) and Uganda (Kiiza and Pederson 2001) shows that income positively and significantly affects household saving. In addition, an improvement in household income was a significant predictor of saving in India, Pakistan (Rehman, Bashir, and Faridi 2011) and the Philippines(Bersales and Mapa 2006). Evidence from Ethiopia also suggests that low-income families have great potential to save with the emergence of microfinance institutions, provided the institutional infrastructure is in place (Amha and Alemu 2014; AYAL 2020; AYALEW 2022; Hailesellasie, Abera, and Baye 2013)

The effect of age varies according to the development of financial institutions and coverage by the social security system (Davis and Schumm 1987). In line with the life cycle hypothesis, empirical data on household saving behavior in Japan show that middle-aged households have a high average saving rate. At the same time, the elderly do not save their assets compared to other age groups (Horioka and Watanabe 1997). However, the evidence from SSA and other developing countries is mixed. Some empirical findings suggest that the age of the household head is not statistically significant; many other studies found a positive relationship between age and saving; as age increases, so does saving e.g. (Johnson and Widdows 1985; Yuh and Hanna 2010). However, families with more dependents are less likely to save than families with fewer dependents.

Surveys in Uganda (Chowa, Masa, and Ansong 2018) and Kenya with middle and low incomes show that the level of education of the head of the family increases the likelihood of saving. In addition, educated heads of households are more likely to have a savings account at the nearest financial institution. A household with sound financial knowledge is more likely to make informed decisions and save than a family without financial knowledge.

There is little empirical evidence in Sub-Saharan Africa, including Ethiopia, on behavioral and institutional determinants of household saving (Amha and Alemu 2014). However, data from low-income families in Uganda suggest that behavioral variables such as self-control and the presence of savings motives are also significant predictors of saving among low-income households (Chowa, Masa, and Ansong 2018). A family with a particular savings motive is more likely to have a high savings balance than a family without a savings motive. Evidence from Malaysia suggests that religious motives are one of the drivers for saving among followers of the Islamic region; about 45 percent of respondents save to go on pilgrimage to the holy land of Mecca and Medina (Horioka and Watanabe 1997). This study will fill this gap by introducing limited behavioral variables such as saving habits and motives for saving.

Due to their low incomes and limited financial infrastructures, poor rural households in developing countries practice informal saving, which is insecure and risky compared to the formal system (Collins 2009). The impact of access to credit on household saving is mixed across East African countries; experience in Kenya (Kibet et al. 2009) shows that access to credit reduces the likelihood of saving. In Uganda (Chowa, Masa, and Ansong 2018; Kiiza and Pederson 2001), the opposite is true: households with access to credit are more likely to save and have higher savings balances. Easy access and immediate availability of financial institutions increased household savings in Kenya (Kibet et al. 2009), Ethiopia (Amha and Alemu 2014) and Uganda (Chowa, Masa, and Ansong 2018).

In summary, the empirical evidence on household saving behavior varies according to two criteria: the definition of saving, the type of data, and the methods used. First, the definition of saving is either income minus consumption (Bae, Hanna, and Lindamood 1993) or accumulated wealth (Canner, Kennickell, and Luckett 1995). In developing countries, many households don’t have a constant income; therefore, they use accumulated wealth as savings. Second, previous studies have used micro or macro data depending on the research objective. However, due to the limited nationwide microdata in developing countries, most empirical studies on saving behavior in sub-Saharan Africa have been conducted using macro data. This approach rarely shows the diversity of saving behavior at the household level. In contrast to previous studies, this study used nationally reprsentative micro-data, included behavioral predictions, and constructed an index of the financial literacy of households. Furthermore, it observed urban and rural households independently.

3. Overview of the Ethiopian Finance Sector

The Ethiopian financial sector consists mainly of banks, microfinance and insurance companies, with the banking sector playing the dominant role. As of August 2022, the Ethiopian financial sector comprised 29 commercial banks, one development bank, 18 insurance companies, one reinsurance company, 40 microfinance institutions, 6 capital goods finance/leasing companies and 8 issuers of payment instruments/system operators. In addition, many conventional and full-fledged interest-free banks are in the process of joining the sector. Except for the leasing sector, the financial industry was closed to foreigners until recently and the state-owned CBE^[3] dominates the banking sector. CBE is responsible for 58% of deposits and have 59% of actual investments in the financial sector. The government has used the state-owned banks to provide loans to state-owned enterprises to finance development projects. Except for the Development Bank of Ethiopia, all state and private banks focus mainly on short- and medium-term trade finance. At the same time, insurance companies are involved in short-term general insurance services. Despite extensive approvals to promote growth, microfinance institutions offer short-term loans without diversification and digital networks. The payment system is still mainly based on the cash model, and the digitalization of finance lags behind regional anchors and lacks diversity of services. Improving financial inclusion through the digitalization of transactions is a top priority for policymakers (NBE 2022).

