UNIVERSITY OF ZIMBABWE

FACULTY OF SOCIAL STUDIES

ECONOMICS

NAME: MUSIYA NATSAI W

REGISTRATION NUMBER: R156862W

COURSE: ECON 370

LECTURER: MR PINDIRIRI

YEAR: 2017

TITLE: THE IMPACT OF FINANCIAL SECTOR DEVELOPMENT ON POVERTY IN ZIMBABWE 1980 TO 2015

CHAPTER ONE

INTRODUCTION AND BACKGROUND

1.0 introduction

The linkage between financial development and poverty reduction has not been the subject of much empirical work. Most studies that have been carried out concentrate on the relationship between financial development and growth (McKinnon, 1973; Goldsmith, 1969; Gurley & Shaw, 1955 and Schumpeter, 1911). Poverty is one of the most prominent problems in the world, regardless of numerous measures which have been taken at both macro and micro level to combat poverty. Since poverty is a burning issue of not only the developing countries but also of the developed countries, financial sector development can be used to alleviate poverty hence leading to the achievement of the first Sustainable development Goal (SDG) which is to end poverty in all forms everywhere.

According to the World Bank (2017), the poverty rate slightly increased from 70.9% in 2001 to 72.3% in 2011/2012, reflecting the structural nature in poverty. United Nation (2015) found out that poverty Zimbabwe is more prevalent in rural areas compared to urban areas with about 76% of the rural households considered poor compared to 38.2% of urban households. However, the financial sector faced problems in its diversification as the country faced a fall in the number of banks from forty to twenty six (Kanyeze et al., 2011). Also the financial sector's performance has been poor as measured by the domestic credit to the private sector which has been falling since 2002 from 103.63% to 10.2% in 2008 (World Bank, 2010).

Both theoretical and empirical studies have shown that there are two channels in which financial sector development (FSD) can impact poverty that is directly and indirectly. The indirect channel is the one in which the financial sector development supports economic growth, whilst the direct channel refers to the one in which financial sector development contribute to poverty reduction by providing or broadening the poor’s access to financial services through the McKinnon conduit effect (Zhuang et al., 2009). The causality between financial sector development and poverty in most studies has been found to have a short run bidirectional causal relationship (Uddin et al., 2012 and Ho ; Odhiambo, 2011). These studies usually apply the autoregressive distributed lag (ARDL).

This study is focusing on the relationship between financial development and poverty in Zimbabwe. The OLS approach will be applied, using annual data for the period 1980 to 2015. Also the granger causality technique will be applied in order to test the causality between financial sector development and poverty.

1.1 Background of study

At independence in 1980, the Government of Zimbabwe inherited the best banking systems in Africa excluding South Africa, at a time when the majority black population had a strong sense of having been excluded from access to modern services, including credit (Brownbridge and Harvey, 1998). The country deregulated the banking sector in the early 1990s which resulted in black owned banking institutions coming up to compete with the traditional institutions (Kanyenze et al., 2011).

Figure 1: GDPPC and Household final consumption per capita

Source: World Bank (2017)

Figure 1 above shows the changes in household final consumption per capita and GDPPC over the period 1980 to 2015. Evidence from the figure below shows that after independence (1980-1983) both GDPPC and household final consumption per capita increased and then declined in 1984 and then the trend followed an inconsistent path between the periods 1988 to 2016. This was influenced by factors such as the increase in population for the past three decades from 7.5 million in 1982 to 13.1 million in 2012 and an estimated economic growth of 1.2 percent in 2016. Unemployment is another factor that is threatening poverty in the country as most people are jobless which constrains household consumption expenditure therefore affecting poverty. In December 2017, 5.3% of the labour force of the country was unemployed (World Bank, 2017).

Figure 2: Domestic credit to private sector as a percentage of GDP and Broad money as a percentage of GDP

Source: World Bank (2017)

From the figure above it can be noted that during the period 2000 to 2002 there was an increase in the ratio which might imply that the higher ratio, the larger the size of the financial sector development. The period 2000 to 2002 lies in the third distinct period known as the financial reform reversal according to Chigumira ; Makochekanwa (2014). This period was characterized by reversal of reforms and its main outcome was currency rationing. From the figure above the domestic credit to private sector ratio to GDP dropped from 14.5% in 2014 to 10.2% (World Bank, 2017). This was due to the challenging macroeconomic environment, thus has constrained credit creation by banks resulting to the lending on short term despite the demand for long term loans to support capital projects (NECF, 2015). As the long term loans are scarce the capital projects close down thereby resulting in unemployment thereby hindering poverty.

According to the Reserve Bank of Zimbabwe (2017) the various types of institutions include 13 commercial banks, 5 building societies, 1 savings bank, 178 microfinance institutions and 2 development financial institutions. Kanyeze et al (2011) stated that although the post reform era saw a flurry of entrepreneurial activity in the financial sector, the indigenous banks were granted licences, the banking crisis began early 2004 which resulted in a number of banks placed under curatorship or closing, quarantining all accounts that were held with them and the number of banks fell from forty in 2002 to twenty six by 2011.

1.2 Problem statement

Poverty reduction is recognizes as one of the pillars of the sustainable development goals, despite the measures put forward by the country to reduce poverty such as the interim poverty reduction strategy (IPRS), the level of poverty is increasing due to the increase in population and income inequality. The household final consumption per capita proxy for poverty has been increasing as shown by figure 1 as the highest value obtained was $900 in 2015. However, domestic credit to private sector as a percentage of GDP proxy for financial sector development is decreasing as it can be denoted from figure 1 that it declined from 14.5% in 2014 to 10.17% in 2015 with a highest of 103.6% in 2002(World bank, 2017). The financial sector of the country has been unstable over the past years ending up in many financial institutions such as banks shutting down. Stakeholders lost confidence in the banking sector due to the hyperinflation which struck the economy in the year 2008 which saw many individuals and businesses losing as people kept some of their wealth in the form of money. Due to the shutting down of financial institutions there is an effect of retrenchment and reduction to financial access thus leading to the effect of increasing poverty. Thereby leads to the need to carry out the study in order to examine if financial sector development causes poverty or poverty causes financial sector development

1.3 Objectives of study

The general objective is to examine the relationship between Financial Sector Development (FSD) and poverty. Whereas the specific objectives are:

To ascertain the direction of causality between financial sector development and poverty reduction in Zimbabwe.

To find out the effect of financial sector development on poverty reduction in Zimbabwe.

1.4 Research Questions

What direction of causality exists between financial sector development and poverty reduction in Zimbabwe?

What is the effect of financial sector development on poverty in Zimbabwe? In particular does an increase in financial sector development lead to a decrease in poverty?

1.5 Justification of study

This study is very important to policy makers, government and others interested in the financial sector (stakeholders). Once the relationship has been determined, it is very important to understand the effect of financial sector development on poverty reduction. According to Honohan (2004), found out that a 10 percentage point increase in private credit to Gross Domestic Product (GDP) reduces poverty ratio by 2.5 to 3 percentage point. With this, the causality of direction will help to recommend for reforms in financial sector development which would help out to support poverty reduction. The Sustained Development Goals (SDGs) also supports the financial sector development as it act as a lubricant engine that drives away poverty reduction, thus fulfilling the SDG number one of no poverty.

1.6 Hypothesis

To achieve the set objectives, the research tests the hypothesis below

There is a positive relationship between financial sector development and poverty.

There is bi direction causality between financial sector development and poverty.

1.7 Organisation of the rest of the study

The study is divided into different chapters that are: Chapter two reviews both empirical and theoretical literature, Chapter three contains the methodology and estimation technique, whilst chapter four presents the estimation of the model and interpretation of results. Finally, chapter five gives the conclusion of the study and policy recommendations.

CHAPTER TWO

LITERATURE REVIEW

2.0 Introduction

In this chapter, both empirical and theoretical literature relating to the impact of financial sector development on poverty reduction are reviewed, and also to have an overview of what theory says in relation to the relationship between FSD and poverty reduction. The theory is divided into two segments as the impact of FSD is in two channels namely the direct and the indirect channel.

2.1 Theoretical literature review

The theory on financial sector development as emphasized today, in developing countries, goes back to Schumpeter (1934) when he stresses the role of banking sector as a financier of productive investments and in that way as an accelerator of economic growth. Modern growth theory however identifies two specific channels which are direct and indirect.

2.1.1Direct Channel

One of the main reasons that explain persistent poverty is due to lack of access to financial services (Levine, 2008). Most theory proposes that access to finance permits the poor to better investments and education (Jacoby ; Skousfias, 1997 and Beegle et al., 2003).

Jalilian and Kirkpatrick (2007) postulated that an increase in the access to financial services to the poor will increase their income growth, thus resulting in the direct channel. Whilst the availability of credit can strengthen the productive assets of the poor by enabling them to invest in productivity, thus enhancing new technologies for instance new and better tools, equipment and fertilizers or to invest in education and health which could provide for a higher income in future. Addressing situations such as financial market failures such as information asymmetry (moral hazard and adverse selection) and the high fixed cost of lending to small scale borrowers the financial sector development can improve the opportunities for the poor to access formal finance (Stiglitz, 1998; Jalilian and Kirkpatrick, 2001). Thus financial development can directly contribute to poverty reduction by improving the opportunities for the poor to access formal finance and enables them to achieve a sustainable livelihood.

