financial asset participation by cypriot households-ucy-m.sc. thesis-2004

66
FINANCIAL ASSET PARTICIPATION BY CYPRIOT HOUSEHOLDS FOCUS ON STOCKHOLDING Course: ECO 698: Master Thesis Prepared for: Associate Professor Mr Michael Haliassos at the University of Cyprus. December 2004 Panayiotis Tilliros

Upload: panayiotis-tilliros

Post on 12-Apr-2017

57 views

Category:

Documents


1 download

TRANSCRIPT

Page 1: FINANCIAL ASSET PARTICIPATION BY CYPRIOT HOUSEHOLDS-UCY-M.SC. THESIS-2004

FINANCIAL ASSET PARTICIPATION BY CYPRIOT HOUSEHOLDS

FOCUS ON STOCKHOLDING

Course: ECO 698: Master Thesis

Prepared for: Associate Professor Mr Michael Haliassos at the University of Cyprus.

December 2004

Panayiotis Tilliros

Page 2: FINANCIAL ASSET PARTICIPATION BY CYPRIOT HOUSEHOLDS-UCY-M.SC. THESIS-2004

2

Dedication

This study is dedicated with all my heart, all my soul and all my gratitude to the

memory of my beloved mother, Maria, who worked very hard and very generously for

me and whose passing from earth lighted my path and offered wings to my spirit.

May her memory live on forever and remain an eternal and universal monument to

humanity, goodness, generosity, kindness, altruism, self-sacrifice and truth.

Panayiotis Christou Tilliros

Αφιέρωση

Αυτή η μελέτη αφιερώνεται με όλη μου την καρδιά, όλη μου τη ψυχή και όλη μου την

ευγνωμοσύνη στη μνήμη της πολυαγαπημένης μου μητέρας, Μαρίας, η οποία

δούλεψε πολύ σκληρά και πολύ γενναιόδωρα για μένα και της οποίας το πέρασμα

από τη γη φώτισε το μονοπάτι μου και πρόσφερε φτερά στο πνεύμα μου. Είθε η

μνήμη της να ζήσει για πάντα στο σύμπαν και να παραμείνει ως αιώνιο μνημείο στην

ανθρωπιά, στην καλοσύνη, στη γενναιοδωρία, στην ευγένεια, στον αλτρουισμό, στην

αυτοθυσία και στην αλήθεια.

Παναγιώτης Χρίστου Τήλλυρος

Page 3: FINANCIAL ASSET PARTICIPATION BY CYPRIOT HOUSEHOLDS-UCY-M.SC. THESIS-2004

3

Acknowledgements

I wish to express my unreserved gratitude and thanks to my study supervisor, the

Associate Professor of Economics at the University of Cyprus Mr Michael Haliassos

for his advice and guidance and to the Assistant Professor of Economics Mr Sofronis

Clerides, for his assistance with the statistical package “Stata” with which the

econometric model regressions were performed.

Page 4: FINANCIAL ASSET PARTICIPATION BY CYPRIOT HOUSEHOLDS-UCY-M.SC. THESIS-2004

4

FINANCIAL ASSET PARTICIPATION BY CYPRIOT HOUSEHOLDS

FOCUS ON STOCKHOLDING

Contents Page

Title 1

Dedication 2

Acknowledgements 3

Index 4

Abstract 6

A. Introduction 7

B. Literature Review and Theoretical Aspects of Asset-holding 10

C. Cyprus Survey of Consumer Finances of 1999 and 2002 – Identity 20

D. Overall Financial Assets Participation - Cypriot Household Behaviour 23

1. Financial and Real Assets 23

2. Overview of Asset Holding 23

3. Differences and Similarities with USA 24

4. Financial Assets Age Participation Profile 24

5. Holding of Life insurance 25

6. Life Insurance – Household Characteristics 26

7. Retirement Accounts and Mutual Funds 26

8. Portfolio Breadth and Popularity of Various Financial Assets 27

9. Real Assets 28

10. Real Assets – Homeownership 28

11. Real Assets - Business 28

12. Risk Attitude – Asset Types 29

13. Participation in Risky Assets 29

14. Risk Attitude - Real Assets 30

15. Risk Attitude - Financial Assets 30

E. Financial Assets Participation: Stocks 31

1. Direct Stockholding 31

2. Indirect Stockholding 31

3. Stockholding: Age Profile 32

4. Stockholding: Diversification and Risk Profile 32

Page 5: FINANCIAL ASSET PARTICIPATION BY CYPRIOT HOUSEHOLDS-UCY-M.SC. THESIS-2004

5

5. Direct Investment in Stocks by Income Category 33

F. Bubbles 34

1. Bubbles – Theoretical Aspects and Insights 34

2. Causes of the Cyprus Stock Exchange Bubble in 1999 36

G. Statistical Cross Tabulations 41

1. Statistical Cross Tabulations – Analytical results 41

H. Econometric Analysis - Probit Model Regression 52

1. Probit Model – Methodology 52

2. Probit Model – Variable Definitions 53

3. Model Estimation and Empirical Results 55

I. Conclusion 62

J. Βibliography 64 K. Tables

Table 1: Descriptive Statistics of Population Sample

66

Table 2: Demographic characteristics of Stockholders and Non-stockholders, %

67

Table 3: Participation in Stocks, Life Insurance and Retirement Plans by Age, %

68

Table 4: Participation in Stocks, Life Insurance and Retirement Plans by Education Level, %

69

Table 5: Participation in Stocks, Life Insurance and Retirement Plans by Income Quartile, %

70

Table 6: Participation in Stocks, Life Insurance and Retirement Plans by Financial Asset Quartile, %

71

Table 7: Probit Model for Direct Stock Participation

72

Table 8: Probit Model for Direct and Indirect Stock Participation

73

Table 9: Probit Model for Indirect Stock Participation via Life Insurance 74

Table 10: Probit Model for Indirect Stock Participation via Retirement Plans 75

Page 6: FINANCIAL ASSET PARTICIPATION BY CYPRIOT HOUSEHOLDS-UCY-M.SC. THESIS-2004

6

Abstract

After a brief introduction outlining the contents as well as the importance of household

portfolio theory, this thesis reviews the theoretical aspects of the literature concerning

household stockholding behaviour with special reference to the major issues and the main

findings. This is followed by an analysis of the stockholding behaviour of Cypriot households

based on the 1999 and 2002 Cyprus Surveys of Consumer Finances, which were a

cooperative project between the University and the Central Bank of Cyprus. Subsequently,

the study compares and reports the main results and findings from statistical cross tabulation

analysis regarding the two surveys. It then proceeds with the use of Probit model regressions

of selected Cypriot household characteristics concerning stockholding behaviour. The

econometric analysis is focused upon the factors driving the stockholding behaviour of

Cypriot households by contrasting and comparing the results of the two surveys, which is the

theme of the thesis. Finally, the study concludes that Cypriot household attitudes and

preferences regarding stockholding behaviour are consistent with mainstream theoretical

predictions and empirical findings based on international trends.

* * *

Page 7: FINANCIAL ASSET PARTICIPATION BY CYPRIOT HOUSEHOLDS-UCY-M.SC. THESIS-2004

7

A. Introduction

Household portfolio theory is of extreme importance for policy makers since it encompasses

issues and examines characteristics that affect not only the macro economy and the micro

economy, but quintessentially reflect household financial behaviour, which impacts on the

financial, the monetary and the real sector.

This thesis reviews in a comprehensive manner the theoretical aspects and empirical

findings of the literature concerning household stockholding behaviour with special reference

to the major issues and the main conclusions. Afterwards, it focuses on the stockholding

behaviour of Cypriots and the major factors behind such behaviour. In particular, the study

compares and contrasts the impact of the relevant variables between the 1999 and 2002

Cyprus Surveys of Consumer Finances (CySCF) and the reasons that explain such

differences or developments.

Following this brief introduction (section A), this study reviews, in section B, the theoretical

aspects of the literature concerning household stockholding behaviour with special reference

to the major issues and the main findings. The purpose here is to offer to the interested

reader a bird’s eye view of the areas in which household portfolio theory with particular

emphasis on stockholding has evolved. The wide and cogent spectrum of coverage has been

judged to be necessary in order to offer the feel of the diverse directions in which the portfolio

and stockholding economic literature has evolved. Moreover, the scope of the analysis

imparts an appreciation of the implications of the existing theory and the empirical findings,

which provide the required background familiarisation for the issues to be examined.

Section C refers to the aims the Cyprus Survey of Consumer Finances, which offers a

comprehensive household-level data base concerning the financial real assets and debts of

Cyprus households and constitutes a unique and pioneering work, placing Cyprus with very

few other countries in the field, with similar statistical databases on household portfolios.

The paper then proceeds, in section D, to describe the overall participation by Cypriot

households in financial and real assets. The main aim of this expository analysis is to

delineate the framework and mark the setting within which the overall financial and real asset

behaviour of Cypriot households evolves. This conveys the breadth and the scope of the

Page 8: FINANCIAL ASSET PARTICIPATION BY CYPRIOT HOUSEHOLDS-UCY-M.SC. THESIS-2004

8

general Cypriot household asset behaviour and provides the required background against

which direct stock market participation as well as indirect via investment insurance and

retirement accounts unfolds.

After describing the general asset-holding framework, section E reports some important

findings regarding the stockholding behaviour of Cypriot households based on the 1999 and

2002 Cyprus Surveys of Consumer Finances. These are particularly related to the main

results of the statistical cross tabulations of section G, which constitute part of the original

work carried out in this thesis, besides the econometric part.

Section F refers to some important theoretical aspects of bubbles owing to their special

relevance with the Cypriot stockholding behaviour and the lightning rapidity with which the

equity culture spread in Cyprus under abnormal conditions in 1999.

This is followed by section G, wherein the study compares and reports the main results and

findings from the statistical cross tabulation analysis of the pertinent data from the 1999 and

2002 CySCF, using the statistical package SPSS. The tabulated data are presented in

Tables 1-6 in a clear format that facilitates comparisons and interpretations and enables an

immediate drawing of conclusions. These include descriptive statistics of the population

sample, the demographic characteristics of Cypriot stockholders and non-stockholders, the

age and educational profile of stockholders, as well as the effect of income and wealth on

stockholding.

The study then proceeds with the most original part of the work performed, which

encompasses the use of Probit model regressions of selected Cypriot household

characteristics concerning stockholding behaviour. The analysis of the CySCF data was

performed using the STATA program. Indeed, the analysis in section H is focused upon the

factors driving the stockholding behaviour of Cypriot households, which is the theme of the

thesis. The econometric analysis (Probit model regression in this case) identifies causal

relationships and makes possible the pinpointing of the causative (marginal probability) effect

of each explanatory variable offering the scientific rigour, which often statistical tables lack. In

fact, this section is the most substantive in that it contains the very essence and the

quintessence of the original contribution of this study, i.e. the contrast and comparison of the

differential impact of the relevant variables by noting the statistically significant differences in

Page 9: FINANCIAL ASSET PARTICIPATION BY CYPRIOT HOUSEHOLDS-UCY-M.SC. THESIS-2004

9

the coefficients and explaining the reasons for such differences. This is performed through

the use of interaction terms in the econometric model, which enable us to compare the

possible net change of the participation probability between the 1st and the 2nd survey, based

on certain household characteristics. The estimated Probit models comprise three interaction

terms, one each for income and wealth and another for college education. Moreover, the use

of a yearly dummy, which takes the value of 1 for all observations in the survey of 2002 and

of 0 for all observations in the survey of 1999, allows the pooling together of all the 1999 and

2002 household survey data and the identification of the constant difference in participation

probability among the two survey years, associated with identical household characteristics.

Indeed, the statistical significance of the yearly dummy, captures the expected big jump in the

intercept term, confirming the upward shift in stock participation in 2002 due to the market

entry of various investment companies following the 1999 survey.

Finally, the study concludes (in section I) that Cypriot household attitudes and preferences

regarding stockholding behaviour are consistent with theoretical predictions and empirical

findings based on international trends. Further, the analysis suggests that the typical portrait

of the Cypriot Stock Exchange direct and indirect participant is that of a middle-aged (in his

40’s), married man, who tends to have higher education and relatively high income and

wealth, belonging mostly to a household comprising up to four members, with two income-

earning members in the family in the majority of cases.

Every effort has been made to define and explain concepts in a simple and crystal clear

manner in order to make this study readable and accessible not only to academics and

professionals but also to the members of the average Cypriot household who might be

interested, since the analysis sheds light on important household financial issues and

because a large number of people have been affected by the Cyprus Stock Exchange bubble

of 1999. Indeed, this constitutes an additional reason why a comprehensive coverage was

considered necessary, as this facilitates understanding for the non-expert and enables the

appreciation of the multiple issues involved.

***

Page 10: FINANCIAL ASSET PARTICIPATION BY CYPRIOT HOUSEHOLDS-UCY-M.SC. THESIS-2004

10

B. Literature Review and Theoretical Aspects of Asset-holding

There are a large number of papers and models looking at the issue of asset holding in

general and stockholding in particular. Lifetime consumption and saving models with and

without frictions have been proposed, while other models give emphasis on the issue of

decision making under uncertainty. However, household behaviour does not always accord

with theoretical predictions, which is inevitable for a human science like economics,

examining often erratic human behaviour. In this respect, the question of asymmetric

information also arises, encompassing the issue of adverse selection occurring prior to a

transaction: bad credit risks concern those who most actively seek out a loan and such

potential borrowers are the ones most likely to be selected and produce an undesirable or

adverse outcome; and moral hazard in financial markets, which is the problem created by

asymmetric information after the transaction occurs: this refers to the risk that the borrower

might engage in undesirable (immoral) activities that are likely to turn him into a defaulter.

The literature review that follows will be covering extensively a few papers from the wide

range of bibliography consulted, which have been judged to encompass the main ideas and

theoretical issues involved, while also drawing upon the complete bibliography mentioned.

This is in order to avoid repetition while focusing on the main issues.

In their paper Haliassos and Hassapis (2002) analyse the spread of an “equity culture”

among households on both sides of the Atlantic and the likely changes in the behaviour of

households that enter the stock market in response to information about how to invest in

stocks. In sum, they look at the following issues: 1. The rise in stock market participation in

the 1990’s, and the development of the equity culture, which encourages households to

increase current consumption and their demand for loans, both for asset purchase and for

consumption. 2. The impact of borrowing constraints on entrants and their willingness to

undertake risk depending on their risk aversion. 3. The interaction between stock acquisition,

earnings risk, borrowing constraints, and the accumulation of precautionary wealth as a

buffer against the riskiness of future income streams.