These financial sector actors in general and banks in particular have made commendable efforts to meet Ethiopia’s unmet demand for financial services, albeit with limited resources. Available data show that the Ethiopian financial system lags in many respects, such as financial technologies, financial infrastructure development and governance, financial sector inclusion and penetration, and domestic capital mobilization (Ababa 2017; NBE 2022). However, there is a new development in the financial sector. After almost half a century, Ethiopian policymakers have decided to reopen the sector to foreign competition. In addition, the financial industry is undergoing intensive digitization of transactions, allowing telecommunication providers to offer financial services and on the way to reviving the capital market. Ethiopian policymakers have decided to revitalize the capital market and lower the barriers to entry for the financial industry.

In this section, we discuss the overall level of financial development in Ethiopia based on a more appropriate measure used by the World Bank in terms of depth, access, efficiency and concentration (Bikker and van Leuvensteijn 2014; Caprio and Honohan 2001)

3.1. Financial Access

Access to finance is measured in three ways: the physical availability of financial institutions, access to the services offered by financial institutions and the variety of instruments.

The World Bank’s 2017 development indicators show that almost 66% of the adult population in Ethiopia is unbanked. In regional anchor countries such as Kenya (82 %), Rwanda (50 %) and South Africa (85 %), the percentage of the adult population with a bank account is much higher. In addition to the limited financial penetration, the branches of financial institutions are also unevenly distributed. Of the total 8944 bank branches (as of June 2022), around 68 % are located in the large regional cities and the capital Addis Ababa (NBE 2022). The capital city alone is home to 35 % of the country’s bank branches. The population living in rural areas (80 %) often has to travel to the nearest city to access the basic banking services such as deposits, transfers and withdrawals. Taking into account the uneven distribution of bank branches, there is at least one branch for around 11,963 people. Informal credit and insurance associations are widely used by both urban and rural households to bridge uncertain events and save for investments.

Although telecommunications operators have entered the financial sector, the spread of digital financial services and mobile money services is limited to urban areas. Less than one percent of the population aged 15 years and older reported having a mobile money account and using the internet to pay bills 12 months prior to the survey. In Kenya, the corresponding figures were 68.7% and 25.7% respectively.

The relationship between banks and customers is not really two-way; although banks receive substantial deposits, access to credit is limited to a few groups of people. In the last two decades, the number of individual borrowers from commercial banks in Ethiopia has been less than 500 (NBE, 2021); this number is very low for a country with a population of more than 100 million. The Credit and Saving Association is far better at lending to its customers than the banks and has already provided loans to more than 6 million people. According to a survey conducted by the World Bank in 2008, only 10.66% of adults surveyed had taken out loans from financial institutions.

This suggests that the Ethiopian financial sector still has much room for improvement, especially in terms of technology adoption, choice of instruments, domestic resource mobilization, financial inclusion and depth of monetization of the economy. The limited financial inclusion, the growing population, and the prospect of economic growth in Ethiopia are assets for operating banks and new entrants who want to maximize their profits without engaging in counterproductive competition.

3.2. Financial Depth

Financial depth is usually measured by private credit, broad money supply, assets and deposits as a percentage of GDP. It shows the position of the domestic economic infrastructure in terms of access to credit, diversity of savings/credit instruments, asset portfolio, and institutional connectivity to the daily lives of individual citizens. Ethiopia’s credit to the private sector (26.8%), bank assets (40.6%), bank deposits (23.9%), and monetary deepening (30.7%), measured as a percentage of GDP, are low compared to the regional anchors and best performing East Asian countries(NBE 2022).