Fields (2001) argued that a lot would be gained by developing credit and finance given that an underdeveloped credit markets contributes to continued poverty, increase in income inequality and stagnant economic growth. He then further emphasized that through the better access to credit, the poor are given the opportunity to take part in more productive activities, which results in increases in their incomes and also financial incomes thereby enabling the poor to respond better to economic or health-related shocks thus reducing the likelihood of falling into poverty when such shocks occur.

Deaton (1991) argued that access to credit and other financial services is likely to decrease the proportion of low risk, low return assets held by poor households for precautionary purposes (such as jewellery), and enable them to invest in potentially higher risk but higher return assets, (such as education, or a rickshaw), with overall long term income enhancing impacts. Eswaran and Kotwal (1990) postulated that just the knowledge that credit will be available to cushion consumption against income shocks if a potentially profitable but risky investment should turn out badly, can make the household more willing to adopt more risky technologies. The behaviour will increase the use of modern technologies with productivity increasing, and hence income enhancing benefits. In the same vein, insurance can offer protection against certain types of shocks. These facilities can reduce the vulnerability of the poor and minimize the negative impacts that shocks can sometimes have on long-run income prospects. Thus the value of financial services in helping the poorest to cope with risks can be as or more important than the expected financial return (DFID, 2004).

Mckinnon (1973) and Shaw (1973) provided the theoretical underpinning for financial liberalization, emphasizing the influence of real interest rates on savings, investment and growth. They argued that financial deepening increases the rate of domestic savings and this lowers the cost of borrowing and thus stimulating investment. The major core of this argument was the claim that developing countries suffer from financial repression, this therefore postulated to that the liberalization of these countries from repressive conditions would induce savings, investment and growth. In this view investment is positively related to real interest rates which are in contrast to the neoclassical theory. This then lead to the “conduit effect” as the rise in interest rates increases the volume of financial savings through the financial intermediaries and as such increases investible funds.

2.1.2 Indirect Channel

This is whereby the financial sector development can also help to reduce poverty indirectly by stimulating economic growth through its impact on capital accumulation and the rate of technological progress (De Gregorio, 1996).

Growth may be denoted as a sufficient condition for sustained poverty reduction. However, there are different views of growth-poverty nexus which include the popular Kuznets inverted U hypothesis (Kuznets, 1955). The theory introduced an increase interest in the contribution that financial deepening can make to income distribution in developing countries. Kuznets suggested that the extent that financial sector development facilitates more migration from low-income but more egalitarian agricultural higher income but more unequal modern (industrial and services) sector it may be expected to increase inequality. The view suggests that economic growth may increase income inequality at the stage of industrialization as the asset-rich classes who have easy access to finance would reap the early harvest of industrialization and thus garner a higher share of economic pie, leaving the poor oppressed. On the other hand, the “trickle down” theory postulated that economic growth would either trickle down to the poor through job creation and other economic opportunities or create the necessary conditions for the wider distribution of the economic and social benefits of growth (Todaro 1997).

Datt ; Ravallion (1992) and Kakwani (2000) attempted to explain changes in poverty in terms of a “growth effect”, stemming from a change in average income, and a “distribution effect”, caused by shifts in the Lorenz curve holding average income constant. They found the growth effect to explain the largest part of observed changes in poverty. Fields (2001) postulated that the extent of the impact of growth on poverty alleviation depends on the growth rate itself and the level of inequality.

2.2 Empirical Literature

On the empirical front, a growing body of literature has examined the interaction between financial development and poverty reduction. The empirical evidence from this literature is however ranked in descending order and controversial across countries, data and methodologies.

Uddin et al. (2014) examined the relationship between financial development and poverty reduction in the case of Bangladesh using quarter frequency data over the period of 1975–2011. Applying autoregressive distributed lag (ARDL) bounds testing approach to cointegration, they found short-run bidirectional causality between the development of the financial sector and poverty reduction. Using the similar estimation technique with structural breaks, Uddin et al. (2012) reached the same finding for Bangladesh over the period of 1975–2010. Also they found out that economic growth is accelerated by financial development and poverty reduction.

Ho and Odhiambo (2011) explored the relationship between financial development and poverty reduction in China for the period 1978 to 2008. The results are sensitive to the finance variable. When the credit ratio is used, a feedback effect exists between financial development and poverty reduction in the short run. However, results for broad money supply ratio show bidirectional causal flow in the short run but poverty reduction causes financial development in the long run.

Honohan (2004a) shows a robust effect of financial depth (measured as the ratio of private credit to GDP) on headcount poverty incidence. The regression results suggest that a 10 percentage point increase in the ratio of private credit to GDP would lead to a 2.5–3.0 percentage point reduction in poverty incidence. Given that per capita GDP is controlled in the analysis, the results suggest that a direct relationship between financial development and poverty reduction exists independent of the indirect effect through growth. Similarly, using data for 58 developing countries over the period 1980 to 2000, Beck, Demirgüç-Kunt, and Levine (2004) discovered that countries with better-developed financial intermediaries (measured as the ratio of private credit to GDP) experience faster declines in both poverty and income inequality by disproportionately boosting the incomes of the poor. Their results are robust to controlling for potential reverse causality. They also hold even when controlling for the average rate of economic growth, which suggests that financial development alleviates poverty beyond its effect on aggregate growth.

Regarding Sub-Saharan African countries:

Zahonogo (2016) investigated how financial development affects poverty indicators in developing countries. Using data from 42 Sub-Saharan countries African countries and covering the period 1980-2012 Zahonogo applied generalized method of moment (GMM) at is appropriate to control country specific effects and the possible endogeneity. The evidence then points an inverted U curve type response and the findings are robust to changes in poverty measures and to alternative model specifications, suggesting thus the non-fragility of the linkage between financial development and poverty for sub-Saharan African countries.

Dauda and Makinde (2014) used vector autoregressive model (VAR) to examine the nexus between financial sector development and poverty reduction in Nigeria over the period 1980-2010. The results showed that economic growth exerts the strongest influence on poverty reduction in the short run but could be detrimental to the poor in the long run due to the adverse effect of income inequality. Furthermore, financial deepening proxied by broad money supply (M2) is negatively related to poverty. Hence, the McKinnon Hence, the McKinnon conduit effect is the likely main transmission channel through which the poor benefit from the financial sector development in the long run. However, contrary to the general belief, credits to private sector do not cause a reduction in the incidence of poverty. The authors attributed this result to the wrong attitude of financial intermediaries that have not adequately channelled savings to the pro-poor sectors of the economy.

Aye (2013) used the Johansen cointegration to examine the dynamic causal relationship between financial deepening, economic growth and poverty in Nigeria over the period 1960-2011.The short and long run causality between these variables is tested using a modified Hsaio-Granger causality within a Vector Autoregressive (VAR) and Vector Error Correction Model (VECM) framework. The results indicate no evidence of long run equilibrium relationship between finance, economic growth and poverty. Therefore, we focus on short-run causality. Our results show a short-run unidirectional causality from growth to poverty conditional on finance. This supports the indirect channel through which finance affects poverty via growth. Thus supporting the indirect channel through which finance affects poverty via growth.

Using annual data over the period 1969–2006, Odhiambo (2010b) investigated causality between financial development and poverty in the case of Zambia. The causality analysis reported that when the ratio of broad money (M2) to nominal GDP is used as an indicator of financial development, poverty reduction proxied by private per capita consumption causes the development of the financial sector. However, when domestic credit to private sector as share of GDP is used, financial development Granger causes poverty reduction. Odhiambo (2009) examined the relationship between finance, growth and poverty reduction in South Africa over the period from 1960 to 2006. Using the Johansen cointegration tests, he found that an increase in economic growth leads to an increase in financial development as measured by broad money ratio to GDP. He also reported that both financial development and economic growth Granger cause poverty reduction in South Africa in the short and long run.

Working with the annual data for Pakistan, Shahbaz (2009) investigated the impact of financial development and financial instability on poverty reduction using the autoregressive distributed lag model (ARDL) for long run relationship between the variables by controlling for economic growth, inflation, agricultural growth, manufacturing and trade openness. The results indicated that all the variables are co-integrated for long run relationship and also found that financial development is negatively related with poverty while financial instability increases poverty. In addition, Agriculture growth, manufacturing and trade openness seem to reduce poverty reduction in Pakistan.

Quartey (2005) investigated the interrelationship between financial sector development and poverty in Ghana. This was done using World Development Indicators over the period 1970 to 2001. Quartey tested causality using Granger causality and found out that financial sector development does not granger cause savings mobilization, it induces poverty and secondly that savings do granger cause poverty. He also obtained the effect of financial sector development on poverty to be positive but insignificant. This was due to the fact that financial intermediaries in Ghana had not adequately channelled savings to the pro-poor sectors of the economy because of government deficit financing, high default rate, lack of collateral and lack of proper business proposals. Another interesting finding is that there is a long-run cointegration relationship between financial sector development and poverty reduction.

2.3 Conclusion

The chapter has provided vital information about the determinants of poverty. The major determinants are economic growth, inflation, agricultural growth, manufacturing and trade openness. The study proxied financial sector development with domestic credit as a percentage of GDP.