The developments in the 1990’s lowered perceived entry costs relative to the expected

benefits from stockholding, throughout the decade, thus assisting in the continuing increase

in participation. The spread of the equity culture raised the percentage of US households that

Page 11: FINANCIAL ASSET PARTICIPATION BY CYPRIOT HOUSEHOLDS-UCY-M.SC. THESIS-2004

11

hold any stock, directly or indirectly, from 31.6% in 1989 to about 48% in 1998. The main

reasons which encouraged households to enrich their portfolios with risky financial assets

included the privatization of public utilities, the proliferation of mutual funds, the introduction

of new types of retirement accounts, and the unusually strong performance of stock markets.

Moreover, direct advertising (e.g. on mutual funds), employer-sponsored seminars (on

pension accounts), or simply watching the financial success of neighbours were crucial in

expanding the perceived asset choice list and spreading equity culture. In fact, the key to the

spread of equity culture was indirect stockholding via mutual funds and retirement accounts.

Haliassos and Hassapis (2002) find that the equity culture tends to encourage households to

increase current consumption and their demand for loans, both for asset purchase and for

consumption. Equity culture is also likely to enhance the tendency of households to make

larger precautionary adjustments to consumption, financial wealth holding and borrowing in

the face of exogenous earnings risk. This is consistent with recently observed upward trends

in household indebtedness. Households are likely to be more eager to safeguard future

consumption prospects when faced with exogenous increases in earnings risk, despite better

wealth prospects due to the equity premium. The effects of equity culture on the tendency to

hold precautionary assets can be reduced, eliminated, or even reversed by binding borrowing

limits.

By creating greater incentives to borrow, equity culture makes the young more susceptible to

any borrowing constraints. Binding borrowing limits tend to reduce stockholding by new

entrants. Young entrants facing borrowing constraints may not be able to increase their

current consumption demand, despite improved future wealth prospects arising from the

equity premium as a result of stock market participation. The manner in which they behave

depends on the tightness of borrowing limits and on their willingness to undertake risk

depending on their risk aversion. However, such increases in consumption will be optimal

when binding borrowing limits are higher, or when either limits are so high or households so

risk averse that borrowing constraints are not binding. Moreover, the effects of a given

change in credit conditions are influenced by whether or not the household perceives stocks

as part of the asset menu.

Further, the models show a tendency among young entrants to lower their net financial

wealth when they are not faced with borrowing constraints. Without borrowing constraints, the

Page 12: FINANCIAL ASSET PARTICIPATION BY CYPRIOT HOUSEHOLDS-UCY-M.SC. THESIS-2004

12

improved prospects arising from the equity premium tend to dominate the increase in

riskiness of future income streams. This encourages entrants to increase consumption and

borrowing, and to reduce net financial wealth. At the same time, they tend to accumulate

more precautionary wealth.

An important finding is that the equity culture creates incentives for entrants to increase loan

demand. The better prospects for future wealth accumulation arising from the equity premium

tend to dominate the associated increase in the riskiness of future income streams and

warrant an increase in current consumption. Consequently, increased stock market

participation has been accompanied by increased indebtedness of households, both for

consumption and for asset acquisition purposes.

Young households, who develop equity culture, experience an upward shift in the policy

function for consumption and a slight increase in the marginal propensity to consume out of

cash on hand. The shift is the net effect of two competing factors. First, access to stocks

creates expectations of higher future wealth because of the equity premium, and facilitates

consumption smoothing, because of the presence of a second asset. In view of higher

expected future wealth, young entrants can afford to increase current consumption at any

given level of cash on hand. The extent to which they choose to do so is governed by their

aversion to intertemporal consumption variability, measured by the degree of absolute risk

aversion. Access to stocks, however, has a second competing effect: it makes future income

streams more risky. Increased riskiness of future income discourages current consumption,

and the intensity of this precautionary motive is measured by the degree of prudence. This is

proven by the behaviour of Cypriot civil servants, who appear to be top stock purchasers

among the professions, since they face less future income uncertainty and prudential saving

is not so imperative in their case.

In all calibration experiments performed by the authors, the wealth effect dominates the

prudence effect. Poorer households tend to curtail their consumption and their borrowing

more than their richer counterparts. Faced with earnings risk in the absence of borrowing

constraints and other frictions, households with preferences characterized by prudence set

aside more wealth (or curtail their net borrowing) in order to buffer future consumption from

income shocks.

Page 13: FINANCIAL ASSET PARTICIPATION BY CYPRIOT HOUSEHOLDS-UCY-M.SC. THESIS-2004

13

The finding that equity culture tends to encourage precautionary responses to given amounts

of earning risk permits the following interpretation: when stockholders are confronted with

income risk, they curtail stock demand but they reduce borrowing (or increase riskless asset

holding) by much more, so as to generate precautionary (net) wealth holdings. Thus, they

combine precautionary wealth accumulation with a portfolio shift away from stocks and

towards bonds, in an effort to reduce total income risk. This portfolio shift, which obviously

does not occur for those without stocks, tends to lower expected future wealth and future

consumption since the household foregoes the equity premium. In order to mitigate these

effects, the stockholder sets aside a larger amount of precautionary wealth than in the

absence of a portfolio shift. As risk aversion increases, precautionary wealth holdings are less

influenced by the equity culture, since less use is made of stockholding opportunities.

Ιn his paper Jiaping Qiu (2002), distinguishes between direct and indirect stockholding. Direct

stockholding refers to the direct purchase of the stocks of publicly-traded companies by the

households themselves. Indirect stockholding includes the investors who delegate the

investment decision to some other organizations, such as mutual funds, trusts, and banks.

The ownership of stock, through indirect holding, has been rising in the past two decades,

while the general trend of household stock ownership through direct holding has been

declining over the same period. The increasing stock ownership through indirect holding

could be attributed to various reasons such as low interest rates, the increasing age of the

population and the transfer of savings from defined-benefit pension plans to defined

contribution.

According to Qiu, each stockholding form may affect differently the stability and efficiency of

the stock market, since individual and institutional investors exhibit different trading

behaviours. Furthermore, differences in the way stock is held have a direct impact on the

diversification of household portfolios. Indirect holding through mutual funds, trusts and

banks, in general, involves much more diversification than direct holding. Such differences in

the holding of equity by households may also provide insight into the theory of household

portfolio choice. For example, high information cost has been argued as one of the most

important reasons for the puzzle of the low stock market participation rate of the US

households. Indirect stockholding certainly provides an effective way of reducing the

information cost for investors who want to access the stock market.

Page 14: FINANCIAL ASSET PARTICIPATION BY CYPRIOT HOUSEHOLDS-UCY-M.SC. THESIS-2004

14

Qiu tests the household’s joint decision regarding direct and indirect stockholding using two

versions of the Probit model: the bivariate (observing a 0/1 dummy variable for direct or

indirect ownership of stock); and multivariate (considering direct stockholding and indirect

stockholding through stock mutual funds or through trusts and retirement accounts, which

latter may entail a different motivation).

The Qiu econometric results show that some characteristics have quite similar effects on the

probability of household direct and indirect stockholding. For example, households with no

risk taking attitude and no shopping habit are significantly less likely to own stock, either

through direct or indirect investment. College education, greater wealth and bequest

significantly increase the probability of direct as well as indirect holding. The marriage status,

gender of the household head and the family size have no effect on the probability of direct

and indirect stock holding. Other characteristics have quite different effects: self employment

significantly reduces the probability of direct stockholding but does not have a significant

effect on indirect holding. This might be due to the higher background risk of the self-

employed people, who try to avoid the additional idiosyncratic risk of direct stockholding.

Households with strong shopping habits as well as households with a retired household head

are significantly more likely to own stock directly. But strong shopping habits and retirement

have no effect on the likelihood of indirect stockholding. One possible interpretation is that

retired people have more time and experience to invest in stock by themselves. People with

strong shopping habits might prefer to search for the best deal by themselves. Lower than

high school education reduces the probability of owing stock directly. However, it has no

significant effect on the probability of indirect stockholding through mutual funds or trusts and

retirement accounts, suggesting that indirect holding is more accessible to the low educated

household. Households with a primary motivation for retirement saving have a significantly

higher probability of holding stock indirectly rather than directly, which might be attributed to a

desire for greater diversification through indirect investment. Other things equal, the people

between ages 65 to 75 are more likely to invest in the stock market directly relative to people

between ages 35-54 (omitted dummy), but less likely to have indirect stockholding through

retirement accounts. This result is reasonable because the elderly might have ceased to

invest in stock through their retirement accounts. The effect of the health status is somewhat

surprising. The sick people are less likely to invest in mutual funds, while sickness has no

significant effect on the probability of direct stockholding. The afore-mentioned results clearly

Page 15: FINANCIAL ASSET PARTICIPATION BY CYPRIOT HOUSEHOLDS-UCY-M.SC. THESIS-2004

15

indicate that the various household characteristics have different impacts on the likelihood of

direct and indirect stockholding.

In their paper, Haliassos and Bertaut (1995) investigate why the majority of United States

households do not hold stocks despite the equity premium and the predictions of expected-

utility models, which constitutes a puzzle. They consider this question to be relevant for

privatisation, asset pricing, and tax progressivity issues. They show that risk aversion itself,

heterogeneity of beliefs, habit persistence, time nonseparability, and quantity constraints on

borrowing do not account for the phenomenon. The difference between borrowing and

lending rates, and minimum-investment requirements are plausible but turn out to be

empirically weak factors. More promising explanations are inertia and departures from

expected-utility maximization. There is also qualified support for nondiversifiable income risk

as a contributing factor.

Ιn a recent paper, Bilias and Haliassos (2004) approach stockholding in a fresh manner and

consider new developments in the field. First, the writers point to the different climate

between the current decade and the previous: the spread of the “equity culture” and investor

euphoria in the 1990’s, documented by empirical studies, led to substantial increases in the

proportion of households holding stocks, either directly or indirectly through mutual funds and

retirement accounts. As already mentioned, this was facilitated by supply-side developments

such as the privatization of public utilities, the growth of the mutual fund industry with

concomitant reductions in participation costs for investors, as well as the demographic

transition and consequent policies to promote individual retirement accounts. Private

retirement funds are invested in stocks, bonds, and similar assets that tend to grow with the

economy.

By contrast, in the current decade, when about one in two US households already holds

stocks and euphoria has been tempered by recession and stock market downturns,

continued growth of the stockholder base cannot be taken for granted. The pressure of

economic downturns might even force a possible exodus of marginal stockholders. At the

same time, there are continued efforts to attract new stockholders on the part of firms,

brokers, and managers of mutual funds and other managed accounts.

Page 16: FINANCIAL ASSET PARTICIPATION BY CYPRIOT HOUSEHOLDS-UCY-M.SC. THESIS-2004

16

Thus, the authors examine which factors gain or lose prominence in household stock market

participation decisions. Moreover, since a systematic analysis of the distribution of net gains

from access to the stock market is lacking, the authors address issues concerning the

demographic and other characteristics of marginal stockholders and the impact of economic

phenomena on the demographic composition of the most likely prospective entrants as stock

market participation expands. Thus, the method of the writers (binary quantile regression

techniques in contrast to standard Probit and Logit) probes into the pertinent issues in more

detail and yields more specialized insights.

Stock market participation is typically regarded as based on expected gains from participation

net of any costs. According to the authors, existing econometric studies employ Probit or

Logit estimation to show that richer households are not scaled-up versions of poorer

households (Carroll, 2001) and that participation is influenced by education, risk aversion,

and employment status, while the role of age, if any, is less clear and less consistent across

countries. However, these techniques estimate the role of each factor on mean net gains,

and assume that it is the same for all households regardless of their position in the

distribution of net gains from entry. By contrast, the Haliassos-Bilias (2004) paper, based on

binary quantile regressions, suggests that there is considerable variation in the influence of

each factor, depending on the position of the household in the distribution of net gains from

access.

Probing into the determinants of gains from stock market access, the authors study model

predictions regarding the influence of household-specific variables as well as of stock market

factors, namely the equity premium and its volatility. By varying one household characteristic

at a time (e.g., education level, age, risk aversion, etc.), while keeping others constant, in

order to isolate the separate effects, as well as differentiating between working life and

retirement, the nature of the impact of each characteristic on gains from access is derived.

Their findings indicate that gains from stock market access have multiple determinants

whose roles are often conflicting and dissimilar at different points in the gains distribution.

The role of education in participation is found to be more limited than usually estimated and

confined to the low end of the gains distribution. This contrasts with the standard result in the

existing empirical literature, based on Probit or Logit estimation, according to which education

increases gains from stock market participation and hence the probability of participation.

Page 17: FINANCIAL ASSET PARTICIPATION BY CYPRIOT HOUSEHOLDS-UCY-M.SC. THESIS-2004

17

The results of the authors show that there are two conflicting effects of education on gains

from stock market access: a negative one arising from the more limited need of more

educated groups to save for the future; and a positive one resulting from lower participation

costs relative to permanent income, which is associated with higher educational attainment.

Risk aversion affects gains through two conflicting channels. First, higher risk aversion

discourages stockholding, ceteris paribus. Second, higher risk aversion generates greater

wealth holding. This is because it implies higher prudence (precautionary motive) and thus

larger precautionary wealth accumulation, as risk aversion increases. There is a wide range

of cash on hand, where the precautionary saving effect dominates the risk effect during

working years, and where it is optimal for higher risk aversion households to hold more

stocks. This ranking is found even for the group of college graduates who tend to face

smaller income risks and thus tend to have more limited precautionary motives. It appears

that by the age of 70, precautionary motives are significantly reduced as a result of reduced

future income risk. Thus, not only portfolio shares but also the amounts of stock drop with

risk aversion among savers. This produces the different rankings of gains during retirement,

compared to working life, as well as the result that highly risk averse retirees with low cash

on hand should not obtain permanent access to the stock market.

Age effects are predicted to be generally negative and non-linear over the life cycle, for given

levels of cash on hand and other characteristics. Age effects on gains from access are small

at the beginning of working life because stockholding tends to be limited early on and there

are still many future periods of life left, over which to enjoy benefits from access to stocks. As

working life continues, progressively more years of high stockholding lie behind and fewer

ahead. Moreover, a given level of cash on hand implies progressively worse future prospects,

and thus lower stockholding, as the household ages. Age effects are more substantial in the

latter part of working life, for both reasons.

