Determinants of Life Insurance Demand in Ethiopia

Advisor - Negus

Name – Michias –Tsega

 ID-NO- 4074/09

2|Page
1.1 Background of the study
Insurance is a contract in which the insured transfers risk of potential
loss to the insurer who promises to compensate the former upon suffering loss. The
insured then pays an agreed fee called a premium in consideration for this promise,
the promisor is called the insurer and the promisee is called the insured. More over
the fundamental purpose of insurance, whether of people or of property, is
protection against possible economic loss that is unintentional and permanent loss
of something which has monetary value (Lowe, 1999). Mean while the worldwide
insurance market has grown rapidly and the internationalization of the insurance
business is becoming more widespread (Nestrova 2008).

In contrast when we came to the case of Ethiopia the insurance industry is

undeveloped and it is untapped industry as it is mentioned by the World Bank
report of 2016. Furthermore the Ethiopian modern insurance industry does not
have a long history of development despite the country’s long history of
civilization. Modern forms of insurance service which were introduced in Ethiopia
by Europeans, trace their origin as far back as 1905 when the bank of Abyssinia
began to transact fire and marine insurance as an agent of a foreign insurance
company (Suheyli Reshid 2015).

Based on a survey made in 1954, there were nine insurance companies that were
providing insurance service in the country. With the exception of Imperial
Insurance Company that was established in 1951, all the remaining of the
insurance companies were either branches or agents of foreign companies. In the
year 1960, the number of insurance companies increased considerably and reached
33. At that time insurance business like any business undertaking was classified as
trade and was administered by the provisions of the commercial code.

Hailu Zeleke (2007) mentioned that the first significant event that the Ethiopian
insurance market observation was the issuance of proclamation No. 281/1970 and
this proclamation was issued to provide for the control & regulation of insurance
business in Ethiopia. With respect to this, it created an insurance council and an
insurance controller's office. The controller of insurance licensed 15 domestic
insurance companies, 36 agents, 7 brokers, 3 actuaries & 11 assessors in

3|Page
accordance with the provisions of the proclamation immediately in the year after
the issuance of the law.

However, in 1975 the under Government “Derg” Monopolized the service of

Insurance under the title “Ethiopian Insurance Corporation” established by
proclamation No. 26/1975 which brought the right for the government to
confiscated the 13 private companies’ ownership and the control under it. After the
collapse of the Marxist regime in mid-1991 a fundamental change has taken place
and there was a shift in political, economic and social orientation from
totalitarianism to that of liberalization of some selected industry.

When we came to the life insurance business from the year 1967 up to 1972
indicates that the share of life insurance in the total gross premium income of the
industry declined from 15.1% in 1967 to 7.9% in 1972. But the current data
collected from National Bank of Ethiopia shows an increase in gross premium for
the years 1996 to 2015 as can be seen from the table 1.1, but in the recent year
since 2016 the gross premium is gradually decreasing.
Table1.1 Life Insurance in Total Gross Premium (1996-2017) in Ethiopia

4|Page
 Year Gross premium
ETB birr
1996 9,288,503.00

1997 10,717,838.00

1998 11,322,272.00

1999 10,454,012.00

2000 14,342,728.00

2001 17,487,438.00

2002 18,794,759.00

2003 23,277,997.00

2004 26,052,178.00
Source -
2005 32,082,275.00

2006 39,627,164.00

2007 52,181,897.00

2008 74,112,459.21

2009 91,797,963.87

2010 110,293,217.33

2011 160,303,821.44

2012 255,895,872.00

2013 299,882,135.63

2014 277,830,182.38

2015 312,486,281.00

2016 205,068,740.00

2017 192,205,100.00

National bank of Ethiopia report

As mentioned by Gebreyes (2011), the insurance market is still undeveloped,

uncompetitive and there exist paucity of information on the kind of life insurance.
The current practice of bulk insurance coverage and business in Ethiopia is
targeting the corporate market and focuses mainly on general insurance with a very
limited coverage in life insurance.

5|Page
 In addition the Ethiopian insurance penetration is lower than the world and
African average which in 2012 was 6.5% and 3.56 respectively while for Ethiopia
it was 0.03%. Even, the 9 penetration of Ethiopia insurance is lower than that of
East African countries such as Kenya 3.4%, Rwanda 2.3%, Uganda 0.85%,
Tanzania 0.9%, and Burundi 0.83 %( African insurance organization report 2014).

In spite the fact most theoretical and empirical studies, identify that the following
factors have influence on demand for life insurance.

According to Nestrova 2008, Determinants of life insurance demand are divided

into: -

Economic: income level, inflation, real interest rate, level of financial

development,

Demographic: life expectancy, dependency ratio, level of education and

urbanization;

Institutional: political stability, Rule of Law, control for corruption, European

Union membership.

However, the researcher will only focus on selected determinants of economic

and demographic factor.

1.2 Statement of the problem

Life insurance is a contract between the policy owner and the

insurer, whereby the insurer agrees to pay to a designated beneficiary a
sum of money upon the occurrence of certain events, such as terminal
illness, death or critical illness or maturity benefit depending on the form
of the contract (Nestrova, 2008).
The function of Life Insurance is not known in the great days of
well-being and happy living days. A sudden set back in the health of
our most loved ones can turn off the life style of the whole family. To
manage things in an uncertain world, it is a wise attempt to choose a life

6|Page
insurance that can support us and our family people in the worse days
most unexpected. It comes as an assurance promised to appear in the
dark days of life.

Especially in modern economy it is huge source of employment and the

insurance companies are an important provider of financial services.
However, many developing economies are experiencing very low levels
of life insurance penetration and low growth level of the financial
industry (Beck and Webb, 2003).
When we came to the case of Ethiopian life insurance business is small
in penetration and density when compared to other developing countries
(African insurance organization report 2016).
Thus, Life insurance gives individuals and the economy many financial
services. In the phase of urbanization, population mobility and
formulization of economic links among individuals, families and
communities, it has been taken as an important element for individuals
and families to overcome risk attached with income. Also, life insurance
items enhance long-term savings and the re-investment in private and
public sector projects Munir and Khan (2012).
Moreover, Life insurance market of transition economies had
experienced a rapid growth over the last decade, indicating the increased
importance of this sector as a financial intermediary (Nestrovil 2008).
The role of life insurance is beyond enhancing individual financial
security it is also useful for economic and financial development this
immense role of life insurance directs the researcher to investigate which
economic and socio-demographic Factors Affecting Life Insurance
Demand. However, according to nestrovil 2008 the factors that drive

7|Page
life insurance demand across countries still remain unclear. The main
purpose of the study is to identify the determinants of life insurance
demand in Ethiopia. In addition, to the best of the researcher‘s
knowledge, there are three previous research works in Ethiopia
concerning the life insurance demand.
The first project is conducted by Roman (2011) she makes her main
emphasis on only general economic factors that determine life insurance
demand. Since there are also demographic and institutional determinants
that must be explored, this study would fill this gap by studying the
remaining variables.
The second project is conducted by Amrot Yilma (2014) examined the
determinants of life insurance for a time series data for the period 1983-
2012. His work focus on the relationship of life insurance and on some
selected independent variables (income, real interest rate, inflation,
dependency ratios and life expectancy). Since this research would fill the
knowledge gap on the time period by this study the time period will
extend from 1983 up to 2017.
The third research is conducted by Abenezer (2016) he take the case of
Ethiopian Insurance Corporation on the determination of life insurance
demand in Ethiopia. Since this is done on only one insurance company
by taking it as a sample and it is difficult to make strong conclusion.
This research would fill this gap.
Generally, this research work would fill existing gap of previous studies
by including omitted variables (like socio demographic variables) and by
using more recent data.

8|Page
9|Page
 1.3 Objective of the study

1.3.1 General objective

The main objective of the study is to investigate the major determinant of life
insurance demand in Ethiopia.

1.3.2 Specific objectives

In order to achieve the general objective of the study, the researchers have
identified the following specific objectives;

1. Identify if income affect life insurance demand.

2. To explore the influence of real interest rate on life insurance demand.
3. To investigate the effect of life expectance on life insurance demand.
4. To examine the effect of dependence ratio on life insurance demand.
5. Identify if inflation affects demand for life insurance;
1.4 Research hypothesis
Based on existing literature, the researcher determines five social and economic
variables which may affect demand for life insurance in Ethiopia.
H1: There is positive and statistically significant relationship between income
level and demand for life insurance policy.
 Previous studies mainly show that there is significant and positive relationship
between level of income and demand for life insurance (Dickinson, Khajuria
(1986), Truett and Truett (1990) and Browne and Kim (1993)).
 H2: There is negative and statistically significant relationship between inflation
and demand for life insurance policy.
 In the literature, it is observed that inflation affects demand for life
insurance in negative way (Beck and Webb (2003), Li [Link] (2007),
Nesterova (2008), Çelik and Kayali (2009)).
H3: There is positive and statistically significant relationship between real

10 | P a g e
 interest rate and demand for life insurance policy.
 Other researches done by Li [Link] (2007), Nesterova (2008) Beck and
webb (2003) identifies that life insurance demand is positively related
with real interest rate.
H4: There is positive and statistically significant relationship between life
expectancy and demand for life insurance.
 Nesterova (2008), Redzuan (2011) identified that life expectance have a
significant and positive impact on the demand for life insurance.
H5: There is positive and statistically significant relationship between
Dependency ratio and demand life insurance.
 Nesterova (2008), Redzuan (2011) identified that life expectance have a
significant and positive impact on the demand for life insurance.