CHAPTER THREE

METHODOLOGY

3.0 Introduction

This chapter presents the methodology used in achieving the objectives of the study. This encompasses model specification (theoretical and empirical framework), definition and justification of variables, data sources and finally the model diagnostic test. Methodology is guided from literature in the previous chapter. The study makes use of the model used by Shabhaz (2009) and some necessary alterations are made to the model to suit Zimbabwe’s situation.

3.1 Model specification

3.1.1 Theoretical framework

Most theory from the indirect channel postulates growth to have a positive impact on reducing poverty prominently from the “growth effect” and “trickledown” (Field, 2001; Kakwani, 2000; Datt & Ravallion, 1992 and Kuznets, 1955). These theories lead to economic growth being one of the factors that affect poverty indirectly.

Povt=fy…………. (1)

Where yis economic growth

3.1.2 Empirical Model

From literature review Dauda & Makinde (2014), Shahbaz (2009) and Odhiambo (2009) carried out related studies in different countries and successfully analyzed variables of the model with an error term to show a stochastic relationship as clearly defined by Gujarati (2004).

Shabhaz (2009) used the following model in their study:

LGR=?1+?2LFD+?3FNS+?iCV+vt ………….. (2)

In order to make the model suitable for the study the measure Income growth of bottom 20 percent population (GR) for poverty is changed to household final consumption per capita as used by Odhiambo (2009) and financial instability (FNS) is removed due to that it is not of relevance to this study. Control variables (CV) include economic growth (GDPPC), agriculture as a share of GDP, manufacturing as a share of GDP, trade openness, investment as a share of GDP and inflation proxy for monetary instability.

Therefore the model becomes

LPOVt=?1+?2LFDt+?iCVt+?t

Where:

POVt is poverty in year t, FDt is financial sector development and CVt are control variables which are economic growth measured by GDPPC is gross domestic product per capita, agriculture as a share of GDP, manufacturing as a share of GDP, trade openness, investment as a share of GDP and inflation proxy for monetary instability and ?t is the white noise error term.

Expected signs all explanatory variables are expected to have positive influence on poverty except inflation which as a negative expected sign.

3.2 Definition and Justification of variable

Poverty

This is the dependent variable of this study measured as household final consumption per capita as used in the studies of Odhiambo (2009), Aye (2013) and Uddin (2014). It is found that consumption expenditures reveals not only what a household is able to command based on its current income, but also whether that household can access credit markets or household savings, Hentschel and Lanjouw (1996). It was previously known as Private per capita consumption expenditure now it is known as household final consumption per capita.

Where household final consumption per capita:

householdfinalconsumptionpercapita=householdexpendituretpopulationtGross Domestic Product Per Capita

This captures the indirect effect as the financial sector development can also help to reduce poverty indirectly by stimulating economic growth which is measured by GDP.Inflation

Mankiw (2004) defined inflation as the increase in the overall levels of prices in the economy. Inflation is measured by GDP deflator. This captures the macroeconomic instability were high and unpredictable inflation is thought to have disproportionally negative impact on poverty because the poor have relatively limited access to financial instruments that hedge against inflation and are more likely to have larger share of cash in their small portfolios (Holden & Prokopenko, 2001).

Financial sector development

Financial development refers to improvement in the quality, quantity or efficiency of the financial systems that are comprised of financial markets, banks and other financial intermediaries (Maskay, 2012). Domestic credit to the private sector (DCPS) is the proxy for FD and it refers to the financial resources that are provided to the private sector by financial intermediaries such as banks.

Trade openness

This is an independent variable measured by the sum of exports and imports as a share of GDP. Since international trade brings in better methods and new ideas, moreover trade openness exposes developing countries to tough competition from developed financial sector of other countries.

Manufacturing as percentage of GDP

Manufacturing value-added increases income for the poor people as it creates employment, generating activities which in turn increase income distribution along with rise in income. Employment opportunities for both skilled and unskilled labor are generated through investment activities. This situation raises the aggregate income and hence improves the economic position of poor segment of population.

Agriculture as a percentage of GDP

Agriculture sector is providing more employment which will increase income of lower segments of population and it will also enhance its share to GDP. This variable was used in a related study by Shabhaz (2009

Investment as a percentage of GDP

This indicator refers to the total share of investment in total production as a share of GDP. It is measured by gross capital formation as a percentage of GDP (GCF_GDP).

3.3 Estimation technique

The study adopts (OLS) Ordinary Least Squares (OLS) methodology because of its effectiveness in the estimation procedure of the Classical Linear Regression Model (CLRM) and also its ability to produce Best Linear Unbiased Estimators (BLUE) (Gujarati, 2004).

3.3.1Unit root test

The unit root test was developed by Dickey- Fuller in 1979, whereas a number of studies postulated that the unit root test has shown that using classical estimation methods, such as the OLS, to estimate relationships with unit root variables gives misleading inferences (Gujarat, 2008; New Bold & Granger, 1974). The presence of non-stationarity might lead to what is known as spurious regression. A spurious regression usually has a high R-squared, and the t-statistics appears to be significant whereas the results have no economic meaning. This test is usually done to detect the order of integration of the variables before estimation. Illustrating the econometric model yt=?+?yt?1+et the ADF test the hypothesis that;

H0: ?=1; H1:?>1 If the null hypothesis is rejected then the series will be stationary at I(0). If not then the differencing method for making them stationary will be applied Gujarati (2008).

3.3.2 Granger-causality

The Granger causality test is used to examine the causality between financial sector development and poverty. The test is chosen for this study because it is suitable technique since it is favourable for both large and small samples (Odhiambo, 2008). This test involves trying to detect if FSD caused poverty reduction or poverty reduction causes poverty. It states tat if poverty causes FSD and FSD does not cause poverty it is uni-directional causality however if poverty causes FSD and FSD causes poverty this implies bi-directional causality.

The hypotheses that are going to be tested are as follows:

H01: FSD does not Granger cause poverty

H02: poverty does not Granger cause FSD

Alternative hypotheses:

H11: FSD Granger causes poverty

H12: poverty Granger causes FSD

Granger causality is sensitive to number of lags included, for lag selection Akaike Information Criterion (AIC) is used to determine the lag length.

3.4 Diagnostic test

3.4.1Multicollinearity

Various diagnostic tests were carried out starting with multicollinearity. To detect the presence of multicollinearity the study used the pairwise correlation test where a partial correlation coefficient exceeding absolute 0.8 indicates the presence of high multicollinearity between variables.

3.4.2 Normality test

Testing for normality is also important in regression analysis. Non normality of errors causes bias in the construction of confidence intervals and significance test (Greene, 2003). The JacqueBera formal test was used to test for normality.

3.4.3 Heteroskedasticity

Heteroskedasticity causes estimators to no longer have the minimum variance. To test the presence of heteroskedasticity the Breusch Pagan Godfrey test was used.

3.4.4 Autocorrelation

In the presence of autocorrelation the OLS estimators remain unbiased, consistent and asymptotically normally distributed, but they are no longer efficient (Ibid, 2004). The study used the Breusch Godfrey Serial Correlation LM test in detecting autocorrelation.

3.4.5 Ramsey RESET test

Ramsey’s Regression Specification Error Test (RESET) was used in order to test for the model specification error. Gujarati (2008) stated that if the model was incorrectly specified the researcher would have encounter model misspecification error.

3.5 Data sources and collection

The study used time series data for Zimbabwe to investigate the impact of financial sector development on poverty for the period 1980 to 2016. Data on financial sector development is obtained from the World Bank. Whilst data for inflation, poverty, trade openness, agriculture as a share of GDP, manufacturing as a share of GDP, trade openness, investment as a share of GDP and gross domestic product per capita is obtained from the World Bank.

3.6 Conclusion

This chapter presented the methodology adopted in this study and various tests carried out to purify the data. Data sources were also provided.

CHAPTER FOUR

PRESENTATION AND INTERPRETATION OF RESULTS

4.0 Introduction

This chapter presents application of the OLS method and diagnostic procedures. Descriptive statistics are presented first while stationarity, model diagnostic and regression results follow respectively. This chapter also gives an interpretation of the results obtained from the regressions carried out. And the last section provides an overview of whether the hypothesis stated in Chapter One are rejected or accepted.