All in all, economic recessions and stock market downturns can have significant effects on

the participation decisions of investors around the participation margin through their influence

on household incomes, wealth, and employment. Factors such as more limited finances and

education, younger age, and above all, an apparently significantly lower willingness to

assume financial risk by those most likely to consider entry, compared to those marginal

investors who have already entered, are likely to pose challenges as regards the continuation

Page 18: FINANCIAL ASSET PARTICIPATION BY CYPRIOT HOUSEHOLDS-UCY-M.SC. THESIS-2004

18

of stock market growth. The authors argue that such factors should not be ignored in

designing financial products or marketing strategies for expanding the stockholder base.

An additional issue examined in the pertinent economic literature concerns the question of

home equity bias. Empirical data point clearly to a strong home equity bias as regards

stockholding. This is so, despite the fact that such behaviour neither maximises return nor

minimises risk. Thus, investment in foreign stock is less than is justified by the covariance

(which is less than 1 and usually less than 0,5) and risk between domestic and foreign

shares, implying that there is room for optimisation and risk diversification by some

combination between the two. Moreover, foreign exchanges often possess greater

capitalisation, liquidity and probably afford more shareholder protection, which add to their

attraction characteristics. In fact, the segmented market approach states that, since countries

around the world exhibit different performance due to their unique characteristics, investment

in international assets can offer attractive risk reduction opportunities. Arguably, emerging

markets, like Latin America and the Asian markets could offer the opportunities for efficient

international diversification, given their relatively low correlations with more developed

markets. Thus, investors should tend to diversify their assets across national borders in order

to achieve and maintain risk reduction of their portfolio.

Risk diversification entails reduction in the portfolio risk level, resulting from the inclusion of

different assets in the portfolio. Diversification usually reduces portfolio risk (measured by

return variability) because the returns (both positive and negative) on various assets are not

perfectly correlated. In particular, equity investment responds to changes in interest rates,

expected appreciation of the exchange rate and of course inflation which determines real

returns. Interest rate changes affect the cost of capital and the discounted value of future

dividends. International capital flows tend to cause real interest rates to move together over

time. Such interest rate linkages may contribute to long-term equity market co-movement or

correlation. In this respect, integrated markets (like the European Union) are very sensitive to

disturbances in each other, which implies that their individual market portfolio returns are

highly correlated. Therefore, higher integration of the world equity markets greatly

undermines incentives for international portfolio diversification. In other words, if stock

markets share long-run equilibrium relationships (they are cointegrated) there are limited

benefits from international diversification for investors with long holding positions, since

almost perfect correlation between the markets is implied. However, there are some

Page 19: FINANCIAL ASSET PARTICIPATION BY CYPRIOT HOUSEHOLDS-UCY-M.SC. THESIS-2004

19

opportunities for diversification in the short-run (divergence from equilibrium), arising from

the expectations regarding inflation, interest rate and exchange rate changes, which alter the

expected real return.

This preference for domestic rather than foreign stock is difficult to explain but the following

probable explanations have been proposed:

1. Domestic stock provides a better hedge against domestic risk arising from inflation or

uncertainty concerning domestic and foreign currency, which makes the purchasing

power theorem fail. Notably, exchange rates affect equity index behavior because part

of the index’s return volatility is induced by monetary phenomena e.g. inflation.

2. It is assumed that there is a positive covariance between the performance of the

domestic stock exchange and expected future income or wealth, which should

logically lead to the holding of foreign stock (that have a different cycle) for the

purpose of risk diversification and keeping income constant when there is a downturn

of the economy at home. However, it is argued that this link is broken since the

assumption may no be so valid, not least because domestic stock exchanges also

reflect the performance of multinational companies (such as coca-cola) registered in

them. Therefore, domestic stock exchanges may already reflect the international

performance and no further diversification is achieved by holding foreign stock.

3. The cost of risk diversification (taxation, information, banking costs, capital movement

restrictions etc) through buying foreign stock is greater than the benefit.

4. There is a problem of estimating the potential return and variance of foreign shares.

However, this also applies to domestic stock.

***

Page 20: FINANCIAL ASSET PARTICIPATION BY CYPRIOT HOUSEHOLDS-UCY-M.SC. THESIS-2004

20

C. Cyprus Survey of Consumer Finances of 1999 and 2002 - Identity

The Cyprus Survey of Consumer Finances (CySCF) is an ongoing joint project between a

research team based at the University of Cyprus and researchers from the Central Bank of

Cyprus. In this framework, two surveys have been conducted so far: one in 1999 and

another in 2002. The purpose of the Cyprus Survey of Consumer Finances was to build a

household-level data base containing extensive information on the financial and real

portfolios of Cypriot households, their labour market status and their attitudes towards

saving, borrowing, risk taking, liquidity and other issues pertinent to financial behaviour and

portfolio choice. In this context, the survey provides a comprehensive source for a very broad

range of the financial and non-financial assets and debts of Cyprus households.

Consequently, the CySCF is not only unique but constitutes pioneering work, placing Cyprus

with very few other countries in the field, with similar statistical databases on household

portfolios. Indeed, the data base is comparable in scope and detailed coverage to that of the

United States Survey of Consumer Finances, the Italian Survey of Household Income and

Wealth, and the Dutch CentER Data Panel.

In fact, the CySCF constitutes a “microeconomic” approach to household economic

behaviour, rather than a “macroeconomic” treatment of aggregate financial and real asset

accounts in the form of assets and liabilities. Apart from the methodological difficulties of the

latter approach (mainly due to lack of data, even though as from 2000 the Ministry of

Finance has started compiling the National Economic Accounts based on sectoral assets

and liabilities and the European System of Accounts (ESA 95)), the former “microeconomic”

approach permits the researcher to pinpoint whether a change in the distributional shares of

the various assets and liabilities is due to a change in household participation or arises from

a change in the percentage share of a certain asset and liability, keeping household

participation constant. For instance, if aggregate financial accounts had been used to

analyse stockholding, which is the focus of our thesis, it would not have been possible to

distinguish whether a change was owing to a change of wealth or the alteration of household

demographic characteristics. Indeed, this is precisely the advantage conferred by the micro

level data base, with the use of interaction terms in the econometric model, which enable us

to compare the possible net change of the participation probability in household behaviour

between the 1st and the 2nd survey. In the light of the above, such surveys are extremely

useful, not only to professionals in the financial industry (in banking, insurance etc) and

Page 21: FINANCIAL ASSET PARTICIPATION BY CYPRIOT HOUSEHOLDS-UCY-M.SC. THESIS-2004

21

academics for analysing financial behaviour, but also to politicians and practitioners for policy

making.

1. Cyprus Survey of Consumer Finances of 1999

The statistical design of the 1999 CySCF (1st Survey) was based on area probability multi-

stage sampling techniques. The data, which refer to 1999, except for incomes which pertain

to 1998, were analysed by the Cyprus University team and compared to those for the United

States and four major European countries (United Kingdom, Germany, Italy, Holland).

Survey interviews were conducted between April 1999 and February 2000, with most

interviews taking place during the second half of 1999. The survey combines portfolio data

with information on the demographic characteristics of each household, and on its attitudes

towards borrowing, lending, risk taking, and liquidity. The survey sample included responses

from 1,097 households in two sub samples. The first sub sample was representative of the

Cyprus population and consisted of 539 households, while the second was confined to

wealthy households and had a size of 558. The over sampling of wealthy households is a

practice followed internationally, in order to handle the highly skewed wealth distribution and

the fact that most of the wealth and the greatest variety of assets are held by the wealthy,

who represent a very small proportion of the population. Since the resulting sample was not

representative of the population, each observation in the sample has been weighted by

appropriate population weights and the statistics reported are accordingly weighted so as to

reflect the average or mean behaviour of the Cyprus households.

2. Cyprus Survey of Consumer Finances of 2002

The 2nd survey interviews were conducted between March 2002 and June 2003. The survey

combines portfolio data with information on the demographic characteristics of each

household, and on its attitudes towards borrowing, lending, risk taking, and liquidity. Thus,

the survey provides a comprehensive source for a very broad range of financial and non-

financial assets and debts of Cyprus households.

Page 22: FINANCIAL ASSET PARTICIPATION BY CYPRIOT HOUSEHOLDS-UCY-M.SC. THESIS-2004

22

The survey sample included responses from 892 households in two sub samples like the 1st

survey. The main sub sample, which is representative of the population comprised of 515

households, while the wealthy households sub sample had a size of 377.

***

Page 23: FINANCIAL ASSET PARTICIPATION BY CYPRIOT HOUSEHOLDS-UCY-M.SC. THESIS-2004

23

D. Overall Financial Assets Participation - Cypriot Household Behaviour

The following section overviews briefly Cypriot financial asset holding (due to their direct

relevance with the topic of the thesis) and real asset holding, and examines the 1999 survey

differences and similarities with the 1998 respective USA data categories. The corresponding

2002 survey results appear in parentheses for comparative purposes. Afterwards, it looks at

the age and risk participation profile, the portfolio breadth and the popularity of the various

financial assets making comparisons and contrasts with the United Kingdom, Germany, Italy,

Holland.

1. Financial and Real Assets

The household portfolio composition is very important in terms of the impact it has on the

macroeconomic environment and variables such as prices and inflation, interest rates,

employment and stock prices.

Eight types of financial assets are considered in the surveys: Liquid accounts such as

checking accounts, deposit and savings accounts, government savings bonds, other (mainly

corporate) bonds, direct holdings of stocks, retirement accounts, and life insurance

investment policies.

Real assets include the primary residence, other real estate that could be used for investment

purposes, equity in businesses, and vehicles.

2. Overview of Asset Holding

Almost 9 out of 10 Cyprus households held some financial asset in 1999, relative to 92,9% in

2002. After checking accounts at 64% (63% in 2002), the most popular financial asset was

government savings bonds at 51%. However, in 2002, government savings bonds are

pushed into third place, at 43.6%, surpassed by direct stock participation at 51,4%. If we

keep constant the definition of retirement accounts across the two surveys, they remain

almost stable in the interval 10-12,5%. Life insurance investment policies exhibit a small

change from 31,1% in the 1st to 32,8% in the 2nd survey.

Page 24: FINANCIAL ASSET PARTICIPATION BY CYPRIOT HOUSEHOLDS-UCY-M.SC. THESIS-2004

24

3. Differences and Similarities with the US

Although the financial sector in Cyprus is quite developed, the United States has a more

competitive financial system with a much longer tradition in the workings of the stock market

and a bigger selection of financial assets that can be held by households. Similarities in

participation rates in the two countries indicate the areas in which Cyprus has already caught

up with the US.

Around 90% (or 93% in 2002), of Cyprus households hold some type of financial asset

regardless of riskiness, comparable to the United States (93%). Specifically, 82.2% (85,2%

in 2002), of Cypriots compared to 90.5% of Americans held liquid accounts (mainly checking

and savings accounts) that facilitate transactions. Half of the household population in Cyprus

(or 43,6% in 2002), compared with only one fifth in the United States participate in

government bondholding (development stock, saving certificates and saving bonds). Also,

significant divergence between Cyprus (10.2% or 12.5% in 2002) and the United States

(48%) is observed on retirement accounts.

4. Financial Assets Age Participation Profile

Participation in most categories of financial assets peaks in the age bracket of 40 to 49 years

(or the 30-39 age group in 2002) and then declines towards retirement. As the number of

asset types increases, the checking account remains the most popular asset with fairly

constant participation rates across the two surveys. Checking accounts constitute by far the

most widely-held financial asset by Cyprus households in all age groups, kept by nearly two

thirds of households (63-64%), with participation peaking at 77.1% in the age range 40 to 49

(or at 79.7% in the age range 30 to 39 in 2002). The widespread use of checking accounts is

explained by their usefulness in transactions. When compared with the much lower

percentage carrying bank, store or other credit cards (41,9% in 1999 and 50,8% in 2002, of

which 38% revolve debt), it demonstrates the tendency of most households to rely on this

older form of effecting transactions.

About one third of households have savings accounts (30,6% in 1999 in relation to 37,4% in

2002), and deposit accounts (36,5% and 35% respectively), that include notice and time

deposits. The high participation rates in checking and savings accounts are not surprising:

Page 25: FINANCIAL ASSET PARTICIPATION BY CYPRIOT HOUSEHOLDS-UCY-M.SC. THESIS-2004

25

combining cash holdings with such forms of highly liquid transaction assets is expected of

households that do not hold rich portfolios.

In 1999, government savings bonds were the second most popular financial asset overall.

Even in the height of stock market fever, more households in all age groups held savings

bonds rather than stocks directly. Savings bonds combine a low denomination, a guaranteed

minimum return, no default risk and a participation in monthly lotteries. All the above features,

together with their introduction long before the stock market was established, explain their

surprising popularity. The percentage of people holding government savings bonds drops

from 50,7% in the 1st survey to 43.6% in the 2nd survey. While in 1999 participation peaks in

the age range 40 to 49, this occurs in the 30 to 39 age group in 2002.

By contrast, in the 2002 survey, stocks turn out to be the second most popular asset, behind

checking accounts, with the direct participation rate reaching 51,4% of the population.

Retirement accounts, including provident funds, constitute the third most popular asset, held

by more than half of the population and reach a high 71,5%, in the 30 -39 age group. Mutual

funds, not yet introduced in Cyprus at the time of the surveys only appear to record an

insignificant rise from 0,4% in 1999 to 1% in 2002.

5. Holding of Life Insurance

Insurance demand is a function of income, age, education, family status and risk. Income

correlates linearly with respect to investment and mixed insurance (preferred by the rich since

it implies a lower marginal loss if investment fails) but non-linearly with the other insurance

types. Insurance constitutes an important element of indirect stockholding. The absence of

organized mutual funds until recently, has allowed this form of indirect stockholding to attract

many households wanting to use professional expertise in portfolio management. More than

one in two households, specifically, 52% have some form of life insurance according to the

1999 CySCF. A considerable percentage, i.e. 31,1% (32,8% in 2002) of households

participate in life insurance schemes with tax exempt premia that invest funds in

professionally managed portfolios and accumulate cash value dependent on risky returns.

Term insurance, which does not accumulate cash value falls by 4,5 percentage points to

13,5% by 2002, while whole life and endowment life insurance, which may or may not

Page 26: FINANCIAL ASSET PARTICIPATION BY CYPRIOT HOUSEHOLDS-UCY-M.SC. THESIS-2004

26

accumulate cash value, rise by almost 2 and 3 percentage points to 10,4% and 12,4%

respectively.