Significance of the study

Since, understanding of the factors that determine life insurance demand in this
country is important for policies at national level and in the industry actors to exert
more effort on working on these factors to expand their knowledge. In addition to
filling this knowledge gap, it will provide important information to those who want
to do similar studies at a larger scale. Therefore the result of this paper would have
great contribution to this specific body of knowledge.

1.7 Scope of the Study

Due to constraints of data; the researcher has limited the analysis only on five
selected determinants of demand for life insurance such as income, inflation,
real interest rate, life expectancy, and dependency ratio. The study used a time
series data for the periods from 1983 to 2018.
1.8 Organization of the Paper

11 | P a g e
 This paper is organized into five chapters. Chapter one w present with
introduction, Chapter two with the review of the related literature, and Chapter
three discusses the methodology employed in the study, including, research
design, data source and type , method of data analysis and model specification.
Chapter four is about data analysis and interpretation of results. Finally, chapter
five will contain summary of findings, conclusions and recommendations.

CHAPTER TWO

LITERATURE REVIEW
In this section the researcher first presents about life insurance basic
policy types. Second theoretical review is presented. Third the theoretical review
is followed by the empirical review. The empirical review presents different
empirical researches done in the area of determinants of life insurance demand
and it is summarized with the most relevant findings in the field of life insurance
demand in table 2.1 and finally at the end researcher develop conceptual
framework of the study.

2.1 Basic Life insurance policy types

As mentioned by Kenneth M. 1992, there are three basic forms of life insurance
products may be distinguished as: (1) whole-life, (2) term life, and (3) annuities.
Within each category, there are numerous varieties and refinements which can
provide almost any form of protection to meet the particular needs of the
insurance client, but the emphasis here will be confined to the standard forms of
insurance.

a. Whole-life Insurance

12 | P a g e
 Sometimes known as permanent insurance or cash-value life
insurance, traditional whole life insurance is based on a level-premium method
of payment, whereby the policyholder pays an unchanging periodic amount
for the entire life of the contract, normally until the death of the insured. In
the early years of the policy, premiums are higher than needed to meet the
average of death, Claims at younger ages; thus, a reserve is accumulated to
meet the higher number of death claims at later ages when the premium
payment remains level. Because there is a sizeable reserve build-up behind
whole-life contracts, such policies have a cash surrender value and they
typically carry a policy loan privilege (Meier, Kenneth (1988)).

According Wright (1991), Mutual companies selling whole life policies

allow the policyholder to participate in any excess earnings on reserves,
through policy dividends paid out to the insured or used to buy additional
coverage. In this way, the net cost of insurance may be less than the gross
premiums required under the contract. Stock companies have also offered
participating policies, but non-participating policies are more standard,
allowing any excess earnings above the assumed rate to remain with the
company as shareholder profits.

In the early 1980s, new forms of whole life policies were developed,
primarily in response to the surge in inflation rates and market interest rates
to double-digit levels during the 1978-1982 periods in the United States. With
fixed dollar amounts on the face value of insurance policies, the real value of
future death benefits was jeopardized by persistent inflation. Reacting to this
threat to insurance purchases, life insurance companies developed three new
forms of "interest-sensitive" policies-universal life, variable life, and flexible
premium variable life. All three had the common element of reflecting
investment performance in the policies, by changing the size of the death
benefit or the annual premium or both over the duration of the policy
(Huggins, Kenneth (1986)).

Under universal life, the policyholder is able to vary his annual premiums as to
the amount and timing of payments. Net premiums (after loading and mortality

13 | P a g e
risk charges) are invested in a floating rate fund, and the earned interest credited
to the policy will vary with investment results. Death benefits cannot fall below
the face value of the policy, but they can expire if the level of premium payments
or the investment experience is not sufficient to carry the policy to maturity.
Thus, the buyer assumes some of the investment risk, but he also can share
directly in the rewards of good performance (Huggins, Kenneth (1986)).

Variable life carries a fixed annual premium but allows the policyholder to
designate investment of his funds into bonds, equities, or a money market
account and to vary his choice during the life of the policy as he sees fit. There is a
guaranteed minimum death benefit, but the size of the benefit will rise or fall
over time, depending on investment performance (Huggins, Kenneth (1986)).

A combined version of the two previous policy types is flexible premium

variable life, wherein premium payments may vary and a choice of investment
media can be made. Death benefits will depend on investment returns on the
assets standing behind the policy.

b. Term insurance

This popular form of ordinary insurance differs in major respects

from whole life, primarily because it is offered for a specified number of years,
usually one, two or five years. During the specified time, the premium is
unchanged, based on the attained age of the insured. At younger ages, the
premium is substantially lower than for cash-value life, but the premium increases
sharply in later years as the likelihood of death increases on average (Kenneth M.
1992;p6).

As mentioned by Sandra G. (1987), term insurance is underwritten

largely on a "pay-as-you-go" basis whereby current claims can be met from the
inflow of current premiums paid by those covered under this system. Thus, little
reserve is accumulated for investment in the financial markets to fund future
death claims. Because term insurance is the least expensive form of coverage, it
appeals largely to younger adults and those with lower incomes. It generates
minimal amounts of savings and provides no policy loan privileges or cash

14 | P a g e
surrender values. Those who purchase term insurance are often urged to
undertake a companion savings program in other forms, but these do not
represent contractual savings on a continuing basis. When the policyholder
reaches later years of life, the cost of term insurance can be prohibitively
expensive, leading to dropping of the policy and a consequent absence of
coverage. In the event of a major health problem, a policyholder may find that his
term insurance cannot be renewed, since a physical examination is typically
required whenever a new policy is purchased. However, some term policies are
written to permit automatic renewal without a health review, but at a higher
premium.

In addition to individual purchases, term insurance is marketed in large volume

through group insurance, whereby a master policy is issues to an employer to
cover all employees as a fringe benefit in addition to wages and salaries (World
Bank 2015).

c. Annuities-Individual and Group

Annuities are designed to provide continuing income to protect

against the possibility of outliving other forms of income, usually beginning
with retirement from active employment. They can be purchased with a single
payment to provide an immediate or deferred stream of income, or with a
series of payments sufficient to provide a stream of income to begin at some
future date. Annuities can provide income for the lifetime of an individual, or a
joint-and-survivorship annuity can cover income for a surviving spouse or
child, usually a reduced amount. Since annuities are related to life
expectancies under a pooling principle, they are a form of insurance which
requires application of underwriting techniques and the use of mortality
tables. If an annuitant dies before the expected life span, the unused annuity
amounts can be applied to payments for those living beyond their calculated
life expectancy. The cost of an annuity is closely related to the age of covered
individuals, but also depends on the projected investment return on amounts
paid to the company (known as annuity considerations), (Kenneth M.

15 | P a g e
 1992;p10).

It should be noted that some corporate pension plans are funded with
bank trust departments or self-administered by the business firm itself. When
an employee retires, a withdrawal is made from such funds to purchase the
necessary annuity from a life insurance company; or, a lump-sum is paid to the
employee which can then be used to purchase a life company annuity or
invested directly by the retire to provide retirement income. For those who are
self-employed. Such as doctors, lawyers, and owners of small firms, special
laws allow them to build pension benefits in other ways, broadly parallel to
those available to employees of corporate businesses. Under the so-called
Keogh Plans, a percentage of compensation up to a fixed dollar amount may
be used to purchase retirement benefits from a life insurance company for the
owner and his employees (Sandra G. 1987).

Other Life Insurance Forms- In addition to the basic policy types described
above, two other insurance forms exist, though in minor amounts. One type is
called industrial life, premiums collected door-to-door on a weekly or monthly
basis primarily sold to low-income groups, such insurance is often meant to cover
burial expenses, and the other minor form of insurance is called credit insurance,
designed to repay a mortgage loan or installment loan in case the borrower dies.
Such insurance is often required as a condition of a loan to an individual, or is
offered as a related option. Credit insurance is term insurance, usually decreasing
in amount in keeping with scheduled repayments of a mortgage or an auto loan
(Huggins, Kenneth (1986)).

