# Determinants of Gross Domestic Saving in Ethiopia: a Time Series Analysis

Economic literatures on economic growth models such as Harrod Domar and Solow growth model suggested that saving is an essential factor to the working of any economic growth.As capital formation is an important factor in economic growth, country that was able to accumulate high level of investment achieved faster rate of economic growth and development (Shini and Kadhikwa, 1999).

Savings determine economic growth through facilitating financial opportunities for investment. Increased investment increases the productive capacity of an economy. Adding to the stock of capital goods increase the nation’s potential out put and promotes economic growth in the long run. That is, for continues growth and future prosperity of a country a high level of investment is essential (Ayele, 2006).The necessary resources for investment are obtained through cumulative savings of income. It fallows that investment also depends upon saving and saving required sacrifice of current consumption to make available resources to finance investment.

Thus, to finance investment required for economic growth, a nation needs to generate sufficient domestic saving or it should borrow abroad and/or develops FDI (Shini and Kadhikwa 1999). However, developing countries still show low performance to attract FDI due to economic and political uncertainly.Further more, the flow of foreign direct investment (FDI) attracted largely by natural resource endowment especially oil and minerals which are scarce in Ethiopia (Getnet and Hirut, 2006). The second alternative to finance domestic investment is through official development assistances (OAD) which are mostly tied up with conditionalities. The main problem of these conditionalities is that donors forced the recipient government to undertake political and economic policies reform which may not accepted by the recipient government (Soubbifina, 2004).Beyond this foreign borrowing may cause exchange rate problem in the long run.

However increased saving does not always corresponding to increase investment, if savings are not deposited in to financial intermediaries like banks there is no chance for those savings to be recycled as investment by business man. This means that saving may increase without increased investment, possibly causing decreased “effective demand” rather than economic growth.This is, therefore, that Keynes (1936) said saving become to be seen as potentially disruptive to the economy and harmful to social warfare as it reduce one component of demand, consumption, with out increase investment and it may cause inadequate demand and hence out put and employment lower than the capacity of the economy. This failure may attribute to variety of reasons including wage rigidity, liquidity preference and fixed capital coefficient in production (Modigliani 1986). Despite the importance of saving for economic growth, saving rate is lower to finance the domestic investment in most developing countries.

Sub Sahara Africa has low gross domestic saving (18% its GDP) when compare to South Asia, 26% and newly industrialized countries 43% in 2005 (IMF, 2007). For Ethiopia, during the imperial era, gross domestic saving as a percent of GDP was 11% on average. After the socialist state took power in 1974 there were expectations to wards the increment of saving by eliminating the luxury life style of the ruling classes. In actual fact the policy of imposing capital ceiling became a serious disincentive to saving class and further encouraged conspicuous consumption (Befekadu and Birhanu, 2000).Instead of increasing, what turned out during the Derg regime was the ratio of gross domestic saving (GDS) as percent of GDP has declined from 11% to 4% on average, while further show a very haphazard rate during the entire Derg regime from high of 7% in 1976-1986 to even less than 1% during the last period of the Derg regime as a consequence of increasing government consumption on military expenditure (Dawit, 2004). Spurred by the sound economic policy and favorable weather condition, the Ethiopia economy witnessed on encouraging over all economic performance for the last six years as real GDP grew by more than two digits.

Despite this promising and sustainable economic growth performance, gross domestic saving still does not show substantial progress in the same years as it was 4. 26 for the period of 2004-2008 with increasing resource gap (NBE, 2008). Ethiopia continues to face a potential shortage of resource to finance public and private investment, which constraints its ability to accelerate economic growth. The chronic resource gap shown is from imbalance between domestic saving and domestic investment as show above.The resource shortage the adversely affect ability of government to under take expenditure in infrastructure and social service to boost domestic demand, encourage private activities and sustain high level of economic growth (ECA, 2006). Unless this trend is quickly reversed, it is bound to have a significant negative effect on the future economic prospects of the country.

The negative saving in 1999/2000 was an obvious indication that the country was consuming more than it was producing and what ever investment took place during this year was a result of foreign grant and loans (Befekadu and Birhanu 2000).Thus, to fill the financing gap and accelerating growth, Ethiopia needs to mobilize more domestic resource. This, therefore, the principal objective of this paper is to investigate the determinants of gross domestic saving in Ethiopia. In addition to this, the study tries to suggest policy recommendations to improve gross domestic saving. The paper focuses on the determinants of gross domestic saving growth in Ethiopia using aggregate time series data.

The explanatory variables included in the model are limited to macro economic and demographic variables given others factors constant.Studies that analyze more disaggregate data and uses non time series model may come up with different result. The study has five chapters. Chapter one provides an introduction and justification of the study. Chapter two reviews the theoretical and empirical literature on the role and determinants of saving.