6. Life Insurance – Household Characteristics

In the 1st survey, all types of insurance follow a hump-shaped distribution pattern, peaking at

the 40-49 age group. This is expectable, since income is high in this bracket and there is also

a desire to protect the family. By contrast, the young lack the financial means to buy

insurance or invest in education. There is a natural decline in the tendency of older groups to

have life insurance. The old probably lack the knowledge and/or the education to go for

insurance policies and may also face refusal by insurance companies. They probably see no

need for holding insurance either. Married households are more likely to hold insurance than

the non-married. The educational level correlates with insurance but non-linearly, except with

the investment type, where a linear relationship seems to prevail, since university degree

insurance policy holders appear to be more (40.4%) than those who went to high school

(37.4%) or have less than high school education (17.4%). In the 2nd survey, life insurance

investment policies and whole life insurance participation peak in the 30-39 age group, while

term and endowment insurance continue to peak in the 40-49 age range.

7. Retirement Accounts and Mutual Funds

The retirement accounts percentage at 12,5% in 1999 (or 10,2% in 2002) is very low and

essentially includes those who participate in private pension schemes, typically run by

insurance companies and linked to life insurance, in order to qualify for tax exemption of the

premium. Low participation in private retirement schemes probably reflects the limited

familiarity of Cyprus households since they are relatively new. In particular, they appear to be

much less popular than other forms of insurance, probably because Cypriots have felt for

many years that they are adequately covered through social security and employer defined-

benefit pension schemes from which employees receive a fixed payment based on salary

and years of service. This is reinforced by the traditional preference of Cypriot households for

real assets, in the form of investment in houses, building plots and enterprises and

exacerbated by the virtual lack of public awareness and debate on the viability of the social

security system and on whether Cyprus will face similar demographically induced problems,

as those projected for the United States and all major European countries. Nonetheless, the

Page 27: FINANCIAL ASSET PARTICIPATION BY CYPRIOT HOUSEHOLDS-UCY-M.SC. THESIS-2004

27

inclusion of provident funds in the 2002 survey raises the percentage to 51,3%. This is

attributed to the fact that several semi-government organisations offer their employees

attractive provident fund schemes. At the same time, the high unionisation rate in Cyprus (at

around 56%, calculated as the number of organised workers to the economically active

population) has made such schemes popular and accessible to many trade union members.

In fact, there are around 1500 provident fund schemes registered at the social security

department of the Ministry of Labour and Social Insurance, offering lump sum benefits at

retirement, sickness benefits and educational and housing loans.

Mutual funds, combining liquidity with appealing interest rates, were not available in Cyprus

until recently due to the absence of a legal framework covering their operation. Indicatively,

as mentioned above, the 2002 survey sets the percentage of mutual fund holders at only 1%,

compared to 0,4% in 1999.

8. Portfolio Breadth and Popularity of Various Financial Assets

According to the 1999 survey about 10% of Cyprus households held no financial assets.

Those who did hold assets tended to hold a limited variety of asset categories. Almost two

thirds of households hold between one and three asset types, while only about 4% hold six

asset types or more. Cyprus experienced an increase in the number of assets held by

households between 1999 and 2002. The 2nd survey reports a reduction in the proportion of

households holding fewer than 4 assets, and a visible increase in the percentages of

households holding 4 asset types or more, with those holding 5 asset types almost doubling.

Thus, the portfolio breadth across asset categories, regardless of riskiness, shows

improvement in 2002. Still, the majority of Cyprus households held 3 assets or less in 2002,

with about 7% of households not holding any financial asset at all.

Portfolio breadth is, of course, not the same as portfolio diversification, since ultimately the

extent of diversification depends on the variance-covariance structure of returns on held

assets, and this is not observed in the data. Thus, since different asset types have different

return properties, the limited spread of financial assets suggests that portfolios are usually

not well diversified across asset categories. Consequently, given the small number of assets

and the absence of mutual funds, it seems safe to conclude that the financial portfolios of

Page 28: FINANCIAL ASSET PARTICIPATION BY CYPRIOT HOUSEHOLDS-UCY-M.SC. THESIS-2004

28

Cyprus households remain fairly undiversified, although the severity of the problem is likely to

have been reduced between 1999 and 2002.

9. Real Assets

The exact nature of the interactions or tradeoffs between financial and real assets is not yet

well understood in economic research because of the technical complexity of the analysis

required. The acquisition of real assets is typically correlated with earnings. Real asset

investments are more lumpy and illiquid and may entail poor diversification properties if a

portfolio is composed mostly of such assets. Such properties may impact on the popularity

and breadth of real asset portfolios. In 1999, only less than 2% of Cyprus households owned

no real asset, compared to 10% in the United States. By 2002, real asset ownership

becomes universal with 100% household participation. Motor vehicles, assisted by the fact

that public transport is almost non-existent, are the most widely held non-financial asset in

Cyprus with 91,6% (90,7% in 2002) owning a car, compared to 86% in the USA. Having a

house and a car creates pride of ownership along with valued services.

10. Real Assets - Homeownership

Housing is a major asset for most of the population, with ownership rates significantly

exceeding those in the US. In 1999, an astonishing 86% (in relation to 83% in 2002) of

Cyprus households owned their primary residence, compared with only 66.3 % in the US.

The majority of homeowners in Cyprus own their home fully compared to only about one third

of United States households. Active intergenerational transfer links probably contribute to the

high home ownership rate. Based on the CySCF, it is estimated that nearly 60% of Cyprus

households receive an inheritance or gift. Family support counteracts high down-payment

requirements, typically 1/3 of the house value compared to 5% or less in the USA.

11. Real Assets - Business

In the 1st survey, about 1/4 of Cyprus households (in relation to 22,5% in 2002), owned

business equity, compared to less than half of that proportion in the United States, a country

famous for fostering private enterprise. This reflects the Cypriot entrepreneurial spirit, while

owning a private business affords an important element of power and control, in addition to

Page 29: FINANCIAL ASSET PARTICIPATION BY CYPRIOT HOUSEHOLDS-UCY-M.SC. THESIS-2004

29

generating an asset return. It also reflects the Cypriot tradition of extensive participation in

real assets owing to the limited variety of financial instruments, which makes holding a well-

diversified portfolio more difficult.

12. Risk Attitude – Asset Types

Four major categories of risky assets are distinguished in the CySCF, namely, “Direct

Stockholding”, “Direct and Indirect Stockholding”, “Risky Financial Assets” and “Total Risky

Assets.” Direct stockholding refers to shares held directly, while the indirect type includes

stockholding through mutual funds, managed investment accounts and retirement accounts.

Since mutual funds have only appeared recently in Cyprus, this category includes those life

insurance policies that invest part of the premium in a risky portfolio and accumulate cash

value dependent on risky returns as well as private pension plans (other than defined-benefit

employer pension funds).

“Risky financial assets” include corporate bonds, in addition to direct and Indirect

stockholding, while “Total Risky Assets” also include investments in real estate (other than

the primary residence) and in private businesses.

13. Participation in Risky Assets

There has been an increased tendency of households in the United States and in major

European countries to invest in risky assets. Cyprus participation in assets with risky returns,

financial or real, far exceeds that in other countries: In 1999, 18.6% of households held risky

real assets but no risky financial assets compared to 21.7% for Italy, 7.7% for the United

States, and 5.1% for the Netherlands. The above finding is partly symptomatic of the

considerable involvement of the Cyprus private sector in economic activity. However, it also

reveals a hesitation of Cyprus households who participate in risky real assets to undertake

financial risk alongside real asset risk. This is confirmed in the attitudes towards risk taking

regarding financial portfolios.

Page 30: FINANCIAL ASSET PARTICIPATION BY CYPRIOT HOUSEHOLDS-UCY-M.SC. THESIS-2004

30

14. Risk Attitude - Real Assets

A strong contingent of households concentrates on risky real assets and abstains from risky

financial assets, even during the stock market frenzy year of 1999. If investments in real

estate and private businesses are also included, the share of households holding the so-

defined “Total Risky Assets” is 69,4% in 1999, which rises to 78,6% in 2002, compared to

56,9% for the US, 43,8% for Italy and 31,5% for Holland. This is by far the highest among all

countries considered, and it is consistent with the popularity of real assets among Cyprus

households.

15. Risk Attitude - Financial Assets

According to the 1999 survey, 50,8% of Cypriot households held “Risky Financial Assets”,

which include corporate bonds besides stock, compared to 49,2% for the US, 32,4% for the

UK, 25,1% for Germany, 22,1% for Italy and 24,8% for Holland. Household participation in

risky financial assets reported an increase to 74,1% in the 2002 survey, mainly coming from

the huge increase in direct stockholding. Risk diversification across categories of financial

assets is limited. The proportion of households that are well diversified across risk categories

(i.e., hold the full range of such categories: safe, fairly safe and fairly risky assets) is only 8%

(9,7% in 2002) in Cyprus, compared to nearly 36% in the US. Interestingly, 30,3% of the

population invested only in safe assets in 1999, compared with less than 18% in 2002. It

seems, therefore, that households have changed their attitudes towards risk taking in 2002

relative to 1999, becoming less risk averse.

Despite the fact that the financial portfolios of Cyprus households remain fairly undiversified,

the 2002 survey reveals a tendency of households to diversify somewhat their financial

portfolio risk by combining assets belonging to different categories of risk: Almost two thirds

(60,1%) of the population in Cyprus combine safe and fairly risky assets, compared with a

participation rate of 41% in 1999 (and only 13% in the USA). In this respect, the implications

of the overall limited portfolio diversification are mitigated by: 1) Significant participation in

managed portfolios, mainly through life insurance investment policies and private retirement

accounts 2) Households with narrow financial portfolios tend not to be exposed to direct

stockholding risk.

***

Page 31: FINANCIAL ASSET PARTICIPATION BY CYPRIOT HOUSEHOLDS-UCY-M.SC. THESIS-2004

31

E. Financial Assets Participation: Stocks

After describing the general asset-holding framework, the following section reports some

important findings regarding the stockholding behaviour of Cypriot households based on the

1999 and 2002 Cyprus Surveys of Consumer Finances. These are particularly related to the

main results of the statistical cross tabulations of section G. The contributions in the

compilation edited by Guiso, Haliassos and Jappelli (2002) have established an increased

tendency of households in financially developed countries to hold risky assets, especially

stocks held indirectly through mutual funds and retirement accounts. One quarter of all

households in Cyprus (25,3%) held stocks directly in 1999, rising to more than half the

population (50,3%) in 2002. Taking together direct and indirect stockholding (50,3% of

households in 1999 and 74% in 2002), stock market participation in Cyprus was, even in

1999, well above that of Germany, Italy, Holland and the UK and comparable to that in the

United States (48.9%) in which mutual funds have been in operation since well before 1990.

1. Direct Stockholding

In 1999, the Cyprus direct stock market participation rate (25.3% of households) compares

favourably even to that of the United Kingdom (21.6%) with its long tradition and is much

higher than that in the three Euro zone countries considered, namely Netherlands (14.4%),

Germany (10%) and especially Italy, where it is only 7.3%. Participation in direct stockholding

is higher than in other countries in general, specifically for households below 50 years, and

unusually high for the very young. This has been further exacerbated by 2002 when the direct

stockholder base more than doubled to 51,4% of the population. However, direct

stockholders hold a limited number of stocks and mostly have limited educational

background.

2. Indirect Stockholding

In 1999, indirect stockholding doubles the proportion of households investing in stocks to half

the population of Cyprus, while a similarly substantial rise is observed for most of the

countries under consideration. By 2002, this expands to reach almost three quarters of the

population. Indirect stockholding provides a number of advantages to households compared

to direct stockholding: risk diversification even at low amounts of investment, lower

Page 32: FINANCIAL ASSET PARTICIPATION BY CYPRIOT HOUSEHOLDS-UCY-M.SC. THESIS-2004

32

informational requirements, book keeping services, as well as delegation of trading decisions

to professionals. The high participation rate of Cyprus households in indirect stockholding can

be attributed partly to the tax exemption of life insurance policy premia and to the ingenuity of

insurance companies that embed life-insurance provisions into managed investment

accounts to ensure that contributions are tax-exempt.

3. Stockholding: Age Profile

Direct participation in the stock market was observed across the age spectrum, even among

those above seventy. The average age of household heads engaged in direct stockholding is

46 in 2002 compared to 44 in 1999. Stockholders are younger on average, as the average

age of households not owning stocks directly exceeds 50 years in 2002, compared with 47 in

the previous survey.

4. Stockholding: Diversification and Risk Profile

Stockholders tend to be poorly diversified in terms of the number of different stocks held.

More than 40% of direct stockholders in 1999 held stock in only one company, while nearly

20% of household heads had not graduated from high school (Lyceum). In 2002, still a

striking 32,6% of Cyprus households, not that much lower than the corresponding 42,4% in

the 1999 survey, owned stocks directly in only one company, while nearly 27,6% of direct

participants had not graduated from high school. Thus, one third of the population appears to

make no attempt to diversify stockholding risk by holding stocks with different return

properties, even after three years of stockholding and a major crash. Nevertheless, we do

observe signs of improved portfolio breadth in the 2002 survey, as almost 50% of the

population held stocks in more than 3 companies in 2002, compared with only a third in

1999. Both the stock concentration and the limited education of a sizable percentage of direct

stockholders appear to be symptoms of the rapidity with which direct stockholding spread,

which led to stock market instability and collapse. The high stockholding ranking occurring in

Cyprus over a very short period, despite extremely limited experience with the Stock

Exchange led to the known bubble problem of 1999.

Page 33: FINANCIAL ASSET PARTICIPATION BY CYPRIOT HOUSEHOLDS-UCY-M.SC. THESIS-2004

33

5. Direct Investment in Stocks by Income Category

In 2002, 86,3% of households in the income bracket £30,000-£35,000 owned stocks directly,

exhibiting the largest participation rate among all income groups, while in 1999 top place

(62,5%), was achieved by households with incomes in excess of £40,000. The largest

percentage increase (249%) in direct stockholding participation was reported among those

with less than £5,000, followed by the £20,000-£25,000 income group (128%). The overall

impression is that participation rates tend to rise with income, as indeed is the case in most

countries for which there are good portfolio data, but participation tends to be high even

among those who report that they belong to low-income classes.

***

Page 34: FINANCIAL ASSET PARTICIPATION BY CYPRIOT HOUSEHOLDS-UCY-M.SC. THESIS-2004

34

F. Bubbles

Some important theoretical aspects of bubbles are exposed below owing to their special

relevance with the Cypriot stockholding behaviour and the lightning rapidity with which the

equity culture spread in Cyprus under abnormal conditions in 1999.

1. Theoretical Aspects and Insights

Bubbles constitute departures in prices (e.g. housing or stock prices) from their fundamental

value. In other words, a bubble is said to exist when a price increases for reasons

unattributable to fundamentals or changes in the underlying supply and demand

determinants. A bubble is by definition, a temporary or transient phenomenon that will be

reversed. A bubble could affect all income levels.