The ratio of private credit to GDP has improved in Ethiopia over the last 15 years. From 2004 to 2022, it grew on average by 25 % per year. Nevertheless, the share of domestic credit to the private sector in Ethiopia (26.8%) is among the lowest, even by SSA standards. In Kenya, South Africa, Vietnam and Thailand, the share of credit to the private sector in GDP was 32.1%, 112%, 116.7% and 160.4% respectively. A significant portion of domestic credit flows to state-owned enterprises, potentially crowding out credit to the private sector. According to the National Bank’s report for 2022, 58.3% / 21.6% / 3.13% of public bank loans are destined for the central government, state-owned enterprises and cooperatives, respectively. Private banks focus primarily on lucrative short- and medium-term loans; they are hardly involved in infrastructure development at all. Ethiopia’s recent development experience shows that despite considerable efforts to mobilize domestic capital, the gap in development financing will widen by 82.8% from 15.2% in 2015 to 27.3% of GDP in 2020. Ethiopia’s financial development is far below the level required to maintain economic growth momentum despite increasing financial intermediation. The low level of credit to the private sector compared to the best performing economies and the widening gap in development finance in Ethiopia indicate that the Ethiopian financial industry has significant potential for expansion. There is also a gap in the market for long-term development finance due to the commercial banks’ commitment to lucrative medium and short-term loans.

The second most important indicator of financial depth is monetary deepening (M2/GDP). It usually measures the degree of monetary integration to facilitate transactions (Liuadty liquidity of the economy). The average monetary deepening in Ethiopia (30.7%) is well below the average for Sub-Saharan Africa. In Kenya, monetary deepening is 42.9 %, South Africa 74.6 %, Egypt 84.0 % and China 218.2 %. This fact shows that a significant part of the Ethiopian population still has no access to financial institutions and alternative savings instruments. According to the Second National Bank Financial Inclusion Strategy 2021, the lack of diversity in savings mobilization tools, financial illiteracy and limited economic infrastructure development have hindered the improvement of monetary deepening.

The assets and deposits of commercial banks as a percentage of gross domestic product are the third and fourth indicators of the financial depth of the domestic economy. Ethiopia’s bank assets to GDP ratio increased by 47 percentage points on average from 2014/15 to 2020/21. Although less diversified, commercial bank assets in Ethiopia reached 40.6% of GDP. On average, net loans alone account for 69.95% of total assets, which means that interest income is an important source of revenue for domestic financial institutions. Most commercial banks in Ethiopia offer three types of deposits (time deposits, savings deposits, and demand deposits). Ethiopia’s average ratio of deposits to GDP was 24% between 2014 and 2020. Supported by a remarkable branch expansion, the deposit liabilities of the banking system increased to 899.6 billion birr; savings deposits grew by 27.4 %, followed by time deposits (25.6 %) and demand deposits (16.5 %). Demand deposits, saving deposits, and time deposits accounted for 54.2%, 35.1% and 10.8% of total deposits respectively. The banks’ aggressive branch expansion in particular contributed to the growth in domestic deposits.

3.3. Competition, Concentration, and Efficiency

Efficiency indicators that affect revenues and costs, such as net interest margin, spread between loans and deposits, non-interest income in relation to total income, overheads (as a % of total assets) and profitability (return on assets, return on equity) are used to identify the strengths and weaknesses of the financial industry as a whole. Over the past six years (2015 to 2021), the efficiency of the Ethiopian banking sector has remained stable. The latest data shows that non-interest income as a percentage of total income was 31.17% in 2020, return on assets and return on equity were 2.43% and 18.16% respectively. The net interest margin of commercial banks in Ethiopia (6.57%) is high compared to the regional anchor countries Kenya (5.98%), Egypt (3.64%) and Nigeria (4.31%). Comparatively, commercial banks in Ethiopia have a better return on assets and equity ratios. This can be explained by the fact that the banking sector in Ethiopia is comparatively well-protected and generates above-average profits

A combination of protection and other domestic factors led to the concentration of the banking sector in Ethiopia. According to the World Bank’s 3-bank concentration measure, the assets of the three largest commercial banks accounted for 81% of total commercial bank assets in 2010 and 76% in 2020. This shows that the banking sector is dominated by a few large banks, above all the Commercial Bank of Ethiopia. However, the share of the three largest commercial banks is declining to around 50%. The segments of the financial sector - insurance and pension contributions - are less developed and are the most neglected. The GDP share of insurance and pension funds is only 0.75 % and 4.3 % respectively. The investments of pension funds in Ethiopia are limited to the purchase of treasury bills, which leads to inflation.