Chapter three focuses on data used and the empirical model employed. Chapter five devoted to discussion of estimation procedure and analysis of empirical results obtained from the empirical exercise. The final chapter summarized the main results of the study and provides recommendations.CHAPTER TWO 2. Literature Review 2.

1 Roles of Saving: Theory and Evidence Many economic literatures on economic growth such as Solow and Harrod- Domar models suggested that saving is an essential factor for economic prosperity. Economic growth is largely related with the rate of investment which in turn related to savings. As capital formation is an important factor in economic growths, countries that were able to achiever high level of capital accumulation enjoy high level of economic growth (Shini and Kadhikula, 1999).Saving can, therefore, be vital to increase the amount of fixed capital formation, which in turn contributes to economic growth. Increased saving, however, does not always corresponding to increase investment if saving stashed in a mattress or other wise not deposited in to financial intermediaries like bank there is no way for those saving to be invested by productive economic agents (Romm, 2003). There are three important stages in process of capital formation that need to be successfully completed if countries hope to achieve the desired level of investment and economic growth.

These are, the stage of mobilizing of saving, stage of canalization saving and stage of putting mobilized saving in to investments on capital good (Befeadu and Biranu, 2000). Thus an economy that have problem of mobilizing and using investable resource is as a result of failures in one of these stages. Empirically there are ample evidences on the relationship between saving and economic growth in developing countries as well as developed countries. These studies also suggested there is bidirectional causality relationship between saving and economic growth.To illustrate, Romm (2003), using time series data for the period of 1946-1992 find that the private saving has a direct as well as indirect effect on economic growth.

The indirect effect is through the private investment. In turn he find also that GDP has positive effect on saving and conclude growth enhance saving which in turn further enhance economic growth. Moreover Remesh, Mohan and Unire (2003) using time series data and Ganger causality test for 25 countries with different income level, found that economic growth cause saving in 13 countries and in five countries it consistent with the theory that saving cause economic growth.Ndikumana (1999) come up with the results that indicate a positive relationship between domestic investments and varies financial development indicators. High levels of financial development lead to high level of implying a potential long run effect of financial development on domestic investment.

Particular study conducted by Dawit (2004) using time series data (1960/61-2002/03) and Granger causality test, supports to the hypothesis that faster growth rate of real gross domestic saving causes faster growth of real growth of GDP in Ethiopia and vice Versace.Therefore, in Ethiopia bidirectional (feed back) causality relation ship between economic growth and domestic saving was found. Generally, the study of aggregate saving has long central to economic growth and development because national saving is the source of the supply of capital which is a major factor of production that control the productivity of labor and it’s therefore that saving regarded as social beneficial act (Modigliani 1986). 2. 2 Determinants of Saving: Theory and Evidence 2. 2.

1 Theoretical DeterminantsIncome variables: The Keynesian saving function postulate a positive relationship between saving and current level of income. Keynesian saving function state that as current income level increase saving also increase (Agarawal, 2000). The permanent income hypothesis develops by Fried man proposed two components of individual income i. e. permanent and transitory income.

This view postulates that consumer spent their permanent incomes. But they save rather than spent most of their transitory income (Mankiw, 2001).The life cycle hypothesis suggested that the aggregate saving will increase in response to increase in income growth via an increase in saving of active worker relative to the dissaving of people out of labor force (Modiglani, 1986). Demographic variables: Modiglani (1986) in his work, entitled by life cycle, Individual thrifty and wealth of nation, postulates several demographic variables that affect saving. The dependence ratio is the most common demographic variable in many empirical studies.

The elderly and young are expected to consume out of past saving while those within the working age are expected to accumulate saving (Quartey and Blankson, 2007). The variation in the age composition of the population may then determine the variation in national saving rate over time and /or across counties (Schultz, 2004). Further more dependent population is likely to reduce labor supply of productive group as they demand labor for their care (Hussein, 2007).Financial variables: The financial variables that have an impact on saving are usually one that captures the degree of development of financial sector. Among these, interest rate is a major and ambiguous variable. Classical economists argue that as interest rate increase people would like to save more because they expect increase in rate of return on saving.

But Keynesian argue that saving is interest inelastic i. e. interest rate has no significant effect on saving, so that to save more large interest rate is needed (Agarawal,2007).There is also an argument that higher interest rate on saving raise the stream of future income and wealth, thus raising the current consumption level in turn decreases current level of saving (Shin and kadhikula, 1999). The relationships between interest and saving depend upon the relative strength of the substitution and income effect of the charge in interest rates. In the case of substitution effect, arise in interest rate is assumed to increase future consumption and depress current expenditure there by increase current saving.