The question then arises naturally, whether a price increase is being driven by fundamentals

or a speculative bubble. The problem with bubbles is that they cannot be identified with any

confidence, since supply and demand determinants change over time, often unpredictably. If

bubbles could be accurately identified, they would never develop in the first place because

people would respond to the emergence of a bubble by selling the asset to avoid future

losses, thereby eliminating the bubble.

Indeed, some economists, who believe markets are always rational and efficient, use this

logic to argue that bubbles can never exist. Economists advocating the efficient market

hypothesis do not assume that prices are efficient because everyone is rational all the time.

Rather, such economists assume that efficient pricing occurs because people do not make

systematic mistakes, and because sufficient people are correct, so that they can take

advantage of others’ mistakes until prices move back to their efficient point. For example, if

an investor realized there was a stock market bubble that would burst soon, he could make

large profits by selling stocks short.

In any case, the efficient market hypothesis is not without its detractors in the economics

profession. A group known as behavioural economists have been trying to use evidence of

non-rational behaviour (irrational exuberance), which is well-documented in psychological

research to explain economic phenomena. In pursuit of this line of thought, it is argued that

for a bubble to emerge and persist, the following criteria would have to occur: 1) most people

Page 35: FINANCIAL ASSET PARTICIPATION BY CYPRIOT HOUSEHOLDS-UCY-M.SC. THESIS-2004

35

are making a mistake which is not quickly corrected; 2) most mistakes have a systematic bias

in the same direction; 3) and those who realize that a mistake has occurred do not or cannot

take actions to profit from it that would reduce or eliminate the bubble.

Some argue that “speculators” are responsible for bubbles. Although theory is ambiguous,

others argue that it is more likely for speculators to prevent or reduce bubbles rather than

cause them. Indeed, speculators are seen as individuals attempting to profit from pricing

mismatches and are therefore expected to disinvest from areas that are overpriced, and the

process of disinvestment would help deflate the bubble before it became serious.

Of course, there is always the possibility that other unidentified “fundamentals” are driving

prices up, rather than a bubble. Although the recent behaviour of the stock market lends

strong support in favour of the existence of bubbles, there are reasons to believe that bubbles

are less likely in housing markets than stock markets. Basically, it is the intangible nature of

certain assets (in particular, stocks among financial assets) that makes their pricing difficult

and opens the possibility of a bubble forming.

For example, corporate equities are difficult to price because their value should equal the

expected future profitability of a company discounted to the present. Since nobody knows

how profitable a corporation will be in the future, the price of its equity is subjective and

imprecise. If enough market participants become “irrationally exuberant,” a bubble can

emerge. On the other hand, houses are easier to price accurately because they are more

tangible. Another difference between housing markets and stock markets is that there are

high transaction costs – financial and time – to buying or selling a house. This means that

buying or selling solely in response to mispricing is less likely to occur. Furthermore, the only

individuals who can take advantage of mispricing are those who are not living in their homes

or are free to move to non-bubble areas, which may be unlikely because of professional,

family, or community ties. Whether high transaction costs make bubbles more or less likely is

unclear: they certainly reduce the opportunity for “rational” traders to correct the mistakes of

others, as economic theory would suggest, but they also prevent the bidding up of prices in

order to profit before a bubble bursts. And another factor that may make it more difficult for

“rational” traders to eliminate a bubble in the housing rather than the financial markets is the

fact that few methods exist in housing markets analogous to selling a stock short. Investors

sell stocks short by selling a borrowed stock that they believe is overpriced in the anticipation

Page 36: FINANCIAL ASSET PARTICIPATION BY CYPRIOT HOUSEHOLDS-UCY-M.SC. THESIS-2004

36

that they will be able to buy back the stock in the future at a lower price, earning a profit on

the difference. Obviously, there is no direct way to sell a borrowed house and then buy it

back in the future.

Stock price declines in international Stock Exchanges and in Cyprus were huge in 2000,

following the unwinding of the Cyprus Stock Exchange index, which scored a record high

849,3 on 29.11.1999, and prices have remained depressed since, thus exacerbating the

macroeconomic consequences of the bubble’s deflating. Price declines may, of course reflect

deflating bubbles but also underlying changes in supply and demand factors exerting

downward pressure on prices.

2. Causes of the Cyprus Stock Exchange Bubble of 1999

Apart from institutional weaknesses and insufficient legislation which were exploited by stock

brokers for their own profit, and misleading statements and actions by people in power, as

well as malpractices by underwriters, in collaboration with the owners of companies seeking

to be listed, who often used creative accounting and other misleading tactics, there was a

series of other factors which assisted in the development of the Stock Exchange bubble of

1999. These included the desire of many private companies to go public following the tax

incentives offered by the government, as well as the ensuing rapid expansion of Stock

Exchange activity, the improved international and domestic environment, the positive

expectations and economic prospects, arising from the GDP achieving an above potential

average growth of 4.8% over 1998 and 1999, and finally the general misconception that

stocks were undervalued, prompted by a misguided and misguiding press and media, which

became actively involved, since in several cases they had a vested interest and an axe to

grind.

Specifically, the main causes of the Stock Exchange bubble were the following:

1. There was a huge desire by private company owners to make their companies

public in order to raise easily cheap finance or financing that was not

otherwise obtainable. Specifically, owners saw the listing of their companies

as a means to cash in on the Stock Exchange frenzy and obtain money by

selling shares at highly inflated prices. The latter was the prevailing motive (i.e.

Page 37: FINANCIAL ASSET PARTICIPATION BY CYPRIOT HOUSEHOLDS-UCY-M.SC. THESIS-2004

37

personal enrichment) rather than any attempt to improve the company debt to

equity (leverage) ratio or to forward any investment plans. In order to further

this end, the company profit forecasting was not based on fundamental

economic and market principles. It was mainly calculated in such a way as to

project a very optimistic view of the company’s prospects. In fact, the accounts

of companies to be listed overvalued the company assets and presented

fictitious or unsustainable profits. This enabled the use of high price to

earnings (P/E) ratios in order to determine a high share valuation for the initial

public offering (IPO). Moreover, the artificially high share valuation made

possible the promise or expectation of a split, which constituted yet another

incentive for people to buy shares, hoping to make a quick profit on rather

cheap capital. Further, the IPO share selling price was often at a premium,

which yielded capital that was used to issue bonus shares, of which the

primary beneficiaries were the initial shareholders and/or managers

2. It was possible for the owners of a company to be listed to retain total

company control while selling a relatively small portion of the company, i.e.

30%, often for more than the initial company value. Even today, only a handful

of companies in the Cyprus Stock Exchange (such as Demetra, the Bank of

Cyprus, the Laiki Bank and the Hellenic Bank and perhaps Sharelink and the

CLR Investment Fund) can be considered as truly public by international

standards. Nevertheless even these companies are controlled in one way or

another by a small group of individuals.

3. There were administrative deficiencies. For instance, the Cyprus Stock

Exchange proceeded with the automation of the trading floor before

automating the back office operations, resulting in an accumulating backlog.

Hence, an upper limit of 2000 daily transactions was imposed. This inevitably

led to an artificially inflated demand during the period concerned, as all

stockbrokers were trying to execute their orders, and a consequential upward

pressure on prices with successive limits up observed for many stocks.

4. The legislation did not oblige stockbrokers to specify the customers for whom

they were executing the transaction (no investor identity code was required

Page 38: FINANCIAL ASSET PARTICIPATION BY CYPRIOT HOUSEHOLDS-UCY-M.SC. THESIS-2004

38

prior to September 1999). Therefore, it was very easy to manipulate the

market for their own profit. In fact, the Cyprus Stock Exchange did not operate

with dematerialized titles until 2002. The resulting backlog prevented the

existence of a proper stockholder depository leading to almost total confusion

as to the title owners and allowing the execution of improper transactions by

stockbrokers who exploited the situation for their own financial benefit. Indeed,

often the companies delayed intentionally the issuing of titles to small

shareholders so that the majority owners themselves and the stockbrokers

could “unload” paper on the unsuspecting public. This brings into light the

special relationship between company directors and the stockbrokers which

was fully exploited by those involved.

5. It was not a prerequisite for the security purchaser to deposit the necessary

funds prior to a transaction. Therefore, it was possible to have much higher

daily transaction volumes than the financing available, resulting in the

execution of transactions without the required funds being available, thus,

leading to the oversaturation of the market by the overissuing of shares.

6. There was a huge public misconception regarding the company action of

splitting its share. The idea was artificially cultivated that the share value after

the split should be at least equivalent to the one before the split and that

security prices could only go up! This led to artificial demand and fully

unjustified share price rises.

7. Certain politicians and “men in power” were involved by acquiring shares at

nominal rather than IPO value and by using a private placement process,

which guaranteed large potential gains by selling the shares on the first day of

trading. Also, certain political parties could not resist and succumbed to a

similar behaviour, exploiting their political leverage for their own profit, at the

expense of the people. This involvement was not only improper but also

misled and encouraged the public to enter the market.

In view of the above circumstances and conditions, the fall was as rapid as the rise. Given

the persistently poor behaviour of the stock market since 2000, the impact on wealth has

Page 39: FINANCIAL ASSET PARTICIPATION BY CYPRIOT HOUSEHOLDS-UCY-M.SC. THESIS-2004

39

become a particular cause for concern in the near future. On the other hand, the poor

performance of the stock market could be expected to lead people to shift more of their

wealth into property, which is seen as a “safe haven”. This, combined with the relatively low

interest rates could be one of the main causes of the increased demand and rising housing

prices over the recent years in Cyprus.

A quantification study by the Central Bank of Cyprus conducted by the Cyprus College in

September-October 2001 examined the Cyprus Stock Exchange bubble phenomenon of

1999. The study was based on a survey of 2017 households and brought out some salient

points.

The Central Bank / Cyprus College survey reveals that prior to 1999 only 7% of households

held shares. This rose to 43% following the bubble (as opposed to 50.3% according to the 1st

Cyprus University survey (carried out between April 1999 and February 2000). This drops to

37% if Demetra is excluded wherein 6% of households invested a sum, which in the majority

of cases did not exceed £1000. Besides direct stockholding, after 1998, 27% of households

held Life Insurance Investment Plans and 14% Provident Funds, both of which constitute

indirect share holding.

The middle income groups entered the Stock Exchange en masse. There was massive entry

even among rural households whose participation rose from 4% prior to 1999 to 38%.

Households spent on average their annual income to buy shares. Rural households invested

approximately equal sums to urban ones. Old investors (those who held shares prior to 1999)

invested less than new ones.

In 1999 and 2000, households directly invested on average £13.130 each, which adds up to

a total of £1.070 m. for the whole population. The above sum does not include indirect

participation through Life Insurance Investment Plans and Provident Funds. Share purchase

was financed 75% (around £825 m.) with own capital, mostly savings, and 25% (around £245

m.) by loans. Around 18.700 households borrowed (£245 m.) to invest directly in stocks.

The mean sum (£13.130) invested was above the median (£9.000) since a significant part of

households (15% of those who invested or 5% of total households) invested over £20.000. At

the time, the average annual household income according to the Central Bank survey data

Page 40: FINANCIAL ASSET PARTICIPATION BY CYPRIOT HOUSEHOLDS-UCY-M.SC. THESIS-2004

40

was £15.900. Hence, the typical household invested 83% of its annual income, whilst new

investors put in 91% of their annual income.

At the time of the Central Bank survey (October 2001) 45% of households did not liquidate

their portfolio, 31% lost from liquidation, 16% gained, while 8% were even. 76% of those who

gained were old investors. 32% of these liquidated their whole portfolio, whilst the majority did

not appear to have losses after the stock price collapse. The 16% of households who made a

profit gained on average £7.000 or a total of £90 m. The 31% of households who lost suffered

an average loss of £5.800 or a total of £150 m.

It seems that households which borrowed to enter the market and were forced to sell were

affected the most. New investors borrowed more heavily than old ones and suffered most of

the losses. High income groups profited most from the rise in stock prices. The Stock

Exchange bubble raised inequality: lower income groups suffered losses disproportionate to

their income.

***

Page 41: FINANCIAL ASSET PARTICIPATION BY CYPRIOT HOUSEHOLDS-UCY-M.SC. THESIS-2004

41

G. Statistical Cross Tabulations

The tabulated data are presented in Tables 1-6 in a clear format that facilitates

interpretations and enables an immediate drawing of conclusions. However, statistical tables

do not offer the complete picture, since they lack the scientific rigour of identifying causal

relationships (by holding constant other causative factors, and thus, isolating their separate

effects (ceteris paribus condition)). For instance, higher income individuals are more likely to

be credit card or stock holders. But high income is also a function of education with which it

is correlated. Hence, the statistical results are reinforced by econometric analysis (Probit

regression in this case), which makes possible the pinpointing of the causative effect of each

explanatory variable.

The data for the statistical and econometric analysis come from the 1999 and 2002 CySCF.

The analysis of the CySCF data was performed using the SPSS and the STATA program.

The SPSS was used for the cross tabulations, while the STATA was used for the Probit

regression. It is important to underline that all the SPSS descriptive statistics cross

tabulations were performed using population weights for the pertinent data categories so as

to reflect the weighted average or mean behaviour of the Cyprus household population.

1. Statistical Cross Tabulations - Analytical Results

Table 1 presents some descriptive statistics which make up a very talkative picture of the

sample population. The average age of the household head is 46,5 in the 1999 CySCF and

48,7 in the 2002 CySCF, compared to 54 in Italy and 51 in Holland, respectively. About 88%

are married in relation to 68% in Italy and 62% in Holland. Single people constitute 4,3% in

the 1st survey and 3,1% in the 2nd, in comparison to 19% and 15% in Italy and Holland

respectively. The male population stood at 65% and 61% in the respective surveys.

As far as the education level is concerned, 27% or 30% by 2002 were university graduates,

38% falling to 33% by 2002 finished high school, while 35% rising to 37% by 2002 possessed

less than high school education. The respective percentages for Italy were 8%, 28%, 65%

(referring to university, high school and no high school) and for Holland 41%, 54%, 5%.

Here, some obvious differences emerge especially with regard to the household heads with

the lowest education level.

Page 42: FINANCIAL ASSET PARTICIPATION BY CYPRIOT HOUSEHOLDS-UCY-M.SC. THESIS-2004

42

The majority of the households, i.e. 67% rising to 68% by 2002 comprised two to four

members, compared to 72% for Italy and 65% for Holland. The distribution of households

into one- two- and three-income families or more was around 35%, 49%, 10%. This changed

to 30%, 53% and 9% by 2002. The corresponding distribution for Italy was 44%, 42%, 14%

and for Holland 60%, 32%, 5%.