Due to the strict regulations and the non-liberalization of the money market, there is no competition based on interest rates and exchange rates. Competition on financial product lines is also limited. Instead, the sector is known for its credit rationing due to the size of the banks and the strict regulations such as the 27% rule, which limit banks’ risk appetite and lending capacity. For these reasons, competition among commercial banks in Ethiopia is limited to a small number of factors. Recent experience shows that competition is limited to aggressive branch expansion, advertising, increasing the capital base, lobbying for term deposits and intensive investment in a digital payment system. Over the past three years, with the exception of a few commercial banks, their aggressive investment in core banking technology and digital payment systems has increased. Growing fintech companies are expected to be observed, which will provide competitive advantages to the Ethiopian banking sector. These recent developments have led to an increase in banks’ cost-to-income ratio. Higher net profits, higher net interest margin, and high cost-to-income ratio indicate that the financial industry in Ethiopia has lived under policy protection.

4. Hypothesis and Measurement of Variables (Table 1)

5. Database and Summary of Essential Variables

The analysis is based on data from a consumer finance survey conducted by the Ethiopian Policy Study and Research Center (recently renamed the Policy Studies Institute, PSI) in 2020. A multi-stage sampling procedure was used to collect the nationally representative data. First, survey areas were purposively selected based on economic activity, safety and geographic importance. Second, proportional sampling techniques were used to put the data in context of the gender and population density of each state and survey area. The researcher intentionally selected cities such as Addis Ababa, Mekele, Dire Dawa and Gonder to represent urban households. Surplus and deficit areas from the regional states of Oromia, SNNPR,^[4] Amhara, and Tigray were purposely selected to represent the rural population. The regional states of Afar and Somali represent pastoralist populations. Of the total number of 3902 (97.5% response rate) randomly selected respondents, 16.7% were urban dwellers; the remaining 83.3% represent rural households, including grain surplus areas (25.1%), chronically food insecure areas (23.3%), cash crop areas (27.8%) and pastoralist areas (7.1%).

Of the households surveyed, only 53.8 % have positive cash deposits, 79.8 % are male and 80 % are married at the time of the survey, 70 % are followers of the Christian faith, 47 % have no specific savings motives. Agriculture (76.35%) and trade (11.36%) are the two most frequently mentioned sources of livelihood. Financial pressure and financial literacy index are higher among urban dwellers. Close to 12.76% of households have converted their cash to in-kind items for fear of rising inflation. Almost 15 % had taken out a loan formally or informally in the 12 months before the survey.

5.1. Saving, rate of saving & saving portfolio

The savings rate and the savings portfolio differ in rural and urban areas. The average savings rate of urban households (16.3%) is higher than the savings rate of rural households (13.6%). Differences in risk, wealth, concentration of financial institutions, literacy and level of monetary deepening contribute to the differences in the higher savings rates of urban households. However, there is one exception to the data: Pastoralists clearly have a higher cash savings rate than urban households. This exception could be due to engagement in border trade and the risk associated with pastoralists’ mobile lifestyles, which increases the need for precautionary savings for pastoralists.

Urban and rural households have different tests for savings instruments. Instead of saving only in cash, the family buys less liquid assets such as livestock, improves the house, and expands the existing business. Of the households, 65.9% save only in cash, 8.7% save only in tangible assets and 25.4% diversify their savings between cash and tangible assets. Accordingly, the average savings in tangible assets are 2.03% higher in rural areas than in cities. The high ratio of real assets to cash in rural areas has two implications. First, Since urban asset prices are overvalued, diversifying savings requires a huge capital outlay. Due to overvalued urban asset prices, switching from cash to non-cash savings is more difficult in urban areas than in rural areas. Second, Most rural households produce family consumption on their farm, whereas urban households need higher precautionary savings to smooth family consumption and therefore hold a lot of cash. Third, Citrus Paribus in rural Ethiopia household savings in rural areas are not optimally monetized or mobilized.

In addition, a significantly large number of families practice informal saving. According to the survey, only 46.1% (44% of rural households and 55% of urban households) rely on commercial banks for savings. In contrast, 20.7 %/9.5 %/1.5 %/1.3 % place their savings at house/MFIs/cooperatives/fixed credit associations. The introduction of various targeted savings instruments, supported by stable macroeconomic conditions, can reduce household savings in kind and deepen the monetization of the economy.