4.1 Descriptive statistics

Table 1: Descriptive Statistics results

HFPC AV_GDP GCF_GDP GDP_DEFLATOR GDPPC PRIVCRE TOT MVA_GDP

Mean 524.6010 16.13843 14.67408 2.153850 1083.545 26.19803 66.71309 18.22576

Median 428.3380 16.25094 15.98748 1.956202 1169.694 24.74932 69.37150 17.44822

Maximum 900.7536 22.67357 23.72906 74.29818 1347.972 103.6323 109.5216 29.53704

Minimum 327.9399 7.413793 1.525177 -27.04865 593.1272 7.476843 35.91686 9.831452

Std. Dev. 174.6733 3.648627 6.040022 14.94728 219.8828 17.61774 19.08893 4.943309

Skewness 0.894086 -0.234857 -0.679745 2.880900 -0.726938 2.492208 0.2372904 0.161845

Kurtosis 2.466571 2.522200 2.527560 16.60458 2.199465 11.54024 2.372904 2.304485

Jarque-Bera 5.223159 0.673386 3.107120 327.4245 4.131917 146.6702 0.927631 0.882767

Probability 0.073418 0.714128 0.211494 0.000000 0.126697 0.000000 0.628880 0.643146

Sum 18885.64 580.9835 528.2668 77.53859 39007.62 943.1289 2401.671 656.1273

Sum Sq. Dev. 1067876. 465.9368 1276.865 7819.737 1692196. 10863.46 12753.55 855.2705

Observations 36 36 36 36 36 36 36 36

Table 1 provides summary of the variables used in the study. The variations in GDPPC are relatively higher than that of the other variables with 219.8828, followed by HFCP with 174.6733, then TOT with 19.08893, PRIVCRE with 17.61774, GDP deflator with 14.94728, GCF_GDP with 6.040022, MVA_GDP with 4.943309 and AV_GDP with 3.648627 as shown by the standard deviations. Measures of skewness show that stock GDP deflator, HFCP, TOT, PRIVCRE and MVA_GDP are positively skewed whereas GDPPC, GCF_GDP and AV_GDP are negatively skewed. None of the variables has a kurtosis closer to three. The Jarque-Bera null hypotheses are not rejected and hence there is normality in all the variables.

4.2 Stationarity tests

Table 2: Stationarity test results

Variable ADF Prob level ADF Prob First Difference ADF Prob Second Difference Order of Integration Level of Stationarity

Log(HFCP) 0.8509 0.0000 – I(1) ***

Log(PRIVCRE) 0.1947 0.0000 – I(1) ***

AV_GDP 0.1086 0.0000 – I(1) ***

MVA_GDP 0.356 0.0000 – I(1) ***

TOT 0.1225 0.0000 – I(1) ***

GDP deflator 0.0005 – – I(0) ***

GDPPC 0.6704 0.0001 – I(1) ***

GCF_GDP 0.4902 0.0000 – I(1) ***

Where *** implies stationary at 1 %, ** stationary at 5% and * stationary at 10

The Augmented Dicky-Fuller (ADF) unit root test has been employed to test for stationarity. It was discovered that GDP deflator is stationary at its level meaning integrated of order zero. Variables HFCP, GDPPC, GCF_GDP, AV_GDP, MVA_GDP, PRIVCRE and TOT are not stationary and they have a unit root, after being differenced once they became stationary thus integrated of order one.

4.3 Diagnostic test results

4.3.1 Multicollinearity

The pairwise correlation test was carried out to check the multicollinearity between independent variable. The results show that the explanatory partial correlation coefficient are less than the absolute 0.8 implying that there is no serious multicollinearity, thus all variables are linearly independent.

4.3.2 Normality test results

The Jacque-Bera was found to 1.751901 with a probability value of 0.416466, the probability is greater than 0.05 therefore we fail to reject null hypothesis that the residuals are normally distributed at 5% level of significance.

4.3.3 Heteroscedasticity test results

Breusch Pagan Godfrey show that the probability value 0.7496 is less than 0.8008 the model’s probability value, thus we fail to reject the null hypothesis that the error variance is homoskedastic and conclude that at 5% level of significance the variances are equal.

4.3.4 Autocorrelation test results

Applying the Breusch Godfrey Serial correlation LM results show that the probability value of 0.9655 is less than 0.9752 the model’s probability value. This implies that we may fail to reject the null hypothesis of no autocorrelation since residuals in one period are not correlated to residuals in periods before.

4.3.5 Model Specification test results

Results from the RAMSEY RESET test show that the probability value of the F-statistic is 0.0805 which is greater than 0.05, thus supporting the null hypothesis that the model is correctly specified. Thereby the results are considered reliable for reporting and interpretation as they have passed most of the diagnostic tests.

4.4 Regression results

Table 3: Regression results

-71755635Dependent Variable: DLHFPC

Variable Coefficient Std. Error t-Statistic Prob.

C -0.006052 0.014351 -0.421695 0.6766

DGDPPC 0.000974 0.000208 4.688549 0.0001

DGCF_GDP -0.021428 0.004268 -5.021022 0.0000

GDP_DEFLATOR 0.011473 0.001414 8.112985 0.0000

DLPRIVCRE 0.031427 0.029801 1.054569 0.3010

DMVA_GDP 0.010804 0.008756 1.233886 0.2279

DAV_GDP 0.005006 0.005404 0.926398 0.3624

DTOT 0.003967 0.001683 2.356898 0.0259

Dependent Variable: DLHFPC

Variable Coefficient Std. Error t-Statistic Prob.

C -0.006052 0.014351 -0.421695 0.6766

DGDPPC 0.000974 0.000208 4.688549 0.0001

DGCF_GDP -0.021428 0.004268 -5.021022 0.0000

GDP_DEFLATOR 0.011473 0.001414 8.112985 0.0000

DLPRIVCRE 0.031427 0.029801 1.054569 0.3010

DMVA_GDP 0.010804 0.008756 1.233886 0.2279

DAV_GDP 0.005006 0.005404 0.926398 0.3624

DTOT 0.003967 0.001683 2.356898 0.0259

R squared=0.832407 F-statisic=19.15783

Adjusted R squared=0.788957 Prob( F-statistic)=0.00000

Durbin Watson stat=1.913509

All the significant variables have the expected signs except for gross capital formation as a percentage of GDP which has a negative sign and GDP deflator which exhibits positive value. DGDPPC, DGCF_GDP and GDP_DEFLATOR are significant at 1% level while DTOT is significant at 5% level and DLPRIVCRE, DMVA_GDP and DAV_GDP are not *statistically significant at all levels. Both the R2 and adjusted R2 are greater than 0.5 with values 0.832407 and 0.788957 respectively. This implies that R2 is a reliable measure of goodness fit as about 83% variations in poverty are explained by combination of explanatory variables. Also the F-statistic probability is 0.0000000006 (614591e-09) which is less than 0.01 implying that the model is significant at 1%.

4.5 Granger Causality test results

Table 3: Granger causality results

Pairwise Granger Causality Tests

Date: 04/06/18 Time: 12:24

Sample: 1980 2015

Lags: 2

Null Hypothesis: Obs F-Statistic Prob.

DLPRIVCRE does not Granger Cause DLHFPC 33 2.69602 0.0850

DLHFPC does not Granger Cause DLPRIVCRE 0.29977 0.7433

There is no variable that granger causes other variable since the first null hypothesis probability is 0.0850 is greater than 0.05 and the second null hypothesis probability of 0.7433 is greater than 0.05. Thus leading to the rejection of first and second hypotheses and concluding that there is no bi directional causality between FSD and poverty.

4.6 Interpretation of results

4.6.1 Financial Sector Development

The coefficient of financial development was found to be positive with value 0.031427 which is statistically insignificant at all conventional levels. Moreover it leads to the rejections of the hypothesis that there is a positive relationship between FSD and poverty since the variable is statistically insignificant.

4.6.2 Economic growth

The coefficient of GDPPC was found to be to be positive with value 0.000974 which is statistically significant at 1% level of significance. This means a unit increase in GDPPC is approximately 0.0974% increase in poverty reduction. This is in line with the sign expectations.

4.6.3 Investment as a percentage of GDP

In this case, the coefficient of GCF_GDP has been observed to have a negative value which is statistically significant 5% of and the coefficient is -0.021428. This means a unit increase in GCF_GDP leads to a 2.1428% increase in poverty reduction meaning that the level of poverty would have increased as there is a negative constant value.

4.6.4 Manufacturing as a share of GDP

The coefficient of MVA_GDP was found to be positive with a value of 0.010804. However, this value was statistically insignificant.

4.6.6 Agriculture as a percentage of GDP

The coefficient AV_GDP was found to have a positive value 0.000506 and the coefficient was statistically insignificant. Meaning that the variable as no economic impact on poverty.

4.6.7 Inflation

The coefficient GDP deflator was found to have a positive value of 0.011473, whereas it contradicts with the theory expectations as it is expected to have a negative value. This means that a unit increase in GDP deflator results in approximately 1.15 percent increase in Poverty reduction.

4.7 Conclusion

The present chapter has presented the regression results after the model had passed most of the diagnostic tests. In addition an interpretation of the results was given. Guided by these results, the next chapter presents policy conclusions and recommendations.

CHAPTER FIVE

CONCLUSIONS AND POLICY RECOMMENDATIONS

5.0 Introduction

This chapter provides a summary of the key findings the study. The chapter is outlined as follows policy recommendations and areas of further study.

5.1 Summary

The main objective of this study is to examine the relationship between financial sector development and poverty using time series data for the period 1980 to 2015. Domestic credit as a percentage of GDP is used as the indicator for financial sector development, the coefficient was found to be statistically insignificant which lead to rejection of the second objective of the study implying that financial sector development does not have an effect to poverty. Most of the independent variable where statistically significant except for manufacturing value added as a percentage of GDP and agriculture value added as a percentage of GDP which are statistically insignificant at all conventional levels. As the study passed all of the all diagnostic tests the model produced parsimonious results. Additionally the coefficient of determination R2 and the adjusted R2 are greater than 0.5 indicating that the estimated model is of good fit. Applying the granger causality none of the variables influenced another variable since both probabilities where less than 0.05 thus resulting in rejection of the null hypothesis of a bi directional causality.