The mean Cypriot household income was £14.194 in 1999 or £15.109 in 2002 (the Cyprus

university research team filled in the missing values by imputation) in comparison to £14.450

for Italy and £14.500 for Holland.

Having observed the descriptive statistics of the population sample in Table 1, the following

Table 2 analyses the demographic and economic characteristics between stock market

participants and non-participants. The results appear to be similar to those in Italy and

Holland. Figures in parentheses indicate the corresponding 2002 survey results.

For instance, those who participated directly in the stock market had an average age of 44,4

(46 in 2002) years compared to 49,4 (58,6) years for the non-participants. In other words,

participants were on average 5 years younger in 1999 and 13 years in 2002 (compared to 6

years in Holland). The percentage ratio of high school to university educated direct

participants (37,5/43 = 0,9:1) is lower than that of non-participants (33/18,2 = 1,8:1), bringing

out the positive impact of education on participation. Similar results, even strengthening

somewhat, come out in relation to the 2002 survey, with the respective ratios being 32,7/40

=0,8:1 and 25,5/11,2 = 2,3:1. Similar outcomes appear in Holland. The educational effect will

also come out in Table 4 that examines it exclusively. Further salient differences among

direct participants and non-participants are that the former are more likely to be male at

70,8% (67,6%) compared to 61,3%, (54,9%) on average; overwhelmingly married (91% rising

to 94% in 2002, in relation to 84% and 73% respectively for non-participants) and rather

belong to bigger households with more income earning members, as is confirmed by intuition.

These results are compatible with the experience of Italy. Being a wage-earner appears to

play a role in the participation decision, since there is income ability and the incentive of an

expected higher return. The main reason for the percentage of self-employed participants

19,3% (24,9%) prevailing over the corresponding percentage for non participation 15,3%

(21,0%) seems to be risk diversification, since the self-employed keep a large part of their

wealth in their own business. Of course, being unemployed and hence with low income does

Page 43: FINANCIAL ASSET PARTICIPATION BY CYPRIOT HOUSEHOLDS-UCY-M.SC. THESIS-2004

43

not encourage participation and such behaviour is consistent with that in Holland and Italy.

Naturally, stock investors have higher income and wealth than non-investors and this again

agrees with the experience of the afore-mentioned European Union countries and accords

with intuition, theory and empirical findings.

The demographic and economic profile of indirect stock market participants through

investment in life insurance is generally similar to that of direct stock participants, when

contrasted to non-participants. Nonetheless, those who invest indirectly through life

insurance appear to be younger, aged 41,8 (43,1) on average compared to 44,4 (46) for

direct stock participants. It is notable that, whereas in 1999, life insurance investors have less

income and wealth and go to college in lower numbers in relation to direct stock participants

(35% compared to 43%), this reverses in 2002 as regards income and college education.

Concerning 1999, this outcome shows that indirect participation requires less knowledge and

specialization than direct participation. The 2002 reversal should reflect the impact of the

insurance funds being placed in investment companies like Demetra. A similar reversal in

characteristics occurs concerning those who participate indirectly through pension plans:

while in 1999 they appear to have lower college education and income (than direct stock

participants), in 2002 they tend to be more educated and have higher income. As expected,

pensioners will be cashing maturing life insurance plans rather than be investing anew in

such plans, hence their percentage is low 2,3% (9,3%) compared to non-participants at

21,7% (17,3%). In this respect, the jump by 7 and almost 6 percentage points in the number

of the retired holding life insurance and pension accounts respectively, is notable and must

be traced to the “Get-rich-quick” attitude that prevailed like mania at the time over the larger

part of the population, including the retired, who often virtually “gambled” their retirement

sums or lifetime savings on the Stock Exchange. In fact, at the time, (according to a finding in

the Central Bank / Cyprus College study) around 4%, out of a total of 190.000 households,

placed large sums of money into single premium insurance investment plans (boosters) and

subsequently lost heavily since they bought at the top unit values through these "boosters",

while their investment was almost annihilated soon afterwards.

Furthermore, wealth for those who held retirement accounts remains lower in both survey

years compared to direct stock participants and insurance policy holders, but higher than the

wealth of non-participants. Nevertheless, the respective percentages of those who held

Page 44: FINANCIAL ASSET PARTICIPATION BY CYPRIOT HOUSEHOLDS-UCY-M.SC. THESIS-2004

44

retirement accounts surpass those for non-participants regarding the educational level,

income and wealth in both surveys. The above results accord with international experience.

Thus, the emerging portrait of the typical Cypriot who invests directly in the Stock Exchange

is that of a married, middle-aged man who tends to have high education, income and wealth

and belongs to a household comprising up to four members 67% (70,9%), two of which are in

the majority of cases 55,2% (57,5%) income earners.

The following Tables 3 -7 look at the major factors or household characteristics that affect the

participation decision, such as age, education level, income and financial wealth.

Table 3 records direct and indirect participation by the age group of the household head. In

1999 around a quarter of households (25,3%) invested directly in stocks in relation to 7,3% in

Italy and 14,4% in Holland. This percentage doubled to 51,4% by 2002. The doubling of

direct participation is attributed to the mass entry of households into the stock market after

1999, following the huge public response to the launch of investment companies like the

Demetra Investment company Ltd (Stock Exchange entry date: 27.4.2000) and the CLR

Investment Fund (Stock Exchange entry date: 14.9.2000), the Eantas Investment Ltd (Stock

Exchange entry date: 28.3.2000) and others, which raised substantially the participation

percentage.

In particular, the increase in stockholding was largely due to the establishment of the Co-

operative Society’s investment company, Demetra, in which almost all clients of the Co-

Operative sector bought shares. Especially, elderly households who had been banking with

the Co-operative sector all their lives, trusted the newly established company and invested

large amounts in its stock. At least some of the increased participation in stockholding must

have come at the expense of participation in government bonds, which decreased by about 7

percentage points by 2002 compared to 1999. Thus, the investment companies have been a

driving force in the expansion of the stockholder base.

Even though the 2002 household survey came after the burst of the stock market bubble, half

of the Cyprus population still held stocks in 2002. In the early stage (1999 to early 2000),

some of these investors were “trapped” by the lack of titles, as analysed in the section listing

the reasons that caused the bubble. At a later stage the “trap” was self-inflicted due to the

Page 45: FINANCIAL ASSET PARTICIPATION BY CYPRIOT HOUSEHOLDS-UCY-M.SC. THESIS-2004

45

desire to avoid the realisation of huge losses (“disposition effect”), following the tremendous

stock market crash. The very limited trading volume in 2002 compared to 1999 is consistent

with this view.

As regards indirect stock market participation, 31,1% invested through life insurance

investment plans (rising slightly to 32,8% by 2002), compared to mutual fund participation of

11,1% and 16,2% for Italy and Holland respectively.

Approved pension plan participation was 12,5% in Cyprus, 7,9% in Italy and 18% in Holland

in 1999. The relatively low 1999 participation in pension funds in relation to 48% in the USA,

demonstrates that Cypriots rely mostly on the social security system, provident funds and

accumulation of real assets (houses, building plots) for their retirement planning. However,

the 2002 survey recorded a quadrupling to 50,3%. This increase by a factor of four to 50,3%

was due to altering the definition in the 2002 survey questionnaire, according to which the

retirement accounts now also included provident funds, which were excluded in 1999.

However, the data which have been adjusted (excluding provident funds) for comparability

purposes show a much closer proximity to the 1999 outcome of 10,3% relative to 12,5% in

the first survey.

It is evident from the above figures that indirect participation (through life insurance and

pension funds) in the stock market is encouraged by the fact that it helps diversify risk,

requires less information, and allows access to professional portfolio management and

accounting services. Tax exemptions of life insurance premiums constitute an additional

factor facilitating such an indirect participation.

Consequently, total participation i.e. both direct and indirect, by Cypriot households rose from

an already high 50,3% (compared to 18,9% in Italy and 33,5% in Holland) to an even higher

73,5%, thus making Cyprus literally a nation of mass participation, since almost three out of

four households participate indirectly and half the households directly in the stock market.

Furthermore, Table 3 indicates an undeniable hump-shaped age pattern emerging,

compatible with international experience: In 1999 total stockholding and investment

insurance both begin at a relatively low percentage in the youngest age group (less than

thirty), and peak at the 40-49 middle age group. Thereafter, a distinct fall is observed and

Page 46: FINANCIAL ASSET PARTICIPATION BY CYPRIOT HOUSEHOLDS-UCY-M.SC. THESIS-2004

46

disinvestment correlates with ageing. Total participation, both direct and indirect peaks at

63,9, while in Italy it peaks at 27.5% between ages 30-39 and in Holland at 40,1% in the 50-

59 age range.

A similar pattern appears concerning direct stockholding and pension plans with some slight

aberrations: As regards direct stockholding, a relatively high 26,9% is observed for the less-

than-30 age group, compared to 26,7% for the 30-39 group. This is almost certainly due to

the bubble episode of 1999, but there is possibly also a wealth effect due to income transfers

from the parents to their kin who co-live with them. The wealth effect seems to reduce risk

aversion with a consequent impetus to direct stockholding. Maximum direct participation

occurs in the 40-49 age group, with 30% of the households directly participating in the stock

market, compared to the average of 25,3%, while in Italy direct participation peaks in the 30-

39 group and in Holland at over 70 years of age. Furthermore, retirement plan participation

peaks at the 30-39, rather than the 40-49 middle age group. Nonetheless, the hump-shaped

pattern still remains.

Again in the 1st survey, indirect participation through life insurance also peaks in the 40-49

age group with 45%, in contrast to Italy and Holland where pension fund participation (used

for comparative purposes) reaches a peak in the 30-39 and 50-59 age groups respectively.

As already mentioned, pension plan participation peaks in the 30-39 group with 15,1%, as in

Italy, whereas in Holland participation peaks in the 50-59 group.

The 2nd survey still preserves the hump shape, with the distribution for both the direct and the

indirect participation categories peaking in the 30-39 age group. Thus, we observe a shift of

the peak towards the younger 30-39 age group and away from the 40-49 group of the 1st

survey. Explanations for this accord with particular Cypriot cultural features, such as income

support for the younger members of the family, which facilitated entry when the stock market

fever was spreading.

Even though an age pattern appears in the statistical cross tabulations, both for Cyprus and

the other countries considered, this result is not robust and is not confirmed by the Probit

regressions. Thus, age has an unclear and inconsistent effect, if any. It is highly probable

that the income factor is hidden behind the middle age groups wherein stockholding and of

course income peak.

Page 47: FINANCIAL ASSET PARTICIPATION BY CYPRIOT HOUSEHOLDS-UCY-M.SC. THESIS-2004

47

Besides risk aversion, liquidity constraints and entry costs push participation in favour of the

middle-aged households when the reverse bell-shaped pattern that emerges concerning

stockholding can be attributed to the “wealth effect”. Under decreasing absolute risk aversion,

households have the highest risk tolerance (or the least risk aversion) in the period of their

lives when they enjoy the maximum income or have amassed their maximum wealth. During

the middle age, income usually peaks and thus we observe the maximum percentage of

direct and indirect participation.

In addition to the wealth effect, which exerts a negative impact on stockholding by younger

households, who have neither amassed wealth nor enjoy high incomes at the beginning,

there is another important force. This is the time diversification effect, which exerts a positive

impact on stock market participation by younger households who have a large time horizon

ahead of them to spread their risk (which might result from a future income shock owing to

the holding of risky assets). However, in the case of young Cypriot households, because of

the “dowry system” and the income support by the parents, the negative impact of the wealth

effect is weakened and this explains their high participation.

Education correlates positively with the participation decision, since it is directly related with

the level of permanent income and wealth and raises the ability to process information of a

financial nature. As already stated, the increase of income and wealth leads to the rise of

intertemporal consumption and under decreasing absolute risk aversion, it encourages

participation in risky assets. Indeed, Table 4 verifies the positive impact of education on

stockholding: the higher the educational level the greater the direct participation in both

surveys. Specifically, 40,4% (65,6% in 2002) of college or university degree holders

participate directly in the Stock Exchange, compared to 25,1% (53,6%) who finished high

school and 13,9% (38,1%) who did not. The participation of households with limited

education is quite pronounced in the group of stockholders. Therefore, the probability of

direct participation of university graduates is 1,3-1,6 times greater than the average and

almost 1,7 - 3 times greater than that for people who did not go to high school. The lower

end of the latter range, i.e. the fact that university degree holders surpassed those who did

not go to high school only by a factor of 1,7 in 2002, rather than 3 as was the case in 1999,

shows the lessening impact exerted by education, as captured in the 2nd survey, since the

bubble phenomenon attracted people of lower education into the stock market. The

Page 48: FINANCIAL ASSET PARTICIPATION BY CYPRIOT HOUSEHOLDS-UCY-M.SC. THESIS-2004

48

proportion of households with less than high school education who directly participate was

relatively high in Cyprus compared to Italy and Holland even at the time of the 1st survey.

This notable shift in the educational composition of the stockholder pool between 1999 and

2002 towards households with less than high-school education and away from the other two

education categories is attributable to the rapid spread of the equity culture and the Stock

Exchange bubble of 1999. In this respect, it is noted that financial education and targeted

advertising by mutual funds are more likely to appeal to educated households but require

time to influence stockholding patterns, as evidenced by the fact that it took American

households about a decade to start participating in mutual funds in large numbers. The

spread of the equity culture among Cyprus households was initially accomplished through

the extensive media coverage and word of mouth, and was then reinforced by the

establishment of investment companies promising hassle-free management of stock

portfolios for the individual household. This contrasted to the foreign experience where the

equity culture spread via indirect stockholding through mutual funds. All of these

developments effectively removed the educational barrier to direct stockholding and drew

less educated households into the stockholder pool. In principle, the existence of a significant

mass of stockholders with limited financial education could cause market volatility, by

inducing abrupt shocks to stock demand and supply in response to limited or misleading

market signals. In practice, a wide range of households (and their funds) seem to have been

miserably entrapped in a static situation of complete lack of confidence, causing the current

inactivity in the stock market.

Similarly, the correlation found between education and direct stockholding by the statistical

cross tabulation, also appears in the case of indirect participation through life insurance and

pension plans. This outcome accords with the experience of Italy and Holland. It should of

course be born in mind that indirect participation through insurance or mutual funds places

management responsibility on professionals and is therefore less demanding in terms of the

educational level.

Table 5 presents household participation in the Stock Exchange by income quartile and the

highest 5% of the income distribution.