5.2. Urban and Rural Household Saving Motives

The results of the financial survey show that 45.5% and 48.8% of the families in urban and rural areas respectively have no specific savings motive. Around 87% of households without a savings motive cited the volatility of their income as an obstacle to having financial savings. The order of savings motives is identical for urban and rural households. The motive of consumption smoothing is the most important driving force behind household saving, followed by motives for larger purchases provision, retirement, religion and investments. However urban dwellers are more likely to have prespecified and concrete saving motives.

Less than 5 % of households hold financial assets or save for investments and retirement provisions. Respondents indicated that they use three strategies to cope with uncertain challenges in old age: having a child early, fulfilling religious obligations, and raising children to avoid unforeseen challenges in old age. In addition, households that save for religious purposes are seen as insurance against unexpected future challenges in life. In contrast to developed countries, around 2% of households surveyed save to fulfill religious obligations, such as pilgrimages to holy lands and celebrating religious festivals. Finally, saving motives of households have elements of randomness: money saved for emergencies can be used for consumption smoothing and vice versa.

6. Econometric model

The microeconomic analysis of household saving behavior usually depends on two basic concepts: the decision to save and the intensity of saving. First, households are asked whether or not they have positive financial savings, then they state their actual saving behavior. Of the 3902 households surveyed, only 53.8% have positive financial savings, the rest, 46.2%, `have no financial savings. In the direct application of ordinary least squares estimation, households without financial savings are not considered; this leads to selection bias and inconsistent estimates. The financial sector in Ethiopia is consistently characterized by negative real interest rates and overvalued exchange rates; under these conditions, zero savings can be a utility-maximizing decision for households. Not including families with zero financial savings in the estimation procedures can therefore lead to self-selection bias and inconsistent parameter estimates (Cameron and Trivedi 2005; Wooldridge 2010). An estimation procedure that takes into account households with zero savings and the two-stage nature of the outcome variable is suitable for the analysis of household saving behavior.

Therefore, the Tobit model is the most commonly used method for estimating left-censored response variables. However, the Tobit estimation of the data violates the assumption of an identical coefficient in the probability equation (probit) and the intensity equation (truncated) of household savings. Therefore, the two-part model is applied to estimate household saving behavior in Ethiopia; Heckman sample selection estimates are also provided for comparison. Thus, we have two possible equations: the participation equation and the intensity equation.

A household has positive savings if the expected benefit of saving outweighs the expected benefit of not having financial savings (Belotti et al. 2015; Cameron and Trivedi 2005; Wooldridge 2010). The equations for participation are

$$p\left(s_i=1 \mid X^h=\left(E\left(U_s\right)-E\left(U_{n s}\right)\right)\right)=\Phi\left(X^{h^`} \sigma\right) \tag{1}$$

$$s_i=X_i^h \sigma_i+\sigma_0 D+\varepsilon_i \quad \varepsilon_i \approx N(0,1) \tag{2}$$

Where $p(.)$ is the probability operator, $s_i$ is the binary variable that takes the value 1 if the head of the household has positive savings and 0 otherwise, $X^{h^`}$ is Vector of socio-economic and demographic variables influencing savings participation, $E(U_s)$ is the expected utility of saving, $E(U_{ns})$ is the expected utility of not saving, $\Phi$ operator of the cumulative probability function and $\sigma$ parameter of the estimates in the participation questions, $D$ stands for dummy variables to identify households from urban and rural areas and $\varepsilon_i$ is random error term of the participation equation. The participation equation is estimated using the probit model for joint urban and rural data and independently for urban and rural data.

The intensity equation is solely dependent on having positive savings,

$$E\left(s_i \mid s_i>0=g\left(X^{h} \beta\right)\right)\tag{3}$$

$$\ln \left(s_i \mid s_i>0\right)=X^h \beta+\beta_0 D+v_i \quad v_i \approx N(0,1)\tag{4}$$

Where $g(.)$ is the density function of households with positive savings, $\beta$ is a vector of parameters associated with the intensity equations $\ln\left( s_{i}|s_{i} > 0 \right)$ is a natural logarithm of the average household savings, $D$ is a dummy variable to identify the households’ place of residence and $v_i$ is the error term. Since the coefficients are additive and separable according to Belotti et al. (2015), the participation equation can be estimated with probit and the intensity equation with OLS.