2.2 Theoretical review

The starting point for any theoretical approach is explaining demand

for life insurance consumption theory in its various forms. Other factors are also
of importance, but these concerns more the supply-side influences on the
activities on the markets for life insurance products. Supply-side factors affect the
cost of life insurance products and thus their prices. These factors include various
attributes of the economic, political and social environments in which the
companies of the insurance industry operate (SWISS RE, 2003).
16 | P a g e
 According to Nesterova (2008), Demand for life insurance has usually been
explained through the life-cycle models where households or individuals
maximize their expected utility of lifetime consumption.

In addition Nestrova (2008), mentioned that the issue of life insurance

demand is not new for researchers and was brought to light beginning from Yaari
(1965), who was the first to work out a theoretical background to explain the
demand for life insurance, Yaari was the first researcher who pointed out the
issue of uncertainty in the life insurance demand, namely, the uncertainty of life
span of a consumer. He developed the life-cycle utility model of a consumer
together with deducing the optimal consumption and optimal saving plans of a
consumer and his results show that individual increases his expected utility by
buying the insurance. In his paper Yaari highlighted the models of utility functions
developed by Fisher (1930) and Marshal (1920), who as he said “were both aware
of the uncertainty of survival, but for one reason or another they did not expound
on how a consumer might be expected to react to this uncertainty if he is to
behave rationally”.

Lewis (1989) makes explicit allowance in his model for the

preferences of dependants and beneficiaries, thereby extending Yaari's (1965)
approach, whose utility considerations took into account only the preferences of
the head of the household or the principal earner. This model approach will be
examined in more detail here, as it forms the theoretical starting point for various
empirical works. Lewis (1989) defines life insurance demand as:

LD= F (p, TC, δ, L, W)

(+),(+),(+)(-)(-)

The demand for life insurance policies LD rises with the probability
of primary income-earner's death p, the present value of consumption by the
beneficiaries TC, and the relative risk aversion of the beneficiaries δ; it declines
with the household's net wealth W and the "policy loading factor" L, which
describes the price of insurance as the ratio between the cost and the actuarial

17 | P a g e
value of the insurance (swiss re 2003).

2.2.1 Concepts for indicators of demand

As mentioned by Kumar (2014) life insurance demand indicators are

very few variables, which were used to measure the life insurance demand or
consumption in an economy by different researchers, such as life insurance
penetration, life insurance density, life insurance in force, life insurance premium
volume, life insurance in savings etc. Out of these, life insurance density and
penetration are being used internationally by IMF, World Bank etc to measure the
global insurance consumption.

According to Swiss re (2003), empirical study of life insurance demand

first needs suitable indicators with which to operationalise the demand. The
simplest approach is to use the number of insurance policies (quantitative
indicator) or the premium volume or the sums of insured.

> As an indicator for demand of life insurance the researcher will use Life
Insurance Density.

Life insurance density is defined as premiums per capita expressed in

constant dollars. It indicates how much each inhabitant of the country spends on
average on insurance in constant dollars. Although both Life Insurance
Penetration and Life Insurance Density use gross premiums, there remain
important differences between both measures; Life Insurance Penetration
measures life insurance consumption relative to the size of the economy, while
Life Insurance Density compares life insurance consumption across countries
without adjusting for the income level of the economy. Consumers that purchase
life insurance policies to insure their dependents against mortality risk will
potentially buy more coverage and thus a higher face value in richer countries,
since the death benefit has to replace a larger income. We therefore expect Life
Insurance Density to be more income elastic than Life Insurance Penetration
(Beck/webb, 2002).

18 | P a g e
2.3 Empirical review on selected studies

In their study, Beck/Webb (2002) investigates the determinants of life

insurance demand in a panel of 68 countries over a period from 1961 to 2000. In
the first stage of the study, five-year averages are analyzed. Random-effects and
fixed-effects models are used for the econometric estimates of the yardsticks for
measurement of life insurance. The per-capita income level, inflation, the ratio of
the elderly to the working population, and the indicator for the development of
the banking system are robust explanatory factors for the international and inter
temporal variation of insurance penetration, insurance density and life insurance
business in force in relation to GDP. The other variables (life expectancy, expected
inflation, human capital, ratio of young people to the working population, real
interest rates, private saving rate, etc) sequentially introduced into the respective
estimates cannot be attributed any further major contribution to the explanation.

In a second stage, the annual observations are analyzed instead of five-year

averages. The study focuses on the insurance penetration as the demand
indicator. The results differ only slightly from those obtained from the five-year
averages. Further to the per-capita income level, the ratio of the elderly to the
working population, the indicator for the development of the banking sector, and
the inflation rate, another significant explanatory factor is the ratio of young
people to the working population. In general, the findings of the study by
Beck/Webb (2002) suggest that the per-capita income, inflation, and the indicator
for the banking sector development level are the most robust factors for
explaining the international and inter temporal variation in life insurance demand.
However, religious and institutional factors also account for some of the
international variation.

Truett/Truett (1990) analyzes the determinants of life insurance

demand on the basis of aggregate time series for Mexico from 1964 to 1979 and
the US in the period 1960 to 1982. As the indicator for life insurance demand in
the US, they use the average spending on life insurance products per family; in
Mexico they use the life insurance in force in relation to the working population.
The demand equations for the US and for Mexico are each dependent on an

19 | P a g e
indicator for age (average age of the population, proportion of earners in the 18
overall population), an indicator for the education level of the population
(average years of school attendance) and the real per-capita income. The demand
equations for the two countries are separately estimated in log-linear form. All
three explanatory variables show the expected positive correlation in the
regression analysis and are statistically significant. The estimates of the demand
equations based on future income are found to be statistically superior to the
variant based on current income, as they exhibit no auto-correlation.

Schwebler (1984) and the GDV (1983) analyze the factors influencing life
insurance demand in the period 1965 to 1980 for Germany. As their indicators for
demand they use the growth rates of the premium volume, the sums insured
under the business in force and the sums insured under new business. The
findings identify three groups of influencing factors: the macroeconomic situation
(economic growth, unemployment rate and price inflation), income trends
(disposable income, private saving rate and private consumption) and socio-
psychological (Allensbach indicator). On the basis of linear multiple regression
equations, the determinants used are capable of explaining almost 90% of the
variation in the respective demand indicator.

Zhou (1998) studies the influencing factors for 29 different regions and 14
large cities in China. The model approach for the regions uses the aggregate life
insurance premiums (sum of premiums from endowment and annuity insurance)
per capita and the separate premium payments as dependent variables. The gross
domestic product per capita is positively correlated both with the aggregate life
insurance premiums and also with the two types of insurance observed. The
dependency ratio of the children likewise has a significant influence on demand
for life insurance policies, and particularly endowment policies. Both the
dependency ratio of the elderly and the size of social insurance, and also the
indicator for human capital, yield no significant contribution to explaining the
regional variation in insurance premiums. For the 14 large cities, the aggregate
life insurance premiums per capita and the endowment premium payments per
capita are used as dependent variables. According to Zhou (1998) the only
significant explanatory variable is the income per capita.
20 | P a g e
 Amrot (2014) investigates the major determinants of life insurance
demand in Ethiopia. The study focuses on the relationship of life insurance with
six selected independent variables (income, inflation, real interest rate, level of
education, life expectance and dependence ratio). The researcher used a time
series data from 1983- 2012. The t-statistics and p-value show that explanatory
variables such as GDP per capita, inflation, real interest rate, level of education
and life expectance are statistically significant at 10 percent significance level.
GDP per capita, real interest rate, level of education and life expectance are
positively related with life insurance demand whereas inflation is negatively
related with life insurance demand. Dependence ratio does not have a statistically
significant relationship with life insurance. Amrot mentioned GDP per capita is the
most important factor that influences demand for life insurance followed by life
expectance and level of Education. According to amrot (2014) Inflation is the least
important factor in influence demand for life insurance.

Sen (2007), evaluated the impact of GDP per capita, GDS (gross domestic
savings) per capita, financial depth, urbanization, dependency ratio, adult literacy,
population, life expectancy at birth, crude death rate, inflation, real interest rate
and insurance price on the demand for life insurance. Sen supports the previous
findings by showing the significant positive relationship between life insurance
consumption and income, financial development, gross domestic savings, and
negative to inflation. Real interest rate was shown to be insignificant in a cross-
country analysis, however, Sen also incorporated time-series analysis on India for
the period 1965-2004 and real interest rate there appeared significant.
Comparing to Beck and Webb (2002) results such demographic factors as life
expectancy and young and old dependency ratio turned to be significant, together
with adult literacy rate and rate of urbanization.

Abenezer (2016), the study is made based on a primary data collected through
self-administered questionnaire from buyers who have purchased life insurance
from Ethiopian insurance corporation and aged 18 years or older. To analyze the
data Abenezer(2016), used combination of descriptive form of data analysis and
multiple regression analysis. Multiple Regression analysis was performed to
investigate the effect of each explanatory variable on life insurance demand.
21 | P a g e
Accordingly except family size and gender factor; income level, age factor,
education level and health status were found to be significant determinants of life
insurance demand. Among the six determinant factors, income level takes the
highest fraction in influencing the demand for life insurance policy followed by
age factor, family size, gender, education level and health status in that order.