Page 49: FINANCIAL ASSET PARTICIPATION BY CYPRIOT HOUSEHOLDS-UCY-M.SC. THESIS-2004

49

As regards the 1st survey, the first income quartile household participation, at almost half the

mean participation (12,2/25,3=0,48) is high compared to Italy and Holland where it is almost

non-existent. This rises even higher in the 2nd survey (30,8/51,4=0,6). Again, this is attributed

to the bubble incident which pushed many lower income households to put their savings and

even borrow (as shown in the Central Bank / Cyprus College study) in order to enter the

stock market. This behaviour occurred in the expectation of a quick and high return,

exhibiting irrational behaviour, not based on fundamentals and ignoring the constant cost of

entry. Also, in the 1999 survey, 23,6% of the population who belonged to the first income

quartile (i.e. almost half the average total participation rate of 50,3%) participated directly and

indirectly, while the respective figure for Italy is 1,5% and for Holland 4,4%. In the 2002

survey, the respective percentage rises to 58% (42,7/73,7=0,58).

The following income quartiles verify the theory and the international experience. Consistent

with theory, direct and total participation rises in the second, third and fourth quartile of the

income distribution in both surveys. The rise continues concerning the top 5% of the income

distribution except in the 2nd survey where total participation stabilizes at around 92%

between the fourth quartile and the top 5% income group.

Similar behaviour is observed as regards the purchase of life insurance, with the correlation

between income and insurance purchase continuing all the way right to the top 5% of the

distribution with respect to both surveys. In the 1999 survey, indirect participation through life

insurance is higher than direct participation in all income quartiles except for the first quartile,

where it is almost similar, confirming that direct participation is more costly in terms of time

and money and probably more troublesome than the indirect type. However, this result

completely reverses in the 2002 survey, as people were enticed by the initial public offerings

of the investment companies and the opportunity for a quick profit, probably viewing

insurance as a less profitable and longer term investment alternative.

Concerning the retirement plans including provident funds in the 2nd survey, a similar pattern

emerges, except for a slight drop from the fourth quartile to the top 5% income group, which

is consistent with intuition. The retirement plans excluding provident funds in 2002, exhibit a

similar pattern with those of 1999 with which they are comparable.

Page 50: FINANCIAL ASSET PARTICIPATION BY CYPRIOT HOUSEHOLDS-UCY-M.SC. THESIS-2004

50

Even though the progressive increase pattern based on income is similar to what is observed

in Italy and Holland, a higher rate of increased participation is seen in the latter countries in

the fourth quartile and the top quintile than is the case in Cyprus. Assuming that income

imputation for missing data in the CySCF was correct, this can be attributed to the spread of

the equity culture in 1999 exerting a higher impact on the lower and especially the middle

income households (the white collar socioeconomic class c1 comprising civil servants, the

teaching profession etc). By contrast, higher income households (socioeconomic classes a

and b, comprising businessmen and professionals), being more educated and informed,

were not deceived by the rumours to such an extent and, in fact, being mostly incumbent

stockholders and/or active traders, (who had a low share acquisition cost prior to the bubble

and moreover had the opportunity to take private stock placements and/or list their own

companies), benefited more in comparison to other categories of the population. Indeed, by

October 1999, the majority of investors belonging to the upper income category, had

liquidated at least part of their portfolio and most of those who did so gained significantly.

This is corroborated in the findings of the Central Bank / Cyprus College study.

Table 6 looks at the participation decision according to financial wealth. Financial wealth, as

estimated, comprises liquid accounts such as checking, saving and deposit accounts,

government saving bonds and certificates, corporate bonds (calculated at their nominal

value, unless not stated, in which case the purchase value was used), warrants (calculated in

the same way as the corporate bonds) and stocks. Retirement accounts, mutual funds and

insurance investment policies have been excluded from the estimation of the wealth total.

The results again confirm the theory and the international experience: wealth is correlated

with participation in the stock market. Specifically, we observe participation rising as we

move to higher levels of the wealth distribution. This is evident in both surveys where the

percentages of both direct and indirect stockholding rise successively in line with wealth. The

same evidence emerges concerning life insurance. The only exception to this pattern, which

is again minor, concerns life insurance for the top 5% of the wealthy in the 1999 survey that

show a fall of a few percentage points compared to the fourth quartile. However, this cannot

detract from such a strong finding.

The retirement plans, when using the comparable definition, display a similar pattern in both

surveys, with rising participation up to the 3rd wealth quartile, thereafter falling. A similar

picture emerges when provident funds are added in the 2002 survey. It is notable that the

Page 51: FINANCIAL ASSET PARTICIPATION BY CYPRIOT HOUSEHOLDS-UCY-M.SC. THESIS-2004

51

differences from the mean (12,5% and 50,3% in 1999 and 2002 respectively) are never that

big across wealth quartiles.

***

Page 52: FINANCIAL ASSET PARTICIPATION BY CYPRIOT HOUSEHOLDS-UCY-M.SC. THESIS-2004

52

H. Econometric Analysis - Probit Model Estimation

In this section, after a brief exposition of the theoretical and methodological background

behind Probit Model estimation (in subsection 1), the model variables are defined (in

subsection 2), while the actual estimation results are set out in section 3.

1. Probit Model - Methodology

In our case, the Probit methodology concerns an econometric model examining discrete

binary choice variables and estimating the various probabilities for different values of the X

variable vector. As mentioned above, Probit econometric analysis separates out the

causative effect of each explanatory variable while keeping other factors constant (ceteris

paribus condition). Consequently, in this case, the model gives an estimate of the probability

that an individual with the given values of the explanatory variables will be a credit card or a

stockholder for example. Specifically, each β coefficient indicates the marginal effect of each

characteristic represented by the x variable on the probability F(X'i β) of participation in the

various categories of stockholding. In other words, the coefficients show the change in the

participation probability.

Assuming that the variable yi* is defined by yi*= β' X i + ui and also imposing the assumption

that ui ~ Ν (0,σ2), i.e. that the error term is normally distributed (implying that maximum

likelihood estimation is possible, since in non-linear models such as probits, logits and tobits

heteroscedasticity can lead to bias), yi* is not observable but the dummy variable yi is, since

there are dichotomous or binary data that distinguish only whether individual observations

are in one category (high values of yi ) or a second category (low values of yi), such that yi = 1

if yi* > 0, or yi = 0 if the value of yi* is otherwise.

In general, the Probit model involves nonlinear maximum likelihood estimation. It assumes

that yi* is an IID random variable in a manner that its probability can be computed from the

standardised cumulative normal probability function:

yi

Pi = F ( yi ) = 1 / 2 Π e – t 2/2

dt

-

Page 53: FINANCIAL ASSET PARTICIPATION BY CYPRIOT HOUSEHOLDS-UCY-M.SC. THESIS-2004

53

Since the Probi ( yi = 1 ) = F ( X'i β), by construction, the probability Pi of the event occurring

lies in the interval (0,1). Hence, this probability is measured by the area under the standard

normal curve, from - to yi, and the larger the value of yi, the more likely will be the

occurrence of the event. We assume that F (the cumulative distribution function) is a

monotonically increasing function of X'i β, therefore as X'i β gets larger the probability of y

being 1 increases. The probability Pi resulting from the Probit model can also be interpreted

as the conditional probability that the event Y will occur, conditional on the given values of the

explanatory variables X vector:

E (yi* I X i) = β' X i

In our case, an individual will be a stockholder, provided she or he possesses the required

characteristics (such as education, income or wealth).

2. Probit Model - Variable Definitions

The main purpose of the specific model variables or the functional form used is to compare

the 1999 and 2002 surveys, by noticing probable significant differences in the coefficients or

their signs. Using the 1999 and 2002 survey data, the model to be estimated, sets

stockholding as the dependent variable, while the independent regressors include age,

education, sex, marital status, income and wealth. These are the fundamental variables or

household characteristics considered widely in the literature as driving household financial

behaviour concerning stockholding. Probit models are estimated for the direct and indirect

(separate categories) as well as for the total participation decision. The results appear in

Tables 7-10, which present an econometric analysis of the stock participation decision

expressed by a function of a multivariable vector of household characteristics in the form of

dummy variables with the exception of income and wealth which enter the equation as

logarithmic continuous variables.

Thus, the Probit models, enable the isolation of the impact of each characteristic, while

controlling for or keeping constrained all the other characteristics. Therefore, the Probit

methodology allows the estimation of the effect of each contributing factor independently, as

explained above. For instance, university education is a characteristic associated with higher

participation but university graduates also tend to have higher income-earning capacity. In

Page 54: FINANCIAL ASSET PARTICIPATION BY CYPRIOT HOUSEHOLDS-UCY-M.SC. THESIS-2004

54

this case, the Probit estimations enable the isolation of the effect of education from that of

income in the stockholding decision.

In order for the Probit methodology to be applied correctly some “control (or “benchmark” or

“base”) variables are excluded, against which the differential effect of the included variables

is compared. In other words, the excluded variable characteristics denote the “base

household”. For example, a household possessing college education has exactly the same

characteristics as the “base household”, with the difference that the latter belongs to the less

than high school education category (excluded). In this respect, the marginal effect of having

college education raises the probability of the head of this household holding stocks directly

by 20 percentage points at a 1% level of statistical significance, associated with a 99%

confidence interval (Table 7).

The estimated Probit models comprise three interaction terms, one each for income and

wealth and another for college education. The interaction terms are variables generated by

multiplication of two other variables. As implied by the definition, a positive and significant

interaction term coefficient indicates the additional effect arising from the interaction of the

two variables. In our case, the interaction terms are composed of the yearly dummy variable

and selected variables such as the income, wealth and college education mentioned above.

The yearly dummy, which takes the value of 1 for all observations in 2002 and of 0 for all

observations in 1999, allows us to pool together all 1999 and 2002 household survey data. It

expresses the constant difference in participation probability among the two survey years,

associated with identical household characteristics. In fact, the yearly dummy represents the

constant or the intercept term and has the following meaning: Given or controlling for all the

household characteristics (i.e. holding the values of all the X regressors constant) it answers

the question whether it is more likely for a household to invest in stock in 1999 or in 2002. In

other words, the yearly dummy is used to check whether coefficients differ across years. By

itself, the year has no significance or makes no difference but the independent variables may

have significance. If the yearly dummy variable is not statistically significant, then there is no

difference between the two survey years. But, if it turns out statistically significant then the

year does indicate an effect. In this case, we expect to obtain a high estimate for the

intercept term confirming the statistical cross tabulation findings (Tables 3-6).

Page 55: FINANCIAL ASSET PARTICIPATION BY CYPRIOT HOUSEHOLDS-UCY-M.SC. THESIS-2004

55

The year effect is demonstrated by examining the other variables (their marginal effects),

provided they are statistically significant. The selected variables indicate the base year

(1999) effects, i.e. they provide an estimation for the effect of income, wealth or education in

the base year of the merged data file (coefficient a1). The other coefficient (a2) disappears for

the 1999 data due to the fact that 1999 takes a zero value in the yearly dummy multiplication

forming the interaction term. For instance, in order to test whether the coefficient on income

has changed across the two years, not only the income variable is included but also the

product of the yearly dummy times income (i.e. the income interaction term). Allowing for

that, the interaction term indicates the slope. Thus, if the income effect (coefficient a1) differs

across the two years, this will be indicated by a statistically significant coefficient on the

pertinent interaction term (a2). More analytically, the coefficient of the interaction term (a2)

shows the differentiation of the participation probability between the two years, i.e. in this

case, the net change in 2002. Adding the two coefficients (a1+a2) indicates the 2002 total

year effect. The following equation makes the above clear:

(a1 + a2 Yearly Dummy) * ln Income or,

a1 * In Income + a2 (Yearly Dummy * In Income)

The bold product term above is the Interaction term. If the interaction term is not statistically

significant, then the effect of the relevant (income) variable is not statistically different across

the two years. However, if the interaction term is statistically significant with a positive sign,

then it indicates that the influence of the variable rose in 2002. If the sign of the interaction

term is negative, then the influence of the variable is indicated to decrease in 2002 compared

to 1999.

3. Model Estimation and Empirical Results

The estimated coefficients can be considered at the 10%, 5% and 1% level of statistical

significance, this being denoted by one, two or three asterisks respectively. Small P-values

provide evidence against the null hypothesis (Ho) since they indicate that the data outcome

occurs with a small probability if Ho is true. Thus, if P-values are small, either the Ho is false

or the sample is unusual. In particular, if the P-value is less than a=5% then the result is

statistically significant at the 5% significance level. Alternatively, if the P-Value is less than

a=1%, then the outcome is statistically significant at the 1% significance level. If the P-value

Page 56: FINANCIAL ASSET PARTICIPATION BY CYPRIOT HOUSEHOLDS-UCY-M.SC. THESIS-2004

56

is greater than a=10% then the result is not statistically significant even at the 10%

significance level and the Ho is accepted.

Adopting the 10%, 5% and 1% significance levels as the reference values for all statistically

significant results, with the 1% level suggesting strong significance and the 10% marginal

significance, we observe the following from the empirical estimations of Table 7: First, the

expected big jump in the intercept term is confirmed by a statistically significant yearly

dummy, showing the upward shift in stock participation due to the market entry of various

investment companies following the 1999 survey, i.e. owing to supply side factors rather than

a yearly trend. Being 70 years old or more reduces the probability of holding stocks directly

by 15 percentage points (at the 1% significance level). The respective Probit estimates for

Italy and Holland, considered in the literature, confirm the poor relationship between age and

participation. Consequently, it may be concluded that, even though an age pattern appears in

the statistical cross tabulations, both for Cyprus and for the other countries considered, this

result is not robust, as it is not confirmed by the Probit estimations. As mentioned before, it is

highly likely that the income factor lies behind the middle age groups and this is the hidden

explanatory variable that drives stockholding behaviour rather than age.

Education also plays a significant role for participation in the Stock Exchange: having college

or high school education raises the probability of the household head holding stocks directly

by 20 and 7 percentage points at a 1% and 5% level of statistical significance respectively.

These regression results confirm the statistical cross tabulation outcome (Table 4) that

education does indeed play a positive role for participation. Indeed, there is theoretical

justification for such an outcome: households with a higher educational level face lower

income uncertainty and borrowing constraints, hence, they are more prone to investing in

stocks. The statistical significance (at the 1% level) of the probability for participation rising

by more than 6 percentage points due to being male rather than female, simply confirms the

empirical fact. Marriage does seem to play a role in direct stock market participation, since

this characteristic comes out as marginally statistically significant at the 10% level.