5.2 Policy Recommendations

Financial sector development coefficient is found to be statistically insignificant at all conventional levels one of the problems faced in this sector is that it faces lack of confidence in the economy due to issues such as the hyperinflation faced in the country in 2008 and the current liquidity crisis in the country. In order for the country to overcome this government should restore confidence of the financial sector by implementing the move to resuscitate RBZ’s role as the lender of last resort. With this function banks can definitely benefit and probably the risk of bank closures would reduce. Additionally the lender of last resort function would lead to the expectation of easing liquidity challenges through improved interbank market which leads to the prospects of increasing lending to the key productive sectors and export sectors of te economy and thus result in poverty reduction.

As a result from the indirect channel economic growth (GDPPC) is statistically significant at 1% level of significance. Since economic growth is sustained through measures and policies that may develop the financial system. Such policies may influence innovation in the sector. Through government intervention it is advocated to encourage financial partnerships with Sub-Suharan countries or other international countries in order to benefit the sector through increase in the information regarding ways/methods to increase financial services and innovation so as to increase access of financial systems to the poor. Also it is imperative for government to encourage for opening of new financial markets as it is likely to go a long way in enhancing competition in the financial sector which should in turn lead to increased investment levels.

Furthermore Zimbabwe’s majority population of the poor people reside in rural areas and mostly they are unskilled labour force with agriculture being their main occupation. The formulation of government policies at both macro and micro level make agricultural reforms favouring the local people and the poor should not be excluded from various development programs rather undertaken on priority basis. Thus both the agriculture and agri-based manufacturing sectors should continue to be labour intensive as they have the potential for employment as well as income generating of the unskilled labour.

5.3 Areas of Further Study

Future studies can use another variable to measure financial sector development such as Broad money as a percentage of GDP in order to determine the relationship between FSD and poverty. Also the study can use different methodologies especially to determine short and long run relationships thus use of ARDL and VECM. Also since financial sector development does not poverty there is need to know what determines financial development so that we may target those variables in order to enhance the prospects of poverty in Zimbabwe.

Reference

Schumpeter, J. 1911. The Theory of Economic Development: An Inquiry into Profits, Capital, Credit, Interest and the Business Cycle. Cambridge: Harvard University Press.

Zhuang, J. et al. (2009).Financial Sector Development, Economic Growth, and Poverty Reduction: A Literature Review.ADB Economics Working Paper Series No. 173,Economics and Research Department, Asian Development Bank, Manila.

McKinnon, R. I. (1973). Money and Capital in Economic Development. Brookings Institution, Washington, DC.

Goldsmith, R. W. (1969).Financial Structure and Development. New Haven, CT: Yale University Press.

Brownbridge, M, and Harvey, C. (1998). Banking in Africa: The Impact of Financial Sector Reform Since Independence. Africa World Press, USA.

Kanyenze, G., Kondo, T., Chitambira, P., and Martens, J. (2011). Beyond the Enclave,

Ministry of Finance, (2011). 2012 Budget Strategy Paper: Building on Our Priorities. Harare, Zimbabwe.

Ministry of Finance, (2015), 2016 Budget Strategy Paper: Building on Our Priorities. Harare, Zimbabwe.

Honohan, Patrick.. 2004. Financial Sector Policy and the Poor: Selected Findings and Issues.World Bank Working Paper No. 43. Washington, D.C.: The World Bank.

Chigumira, G and Makochekanwa, A. (2014). Financial Liberalization and Crisis: Experience and Lessons for Zimbabwe, ZEPARU, Zimbabwe.

Uddin, G. S, Kyophilavong, P., ; Sydee, N. (2012). The causal nexus between financial sector development and poverty reduction in Bangladesh. International Journal of Economics and Financial Issues, 2(3), 304-311.

Uddin, G. S., Shahbaz, M., Arouri, M., ; Teulon, F. (2014). Financial development and poverty reduction nexus: a cointegration and causality analysis in Bangladesh. Economic Modelling, 36, 405-412.

Shahbaz, M. (2009). Financial performance and earnings of poor people: a case study of Pakistan. Journal of Yasar University, 4, 2557-2572.

Odhiambo, N. M. (2009). Finance–growth–poverty nexus in South Africa: a dynamic causality linkages. Journal of Socio-Economics, 38, 320-325.

Odhiambo, N. M. (2010a). Is financial development a spur to poverty reduction? Kenya;s experience. Journal of Economic Studies, 37(3), 343-353.

Odhiambo, N. M. (2010b). Financial deepening and poverty reduction in Zambia: an empirical investigation. International Journal of Social Economics, 37(1), 41-53.

APPENDIX LIST

Appendix 1: Descriptive statistics

HFPC AV_GDP GCF_GDP GDP_DEFLATOR GDPPC PRIVCRE TOT

Mean 524.6010 16.13843 14.67408 2.153850 1083.545 26.19803 66.71309

Median 428.3380 16.25094 15.98748 1.956202 1169.694 24.74932 69.37150

Maximum 900.7536 22.67357 23.72906 74.29818 1347.972 103.6323 109.5216

Minimum 327.9399 7.413793 1.525177 -27.04865 593.1272 7.476843 35.91686

Std. Dev. 174.6733 3.648627 6.040022 14.94728 219.8828 17.61774 19.08893

Skewness 0.894086 -0.234857 -0.679745 2.880900 -0.726938 2.492208 0.2372904

Kurtosis 2.466571 2.522200 2.527560 16.60458 2.199465 11.54024 2.372904

Jarque-Bera 5.223159 0.673386 3.107120 327.4245 4.131917 146.6702 0.927631

Probability 0.073418 0.714128 0.211494 0.000000 0.126697 0.000000 0.628880

Sum 18885.64 580.9835 528.2668 77.53859 39007.62 943.1289 2401.671

Sum Sq. Dev. 1067876. 465.9368 1276.865 7819.737 1692196. 10863.46 12753.55

Observations 36 36 36 36 36 36 36

Appendix B: Correlation matrix

AV_GDP GCF_GDP GDP_DEFLATOR GDPPC PRIVCRE TOT MVA_GDP

AV_GDP 1.000000 -0.149487 -0.100146 0.148241 0.128590 0.090144 0.036710

GCF_GDP -0.149487 1.000000 -0.147222 0.549109 -0.255357 -0.271429 0.463574

GDP_DEFLATOR -0.100146 -0.147222 1.000000 -0.405403 -0.101286 0.053018 -0.260668

GDPPC 0.148241 0.549109 -0.405403 1.000000 0.185244 -0.585773 0.556157

PRIVCRE 0.128590 -0.255357 -0.101286 0.185244 1.000000 0.028521 -0.003929

TOT 0.090144 -0.271429 0.053018 -0.585773 0.028521 1.000000 -0.610347

MVA_GDP 0.036710 0.463574 -0.260668 0.556157 -0.003929 -0.610347 1.000000

Apendix C: Stationarity tests

Variable HFPC

Null Hypothesis: LHFPC has a unit root

Exogenous: Constant, Linear Trend

Lag Length: 0 (Automatic – based on SIC, maxlag=9)

t-Statistic Prob.*

Augmented Dickey-Fuller test statistic -1.374741 0.8509

Test critical values: 1% level -4.243644

5% level -3.544284

10% level -3.204699

*MacKinnon (1996) one-sided p-values.

Augmented Dickey-Fuller Test Equation

Dependent Variable: D(LHFPC)

Method: Least Squares

Date: 04/03/18 Time: 10:00

Sample (adjusted): 1981 2015

Included observations: 35 after adjustments

Variable Coefficient Std. Error t-Statistic Prob.

LHFPC(-1) -0.135021 0.098215 -1.374741 0.1788

C 0.765068 0.608141 1.258045 0.2175

@TREND(;1980;) 0.004512 0.002831 1.593713 0.1208

R-squared 0.115084 Mean dependent var 0.009536

Adjusted R-squared 0.059777 S.D. dependent var 0.174095

S.E. of regression 0.168811 Akaike info criterion -0.638258

Sum squared resid 0.911909 Schwarz criterion -0.504942

Log likelihood 14.16951 Hannan-Quinn criter. -0.592237

F-statistic 2.080815 Durbin-Watson stat 1.849327

Prob(F-statistic) 0.141394

Variable D(LHFPC)

0635Null Hypothesis: D(LHFPC) has a unit root

Exogenous: None

Lag Length: 0 (Automatic – based on SIC, maxlag=9)

t-Statistic Prob.*

Augmented Dickey-Fuller test statistic -5.553387 0.0000

Test critical values: 1% level -2.634731

5% level -1.951000

10% level -1.610907

*MacKinnon (1996) one-sided p-values.

Augmented Dickey-Fuller Test Equation

Dependent Variable: D(LHFPC,2)

Method: Least Squares

Date: 04/03/18 Time: 10:15

Sample (adjusted): 1982 2015

Included observations: 34 after adjustments

Variable Coefficient Std. Error t-Statistic Prob.