Li [Link] (2007) analyzed the determinants of aggregate life insurance

demand on a cross section of developed economies. They found that income
plays a major role in the consumption of life insurance products. An increase of 1
percent in aggregate income can be expected to induce at least a 0.6 percent
increase in aggregate life insurance demand. In particular, the demand for life
insurance decreases with the average life expectancy and increases with the
dependency ratio. Education level is positively related to life insurance demand,
whereas the influence of social security expenditure is significantly negative. On
the other hand, inflation significantly decreases the demand for life insurance.
High real interest rates do not persuade households to purchase more insurance,
but actually stimulate them to reduce their purchase either because of higher
expected benefits for the same invested amount or because of higher preference
for immediate consumption relative to deferred consumption.

Munir and Khan (2012) on their study on impacts of macroeconomic &

demographic variables on the demand of life insurance in Pakistan identified that
financial development, gross savings, income level are directly linked to life
insurance demand while, price of insurance are inversely linked with life
insurance demand and the demographic variables of crude birth rate, crude death
rate, old age dependency ratio, urbanization are positively related with life
insurance demand for Pakistan.

Mahdzan & Victorian (2013) investigated the determinants of life

insurance demand among life insurance policyholders of five major life insurance
companies in Kuala Lumpur, Malaysia their results reveals that demographic
variables and saving motives were significantly related to life insurance demand.
Financial literacy, however, was found to be insignificant in determining life
insurance demand. Also they showed that education level is significantly related

22 | P a g e
to life insurance demand, where individuals with higher levels of education have
higher life insurance demand.

Summing up the previously mentioned theoretical and empirical

researches, life insurance demand is mainly determined by such macroeconomic
variables as income level, interest rate, wealth, inflation and financial
development also various demographic and social factors. In addition to this it is
summarized under table 2.1 with statistical result and Operationalisation.

23 | P a g e
 Table 2.1> Determinants of life insurance demand in empirical studies with their
statistical results

Categories Variables Operationalisation Expected Statistical Source

sign results (research papers)
Demographic Population •Ratio of young + *** Beenstock/Dickinson/
variable dependents to working Khajuria (1996),
population
•Ratio of old dependents Browne/Kim (1993),
to working population ***
+ Beck/Webb (2002),
• Number of marriages
+ # Truett/Truett (1986)
• Birth rate
Diacon (1980)
+ **

Education • Average years of school + *** Beck/Webb (2002),

attendance Truett/Truett (1986),
• Proportion of persons Browne/Kim (1993)
with (tertiary) education + **
in the population
between 20 and 24
(UNESCO)

Religion • Dummy variables for - ** Browne/Kim (1993),

predominantly Islamic, Beck/Webb (2002)
Protestant and Catholic
countries
Urbanity • Ratio of urban + # Beck/Webb (2002)
population to overall
population

24 | P a g e
 Life • Life expectancy at birth + *** Beck/Webb (2002),
(UN statistics) Beenstock/
expectancy • Life expectancy of Dickinson/Khajua
persons aged over 40 in (1996), Outreville
1975 (1996),

25 | P a g e
Macroeconomic •Economic • Growth of (per + ** Schwebler (1994), Beck/Webb
variable growth capita) GDP (2002), Zhuo (1998)

• Income •Disposable + *** Schwebler(1984),Lewis (1989),

income of private Beenstock/Dickinson/Khajuria
(1986), Truett/Truett (1990),
households Browne/Kim(1993), Outreville
(1996), DePamphilis (1977)

Price Aggregate price + ** Beenstock/Dickinson/Khajuria

level (1986)

Monetary Inflation - ** schwebler (1994), Beck/Webb

stability (2002), Browne/Kim (1993),
Diacon (1980)
Interest rate •Long-term interest ? # Beenstock/Dickinson/Khajuria
rate (real or nominal (1986), Outreville (1996)
terms)
• Ratio of quasi- + *** Outreville (1996)
Development of the money (M2-M1) to
financial market and the broad definition
banking sector of money (M2)
• Total claims of
deposit money banks
on domestic non-
financial sectors as a Beck/Webb (2002)
share of GDP + ***

Labor market •Unemployment rate - ** Schwebler (1984)

•Government social
security expenditure - ** Browne/Kim (1993), Beck/ Webb
in relation to GDP (2002), Beenstock/
• Average income tax Dickinson/Khajuria (1986),
rate DePamphilis (1977),
• Ratio of net to - *** Diacon (1980)
gross income
+ ** Müller (1998)

Social security • Gini coefficient ? (-),

*** DePamphilis (1977),
• Theil's inequality ? # Diacon (1980)
coefficient

•Private saving + ** Schwebler(1984),

Saving Beck/Webb (2002)

•Competitive • Dummy variable ? (-),

Macroeconomic situation indicating whether the ***
market is a Outreville (1996)
variable monopolistic one
• Dummy variable ?
26 | P a g e indicating whether #
foreign companies are
writing business in the
market
Socio-psychological •Moods(present) •Allensbach indicator + *** Schwebler (1984), GDV (1983)
 Source ;( Swiss Re Economic Research & Consulting, (2003) Page; 8-9)

(A) +, represent positive correlation,

-, represent negative correlation,

?,represents Direction of correlation not clear from theory

(B) ***, represent statistically significant and coincides with the expected sign, robust result in conjunction with
further explanatory factors

**, represents statistically significant and coincides with the expected sign, but not very robust result in conjunction
with further explanatory factors,

# represents not statistically significant. The result is given in parentheses if the expected sign cannot be
unambiguously determined.

2.4 Conceptual framework

The conceptual framework of this study is developed based on the two

major determinants. Which are economic and socio demographic, Different
studies that are mentioned above suggests that several factors such income,
inflation, real interest rate, banking sector development, savings, unemployment,
pension, price of insurance, education, life expectance, dependency ratio and age
are considered important factors that determine life insurance demand.

However due to constraints of data and time, the researcher selects

only five main determinants to test empirically in Ethiopian context, these are
income/ GDP per capita, inflation, real interest rate, life expectancy and
dependency ratio. So, by taking this work and clusters as a base, and to test in the
Ethiopian context, the researcher developed the following conceptual framework.

Dependant variable

Demand for life insurance

27 | P a g e
 Independent variable

Economic Variables

• Income • Inflation • Real Interest Rate

Socio-Demographic Variables

• Life expectancy • Dependency Ratio

Figure 2.1: Conceptual frameworks for the study

28 | P a g e
CHAPTER THREE
3. DATA SOURCES, METHODOLOGY AND
MODEL SPECIFICATION

3.1 Data Type and Source

This study fully relies on secondary data. Data is collected from annual
reports and publications of National Bank of Ethiopia (NBE) and Central
Statistics Authority (CSA). Annual time series data from 1983 to 2018 has been
used to analyze the relationship between life insurance demand and its
determinants.
3.2 Model specification
To find the determinant factors for life insurance demand, the researcher
adapt the following log linear model from Browne and Kim (1993), Li
et al (2007) and Nestrova (2008).
The following regression model will be used in estimating the
parameters for the variables in this study.

𝑙𝑜𝑔𝐿𝐼𝑃 = 𝛼 + 𝛽1 𝑙𝑜𝑔𝐺𝐷𝑃 + 𝛽2 𝐶𝑃𝐼 + 𝛽3 𝑅𝐼𝑅 + 𝛽4

𝑙𝑜𝑔𝐿𝐸𝑋𝑃 + 𝛽5𝑙𝑜𝑔𝐷R + Ɛ
Where
LIP - Life insurance premium per capital (life insurance density)
GDP – GDP per capital
CPI – inflation
RIR– Real interest rate
 DR – dependency ratio
LEXP – life Expectancy
Ɛ - Error term
As mentioned by Gujarati (2004), estimating log-linear
model; the coefficient can be used to determine the impact of your independent
variable(X) on your dependant variable(Y). The coefficient in a log-linear
model represents the estimated percent change in your dependant variable for a
unit change in your independent variable.
According to Li et al (2007) and Nestrova (2008) the regression model
used in estimating the parameters for the variables Inflation and real interest rate
are not specified in log form because it is not possible to make them log since there
are negative values.

3.3Method of analysis

In order to minimize the sum of squared errors, the researcher applied

OLS method of estimation. OLS method of estimation was applied by various
researchers in order to minimize squared error.

Different tests has been used to make the data ready for analysis and
to get reliable output from the research. The tests are intended to check whether
classical linear regression model (CLRM) assumptions, i.e. the ordinary least
squares (OLS) assumptions are fulfilled or not when the independent variables are
regressed against the dependent variables.