In the case of income effect, a rise in interest rate will increase future income which encourage economic agent to consume more currently at the expense of current saving (Quartey and Blank son 2007). Other than interest rate, financial development could be measured in variety of ways including the ratio of financial asset to the national wealth, distribution of financial wealth, distribution of financial assets and various financial instruments, number of different type of financial institution, capital and money market etc.Development of money supply is one of the indicators of development in financial institution. In general the higher ratios and the number of different instrument and institutions the more developed in financial sector which may improve the opportunity to save among the society (Befkadu and Birhanu 2000). Therefore, financial development in a country is important in enhancing the mobilization of saving and channeling of saving into productive investment sector (Ndikumana, 1999). Uncertainty variables: There is consensus that macro economic stability is good for saving.

Stability increase confidence of in the financial system and encourage house holds to place their saving in financial institution that channel resource fund to productive use. Household saving appear to have benefited form the stability offered by low inflation with out suffering much low interest rate. If interest rates do not adjust in simultaneously to change in inflation rate, the relative rate of return on financial saving and real assets are at the expense of financial saves (Modiglani, 1986).Similarly lower real interest rate causes a portfolio adjustment from real money balance to ward real capital. However, this may not be the case in developing countries (such as Ethiopia) with relatively under developed capital and money markets.

In such case portfolio adjustment would likely from real money balance to real asset like land, live stock, grain and consumer durables (Shini and Kadhikula, 1999). Further, Political uncertainty like war is a major source of the distraction of national wealth and income which leave little portion of national income that would like to save.Further more political uncertainty especially war increase the government expenditure on unproductive activity which directly deteriorate public saving and hence gross domestic saving. In contrast to this there is an argument that war may increase private saving due to the fear to invest in productive activities rather they prefer to save and/or hoard their money (Befekadu and Birhanu, 2001). Policy variables: Various action of the government can have a bearing on the gross domestic saving. Among these, the effect of fiscal policy has been the center of debate (Guyan and Ertac, 2000).

The relation ship between tax rate and private saving is based on the premises for reduction of disposable income through the direct or indirect taxation levied on reduces personal consumption and the capacity to save as well, regarding marginal propensity to save is higher income classes, then a highly progressive income tax or direct tax reduce the over all rate of saving in the private sector, considering this reduction in private saving would be offset through the increase in public saving at a net effect while assuring this increase revenue is not used to increase government consumption on defenses or civil service (Stchutz, 2004).With regarded to private saving, the effect of tax on saving depends on the duration of tax imposition. A temporary rise in tax rate is assumed to reduce disposable income, which intern depresses private savings. This is because people are aware of the fact that the tax rate is temporary and hence smooth their consumption by borrowing against the future rather than saving. In the case of permanent tax household no more smooth their consumptions since they know that the tax imposition is permanent.

In this case they reduce their consumption by a greater margin than they use to do for the temporary tax impose (Agarawal, 2000). The Ricardian Equivalence hypothesis states that it does not matter whether government finances its expenditure through taxes or by borrowing. Therefore, only the time path of government expenditure affects gross domestic saving and not the time path of taxes that finance such expenditure and at the end the tax imposition have zero sum game effect in gross domestic saving (Nwachukwu and Egwaikhide, 2000). 2. 3 Empirical Determinants of SavingThe hypothesis that saving and income has positive relation ship is supported by numbers of studies conducted at aggregate as well as micro levels. Horioke and junminwan (2007) found one percent increase income growth rate causes a 0.

20 to 0. 54 percentage increase in the house hold saving rate. Moreover the long run impact of the income growth rate was 3. 58 to 7. 63 time the above figure in Chaina.

In an attempting to identify the determinants of saving in Nigeria, Nwanchukwu and Egwaikhide (2000) fund that income has a positive and significant effect on the private saving in Nigeria.Also Hussein and Thiwal (1999) using panel data across counties found that 75% difference saving ratio between counties is in terms of the level of income. Using cross sectional data Hussein (2007) reveal that both farm and nonfarm income were statistically significant. He shows that one birr increase in non farm income causes 0. 52 birr increase in saving in South East Ethiopia. However, Guyen and Erac (2000) examine the determinants of saving in Turkey and found that growth of per capital income had insignificant impact on saving rate for the period of 1964-1994.