The estimation was conducted in two ways to observe the variation in determinants between urban and rural households. First, the data from urban and rural areas are merged by introducing a dummy variable for the geographical location of households. Second, depending on the statistical significance of the dummy variable introduced, the intensity and participation equations are estimated independently for urban and rural families. For the intensity equations, however, a hierarchical OLS is used to check the relative importance of the predictors. Stepwise addition to R-squared is used to observe the relative importance of each variable.

For a detailed measurement and detailed hypothesis of the effects of the included socioeconomic and institutional predictors. The Heckman estimation of each model is retained in the analysis for comparison purposes. The hypotheses and measurements of the socioeconomic and demographic variables used in the analysis are presented in section 4 of the table 1.

7. Two part and Heckman estimation of the urban, rural, and joint model

The Heckman sample selection and the two-part models meet the basic regression diagnostic tests, the variables included in the model are jointly statistically significant, and the estimates of the two models show almost identical empirical results. The F-statistics and chi-square statistics are statistically significant at a one percent significance level. The R² of the participation and intensity equation is low, which is common for cross-sectional analyses. The Mill’s Ratio from the Heckman sample selection is not statistically significant; therefore, following the methodological hypothesis, the estimation result favors the two-part model, the dominance of the participation equation.

The determinants of household savings are almost identical among rural and urban households. The income of the household head, financial literacy, education level, financial pressure, inflation, access to credit and social capital are statistically significant in shaping the saving patterns of urban and rural households. The difference lies in marital status, religion and ethnicity. Where ethnic background, marital status, and religious affiliation have differential effects on household saving patterns.The effects of ethnicity and marital status are strong in rural areas, while religious affiliation is more robust in urban households. A significant proportion of families in the Muslim community consider bank interest a sin; therefore, instead of saving in banks, families prefer to join informal credit rotation schemes using ethnic and religious contact lines. Unlike banks, informal credit rotation has two advantages: members must save continuously, and for the winners, it serves as a source of access to credit (you keep paying until the money is distributed to all members). Therefore, instead of following bank interest rates, they prefer informal lottery-based credit rotation to save continuously and overcome the cost of a religious festival and pilgrimages. The determinants of household saving behavior in urban and rural areas are thus identical, except for some irregularities. However, actual savings are higher in urban and semi-urban areas that produce cash crops.

The two most important conventional economic variables, income and age, are statistically significant in each model. Taking the joint model as a reference, the probability of saving is higher for high-income households. Households in the fourth income quartile are 2.3 times more likely to save than households in the first quartile, and their average savings are 1.27 logarithms higher than those in the first quartile. In developing countries, the willingness and ability to save depends on having more than the resources needed to meet basic needs. This study’s results align with empirical expectations, as the regression result shows that income plays a crucial role in shaping household savings(Carpenter and Jensen 2002). In line with theoretical expectations, saving varies between different age groups. Households of older age are less likely to save; households aged 20 to 30 are 0.88 times more likely to save than households aged 65 and older.

The demographic and sociological factors included, such as religious affiliation, ethnic origin, marital status, level of education and gender of the head of household, are all statistically significant, except the last factor. Compared to all other ethnic groups, households with Gurage and Siltie ethnicity are more likely to save, followed by Amara; this is in line with the research hypothesis. A household of Gurage and Silitie ethnicity is known to have a family familiar with the art of economizing and managing savings portfolios.

In contrast to the hypothesis investigated, the savings of Muslim households are 0.525 logarithms lower than those of Orthodox Christian households. This may be related to the recent increasing phenomenon of pilgrimage to the holy land of Mecca and Medina as a religious duty, the consideration of interest earned on savings as a sinful act and the low priority of cash as a savings portfolio. Although limited to rural areas, married households have higher average savings compared to single and divorced households. In line with the hypothesis, the household head’s education level contributes positively and significantly to household savings. One more year of schooling increases the probability of saving by 11.6 and 8.3 percent for urban and joint rural/urban households, respectively. Membership in the local insurance rotation scheme and income from remittances are used as proxies for measuring the social capital of the household head. Accordingly, households with social capital are 1.8 times less likely to save than households without social capital.

Of the institutional variables included, access to credit and the financial literacy index are statistically significant in influencing the savings of urban and rural households. The result of the estimation shows that a one-unit improvement in the financial literacy index of the household head increases the probability of saving for rural and urban households jointly by 3.6 percent. The coefficient of easy access to credit has a different sign in the equation for participation and intensity of saving. A family with easy access to credit has a high potential for cash savings with a lower probability of saving compared to households without access to credit. A household with access to credit is willing to take the risk of investing its cash to build up savings; therefore, a low savings intensity is justifiable for a household with access to credit.