30 | P a g e
 CHAPTER FOUR

DATA ANALYSIS and INTERPRETATION

In the fourth chapter of the research, the data that are defined in chapter three
are tested for tpresence of econometric problems, presented and analyzed. Besides,
in each sub section brief interpretations are enclosed to explain the results
obtained. The first section deals with the tests for unit root, fulfillment of basic
OLS assumptions. secondly, the discussion of the summary of descriptive statistics
results of all variables, thirdly the illustration and discussion of the correlation
analysis among basic variables, and finally detail discussions on the regression
results of various determinants of life insurance demand would be presented.

The data and method of analysis

The main data that are used in this study are secondary which are obtained from
National Bank of Ethiopia (NBE), Ministry of finance and economic
development(MOFED) and Central Statistics Authority (CSA).Annual time
series data from 1983 to 2017 will be used to analyze the relationship between
life insurance demand and its determinants.

To estimate the life insurance demand equation in Ethiopia , this study uses
secondary data by considering life insurance premium as dependant [Link]
explanatory variables include gross domestic per capita , life expectancy
represented , dependency ratio, inflation and real interest rate.

To manage the magnitude of the figure to reduce the effect of hetroscedasticy and
to see the effect each independent variable on the dependant using percentage the
data series are transformed to their natural logarithm form.

4.1 Unit Root Test Results

Stationary means a process whose mean and variance are

constant over time. Unit root tests are used to test whether a series is stationary or
not. If a series is not stationary, it is said to have a unit root. For a series to be
stationary its mean and variance have to be constant over time and the value of the
co variance between the two time periods must depends only on the distance or gap
or lag between the two time periods and not the actual time at which the covariance
is computed. As mentioned by Gujarati (2004), a study on the stationarity of

31 | P a g e
variables is important for the reason that it makes it easier the study for the
behaviour of variables in the long run. If a time series is non stationary, the
behaviour of the series can be studied only for the time period under
consideration. In general, it is not possible to generalize it to other time periods.
Therefore, for the purpose of forecasting, non stationary time series are of little
practical value. These call for the need to test for stationarity of the series prior to
detail analysis of the variables.

There are several test that are used to test for unit root. Dick
fuller and Augmented dicky fuller(ADF) have been widely used applied to test for
stationary. As mentioned by gujarati 2004, the DF assume that the error term ei is
uncorrelated which is to restrictive for econometric application however the
augmented dicky fuller test account for possibilities that ei is not white noise. For
this reason the researcher use ADF test.

Variable with intercept with trend and intercept

At level At first difference At level At first difference

GdpPci -2.551*** -6.2331 *** - 2.689* -5.6499***

LIEX -1.0365 -4.5289*** -3.5559** -4.2356**

DP -1.379 -4.459 *** -2.2350 -6.2315***

CPI -0.552 - 2.531 ** -2.0215 -5.2356***

RIR -1.636 * -2.0568 ** -1.0689 -3.2164**

*, ** and *** indicates the rejection of the null hypothesis i.e existence of unit root
at 10%, 5% and 1% respectively. The null hypothesis is that variable is not
stationary (has a unit root).

If t* > ADF critical value, ==> not reject null hypothesis, i.e., unit root exists.

If t* < ADF critical value, ==> reject null hypothesis, i.e., unit root does not exist

32 | P a g e
The table above indicates, the computed ADF test-statistic value is smaller than the
critical values at 10%, 5%, 1% significant level, respectively), therefore we can
reject Ho that variables have unit root. It means the GDP, CPI, RIR, LEXP and DR
series do not have unit root problem and the series are stationary at 1%, 10% and
5% significant level.

4.2 Tests for the CLRM Assumptions

Different tests such as normality, multicollinarity, hetroscedasticity and

auto correlation were run to make the data ready for analysis and to get reliable
output from the research. This tests are intended to check whether classical linear
regression model (CLRM) assumptions,which is the OLS assumptions, are
fulfilled or not when the independent variables are regressed against the dependent
variables. The implication of the test, decision rules and test results are discussed
in the upcoming subsection.

Test of normality

We test normality because of an outlier in an observation that lies

an abnormal distance from other values in a random sample from a population. We
must be extremely mindful of possible outliers and their adverse effects during any
attempt to measure the relationship between two continuous variables. For this
reason, the researcher use Bera-Jarque probalbility statistics in order to test for
normality.

The figure below shows a bell-shaped distribution of the residuals but not a perfect
bell-shaped. X-axis shows the residuals, whereas Y-axis represents the density of
the data set. Thus this histogram plot confirms the normality test results from the
two tests in this article.

In addition to this the normality test for this study shown in Appendix B
shows the Bera-Jarque statistic p-value was found to be 0.230225 which is greater
than 0.05 not reject the null hypothesis that indicates the residual values is
normally distributed.

33 | P a g e
 1
Density
.5 0

-1 -.5 0 .5 1
Residuals

Fig; histogram for normality test generated using STATA 13

4.2.2 Test of Multicollinearity

Multicollinearity (MC) is a situation that may be observed in

multiple linear regressions. There are basically two reasons for Multicollinarity.
Either there is little variation in the explanatory variables, or there is high
correlation among the explanatory [Link], the impact of multicollinearity
is to reduce any individual explanatory variable‘s predictive power by the extent to
which it is associated with the other explanatory variables. That is, none of the
predictor variables may contribute uniquely and significantly to the prediction
model after the other independent variables is included.

As a measure of the Muticollinerity in this model the researcher use both

Variance Inflation Factor (VIF) and correlation matrix.

34 | P a g e
 Logpci CPI RIR LogDR LogLEXP

Logpci 1.0000

CPI 0.3024 1.0000

RIR -0.3554 -0.1098 1.0000

LogDR -0.6873 0.0563 0.2329 1.0000

LogLEXP 0.6029 0.2364 -0.2771 -0.6030 1.0000

Source; generated from STATA 13

The above table presents the correlation matrix for all explanatory variables used
in the analysis. Low inter correlations among the explanatory variables used in the
regressions indicate no reason to suspect serious multicollinearity. According to
Anderson W. (2010) (as cited by amrot yilma) multicollinearity is a potential
problem if the absolute value of the sample correlation coefficient exceeds .7 for
any two of the independent variables. There for the explanatory variables included
in the study were not substantially correlated with each other.

As an alternative test for multicollinearity the researcher use variance influence

factor (VIF)

Variable VIF 1/VIF

logpci 6.45 0.155039

loglife 5.88 0.169926
logde 2.37 0.422618
CPI 1.29 0.776171
RIR 1.16 0.863148

Mean VIF 3.43

35 | P a g e
As mentioned by Gujarati 2004, if the VIF > 10 indicates that MC is unduly
affecting the least squares estimates of the regression coefficients. That is, as a rule
of thumb, if VIF of a variable exceeds 10 the variable is said to be highly collinear
and the reverse is true. Based on the above table we can say that their is no
multicollinearity because the VIF is 3.43

Autocorrelation

Autocorrelation refers to the existence of a relationship between error terms across

observations of a time series. Error covariance are therefore different from zero.
This constitutes a violation to one of the assumptions of the classical linear model.

Most of the CLRM assumptions that allow econometricians to prove the

desirable properties of the OLS estimators (the Gauss-Markov theorem) directly
involve characteristics of the error term. One of the CLRM assumptions deals with
the relationship between values of the error term. Specifically, the CLRM assumes
there’s no autocorrelation. i.e No autocorrelation refers to a situation in which no
identifiable relationship exists between the values of the error term. .

There are a very popular test for autocorrelation which is the Durbin
Watson and breusch Godfrey,Whereas the Durbin-Watson Test is restricted to
detecting first-order autoregression, the Breusch-Godfrey (BG) Test can detect
autocorrelation up to any predesignated order [Link] this reason the researcher uses
breusch-Godfrey test to detects the presence of autocorrelation.

The decision rule is to accept null hypothesis of no autocorrelation among the

error term across the observation of time series if the probabilities of the F
statistics of the intermediary equation are greater than 0.05, which depict the
absence of auto correlation.
Breusch-Godfrey LM test for autocorrelation
---------------------------------------------------------------------------
lags(p) | chi2 df Prob > chi2
-------------+-------------------------------------------------------------
1 | 1.820 1 0.1773
---------------------------------------------------------------------------
H0: no serial correlation

For this reason we failed to reject the null hypothesis which is no serial correlation,
since probability value is greater than 0.05. Therefore, Ho is not rejected, meaning
autocorrelation is absent.

36 | P a g e
Heteroskedasticity test

The other important assumption on which OLS is based is on the assumption of

constant variance of the random variable (error term).

In order to ensure that the residuals are randomly dispersed throughout

the range of the independent variable, Heteroskedasticity test was used. The
variance of the error should be homosekedastic i.e constant for all values of the
independent variable.

Heteroskedasticity in this study is diagnosed using the Breusch

Pagan-Godfrey [Link] decision rule is to reject the null hypothesis if
the probability of the F-statistic and observed R2 are less than 0.05,
meaning Heteroskedasticity is present.