D(LHFPC(-1)) -0.953552 0.171706 -5.553387 0.0000

R-squared 0.483014 Mean dependent var -0.002758

Adjusted R-squared 0.483014 S.D. dependent var 0.241858

S.E. of regression 0.173900 Akaike info criterion -0.631705

Sum squared resid 0.997957 Schwarz criterion -0.586812

Log likelihood 11.73898 Hannan-Quinn criter. -0.616395

Durbin-Watson stat 1.980307

Null Hypothesis: D(LHFPC) has a unit root

Exogenous: None

Lag Length: 0 (Automatic – based on SIC, maxlag=9)

t-Statistic Prob.*

Augmented Dickey-Fuller test statistic -5.553387 0.0000

Test critical values: 1% level -2.634731

5% level -1.951000

10% level -1.610907

*MacKinnon (1996) one-sided p-values.

Augmented Dickey-Fuller Test Equation

Dependent Variable: D(LHFPC,2)

Method: Least Squares

Date: 04/03/18 Time: 10:15

Sample (adjusted): 1982 2015

Included observations: 34 after adjustments

Variable Coefficient Std. Error t-Statistic Prob.

D(LHFPC(-1)) -0.953552 0.171706 -5.553387 0.0000

R-squared 0.483014 Mean dependent var -0.002758

Adjusted R-squared 0.483014 S.D. dependent var 0.241858

S.E. of regression 0.173900 Akaike info criterion -0.631705

Sum squared resid 0.997957 Schwarz criterion -0.586812

Log likelihood 11.73898 Hannan-Quinn criter. -0.616395

Durbin-Watson stat 1.980307

Variable PRIVCRE

Null Hypothesis: LPRIVCRE has a unit root

Exogenous: Constant, Linear Trend

Lag Length: 0 (Automatic – based on SIC, maxlag=9)

t-Statistic Prob.*

Augmented Dickey-Fuller test statistic -2.836192 0.1947

Test critical values: 1% level -4.243644

5% level -3.544284

10% level -3.204699

*MacKinnon (1996) one-sided p-values.

Augmented Dickey-Fuller Test Equation

Dependent Variable: D(LPRIVCRE)

Method: Least Squares

Date: 04/03/18 Time: 10:01

Sample (adjusted): 1981 2015

Included observations: 35 after adjustments

Variable Coefficient Std. Error t-Statistic Prob.

LPRIVCRE(-1) -0.379742 0.133892 -2.836192 0.0079

C 1.324908 0.435822 3.040017 0.0047

@TREND(;1980;) -0.007442 0.007542 -0.986702 0.3312

R-squared 0.228775 Mean dependent var 0.006966

Adjusted R-squared 0.180574 S.D. dependent var 0.496397

S.E. of regression 0.449350 Akaike info criterion 1.319785

Sum squared resid 6.461281 Schwarz criterion 1.453101

Log likelihood -20.09624 Hannan-Quinn criter. 1.365806

F-statistic 4.746228 Durbin-Watson stat 2.059081

Prob(F-statistic) 0.015664

Variable D(PRIVCRE)

Null Hypothesis: D(LPRIVCRE) has a unit root

Exogenous: None

Lag Length: 1 (Automatic – based on SIC, maxlag=9)

t-Statistic Prob.*

Augmented Dickey-Fuller test statistic -6.148211 0.0000

Test critical values: 1% level -2.636901

5% level -1.951332

10% level -1.610747

*MacKinnon (1996) one-sided p-values.

Augmented Dickey-Fuller Test Equation

Dependent Variable: D(LPRIVCRE,2)

Method: Least Squares

Date: 04/03/18 Time: 10:02

Sample (adjusted): 1983 2015

Included observations: 33 after adjustments

Variable Coefficient Std. Error t-Statistic Prob.

D(LPRIVCRE(-1)) -1.586563 0.258053 -6.148211 0.0000

D(LPRIVCRE(-1),2) 0.356738 0.168960 2.111379 0.0429

R-squared 0.634632 Mean dependent var -0.005856

Adjusted R-squared 0.622846 S.D. dependent var 0.778400

S.E. of regression 0.478038 Akaike info criterion 1.420438

Sum squared resid 7.084123 Schwarz criterion 1.511135

Log likelihood -21.43722 Hannan-Quinn criter. 1.450955

Durbin-Watson stat 2.075656

Variable GDPPC

Null Hypothesis: GDPPC has a unit root

Exogenous: Constant, Linear Trend

Lag Length: 1 (Automatic – based on SIC, maxlag=9)

t-Statistic Prob.*

Augmented Dickey-Fuller test statistic -1.812753 0.6764

Test critical values: 1% level -4.252879

5% level -3.548490

10% level -3.207094

*MacKinnon (1996) one-sided p-values.

Augmented Dickey-Fuller Test Equation

Dependent Variable: D(GDPPC)

Method: Least Squares

Date: 04/03/18 Time: 10:06

Sample (adjusted): 1982 2015

Included observations: 34 after adjustments

Variable Coefficient Std. Error t-Statistic Prob.

GDPPC(-1) -0.139887 0.077168 -1.812753 0.0799

D(GDPPC(-1)) 0.387895 0.161664 2.399395 0.0228

C 179.6459 109.7959 1.636180 0.1123

@TREND(;1980;) -1.901686 1.712281 -1.110615 0.2756

R-squared 0.201632 Mean dependent var -10.03471

Adjusted R-squared 0.121796 S.D. dependent var 71.75142

S.E. of regression 67.24009 Akaike info criterion 11.36455

Sum squared resid 135636.9 Schwarz criterion 11.54412

Log likelihood -189.1973 Hannan-Quinn criter. 11.42579

F-statistic 2.525558 Durbin-Watson stat 2.043098

Prob(F-statistic) 0.076343

Variable D(GDPPC)

Null Hypothesis: D(GDPPC) has a unit root

Exogenous: None

Lag Length: 0 (Automatic – based on SIC, maxlag=9)

t-Statistic Prob.*

Augmented Dickey-Fuller test statistic -4.190570 0.0001

Test critical values: 1% level -2.634731

5% level -1.951000

10% level -1.610907

*MacKinnon (1996) one-sided p-values.

Augmented Dickey-Fuller Test Equation

Dependent Variable: D(GDPPC,2)

Method: Least Squares

Date: 04/03/18 Time: 10:07

Sample (adjusted): 1982 2015

Included observations: 34 after adjustments

Variable Coefficient Std. Error t-Statistic Prob.

D(GDPPC(-1)) -0.666567 0.159064 -4.190570 0.0002

R-squared 0.346403 Mean dependent var -3.112504

Adjusted R-squared 0.346403 S.D. dependent var 84.20985

S.E. of regression 68.07976 Akaike info criterion 11.30821

Sum squared resid 152950.2 Schwarz criterion 11.35310

Log likelihood -191.2395 Hannan-Quinn criter. 11.32352

Durbin-Watson stat 1.955252

Variable AV_GDP

Null Hypothesis: AV_GDP has a unit root

Exogenous: Constant, Linear Trend

Lag Length: 0 (Automatic – based on SIC, maxlag=9)

t-Statistic Prob.*

Augmented Dickey-Fuller test statistic -3.161611 0.1086

Test critical values: 1% level -4.243644

5% level -3.544284

10% level -3.204699

*MacKinnon (1996) one-sided p-values.

Augmented Dickey-Fuller Test Equation

Dependent Variable: D(AV_GDP)

Method: Least Squares

Date: 04/03/18 Time: 10:09

Sample (adjusted): 1981 2015

Included observations: 35 after adjustments

Variable Coefficient Std. Error t-Statistic Prob.

AV_GDP(-1) -0.488500 0.154510 -3.161611 0.0034

C 8.651967 2.858132 3.027141 0.0048

@TREND(;1980;) -0.045672 0.054534 -0.837504 0.4085

R-squared 0.241905 Mean dependent var -0.117142

Adjusted R-squared 0.194524 S.D. dependent var 3.604692

S.E. of regression 3.235150 Akaike info criterion 5.267844

Sum squared resid 334.9183 Schwarz criterion 5.401160

Log likelihood -89.18727 Hannan-Quinn criter. 5.313865

F-statistic 5.105532 Durbin-Watson stat 1.950946

Prob(F-statistic) 0.011901

Variable D(AV_GDP)

Null Hypothesis: D(AV_GDP) has a unit root

Exogenous: None

Lag Length: 0 (Automatic – based on SIC, maxlag=9)

t-Statistic Prob.*

Augmented Dickey-Fuller test statistic -7.226290 0.0000

Test critical values: 1% level -2.634731

5% level -1.951000

10% level -1.610907

*MacKinnon (1996) one-sided p-values.

Augmented Dickey-Fuller Test Equation

Dependent Variable: D(AV_GDP,2)

Method: Least Squares

Date: 04/03/18 Time: 10:09

Sample (adjusted): 1982 2015

Included observations: 34 after adjustments

Variable Coefficient Std. Error t-Statistic Prob.

D(AV_GDP(-1)) -1.221133 0.168985 -7.226290 0.0000

R-squared 0.612693 Mean dependent var -0.075791

Adjusted R-squared 0.612693 S.D. dependent var 5.708495

S.E. of regression 3.552623 Akaike info criterion 5.402220

Sum squared resid 416.4974 Schwarz criterion 5.447113

Log likelihood -90.83774 Hannan-Quinn criter. 5.417530

Durbin-Watson stat 2.139692

Variable GCF_GDP

Null Hypothesis: GCF_GDP has a unit root

Exogenous: Constant, Linear Trend

Lag Length: 0 (Automatic – based on SIC, maxlag=9)

t-Statistic Prob.*

Augmented Dickey-Fuller test statistic -2.170986 0.4902

Test critical values: 1% level -4.243644

5% level -3.544284

10% level -3.204699

*MacKinnon (1996) one-sided p-values.