As far as I know, this paper is the first to include behavioral variables such as the savings habit index, motives for saving, and perceptions of inflation in the microeconomic analysis of household saving in Ethiopia. Frequent visits to the nearest branch measured households’ saving habits or ability to postpone consumption. A household that frequently visits the nearest financial institution has a low level of savings but a high level of participation. Most empirical work with macroeconomic data concludes that the real interest rate does not have a statistically significant effect on saving; however, the negative real interest rate (rising general price level) is reflected in the behavioral change to diversify the savings portfolio from cash to tangible assets. Inflation (negative real interest rate) significantly impacts cash savings by influencing the choice of a household`s savings portfolio.

8. The Hierarchical estimation of the intensity of saving

The stepwise increase in R² is statistically significant, with the two most empirically studied conventional economic variables (income and age) contributing about 19.55 percent of the explained variation in the dependent variable, followed by institutional variables (6.06%), behavioral factors (4.69), and sociological factors (2.25). Thus, although no single theory is able to explain the existing reality of household saving, income (permanent income hypothesis) and institutional theories play an important role in explaining the saving behavior of Ethiopian households.

The stepwise increase in R² is statistically significant, implying that improving household savings in Ethiopia requires multidimensional and disaggregated strategies. Improving household income, family planning, education and financial literacy must go hand in hand to improve household savings in Ethiopia.

9. Conclusion and recommendation

Two parts, the Hekchman and Hierarchical models, were applied to determine whether household savings’ determinants differ in urban and rural areas. The result of the estimation shows that there is no significant difference in the determinants of savings between the two areas. The difference is limited to variables such as marital status, ethnic origin and religious affiliation. Consistent with many empirical findings, income and age are the two most important determinants for urban and rural households, followed by behavioral variables such as access to credit and financial knowledge. In addition, behavioral and sociological factors such as saving habits, reaction to inflation, financial pressure and education are also essential in urban and rural areas. However, religious denomination is more important in urban households, while both religion and ethnicity play a role in rural households. Religion and ethnicity particularly influence spending behavior, the perception of interest income and the choice of savings instruments. Many families consider interest income a sin; they save in kind to fulfill religious obligations and rarely prefer cash and savings instruments. The motive of consumption smoothing is the main driver for saving in both rural and urban areas.

As this paper is the first to include variables such as savings habits, inflation perceptions, financial literacy index and religious ethnicity, it has new implications. First, most empirical work using macroeconomic data concludes that the real interest rate does not statistically affect household savings in Ethiopia. In line with the empirical prediction based on macroeconomic data, interest rate as an instrument to attract savings may not perform well due to the existing religious beliefs. However, the estimation results from the survey data suggest that a persistent negative real interest rate has a behavioral effect in diversifying a savings portfolio. Due to the negative real interest rate and the overvalued exchange rate, many of the surveyed households shift their savings from cash to real assets; sometimes they even save by illegally buying hard currencies. Second, a triangular analysis of the income and financial literacy index coefficient shows that household heads with a high disposable income and a high financial literacy index rarely prefer cash to save a portfolio. Third, half of the households surveyed have no cash deposits, and savings in kind are twice as high as cash deposits. Thus, the future of domestic capital mobilization in Ethiopia depends on economic growth, digitalization of the financial sector, introduction of new savings instruments, management of artificially overvalued urban land prices, population dynamics and stability of market prices and exchange rate.

Policy recommendations

• Considering the stability of land and property prices as an integral part of the saving policy. The artificially overvalued land and property prices in urban areas harmed the demand for monetary saving instruments; households preferred to save in tangible assets instead of going to the nearest bank.

• The inclusion of financial literacy in primary and secondary school textbooks improves savings and enables individuals to rationalize their choice of savings instruments.

• Introducing savings products that take into account the ethnic, cultural and religious aspects of households. For example, savings products are offered for pilgrimages, funerals, and important religious and ethnic festivals.

• Linking savings with access to credit: Most commercial banks receive demand deposits from their customers but do not offer access to credit based on savings. For example, if you have saved 80% of the cost of the pilgrimage, banks can provide the rest as credit to fulfill your religious obligation.