Schultz (2004) in analyzing the demographic determinants of saving in a group of Asian countries found that dependence ratio of old as well as youth population has a significant negative effect on saving across counties. Hussein’s (2007) work on the behavior of form house hold saving in Ethiopia (South East) show the dependence ratio is highly significant in determining the saving rate of farmers. However there are studies that are not confirm with the hypothesis of Modigliani that the age structure of a nation can affect saving in the nation with high dependence ratio.The work of Guyan and Erac (2000) indicate increase independence ratio of old and youth has no significant effect on saving. Further Horioka and Juminwan (2004) indicate that variable relatively to age structure of the Chinese population usually do not have a significant impact on house hold saving rate. Contrast to Guyan and Erac (2000), and Harioka and jimniwan (2007), Cardenas (1997) study indicate that increase independence ratio have had a significant negative effect on the private saving in Colombia.

In most empirical study, the direction of the effect of interest rate change on saving is inconclusive due to in the temporal consumption decision (Quartey and Blank son, 2005). Horioka and jimnwan (2007) came up with the result that increase in interest rate has a significant positive impact on the house hold saving rate for every pooled sample, which suggests that the interest elasticity of saving is positive in Chaina. However, there are many studies that come up with in significant result (Nwachukwu and Egwaihide (2000), Hussein (2007) and Guyan and Erac (2000)).Nwanchukwu and Egwaihide (2000) found that there is a negative and insignificant effect on saving due to charge in interest rate on bank deposits. Quartay and blankson (2005) on their work suggested that the effect of real interest rate may also be explained by the inflation effect that it, they assume constant interest rate and a rise in inflation rate lower the real cost of borrowing which in turn increase in consumption expenditure and lower saving. Many studies use different variables to capture the development of financial market and its effect on saving at aggregate as well as house hold level.

The result of Guyan and Erac (2000) regarding financial variable clearly shows the money to GDP ratio is positive and significant effect on saving. This finding confirms the papers hypothesis that an increase in growth of broad money is likely to increase saving rate. Cardenas (1997) found that increase in bank density will increase saving habit among Colombian economic agents. Turn to Africa, Nwachukwu and Egwaihide (2000) work on Nigerian economy show that increase money to GDP ratio associated with increase saving in Nigeria.However, Quartey and Blankson (2005) show that money to GDP ratio does not positively associated with saving in Ghana. Empirically, Hussein and Thirlwan (1999) in explaining the difference in domestic saving across countries of the world using panel data found inflation exert a mild positive effete on saving but soon turn to negative.

Guyan and Erac (2000) estimation result show that inflation had a positive coefficient and statistically significance when inflation rate regress with income, lagged saving rate and demographic variables in Turkey.However, Cardenas (1997) for the sample period of 1925-1994 the volatility of inflation does not seem to affect private saving in the long run in Colombia. On the other hand Nwachukuwa and egwaikhide (2000) found that inflation has a positive impact on the private saving in Nigeria. Various action of the government can have significant effect on saving. Among these the effect of fiscal policy is center of debate. This is because the impact of fiscal policy on aggregate saving depends on the responses of economic agents to change in fiscal policy instruments.

Hussein and Thriwal (1999) found that change in government budget deficit associated positively with change in aggregate saving in two counties. However the positive effects that change in budget defect affect negatively aggregate saving holds for four counties. However, the Recardian equivalence theory that change in budget deficit maintaining aggregate saving un affected holds true for only one counties in the group of seven African counties. CHAPTER THREE 3. Data and Methodology 3. 1 DataThe data include the wide range of macro economic and demographic variables for the period of 1971-2009.

Such time series data have an advantage to show the long run effect of explanatory variables on the dependant variable in macro economic variables (Gujarati, 2003). Table: 1 Data sources and definition of variables |Symbol |Definition of variable |Source | |GGDS |Annual growth rate of real gross domestic saving MOFED | |PCI |Real per capital income computed by dividing real gross domestic product to |MOFED | | |total population | | |GPCI |Annual growth rate of real per capital income |MOFED | |INF |Annual growth of consumer price index |CSA and NBE | |DR |Dependence ratio which is computed as the proportion population under the age |World development indicators | | |of 15 years and above the age of 65 years old to the total population |data base | |RIR |Real deposit rate which is equal to nominal deposit interest adjusted to the |NBE | | |rate of inflation. | | |GM2 |Annual growth of broad money supply |NBE | |GTR |Annual growth of total tax revenue including direct and indirect tax revenue |MOFED | 3. Model Specification Both theoretical and empirical work on saving have consistently out lined major potential determinants of saving that can be grouped under the head of: • Income variables • Demographic variables • Financial variables • Uncertain variables • Policy variables (Guyan and Erac, 2000). In line with potential saving determinants out lined above and in literature review as well as other empirical studies conducted at aggregate level as well as the availability of data the following specification is used in this study to examine the impact of macro economic and demographic variables on the growth of real gross domestic saving in the Ethiopia.