Augmented Dickey-Fuller Test Equation

Dependent Variable: D(GCF_GDP)

Method: Least Squares

Date: 04/03/18 Time: 10:10

Sample (adjusted): 1981 2015

Included observations: 35 after adjustments

Variable Coefficient Std. Error t-Statistic Prob.

GCF_GDP(-1) -0.256390 0.118098 -2.170986 0.0375

C 5.003960 2.680321 1.866926 0.0711

@TREND(;1980;) -0.075391 0.070467 -1.069875 0.2927

R-squared 0.128381 Mean dependent var -0.132844

Adjusted R-squared 0.073905 S.D. dependent var 3.817016

S.E. of regression 3.673261 Akaike info criterion 5.521853

Sum squared resid 431.7711 Schwarz criterion 5.655169

Log likelihood -93.63243 Hannan-Quinn criter. 5.567874

F-statistic 2.356642 Durbin-Watson stat 1.715612

Prob(F-statistic) 0.110976

Variable D(GCF_GDP)

Null Hypothesis: D(GCF_GDP) has a unit root

Exogenous: None

Lag Length: 0 (Automatic – based on SIC, maxlag=9)

t-Statistic Prob.*

Augmented Dickey-Fuller test statistic -5.724293 0.0000

Test critical values: 1% level -2.634731

5% level -1.951000

10% level -1.610907

*MacKinnon (1996) one-sided p-values.

Augmented Dickey-Fuller Test Equation

Dependent Variable: D(GCF_GDP,2)

Method: Least Squares

Date: 04/03/18 Time: 10:12

Sample (adjusted): 1982 2015

Included observations: 34 after adjustments

Variable Coefficient Std. Error t-Statistic Prob.

D(GCF_GDP(-1)) -0.981274 0.171423 -5.724293 0.0000

R-squared 0.498053 Mean dependent var -0.100503

Adjusted R-squared 0.498053 S.D. dependent var 5.387412

S.E. of regression 3.816887 Akaike info criterion 5.545718

Sum squared resid 480.7646 Schwarz criterion 5.590611

Log likelihood -93.27720 Hannan-Quinn criter. 5.561028

Durbin-Watson stat 1.957980

Variable GDP _deflator

Null Hypothesis: GDP_DEFLATOR has a unit root

Exogenous: Constant, Linear Trend

Lag Length: 0 (Automatic – based on SIC, maxlag=9)

t-Statistic Prob.*

Augmented Dickey-Fuller test statistic -5.432800 0.0005

Test critical values: 1% level -4.243644

5% level -3.544284

10% level -3.204699

*MacKinnon (1996) one-sided p-values.

Augmented Dickey-Fuller Test Equation

Dependent Variable: D(GDP_DEFLATOR)

Method: Least Squares

Date: 04/03/18 Time: 10:25

Sample (adjusted): 1981 2015

Included observations: 35 after adjustments

Variable Coefficient Std. Error t-Statistic Prob.

GDP_DEFLATOR(-1) -0.942469 0.173478 -5.432800 0.0000

C -5.433836 5.187703 -1.047446 0.3027

@TREND(;1980;) 0.397735 0.256720 1.549290 0.1311

R-squared 0.480107 Mean dependent var -0.338368

Adjusted R-squared 0.447614 S.D. dependent var 19.97522

S.E. of regression 14.84612 Akaike info criterion 8.315170

Sum squared resid 7053.032 Schwarz criterion 8.448486

Log likelihood -142.5155 Hannan-Quinn criter. 8.361191

F-statistic 14.77558 Durbin-Watson stat 2.019630

Prob(F-statistic) 0.000028

Variable MVA_GDP

Null Hypothesis: MVA_GDP has a unit root

Exogenous: Constant, Linear Trend

Lag Length: 0 (Automatic – based on SIC, maxlag=9)

t-Statistic Prob.*

Augmented Dickey-Fuller test statistic -2.432419 0.3576

Test critical values: 1% level -4.243644

5% level -3.544284

10% level -3.204699

*MacKinnon (1996) one-sided p-values.

Augmented Dickey-Fuller Test Equation

Dependent Variable: D(MVA_GDP)

Method: Least Squares

Date: 04/03/18 Time: 10:29

Sample (adjusted): 1981 2015

Included observations: 35 after adjustments

Variable Coefficient Std. Error t-Statistic Prob.

MVA_GDP(-1) -0.289224 0.118904 -2.432419 0.0208

C 7.457892 3.087387 2.415600 0.0216

@TREND(;1980;) -0.136270 0.055678 -2.447461 0.0201

R-squared 0.169865 Mean dependent var -0.335672

Adjusted R-squared 0.117982 S.D. dependent var 2.034819

S.E. of regression 1.911017 Akaike info criterion 4.214965

Sum squared resid 116.8636 Schwarz criterion 4.348280

Log likelihood -70.76188 Hannan-Quinn criter. 4.260985

F-statistic 3.273974 Durbin-Watson stat 2.067620

Prob(F-statistic) 0.050860

Variable D(MVA_GDP)

Null Hypothesis: D(MVA_GDP) has a unit root

Exogenous: None

Lag Length: 0 (Automatic – based on SIC, maxlag=9)

t-Statistic Prob.*

Augmented Dickey-Fuller test statistic -6.468006 0.0000

Test critical values: 1% level -2.634731

5% level -1.951000

10% level -1.610907

*MacKinnon (1996) one-sided p-values.

Augmented Dickey-Fuller Test Equation

Dependent Variable: D(MVA_GDP,2)

Method: Least Squares

Date: 04/03/18 Time: 10:30

Sample (adjusted): 1982 2015

Included observations: 34 after adjustments

Variable Coefficient Std. Error t-Statistic Prob.

D(MVA_GDP(-1)) -1.118765 0.172969 -6.468006 0.0000

R-squared 0.559012 Mean dependent var -0.019905

Adjusted R-squared 0.559012 S.D. dependent var 3.130841

S.E. of regression 2.079095 Akaike info criterion 4.330713

Sum squared resid 142.6470 Schwarz criterion 4.375606

Log likelihood -72.62212 Hannan-Quinn criter. 4.346023

Durbin-Watson stat 2.018321

Variable TOT

Null Hypothesis: TOT has a unit root

Exogenous: Constant, Linear Trend

Lag Length: 0 (Automatic – based on SIC, maxlag=9)

t-Statistic Prob.*

Augmented Dickey-Fuller test statistic -3.097679 0.1225

Test critical values: 1% level -4.243644

5% level -3.544284

10% level -3.204699

*MacKinnon (1996) one-sided p-values.

Augmented Dickey-Fuller Test Equation

Dependent Variable: D(TOT)

Method: Least Squares

Date: 04/03/18 Time: 10:31

Sample (adjusted): 1981 2015

Included observations: 35 after adjustments

Variable Coefficient Std. Error t-Statistic Prob.

TOT(-1) -0.513751 0.165850 -3.097679 0.0040

C 21.71849 7.403997 2.933347 0.0062

@TREND(;1980;) 0.726365 0.313374 2.317881 0.0270

R-squared 0.233582 Mean dependent var 0.559764

Adjusted R-squared 0.185681 S.D. dependent var 11.83856

S.E. of regression 10.68307 Akaike info criterion 7.657014

Sum squared resid 3652.096 Schwarz criterion 7.790330

Log likelihood -130.9977 Hannan-Quinn criter. 7.703035

F-statistic 4.876338 Durbin-Watson stat 2.090722

Prob(F-statistic) 0.014173

Appendix D: Regression results

Dependent Variable: DLHFPC

Method: Least Squares

Date: 04/03/18 Time: 10:47

Sample (adjusted): 1981 2015

Included observations: 35 after adjustments

Variable Coefficient Std. Error t-Statistic Prob.

C -0.006052 0.014351 -0.421695 0.6766

DGDPPC 0.000974 0.000208 4.688549 0.0001

DGCF_GDP -0.021428 0.004268 -5.021022 0.0000

GDP_DEFLATOR 0.011473 0.001414 8.112985 0.0000

DLPRIVCRE 0.031427 0.029801 1.054569 0.3010

DMVA_GDP 0.010804 0.008756 1.233886 0.2279

DAV_GDP 0.005006 0.005404 0.926398 0.3624

DTOT 0.003967 0.001683 2.356898 0.0259

R-squared 0.832407 Mean dependent var 0.009536

Adjusted R-squared 0.788957 S.D. dependent var 0.174095

S.E. of regression 0.079978 Akaike info criterion -2.016499

Sum squared resid 0.172705 Schwarz criterion -1.660991

Log likelihood 43.28873 Hannan-Quinn criter. -1.893778

F-statistic 19.15783 Durbin-Watson stat 1.913509

Prob(F-statistic) 0.000000

Appendix E: Heteroccedasticity results

Heteroskedasticity Test: Breusch-Pagan-Godfrey

F-statistic 0.534309 Prob. F(7,27) 0.8008

Obs*R-squared 4.258457 Prob. Chi-Square(7) 0.7496

Scaled explained SS 1.244740 Prob. Chi-Square(7) 0.9899

Test Equation:

Dependent Variable: RESID^2

Method: Least Squares

Date: 04/03/18 Time: 10:52

Sample: 1981 2015

Included observations: 35

Variable Coefficient Std. Error t-Statistic Prob.