Breusch-Pagan / Cook-Weisberg test for heteroskedasticity

Ho: Constant variance
Variables: fitted values of lnpre

chi2(1) = 0.35
Prob > chi2 = 0.5564
From our test result both are above 0.05 indicating the absence of
Heteroskdasticity.

37 | P a g e
3.2 Descriptive Statistics

In order to understand the structure,pattern and disrbution of the data, it is

important to look at summery statistics which mainly include mean,standard
deviation, coefficient of variation,skewness and kurtosis.

The following summary of descriptive statistics of all dependent and

independent variables gives the general distribution of the data set to examine
the determinants of life insurance demand in Ethiopia from 1983 to 2018. The
table reports the mean, maximum, minimum, standard deviation, and the
number of observations for life insurance per capita and five explanatory
variables for 36 observations. The mean is the sum of the observations divided by
the total number of observations. The standard deviation is the squared root of
the variance, Indicates how close the data is to the mean. Max is the largest value
in the variable. Min is the lowest value in the variable.

Table 3.1 Summary of descriptive statistics

Variable Obs Mean [Link] Min Max

LIP 36 .8316 1.02 .09325 3.478

RIR 36 -3.87 14.42 -51.2 19

DR 36 89.12 5.28 75 99.13

CPI 36 9.89 15.17 -11.823 55.24

GDP PCI 36 304.72 198.2 117 783

Life EX 36 53.21 6.19 43.84 66.26

Source; generated using STATA 14.1

38 | P a g e
 The dependent variable in this study is life insurance demand, the variable
used in examining the determinants of the demand for life insurance is life
insurance density or Premium per capita. The table above shows that the average
life insurance density is $0.8316. i.e. from the total population each individual on
average spends $0.8316, a maximum of $3.478 and a minimum of $
0.09325annually on life insurance. The standard deviation of life insurance per
capita was 1.02 percent, suggesting that LIP was not highly dispersed or not far
from the mean value.

GDP per capita is used as a proxy for income and it is measured as the GDP
at market price divided by the number of population that represents disposable
personal income. The table above shows that the average GDP per capita for 36
years is $304.72 the maximum amount of GDP per capita is $783 and a minimum
amount of GDP per capita is $117. The standard deviation of GDP per capita was
198.2 percent, suggesting that GDP per capita was highly dispersed or far from
the mean.

Consumer Price Index(CPI) rate for 36 years is 9.89, the maximum amount of
CPI is 55.24 and a minimum amount of CPI is negative 11.82. The standard
deviation of CPI was 15.17 percent, suggesting that CPI was not highly dispersed
or far from the mean. Real interest rate is calculated by subtracting inflation from
deposit interest rate.

Real interest rate(RIR) is calculated by subtracting inflation from deposit

interest rate, the average for 36 years is -3.87, the maximum amount of RIR is 19
and a minimum amount of RIR is - 51.2. The standard deviation of RIR was 14.42
percent, suggesting that RIR was not highly dispersed or not far from the mean.

Life expectancy is measured by the number of years that the average

individual in a country is expected to live. The table above shows that, the
average Life expectancy was 53.21 years, the maximum was 66.26 years and a
minimum LIFE EX was 43.84 years. The standard deviation of LIFE EX was 6.19,
suggesting that Life EX was not highly dispersed or not far from the mean.

Dependence ratio is measured by the Number of family members

dependent on the main source of income. The table above shows that, the
39 | P a g e
average DR was 89.12, the maximum DR was 99.13 and a minimum DR was 75.
The standard deviation of DR was 5.28, suggesting that DR was not highly
dispersed.

As mentioned by gupta (1973), that the larger standard deviation implies the
more the data variable are dispersed as a result the variable are less
representative by their mean. Gupta also argues that if the coefficent of variation
is above 100 percent indicates each values are more disprsed and the cofficent of
variation means the more the more consistent of measure of the central
tendency.

In order to apply the statistical test the distributional values should be fulfill
the normal distribution assumption with the stated that kurtosis and skewness 3
and 0 respectively. Gupta(1983) argues that, kurtosis and skewness exhibits the
extent to which the curve is more peaked or more flat topped than, the normal
curve skwness on the other hand measures the scaterdeness or symmetry of the
distribution.

Trends in gross premium from 2000 up to 2018

350,000,000.00

300,000,000.00

250,000,000.00

200,000,000.00
GROSS PREMIUM

150,000,000.00

100,000,000.00

50,000,000.00

0.00
00 001 002 003 004 005 006 007 008 009 010 011 012 013 014 015 016 017 018
20 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2
YEAR

Source; from national bank data

From the past decades gross premium has been increasing as the graph shows
above. However starting from 2016 to 2018 their were a decrease on gross
40 | P a g e
premium. The main objective of the researcher is to determine the factor that
affect the gross premium, i.e factors determine life insurance deman

41 | P a g e
Autocorrelation

Autocorrelation refers to the existence of a relationship between error terms across

observations of a time series. Error covariance are therefore different from zero.
This constitutes a violation to one of the assumptions of the classical linear model.

Most of the CLRM assumptions that allow econometricians to prove the

desirable properties of the OLS estimators (the Gauss-Markov theorem) directly
involve characteristics of the error term. One of the CLRM assumptions deals with
the relationship between values of the error term. Specifically, the CLRM assumes
there’s no autocorrelation. i.e No autocorrelation refers to a situation in which no
identifiable relationship exists between the values of the error term. .

There are a very popular test for autocorrelation which is the Durbin
Watson and breusch Godfrey,Whereas the Durbin-Watson Test is restricted to
detecting first-order autoregression, the Breusch-Godfrey (BG) Test can detect
autocorrelation up to any predesignated order [Link] this reason the researcher uses
breusch-Godfrey test to detects the presence of autocorrelation.

The decision rule is to accept null hypothesis of no autocorrelation among the

error term across the observation of time series if the probabilities of the F
statistics of the intermediary equation are greater than 0.05, which depict the
absence of auto correlation.
Breusch-Godfrey LM test for autocorrelation
---------------------------------------------------------------------------
lags(p) | chi2 df Prob > chi2
-------------+-------------------------------------------------------------
1 | 1.820 1 0.1773
---------------------------------------------------------------------------
H0: no serial correlation

For this reason we failed to reject the null hypothesis which is no serial correlation,
since probability value is greater than 0.05. Therefore, Ho is not rejected, meaning
autocorrelation is absent.

Heteroskedasticity test

The other important assumption on which OLS is based is on the assumption of

constant variance of the random variable (error term).
 In order to ensure that the residuals are randomly dispersed throughout
the range of the independent variable, Heteroskedasticity test was used. The
variance of the error should be homosekedastic i.e constant for all values of the
independent variable.

Heteroskedasticity in this study is diagnosed using the Breusch

Pagan-Godfrey [Link] decision rule is to reject the null hypothesis if
the probability of the F-statistic and observed R2 are less than 0.05,
meaning Heteroskedasticity is present.

Breusch-Pagan / Cook-Weisberg test for heteroskedasticity

Ho: Constant variance
Variables: fitted values of lnpre

chi2(1) = 0.35
Prob > chi2 = 0.5564
From our test result both are above 0.05 indicating the absence of
Heteroskdasticity.

One of the CLRM assumption is that, the disturbance term related to any other
observation is not influenced by the disturbance term to any other observation
E(ui,uj)=0

4.5 Regression analysis

Table 4.7 displays the results of Ordinary Least Square (OLS) estimation for the
initial test equation. From the Table OLS estimation indicates that GDP per capita,
inflation, real interest rate and life expectance appear to be important variables
associated with the demand for life insurance. However, based on the OLS
estimation Depndency Ratio is not statistically significant variable in determining
life insurance demand in Ethiopia.

43 | P a g e
Independent variable coefficient Std error T-statistic prob

LogPCI .7675868 .2297767 3.34 0.003

LogDR -2.152114 1.317024 -1.63 0.113

LogLIEX .014259 .0067075 2.13 0.062

CPI -.1016109 .0389755 -2.61 0.004

RIR .1000269 .0374267 2.67 0.002

cons -0.423636 - 0.943126 -2.354023 0.0012

R-squared = 0.9425 standard error of the regeression=0.005602

Adjusted r squared = 0.9255

Source ; generated from STATA 13

The adjusted R squared for the model is 0.9255 which indicates that
about 92.55 percent of demand for life insurance is explained by the selected five
factors ( GDP per capita, inflation, real interest rate,life expectancy and
dependence ratio). In other words, about 92.55 percent of the change in the life
insurance demand is explained by the independent variables that are included in
the model and the remaining 7.45 % change in life insurance demand, is because of
other factors that outside the model. From Table above the researcher have
derived the following estimated regression equation.

LPI = -0.423636 +0 .7675868Gdppci - 0.1016109 CPI + 0.1000269 RIR

+.014259 LIEX

The t-statistics and (p-value) show that the independent variables such as GDP per
capita, inflation(CPI) and real interest rate are statistically significant at 5 percent
significance level. The coefficient for LIEX is statistically significant at 10 percent
significant level.