GGDSt =(0+(1PCIt+(2GPCIt+( 3INFt+(4DRt+(5RIRt+(6GM2t+(7GTR7+(8GGDS t-1+Ut ……… (1) + + – – + + – – Where,(‘s – coefficient of determinants Ut – stochastic error terms. Note: The sign under equation (1) is the expected (hypothesized) relation of explanatory variables with the dependant variable. Where, GGDS is growth of real gross domestic saving, PCI is per capital income, GPCI is growth of per capital income, INF is inflation rate, DR is dependence ratio, RIR is real deposit rate, GM2 represent growth of broad money, GTR represents growth of tax revenue and GGDS-1 is define as lagged growth of real domestic saving .In equation (1), the study includes the lagged growth of gross domestic saving (GGDSt-1) in order to capture inter-temporal behavior of economic agents. Since all variables except per capital income (PCI) in equation (1) are in rate term it is not need may (be impossible if it has negative value) to transfer into log form (Asteriou and Hall, 2007). CHAPTER FOUR 4.

Estimation Procedures and Analysis of Empirical Result 4. 1 Estimation procedures As noted by Asteriou and Hall (2007), Engle and Granger proposed a straight forward method which involves four steps in order to get the co integration (long run) and error correction (short run) results. Step 1: Test for Stationary and Order of IntegrationOne of the classical linear regression assumptions is that the series should be stationary in order to estimate on the base of ordinary least square (OLS) procedure. The problem with non stationary or trended data is that the standard OLS regression procedures can easily lead to incorrect conclusion (Spurious regression). It can be shown that in this case the norm is to get very high value of R2 and very high value of t-ratio while the variable used in the analysis have no interrelationship (Asteriou and Hall, 2007). Empirical studies usually use the Augmented Dickey Fuller (ADF) and Phillips-person (PP) test to check for stationary in data series.

For this study, it employed ADF test and applied the test to each variable that are included in the analysis. ADF test requires running a regression of first difference of series against the series lagged once, lagged first difference term and a constant with a time trend of the following form (Gujarati 2003). (Yt = a + bt +cYt-1+ (d(Yt-1+vt…………………….

……

…..

…..

…..

. (2) Where, Yt is the variable to be tested, t is time trend and vt is the error term. The ADF test concerned on the OLS estimate of C, where the null hypothesis (HO: C=0) is that the series are integrated of order one, I(1) or unit root (Gujarati, 2003). Table: 2 Augmented Dickey Fuller (ADF) unit root test result Variable |Trend an constant |(ADF statistic) |I(d) |No: of lags | | | |ADF critical value | | | |GGDS |Constant |(-3. 7985)* |I(0) |1 | | | |-3. 6289 | | | |PCI |Constant |(-3.

7422)* |I(1) |1 | | | |-3. 6289 | | | |GPCI |Constant |(-4. 8529)* |I(0) |1 | | | |-3. 289 | | | |DR |Constant |-3. 6607)* |I(1) |1 | | | |-3. 6289 | | | |R/R |Constant |(-4.

9027)* |I(1) |1 | | | |-3. 6289 | | | |GM2 |Constant trend |(-4. 7522)* |I(1) |1 | | | |-3. 289 | | | |GTR |Constant and trend |(-5. 0301)* |I(1) |1 | | | |-3.

6289 | | | *1% significance level If the estimated ADF statistic is larger (in absolute term) than its ADF critical value the null hypothesis is rejected suggesting that the series are stationary (Asterious and Hall 2007). The out put of ADF test in table: 2 indicated that GGDS and GPCI are stationary at level, I(0). The remaining variables (PCI, DR, RIR, GM2, and GTR) are non stationary at level.But after differencing the variables once, they all obtain stationary behavior at 1% significance level. Step 2: Estimate the Co-integration Model By definition, co-integration regression necessitates that the variable should be integrated of the same order.

As discusses in step 1 test for stationary section, all variables are integrated of order one except GGDS and GPCI which are integrated of order zero. However the robustness of ADF test allows us to treat the variables (GGDS and GPCI) as I(1) and proceed with co-integration analysis. Therefore, if all variables are integrated of same order then we proceed to estimate the long run (co-integration) relation ship (Asteriou and Hall, 2007).The result of co-integration model (equation (1)) is presented in table: 4 in section 4. 2 below and will be discussed after testing the existence of co-integration between variables. Step 3: Test for Co-integration If residual series (ut) obtained from long run equilibrium relationship, are found to be stationary I(0) we conclude that saving growth rate and its determinant are co-integrated and result from regression of equation (1) is not spurious rather it is long run relation.