C 0.005128 0.000936 5.476459 0.0000

DGDPPC 1.51E-05 1.36E-05 1.113268 0.2754

DGCF_GDP -4.40E-05 0.000278 -0.158028 0.8756

GDP_DEFLATOR -3.28E-05 9.23E-05 -0.355365 0.7251

DLPRIVCRE 0.002005 0.001945 1.031299 0.3116

DMVA_GDP 0.000168 0.000571 0.294470 0.7707

DAV_GDP 0.000144 0.000353 0.408705 0.6860

DTOT 4.43E-05 0.000110 0.403571 0.6897

R-squared 0.121670 Mean dependent var 0.004934

Adjusted R-squared -0.106045 S.D. dependent var 0.004962

S.E. of regression 0.005219 Akaike info criterion -7.475562

Sum squared resid 0.000735 Schwarz criterion -7.120054

Log likelihood 138.8223 Hannan-Quinn criter. -7.352840

F-statistic 0.534309 Durbin-Watson stat 2.405918

Prob(F-statistic) 0.800782

Appendix F: Autocorrelation

Breusch-Godfrey Serial Correlation LM Test:

F-statistic 0.025118 Prob. F(2,25) 0.9752

Obs*R-squared 0.070188 Prob. Chi-Square(2) 0.9655

Test Equation:

Dependent Variable: RESID

Method: Least Squares

Date: 04/03/18 Time: 10:53

Sample: 1981 2015

Included observations: 35

Presample missing value lagged residuals set to zero.

Variable Coefficient Std. Error t-Statistic Prob.

C -0.000184 0.014924 -0.012341 0.9903

DGDPPC -2.03E-05 0.000234 -0.086864 0.9315

DGCF_GDP 0.000329 0.004766 0.069016 0.9455

GDP_DEFLATOR -1.61E-05 0.001493 -0.010778 0.9915

DLPRIVCRE -0.001950 0.032306 -0.060352 0.9524

DMVA_GDP 0.000223 0.009271 0.024027 0.9810

DAV_GDP -0.000106 0.005692 -0.018557 0.9853

DTOT -5.00E-05 0.001762 -0.028362 0.9776

RESID(-1) -0.035947 0.234110 -0.153548 0.8792

RESID(-2) -0.043056 0.236682 -0.181915 0.8571

R-squared 0.002005 Mean dependent var -3.97E-18

Adjusted R-squared -0.357273 S.D. dependent var 0.071271

S.E. of regression 0.083032 Akaike info criterion -1.904221

Sum squared resid 0.172359 Schwarz criterion -1.459836

Log likelihood 43.32386 Hannan-Quinn criter. -1.750819

F-statistic 0.005582 Durbin-Watson stat 1.888531

Prob(F-statistic) 1.000000

Appendix G: Model specification results

Ramsey RESET Test

Equation: EQ01

Specification: DLHFPC C DGDPPC DGCF_GDP GDP_DEFLATOR

DLPRIVCRE DMVA_GDP DAV_GDP DTOT

Omitted Variables: Squares of fitted values

Value df Probability

t-statistic 1.818887 26 0.0805

F-statistic 3.308350 (1, 26) 0.0805

Likelihood ratio 4.192157 1 0.0406

F-test summary:

Sum of Sq. df Mean Squares

Test SSR 0.019495 1 0.019495

Restricted SSR 0.172705 27 0.006396

Unrestricted SSR 0.153210 26 0.005893

LR test summary:

Value df

Restricted LogL 43.28873 27

Unrestricted LogL 45.38481 26

Unrestricted Test Equation:

Dependent Variable: DLHFPC

Method: Least Squares

Date: 04/03/18 Time: 10:55

Sample: 1981 2015

Included observations: 35

Variable Coefficient Std. Error t-Statistic Prob.

C 0.010619 0.016545 0.641845 0.5266

DGDPPC 0.000935 0.000201 4.660716 0.0001

DGCF_GDP -0.017404 0.004655 -3.738670 0.0009

GDP_DEFLATOR 0.011709 0.001363 8.587283 0.0000

DLPRIVCRE 0.040842 0.029068 1.405045 0.1718

DMVA_GDP 0.009500 0.008435 1.126213 0.2704

DAV_GDP 0.003192 0.005281 0.604417 0.5508

DTOT 0.002802 0.001738 1.612550 0.1189

FITTED^2 -0.687325 0.377882 -1.818887 0.0805

R-squared 0.851325 Mean dependent var 0.009536

Adjusted R-squared 0.805579 S.D. dependent var 0.174095

S.E. of regression 0.076764 Akaike info criterion -2.079132

Sum squared resid 0.153210 Schwarz criterion -1.679186

Log likelihood 45.38481 Hannan-Quinn criter. -1.941071

F-statistic 18.60980 Durbin-Watson stat 1.972405

Prob(F-statistic) 0.000000

Appendix H: Normality results

Appendix I: Raw Data

Year HFPC GDPPC GDP deflator TOT PRIVCRE GCF/GDP MVA/GDP AV/GDP

1980 645,1463 1175,135 12,7409 49,8904 7,97441 16,937 21,58 15,6971

1981 774,7579 1274,683 6,59908 45,3306 8,82371 20,8159 21,7679 17,7366

1982 771,4266 1259,196 3,85876 39,1453 7,47684 19,0537 20,8819 16,1197

1983 714,4797 1230,394 -10,502 35,9169 11,9813 14,3052 23,0131 11,2377

1984 483,843 1161,406 -16,595 41,3661 20,8859 17,0355 22,6512 14,8583

1985 402,154 1196,635 -17,017 44,2137 24,4871 17,82 19,85 22,6736

1986 407,0306 1178,559 8,02588 45,5704 25,0116 18,0564 21,4468 17,7608

1987 427,5508 1151,447 7,18936 45,2906 29,9427 14,9362 22,6894 14,4075

1988 410,2686 1198,305 7,78512 44,1003 25,8203 18,7017 21,5114 16,3821

1989 541,0039 1222,647 0,79293 45,0625 44,3659 15,038 25,5966 14,9303

1990 544,3438 1272,051 -0,9204 45,6593 23,0399 17,3769 22,7557 16,4763

1991 563,1761 1309,005 -6,7773 51,0515 26,1662 19,1034 27,1562 15,2673

1992 409,9339 1164,254 -14,13 63,7125 28,7709 20,2373 29,537 7,41379

1993 385,2112 1152,45 -3,7911 63,1671 29,8406 22,7749 23,0119 15,0389

1994 381,1789 1234,966 -3,8957 71,1195 28,4084 23,7291 21,167 18,9734

1995 407,2875 1214,693 3,03854 79,1568 33,8378 19,6602 21,7952 15,2352

1996 470,8001 1317,51 8,98438 72,0696 31,2324 18,5419 18,7807 21,7711

1997 535,7287 1330,676 -2,879 82,2051 38,5965 18,1339 18,0075 18,9341

1998 350,926 1347,972 -27,049 88,514 34,7091 20,7505 16,6279 21,7885

1999 363,3399 1317,969 8,00681 70,9227 22,2877 14,3963 16,3525 19,1767

2000 327,9399 1261,163 0,6279 74,0674 27,1112 13,5694 15,6051 18,2616

2001 383,7556 1264,431 -0,1309 67,8979 34,5213 10,2665 14,5585 17,307

2002 406,9653 1139,59 2,71295 66,8074 103,632 5 13,2514 14,029

2003 361,5095 935,9302 8,80128 70,452 57,03 8 13,6472 16,5934

2004 370,7141 871,6671 7,61152 76,0396 18,0247 4,50911 15,1165 19,575

2005 410,1135 811,563 5,1366 76,0437 15,7937 1,52518 16,3831 18,5773

2006 429,1253 772,4726 -2,0177 82,8206 46,4901 1,57116 16,889 20,2818

2007 390,1866 732,77 0,89489 84,1729 27,3012 7,10975 16,4013 21,5979

2008 388,9023 593,1272 1,34922 109,522 10,2381 5,12791 16,659 19,3989

2009 628,7098 652,2887 74,2982 69,2609 17,5438 14,2907 14,2793 13,9065

2010 632,4728 719,9795 4,48786 98,6952 9,629 22,2781 12,5889 13,1382

2011 681,1096 813,834 3,33187 104,282 18,1924 20,2788 12,2638 11,5865

2012 898,0457 913,5306 2,56318 89,1193 16,4203 11,8449 11,3696 11,0252

2013 861,9478 942,0387 2,8057 72,4715 12,7963 11,3785 10,7828 10,0944

2014 823,7967 939,7803 0,70092 67,0724 14,5693 11,8256 10,3202 12,1346

2015 900,7536 933,5033 0,89807 69,4821 10,1762 12,2874 9,83145 11,5971