44 | P a g e
 In conclusion, statistically significant variables (GDP, CPI, RIR and
LEXP) could determine life insurance demand for Ethiopia. While dependence
ratio do not have a statistically significant relationship with life insurance demand
because its p-value is greater than 10 percent

4.5.1 Hypothesis testing and discussion of results

1. Life Insurance and Level of Income

Research hypothesis 1 (H1), predicts a positive relationship between life

insurance demand and level of income. Similar to the hypothesis, the regression
output showed positive and strongly statistically significant relationship between
life insurance demand and level of income. So we can not reject H1 that there is
positive and statistically significant relationship between income level and demand
for life insurance. This positive relationship implies that as income of the society
arises, life insurance become more affordable. The coefficient at 5 percent
significant level for GDP per capita indicated that .7675868. This implies that a
1% increase in GDP per capita would result on 76.75 percent increase on LIP. This
result agree with various previous research findings like Dickinson, Khajuria
(1986),, Browne and Kim (1993), and Outreville (1996), Beck and Webb, (2003),
Abenezer(2015), Celik and amrot yilma(2012).

They confirmed the significant positive relationship of income and demand

for life insurance. In this situation, as PCI rises, insurance become more affordable.
This study find out that GDP per capita (t-stat =3.34), which is the highest value in
this test. This implies that GDP per capita is the most important factor that
influences demand for life insurance in Ethiopia.

B. Life Insurance and Inflation

Research hypothesis 2(H2), predicts a negative relationship between inflation

and Life insurance demand. Similar to the hypothesis, the regression showed
significant negative relationship between inflation and Life insurance demand. So
we can not reject H2 that there is negative and statistically significant relationship
between inflation and demand for life insurance. It implies that when inflation
45 | P a g e
increase, it tends to discourage people from owning life insurance policies or they
do not afford to pay the cost of life insurance and the cost of life insurance
protection would increase with anticipated inflation resulting in a decline in life
insurance sales. The coefficient is significant at 5 percent level for inflation
(-.1016109 ). This implies that a 1% increase in inflation would result on 10.16
percent decrease on [Link] agree with various previous research findings like
Beck and Webb (2003), Nesterova (2008), Kayali (2009), This study found out that
inflation (t-stat =-2.61) which is significant.

C. Life Insurance and Real Interest Rate

Research hypothesis 3 predicts a positive relationship between life insurance

demand and Real interest rate. Similar to the hypothesis, the regression showed
positive and statistically significant relationship between life insurance demand
and RIR. So we can not reject H3 that there is positive and statistically significant
relationship between real interest rate and demand for life insurance. The
positive relationship between real interest rate and life insurance demand implies
that a higher real interest rate increases life insurer‘s investment returns and
profitability, it ultimately leads to high life insurance demand. The coefficient at 5
percent significant level for RIR indicated that .1000269 . This implies that a 1%
increase in RIR would result on 10 percent increase on LIP. This result agrees with
various previous research findings like Beck and Webb (2003)and Schwebler
(1984) , both of them confirmed the significant positive relationship of RIR and
demand for life insurance. In this situation, as RIR arises, insurance become more
affordable. This study found out that RIR (t-stat =2.67), High real interest rate
increases life insurer‘s investment returns and so profitability, in turn offering
improved profitability of financial relative to real investments for potential
purchasers of life insurance policies.

D. Life Insurance and Life Expectancy

Research hypothesis 4 shows a positive relationship between life

insurance demand and life expectancy. Similar to the hypothesis, the regression
output showed positive and statistically significant positive relationship between
life insurance demand and life expectancy. So we can not reject H5 that there is
positive and statistically significant relationship between life expectancy and
demand for life insurance. This positive relationship implies that Societies with

46 | P a g e
longer life expectancies should have higher savings through life insurance. The
coefficient at 1 percent significant level for life expectancy indicated that .014259 .
This implies that a 1 year increase in life expectancy would result on 1 percent
increase on Life insurance per capita. This result agree with various previous
research findings like Beck/Webb (2002), Beenstock/ Dickinson/Khajua (1996),
Outreville (1996).They confirmed the significant positive relationship of life
expectancy and demand for life insurance. In this situation, as LEXP arises,
insurance become more affordable. The study found out that life expectancy (t-stat
=2.13) which is the lowest value in this test. It means that life expectance is the
least important factor that influence demand for life insurance.

E. Life Insurance and Dependence ratio

The regression output shows no statistically significant relationship between

Dependence ratio and life insurance demand. However some researcher like
kedir(2015) accept that dependcy ratio have statistical relationship, the reason for
the difference and no evidence has been found that the life insurance industry
suffered an adverse impact over the periods of high DR in Ethiopia. So we can
reject H6 that there is

positive and statistically significant relationship between Dependency ratio and

demand for life insurance. This implies that an increase or decrease in DR have no
impact on Life insurance per capita.

47 | P a g e
 CHAPTER FIVE

Conclusions and Recommendations

The main objective of this chapter is to present the overall overview of the research
by summing the main findings of the analysis part and give recommendations for
future researchers. Section one briefly sum up the overviews of the study and its
main findings. In section two based on the study finding the researcher highlight
some recommendations to concerning stakeholders. At the end, recommendations
for future researchers are presented in section three.

5.1 Conclusions

The main objective of the study is to examine the relationship between life
insurance demand and its main determinants.

Based on the analysis made in previous chapter on factors that determine life
insurance demand the following conclusions are drawn.

  There is significant positive relationship between income and demand

for life insurance. In this situation, as income rises, insurance become more
affordable. This study found out that GDP per capita is the most important
factor that influence demand for life insurance. Similar to the hypothesis, the
regression showed positive and statistically significant relationship between
life insurance demand and level of income. The coefficient at 1 percent
significant level for GDP per capita would have This implies that a 1%

48 | P a g e
 increase in GDP per capita would result on 76 percent increase on life
insurance per capita.

  There is a significant positive relationship between Real interest rate and

demand for life insurance. Similar to the hypothesis, the regression showed
positive and statistically significant relationship between life insurance demand
and RIR. The positive relationship between real interest rate and life insurance
demand implies that higher real interest rate increases life insurer‘s investment
returns and profitability, it ultimately leads to high life insurance demand. The
coefficient at 5 percent significant level for RIR indicated that .1000269. This
means that a 1% increase in RIR result on 10 percent increase on Life
insurance per capita.

 It is found out that there is a significant negative relationship

between inflation and demand for life insurance. Similar to the hypothesis, the
regression showed significant negative relationship between inflation and Life
insurance demand. It means that when inflation increase through time, it tends
to discourage people from owning life insurance [Link] result shows that a
1% increase in inflation would result on 10.16 percent decrease on Life
insurance per capita.

  Life expectancy is the least important factor that influence demand for life
insurance. Similar to the hypothesis, the regression showed positive and
statistically significant relationship between life insurance demand and life
expectancy. The implies that a 1 year increase in life expectance would result
on 1.4 percent increase on Life insurance per capita.

 The study found that their is no significant relationship between DR and

life insurance demand. However according to kedir(2014), found out that their
is positive and statistical significant relationship between dependency ratio and
life insurance demand. The difference in result exist because researcher kedir
uses IMF data on the projection of dependency [Link] this research
uses CSA data.

49 | P a g e


Reccomendation

This study has tried to identify factor that determine the demand for life
insurance based on the result obtained from the study. The following
recommendations may be drawn from it.

The study reveals that the impact of income that have a significant impact
on determining life insurane demand in Ethiopia. Therfore the government have to
give much more emphasis for increasing GDP per capita through more
investment; and decreasing unemployment rate of the country so that the per
capita GDP will be increased.

The negative relation between inflation and life insurance demand implies that
the government should have to decrease the level of inflation through direct and
indirect intervention in the price mechanism of the country. Because as inflation
becomes more and more, it tends to discourage people from owning life insurance
policies.

Based on the regression result it shows that their is a postive relation between life
insurane demand and life expectancy as societies with longer life expectancies
should have higher savings through life insurance vehicles and more demand for
annuities. Therefore, it will be better if the government will develop a nationwide
structural plan that gives much emphasis for health promotion and disease
prevention activities through hospitals and health centers expansion that facilitate
especially preventive mechanisms by giving training and professional advice to the
community at largegeneral public. Thus, the National bank of Ethiopia and the
managements of insurance

companies should further enhance public‘s awareness and financial literacy by

using

different approaches. For example they should use media, internet and billboard as
a

medium of promoting the life insurance and made a program such true story
regarding

the life insurance. This program will give big picture to society about how
important

50 | P a g e
51 | P a g e
Reference
Abenezer shiferaw(2016). Factors Affecting Life Insurance Demand: A case study on
(Ethiopian Insurance Corporation) EIC, St. Mary University.

Amrot yilma (2014).Determinants of Life Insurance Demand in Ethiopia, college of

business and Economics, department of Accounting Addis Ababa University.