For this purpose, the study conducted an ADF test on the residual series to determine their order of integration using the following ADF specification form: ([pic]t = a[pic]t-1+[pic] -1+ et…………………………………………………………………………… (3)The decision rule is to reject the null hypothesis of no co-integration if the t-value associated with a in equation (3) is more negative than the computed one. Table 3: Unit root test result on residual |Variable |No lags (ADF) |One lag (ADF(1)) | |RESD |(-6. 4284)* |(-476357)* | | |-4. 07 |-3. 73 | *1% significance level From table 2 the decision rule is to reject the null hypothesis that the residual is non-stationary at the 1% level of significance. This indicates there is co-integration (long run) relationship between growths of real grows domestic saving and the selected explanatory variables in the model, (Gujarati, 2003).

Step 4: Develop the error correction model (ECM)Error correction technique helps to correct the potential bias in the estimation of the coefficients in the model with differences that do not take into account co-integration relationships (Asteriou and Hall 2007). Given the variables that include in equation (1), the distinct ECM can be specified as follow using differencing method (Gujarati 2003): D(GGDS) = (0+(1D(PCIt)+(2D(GPCIt)+(3D(INFt)+( 4D(DRt)+(5D(RIRt) +(6D(GM2t)+ ( 7D(GTRt) +(8 D(SGR t-1) +[pic]( RESD-1+Ut…………………………………………. (4) In equation (4), ( is the error correction coefficient and is also called the adjustment coefficient which tell us how much of the adjustment to equilibrium takes place each period or how much of the equilibrium error is corrected.The spurious regression problem arise because we are using non stationary data but in equation (4) every thing is stationary, the change in dependent and change in independent variables are stationary because they are integrated of order one variable and the residual (RESD) obtained from the level regression in equation (1) is also stationary by the assumption of co-integration (Asteriou and Hall, 2007). The result of error correction model is presented in table: 5 and discussed under section 4.

2 below. D in equation (4) represents difference. 4. 2 Analysis of Empirical Result Estimations of equation (1) (long run) and equation (4) (short run) are reported in table: 4 and table: 5 below respectively. In addition, both tables also include the survey of statistical and diagnostic tests.

Durbin-h test (auto correlation test), ARCH test (auto regressive conditional heteroskedasticity and RESET (functional form) detects no problem since their probability value is not significant to reject the null hypothesis of no auto correlation, no heteroskedasticity, and correct specification respectively. Growth of per capital income (GPCI) exert strong positive pressure on growth gross saving rate(GGDS) in Ethiopia in the long run as well as in short run model which support the Modigliani Life Cycle Hypothesis. However, the Keynesian theory does not hold true for Ethiopia since current level of per capital income(PCI) has significant negative effect on GGDS in co-integration model, but turn to in significant in error correction model.This suggest that more of current income level goes to for the purpose of current consumption rather than to ward saving in Ethiopia and confirms with the study of Guyan and Erac (2000) in Turkey. Table: 4 Co-integration regression result of equation (1).

|Dependent Variable: GGDS | |Variable |Coefficient |Std. Error |t-Statistic |Prob. | |C |1251. 775 |402. 0986 |3. 113104 |0.

0042* | |GPCI |1. 654026 |0. 744492 |2. 221684 |0. 0346** | |PCI |-0. 67558 |0.

212201 |-2. 674622 |0. 0123** | |INF |-2. 203717 |0. 737010 |-2. 990076 |0.

0058* | |DR |-11. 67030 |3. 847951 |-3. 032860 |0. 0052* | |RIR |-0. 660835 |0.

599412 |-1. 102473 |0. 2796 | |GM2 |-0. 505200 |0. 587620 |-0. 859739 |0.

3972 | |GTR |1. 270448 |0. 329979 |3. 850085 |0. 0006* | |GGDS(-1) |-0. 809771 |0.

108414 |-7. 69244 |0. 0000* | |R-squared |0. 771075 |ARCH prob |0. 6755 | |Adjusted R-squared |0. 705667 |RESER prob |0.

2773 | |Durbin-Watson stat |2. 150198 |F-statistic |11. 78883 | |Dutbin-h stat |-0. 6077 |Prob(F-statistic) |0. 000000 | *1% significance level, ** 5% significance level Regarding the uncertain variable, inflation rate (INF) has negative impact on GGDS in both the short run and long run regression results as shown in table 4 and 5 respectively.Given under developed financial market in Ethiopia, the negative effect of inflation rate on saving rate seem to takes place through portfolio adjustment from real money balance toward real assets like land, stock, grain and consumer durable as in the case of most developing countries.