African insurance organization report 2014.

Baldwin, B.G. (2002). The new life insurance investment,.

Beck, T. and Webb(2002). Economic, Demographic, and Institutional Determinants of

Life Insurance Consumption across Countries.

Browne M. J. and K. Kim( 1993). an International Analysis of Life Insurance Demand,

Journal of Risk and Insurance.

Cummins(1973). An Econometric Model of the Life Insurance Sector of the U.S.

Economy, Journal of Risk and Insurance.

Curak [Link] (2013). The Effect of Social and Demographic Factors on Life Insurance
Demand in Croatia , Journal of Risk and Insurance.

Depamphilis(1977). Variation in Individual Life Insurance Premium Revenues.

Fischer (1973). A Life Cycle Model of Life Insurance Purchases, International Economic.

Huggins,Kenneth (1986). Customer service in insurance principle and practice.

Hailu Zeleke (2007). Insurance in Ethiopia: Historical Development, Present Status and
Future Challenges.

Hanksson, Nils H. (1969). Optimal Investment and Consumption Strategies Under Risk,
and Under Uncertain Lifetime and Insurance.

Lewis, Frank, D. (1989). Dependents and the Demand for Life Insurance.

Madheswaran(2007).Life Insurance Demand Determinants for Selected Asian Economies

and India.

Mahdzan & Victorian (2013). determinants of life insurance demand among life
insurance policyholders of five major life insurance companies in Kuala Lumpur and
Malaysia.

Nesterova (2008). Determinants of the demand for life insurance: Evidence from
selected CIS and CEE Countries. National University “Kyiv-Mohyla Academy”.

Outreville, Francois J. (1992). The Relationship between Insurance, Financial

Development and Market Structure in Developing Countries.

Pissariades, 1980. The Wealth-Age Relation with Life Insurance.

Roman Gebreyes (2011). Determinants of life insurance demand in Ethiopia, college of

business and Economics.

Schwebler, R. (1984). Identification and quantification of the overall impact of the

national economy on life insurance in the Federal Republic of Germany, Papers on Risk
and Insurance.

Sandra G. and Robert (1987). Life insurance theory and practice.

Swiss Re (2003). Determinants of life insurance demand,.

Truett, D. B. and Truett, (1990). The Demand for Life Insurance in Mexico and the United
States.

Ward, D. and R. Zurbruegg (2002). Law, Politics and Life Insurance.

Yaari, Menahem E. (1965). Uncertain Lifetime, Life Insurance, and the Theory of the
Consumer.

Zhou (1998). Determinants of life insurance demand for 29 different regions and 14
large cities in China.

53 | P a g e
Appendix A: Unit root tests (Dickey-Fuller test for unit root)
Unit root test at level and intercept

Null Hypothesis: GDP has a unit root

Dfuller gdp, noconstant lags(0)

---------- Interpolated Dickey-Fuller ---------

Test 1% Critical 5% Critical 10% Critical
Statistic Value Value Value
------------------------------------------------------------------------------
Z(t) -2.551 -2.644 -1.950 -1.604

Null hypothesis: life expectancy has a unit root

Dfuller lexp, noconstant lags(0)
---------- Interpolated Dickey-Fuller ---------
Test 1% Critical 5% Critical 10% Critical
Statistic Value Value Value
------------------------------------------------------------------------------
Z(t) -1.0365 -2.646 -1.950 -1.604

Null hypothesis ; dependence ratio has a unit root

dfuller dp, noconstant lags(0)
---------- Interpolated Dickey-Fuller ---------
Test 1% Critical 5% Critical 10% Critical
Statistic Value Value Value
------------------------------------------------------------------------------
Z(t) -1.379 -2.644 -1.950 -1.604

null hypothesis ; inflation has a unit root

dfuller cpi, noconstant lags(0)

---------- Interpolated Dickey-Fuller ---------

Test 1% Critical 5% Critical 10% Critica

Statistic Value Value Value

Z(t) -0.552 -2.644 -1.950 -1.604

Null hypothesis; real interest rate has a unit root

dfuller cpi, noconstant lags(0)

---------- Interpolated Dickey-Fuller ---------

Test 1% Critical 5% Critical 10% Critical
Statistic Value Value Value
------------------------------------------------------------------------------
Z(t) -1.636 -2.646 -1.950 -1.604

Unit root test at 1st difference and Intercept

Null Hypothesis: d(GDP) has a unit root

dfuller dgdp, noconstant lags(0)

---------- Interpolated Dickey-Fuller ---------

Test 1% Critical 5% Critical 10% Critical
Statistic Value Value Value
------------------------------------------------------------------------------
Z(t) -6.2331 -3.646 -2.960 -2.677

Null hypothesis ; d(LIEX) has a unit root

dfuller dgdp, noconstant lags(0)

---------- Interpolated Dickey-Fuller ---------

Test 1% Critical 5% Critical 10% Critical
Statistic Value Value Value
------------------------------------------------------------------------------
Z(t) -4.5289 -3.646 -2.960 -2.677

Null hypothesis ; d(DR) has a unit root

---------- Interpolated Dickey-Fuller ---------
Test 1% Critical 5% Critical 10% Critical
Statistic Value Value Value
------------------------------------------------------------------------------
Z(t) -4.459 -3.646 -2.960 -2.677

Null hypothesis ; d(CPI) has a unit root

---------- Interpolated Dickey-Fuller ---------
Test 1% Critical 5% Critical 10% Critical
Statistic Value Value Value
------------------------------------------------------------------------------
Z(t) -2.531 -2.646 -2.960 -2.677

59 | P a g e
 Null hypothesis ; d(RIR) has a unit root
---------- Interpolated Dickey-Fuller ---------
Test 1% Critical 5% Critical 10% Critical
Statistic Value Value Value
------------------------------------------------------------------------------
Z(t) -2.0568 -2.646 -2.960 -2.677

Unit root test at level with trend and intercept

Null hypothesis ; GDP has a unit root

---------- Interpolated Dickey-Fuller ---------

Test 1% Critical 5% Critical 10% Critical
Statistic Value Value Value
------------------------------------------------------------------------------
Z(t) -2.2546 -4.524 -3.560 -2.677

Null hypothesis ; LIEX has a unit root

---------- Interpolated Dickey-Fuller ---------

Test 1% Critical 5% Critical 10% Critical
Statistic Value Value Value
------------------------------------------------------------------------------
Z(t) -3.5559 -4.524 -3.560 -2.677

Null hypothesis ; DR has a unit root

---------- Interpolated Dickey-Fuller ---------
Test 1% Critical 5% Critical 10% Critical
Statistic Value Value Value
------------------------------------------------------------------------------
Z(t) -2.2350 -4.524 -3.560 -2.677

60 | P a g e
 Null hypothesis ; CPI has a unit root
---------- Interpolated Dickey-Fuller ---------
Test 1% Critical 5% Critical 10% Critical
Statistic Value Value Value
Z(t) -2.0215 -4.524 -3.560 -2.677

Null hypothesis ; RIR has a unit root

--------- Interpolated Dickey-Fuller ---------
Test 1% Critical 5% Critical 10% Critical
Statistic Value Value Value
Z(t) -1.0689 -4.524 -3.560 -2.677

Unit root test at 1st difference and trend and intercept

Null hypothesis ; d(GDP) has a unit root
---------- Interpolated Dickey-Fuller ---------
Test 1% Critical 5% Critical 10% Critical
Statistic Value Value Value
Z(t) -5.6499 -5.1243 -3.960 -2.987
Null hypothesis ; d(LIEX) has a unit root
---------- Interpolated Dickey-Fuller ---------
Test 1% Critical 5% Critical 10% Critical
Statistic Value Value Value
Z(t) -4.2356 -5.1243 -3.960 -2.987
Null hypothesis ; d(DR) has a unit root
---------- Interpolated Dickey-Fuller ---------
Test 1% Critical 5% Critical 10% Critical
Statistic Value Value Value
Z(t) -6.2315 -5.1243 -3.960 -2.987
Null hypothesis ; d(CPI) has a unit root
---------- Interpolated Dickey-Fuller ---------
Test 1% Critical 5% Critical 10% Critical
Statistic Value Value Value

Z(t) -5.2356 -5.1243 -3.960 -2.987

Null hypothesis ; d(RIR) has a unit root

---------- Interpolated Dickey-Fuller ---------
Test 1% Critical 5% Critical 10% Critical
Statistic Value Value Value

Z(t) -3.2164 -5.1243 -3.960 -2.987

61 | P a g e
Appendix B
Series: RESID

Sample 1983 2018

Observations 36

Mean 2.05e-11

Median 0.000458

Maximum 0.069452

Minimum -0.050509

Std. Dev. 0.014014

Skewness -0.15689

Kurtosis 4.458567

Jarque-Bera 2.625992

Probability 0.230225

C.

62 | P a g e