The coefficient of dependence ratio (DR) is significant and negative as expected in the long run model. The insignificant effect of dependence ratio in the short run model is possibly because of that the population structure of Ethiopia does not dramatically change within a year. The result confirms with the theory that the age composition of the population of Ethiopia is associated with its saving rate and has consequence for its economic growth over time.The two financial variables, real interest rate (RIR) and growth of broad money supply (GM2) included in both model to use as proxy for the development financial market, are insignificant and negatively related with growth of gross domestic saving in Ethiopia. The insignificant effect of real deposit rate is caused by insufficient devolvement of financial and capital market in Ethiopia. The tax revenue growth (GTR) show there is positive and significant correlation with growth of real gross domestic saving (GGDS) in both co-integration and error correction models.

The result evidenced that the Ricardian equivalence theory does not hold true for Ethiopia.This implies that increase in government saving due to tax increase offset decrease in private saving due to tax imposition. Table: 5 Error correction regression result of equation (4) |Dependent Variable: D(GGDS) | |Variable |Coefficient |Std. Error |t-Statistic |Prob. | |C |-0.

045035 |3. 009408 |-0. 014965 |0. 9882 | |D(GPCI) |1. 139216 |0. 544904 |2.

090675 |0. 0465** | |D(PCI) |-0. 298290 |0. 262398 |-1. 136783 |0. 2660 | |D(INF) |-1.

661325 |0. 577889 |-2. 874819 |0. 080* | |D(DR) |-8. 257457 |6.

047722 |-1. 365383 |0. 1838 | |D(RIR) |-0. 086483 |0. 451918 |-0. 191369 |0.

8497 | |D(GM2) |-0. 402607 |0. 410730 |-0. 980224 |0. 3360 | |D(GTR) |1. 278966 |0.

278540 |4. 591681 |0. 0001* | |D(GGDS(-1)) |-0. 813177 |0. 069303 |-11.

73359 |0. 0000* | |RESD(-1) |-1. 113589 |0. 210094 |-5. 300439 |0. 0000* | |R-squared |0.

34822 |ARCH prob |0. 9341 | |Adjusted R-squared |0. 912261 |RESET prob |0. 5655 | |Durbin-Watson stat |2. 003126 |F-statistic |41.

43437 | |Durbin-h stat |-0. 0103 |Prob(F-statistic) |0. 000000 | *1% significance level, ** 5% significance level The lagged saving rate variable (GGDS-1) negatively related with the growth of gross domestic saving at all. This show there is inertia among economic agents to show persistence toward saving growth. The magnitude of error correction term (RESD-1) coefficient (-1.

136) implies that after shock is given to the system it takes approximately 11 years for growth of real gross domestic saving to restore its equilibrium level (Ndikumana, 1999). The constant variable (C) is significant at 1% level of significant which suggest that the variable which are not include in the model jointly affect saving growth rate positively in the long run. Chapter Five 5. Conclusion and Recommendation 5. 1 Conclusion This study attempts to investigate the determinants of gross domestic saving in Ethiopia in short run and long run perspectives. For this purpose, the study employed aggregated time series data for about last four decades and reviewed theoretical and empirical relating to the determinants of saving focusing on developing countries.

Further more; the study tries to address the role of saving in its theoretical and empirical reviews. The study findings show that growth of per capital income and tax growth rate have a significant positive impact on growth of gross domestic saving in both models, but level of per capital income have significant negative impact in the long run then turn to insignificant in the short run. Dependency ratio and inflation have a negative impact on the GGDS. However, the financial variables, interest rate and growth broad money does not have significant impact on growth of gross domestic saving at all. The growth of tax revenue has a positive significant impact on GGDS by increasing public saving.Beyond these, growth of gross domestic saving series does not show consistence over the time under consideration in Ethiopia 5.

2 Recommendation • The result implies that it is important to further improve and sustain the current economic growth which can help to improve saving and then increase the potential of capital formation. • To have stable and sustainable saving rate, Ethiopia need have stable its economy in general inflation rate in particular, government can achieve this through restrictive as well as technical measures on the factor that expected to aggravate inflation rate. • It is important to design effective demographic policy that can control youth dominant population structure of Ethiopia. This can be done by roviding the purpose of family planning with adult education and creating awareness in the society for immediate and improve the spread of education in the way to address rural area for long run. • The under development of financial market may further deteriorate domestic saving.

Therefore, the financial market needs to improve there by enhance the economy to mobilize the potential source of the country. The spread of financial institution should try to address the farming sector of the economy which is still major sources of national income. Reference Ayele Kuris (2006) the Ethiopian Economy. 2nded CPP. Aylit Tina Romm, (2003) the Relation Ship Between Saving and Economic Growth in South Africa.

An Empirical Study Asteriou.D and Hall . E. Stephen, (2007) Applied Econometrics: A Modern Approach. Palgrave Publisher Agarawal B. G (2001), Text Book of Economics 8th Edition.

Befekadu D and Birahanu N (1999/2000): Annual Report on the Ethiopian Economy, EEA Vol. I Befkadu . D And Birhanu. N 2000/2001 Annual Report on the Ethiopian Economy, EEA. Vol. II Dawit Shugu (2005), Causal Relation Ship between Economic Growth and Gross Domestic Saving: Case of Ethiopia.

Emmanuel Anouruo and Yusuf Ahmed (2003), Causal Relation Ship between Domestic Saving and Economic Growth: Evidence from Seven African Countries. Fran Co Modiglini, (1987) Life Cycle, Individual Thrift and the Weath ofNation. Gujarati N. D (2003), Basic Econometrics, 4th Edition Hussein Hamda (2007), Farm House Hold Economic Behavior in Imperfect Market Structure. Doctoral Thesis, Simidishee University of Agriculture.

Horioka C. Y and Jumni Wan (2007), the Determinants of House Hold Saving In China: A Dynamic Panel Analysis IMF (2007), World Economic Out Look: Spill Over And Miles In Global Economy, Washing Ton DC. IMF, 2007 Regional Economic Out Look: Sub Sahara Africa Washington DC April. Ipumbuw Shini and Gerso Kadhikula (1999), Saving and Investment in Namibia Jinghan M. L (2001), Money Banking and International Trade Risk Publisher. Kahaled.

A Hussein and A.P Thirlanl (1999), Explaining the Deference in the Domestic Saving Ratio across Counties: A Panel Data Study. Keynes J. M (1936), General Theory of Employment, Interest Rate and Money. Leone Ndikumana (1997), Financial Determinants of Domestic Investment in Sub Sahara Africa.

Evidence from Panel Mausio Carden. (1997), Determinants of Saving In Colombia 1925-1994. Feda Sartoll. AA. 770774 Bagota Colombia Mankiw N.

G (2001) Macro Economics 5th Edition. Nwachukwu E. Egwaikhide (2000), An Error Correlation Model of Determinants of Private Saving In Nigeria. Quartey Blankson (2005), Low Saving In Ghana Does Policy. Remesh, Mahan and Braynt Unire (2003), Causal Relation Ship between Saving and Economic Growth in Countries with Deferent In Come Levels.

T.Poul Schultz (2004), Demographic Determinate Of Saving, “Estimation and Inter Pretaiton the Aggregate Association in Assia, Yale University Tolyan . P Soubbifina (2004), Beyond Economic Growth an Introduction to Sustainable Development 2nded Appendix Table: 6 Correlation Matrixes | |GGDS |GPCI |PCI |INF |DR |RIR |GM2 |GTR | |GGDS | 1. 00 | 0. 44 | 0. 08 |-0.

25 |-0. 12 | 0. 18 |-0. 04 | 0. 21 | |GPCI | | 1.

00 | 0. 52 |-0. 17 |-0. 34 |-0. 01 | 0. 22 | 0.

45 | |PCI | | | 1. 00 | 0. 16 |-0. 84 |-0. 18 | 0.

43 | 0. 33 | |INF | | | | 1. 00 |-0. 23 |-0. 77 | 0.

43 | 0. 3 | |DR | | | | | 1. 00 | 0. 25 |-0. 38 |-0. 21 | |RIR | | | | | | 1.

00 |-0. 38 |-0. 04 | |GM2 | | | | | | | 1. 00 | 0. 36 | |GTR | | | | | | | | 1. 00 | Table: 7 Summery statistics | |GGDS |GPCI |PCI |INF |DR |RIR |GM2 |GTR | | Mean | 9.

34 | 1. 08 | 254. 83 | 7. 67 | 93. 05 |-2.

23 | 14. 28 | 13. 12 | | Median | 6. 6 | 0. 89 | 250.

01 | 7. 00 | 93. 00 |-2. 00 | 12. 80 | 12.

70 | | Maximum |105. 19 | 10. 92 | 363. 08 | 25. 30 | 95. 00 | 12.

00 | 31. 39 | 41. 48 | | Minimum |-55. 41 |-12. 35 | 208.

11 |-9. 50 | 89. 00 |-21. 10 | 3. 20 |-21.

12 | | Std. Dev. | 32. 99 | 5. 72 | 29. 77 | 8.

00 | 1. 49 | 8. 96 | 6. 51 | 11. 74 | | Skew ness | 0.

99 |-0. 13 | 1. 69 | 0. 14 |-0. 94 |-0. 13 | 0.

92 |-0. 09 | | Kurtosis | 4. 60 | 2. 41 | 7. 00 | 2.

50 | 3. 65 | 1. 99 | 3. 51 | 4. 51 |