Wealth also has a positive marginal effect on direct stock participation, raising the probability

of this event by more than 15 percentage points at a 1% level of statistical significance,

backed by a very small standard error and a consequent strong Z value. The positive impact

of wealth on stock participation is confirmed by the experience of Italy and Holland as well. It

Page 57: FINANCIAL ASSET PARTICIPATION BY CYPRIOT HOUSEHOLDS-UCY-M.SC. THESIS-2004

57

is noted that wealth is used here as an exogenous variable, even though it could also be

considered as endogenous, since a household investing in risky assets such as stocks has a

higher probability of ending up with increased wealth at the beginning of the next period.

Income just falls within the acceptance region, at the limit of the confidence interval with 95%

probability (Z value=1.96) but with a negative sign. This result goes against theory but tax

evasion may have led households to declare lower incomes and explain this outcome. By

contrast, the income interaction term is statistically significant, at the 5% significance level,

with a positive sign, indicating that the influence of the income variable rose in 2002 with a

net change of 4 percentage points. In other words, income became more important for

continuing stock market participation, since poorer income households tended to exit rather

than remain in the market, following their rushed entry in the period 1999-2000 and the

losses suffered with the unwinding of the bubble. Nonetheless, adding the base year log

income effect (minus 3,4 percentage points) to the net change recorded by the income

interaction term (4 percentage points), indicates a positive overall income effect on direct

stock participation of 0,6 percentage points.

By contrast, the sign of the wealth interaction term is negative, indicating that the effect on

direct stockholding of a unit change in log wealth decreased by almost 10 percentage points

(at the 1% significance level) in 2002 compared to 1999. Hence, in 2002 wealth does not

seem to be equally important for participation as it was in 1999. Specifically, adding the base

year log wealth influence (15,3 percentage points) to the net change recorded by the wealth

interaction term (minus 9,7 percentage points) still indicates a positive total 2002 effect on

direct participation of 5,6 percentage points. However, households with the same

characteristics except financial wealth are differentiated to a lower extent in their participation

behaviour in 2002 relative to 1999. The fact that wealth appears to play a less significant role

in 2002 is attributed to the optimistic outlook prevailing in 1999 when more wealthy

households tended to enter the market than poorer ones. In 2002, when the situation turned

sour and there was widespread disappointment with the Cyprus Stock Exchange, the

differentiation across wealth groups with respect to the poor declined, as evidenced by the

negative sign of the wealth interaction term. The college education interaction term is not

statistically significant, hence university training did not exert a statistically different role

across the two years covered by the surveys.

Page 58: FINANCIAL ASSET PARTICIPATION BY CYPRIOT HOUSEHOLDS-UCY-M.SC. THESIS-2004

58

Table 8 presents the results of the model estimation regarding total stockholding. As

predicted, the yearly dummy verifies the upward shift in stock participation found in the

statistical cross tabulations and turns out statistically significant at the 1% level. Stock

disinvestment (since indirect participation, as defined, comprises insurance and retirement

accounts) begins to occur at the earlier age of 60 with retirement, with a reduction in the

probability of holding stocks indirectly by almost 22 percentage points, while at 70 this rises

to almost 47 percentage points. High school and college education remain strong

explanatory factors for total stock participation, raising the probability of this outcome by

around 10 and 19 percentage points respectively at a 1% level of statistical significance. The

total stock participation probability for men rather than women is even higher (at 8

percentage points, at a 1% level of statistical significance than in the case of direct

stockholding. Marriage also has a positive marginal effect on total stock participation, raising

the probability of this event by 17 percentage points at a 1% level of statistical significance.

This is a much stronger result than in the case of direct stock participation and should not be

surprising, since indirect stockholding, being regarded as safer, comes with a promise of

raising the family welfare, in addition to the security of the living standard afforded by the

insurance element. Income and wealth raise the probability for participation by 4 and 5

percentage points at a 5% and 1% levels of statistical significance respectively.

The wealth interaction term is the only one among the interaction terms used that appears

marginally statistically significant (at the 10% significance level) with a negative sign,

indicating that the effect on total stockholding of a unit change in log financial assets

decreased by around 1,5 percentage points in 2002 compared to 1999. This result, even

though much weaker, is in the same direction as the one obtained in the case of direct

stockholding above, and, confirms the tendency for a lower differentiation across wealth

groups regarding household stock participation behaviour in 2002 relative to 1999. As

already noted, none of the other interaction terms is statistically significant, indicating that the

impact of the relevant variables (income, college education) is not statistically different

between the two survey years.

Table 9 shows that the only evident relationship between age and indirect participation

through life insurance concerns the age groups 60-69 and over 70, both of which display a

lower probability for participating by 24 and 28,8 percentage points respectively at 1% level of

statistical significance. The coefficient signs for these age groups have come out negative,

Page 59: FINANCIAL ASSET PARTICIPATION BY CYPRIOT HOUSEHOLDS-UCY-M.SC. THESIS-2004

59

indicating that they are less likely to participate in any manner, either direct or indirect. This

confirms the intuitive reasons suggested previously: older people probably lack the

knowledge and /or the education to go for insurance policies and most probably perceive no

such need, besides facing refusal by insurance companies. As for the rest, being married and

having college education also prompts insurance purchase at a 5% level of statistical

significance. Moreover, the higher the wealth the higher (by 2 percentage points) appears to

be the probability of insurance purchase at a 1% significance level. An even stronger

probability association is revealed between rising income and the holding of insurance: the

pertinent coefficient shows a marginal rise of 10,7 percentage points at 1% level of statistical

significance. Linking this result with the theory presented in the literature review section, the

correlation between income and investment insurance purchase is easily rationalized: lower-

income households are more likely to face borrowing or income constraints and therefore

cannot proceed with an optimal allocation of their life cycle earnings across periods. This

renders insurance purchase perhaps an unnecessary luxury and lower income households

tend to prefer to invest in safer financial or real assets. According to Heaton-Lucas (1997),

the very poor households tend not to buy insurance or invest in stocks in any manner, not

least because of the relatively high entry or informational costs. The yearly dummy is also

strongly statistically significant at the 1% significance level, confirming the expected big jump

in the intercept term. The college education interaction term appears to be statistically

significant (at the 10% significance level), indicating a net change in 2002 in the probability of

insurance purchase by almost 9 percentage points compared to 1999. This development

corroborates the influence of education on indirect participation.

The income interaction term turns out to be statistically significant (at the 1% significance

level), with a negative sign, which implies a decrease in the probability of insurance purchase

(and thus indirect stock participation) by 11 percentage points in 2002 compared to 1999.

Adding the base year log income influence (10,7 percentage points) to the net change

recorded by the income interaction term (minus 10,9 percentage points) indicates a slightly

negative, close to zero total income 2002 effect on insurance purchase, pointing to the

corresponding decline in the influence of income on indirect stockholding via insurance: while

in 1999 we observe a positive correlation between income and insurance participation, this

disappears in 2002. Hence, in 2002, high-income households are not more interested than

low-income households to buy insurance, as they were in 1999. In other words, the

differentiation across income groups with respect to indirect stock market participation via

Page 60: FINANCIAL ASSET PARTICIPATION BY CYPRIOT HOUSEHOLDS-UCY-M.SC. THESIS-2004

60

insurance purchase is obliterated, as evidenced by the negative sign of the income

interaction term. This development is similar to the one observed above with regard to the

decline of the effect of wealth concerning stockholding. It seems that the collapse of the

Stock Exchange, which caused the insurance company investment funds to suffer great

losses prompted a general disappointment and households with higher incomes or available

cash on hand, altered their investment behaviour, turning towards safer real assets in

building their asset portfolios. Evidently, the shock of the Stock Exchange collapse has

exerted a homogenising or harmonising influence on household financial behaviour at least in

some respects. Cypriot households now seem to be making their portfolio choices with a

longer horizon in mind (i.e. purchasing land, which is a rather illiquid but safer asset). This

was a natural and rather foreseeable development following the collapse of the Stock

Exchange, since buying land and property is in the Cypriots’ mentality and goes back in the

tradition of family economic culture. Of course, this pushed property prices up, raising the

return on this type of investment, even if the very high current prices could perhaps

eventually entail a lower (but nonetheless safer) expected return in the medium and the long

term. Hence, the construction and land price boom that we have been witnessing over the

last few years. This behaviour revokes the moral of the English proverb “Once bitten twice

shy”! Hence, the current complete lack of confidence in the Cyprus Stock Exchange, which

renders its revival an almost impossible task.

In the case of the Probit model for indirect stock participation via retirement accounts, the

sample has been adjusted, excluding household heads over the age of 60, since these

people are already in retirement age. As revealed in Table 10, high school and university

graduates are around 6,3 and 9,7 percentage points respectively more likely to hold

retirement accounts at a 1% significance level. Being male, rather than female raises the

marginal probability of participation through pension plans by about 9,4 percentage points at

the 1% significance level. This, of course, is largely related to the differing employment rate

(Gainfully Employed Aged 15-64 / Population Aged 15-64), which was 78,8% for males and

60,2% for females in 2003. None of the age groups up to the age of 60 record statistically

significant results, indicating that age is not a significant characteristic for holding a

retirement account. Marriage, rather surprisingly, does not appear to play a role for pension

fund participation even though this characteristic comes very close to being statistically

significant at the 10% level. In the relevant model regression, income comes out as a

significant explanatory variable, exerting a positive marginal effect that raises retirement

Page 61: FINANCIAL ASSET PARTICIPATION BY CYPRIOT HOUSEHOLDS-UCY-M.SC. THESIS-2004

61

account participation by about 3,3 percentage points at the 5% significance level. By

contrast, financial wealth does not emerge as a significant factor, as the richer households

are more likely to prefer and actually possess other investments to cover their retirement age

needs.

Further, the yearly dummy (the intercept term) is not statistically significant either, since there

was no notable shift in behaviour concerning pension plans between the 1st and 2nd survey.

This result is also confirmed in the statistical cross tabulations. The college education

interaction term indicates a net change, in 2002, in the probability of holding a retirement

account by 7,4 percentage points, at the 10% significance level. On the contrary, the income

interaction term records a net decrease in 2002 in the probability of holding a retirement

account by around 2,8 percentage points, at the 10% significance level, compared to 1999.

This is a much weaker result than the corresponding one obtained in the case of participation

via insurance, but nevertheless, both the direction and the explanations (relating to the

decline of the differentiation across income groups) remain the same. In any case, adding the

base year log income effect (3,3 percentage points) to the net change recorded by the

income interaction term (minus 2,8 percentage points), indicates a still positive overall 2002

income effect on indirect stock participation through retirement accounts of half a percentage

point.

***

Page 62: FINANCIAL ASSET PARTICIPATION BY CYPRIOT HOUSEHOLDS-UCY-M.SC. THESIS-2004

62

I. Conclusion

This study has examined the financial asset behaviour of Cypriot households, with a special

focus on stockholding, drawing on data from the CySCF of 1999 and 2002. The pertinent

years turned out to be pivotal for the structure of household portfolios in Cyprus. Perhaps the

most striking difference between the two surveys lies in the enormous increase in

stockholding. The timing of the 1st survey coincided with a period of stock market euphoria

that prompted a massive entry of households, raising participation to levels comparable to

those of developed countries, which had a much smoother and longer transition into the era

of the “equity culture”. Thus, about a quarter of Cypriot households found themselves holding

stocks directly, and half the households participating either directly or indirectly. The 2nd

survey captured the equally massive effect on participation exerted by the investment

companies, which increased direct participation to over half the population and indirect to

almost three quarters of the households. The subsequent instability and collapse of the stock

market confirmed the existence of a bubble and of stock mispricing way beyond justification

by the fundamentals.

Participation in direct stockholding is higher than in other countries in general and is

unusually high for the very young. This has been further exacerbated by 2002, when the

direct stockholder base more than doubled relative to 1999 to exceed half the population of

Cyprus.

In particular, direct stockholding is observed across age groups, with the exception of those

over seventy for whom the percentage diminishes drastically from the average. Even though

a hump-shaped age pattern appears in the statistical cross tabulations, both for Cyprus and

the other countries considered, this result is not robust and is not confirmed by the Probit

estimates. Thus, age plays an unclear and inconsistent role, if any. It is highly probable that

the income factor, through its correlation with age, is hidden behind the middle age groups

wherein stockholding and of course income peak.

Education has also been shown econometrically to play a positive role, confirming what is

observed in the relevant cross tabulation (Table 4). The high participation rate of persons with

less than high school education is notable and is attributed to the stock market frenzy, the

family income support as well as the less stringent borrowing constraints due to bank policy.

Page 63: FINANCIAL ASSET PARTICIPATION BY CYPRIOT HOUSEHOLDS-UCY-M.SC. THESIS-2004

63

Income and wealth have clearly emerged econometrically as major factors for stock market

participation and these accords both with theory and the international experience. However,

there has been a significant development between the 1st and 2nd survey: the econometric

regressions indicate that in 2002, the differentiation across income and wealth groups with

respect to stockholding behaviour has declined, on the whole, in comparison to 1999. It is

evident that the shock of the Stock Exchange collapse, following the 1999 bubble, has

exerted a homogenising or harmonising influence on household financial behaviour, at least

in some respects.

Marriage and being male constitute additional household characteristics that exert a positive

effect on direct and total stock participation. This is not surprising, since stockholding comes

with a promise of raising the family welfare in the longer term.

The analysis suggests that the typical portrait of the Cypriot Stock Exchange direct and

indirect participant is that of a middle-aged (in his 40’s), married man, who tends to have

higher education and relatively high income and wealth, belonging mostly to a household

comprising up to four members, with two income-earning members in the family in the

majority of cases.

In the light of the above, and provided we take account of the 1999 Stock Exchange bubble,

it is concluded that Cypriot household attitudes and preferences regarding stockholding

behaviour are consistent with mainstream theoretical predictions and the empirical findings

based on international trends.

***

Page 64: FINANCIAL ASSET PARTICIPATION BY CYPRIOT HOUSEHOLDS-UCY-M.SC. THESIS-2004

64

J. Βibliography:

1. Antoniou Georgia, Christiana Argyridou, Michael Haliassos, Alex Karagrigoriou,

George Kyriacou, Michalis C. Michael, Maria Papagheorgiou, and George Syrichas

(2004). “Assets and debts of Cyprus Households: Changes between the 1999 and

2002 Cyprus Survey of Consumer Finances”, mimeo, University of Cyprus.

2. Bertaut Carol (2002). “Equity Prices, household Wealth and Consumption Growth”,

International Finance Discussion Papers Number 724, Board of Governors of the

Federal Reserve System, USA.

3. Bilias Yiannis and Michael Haliassos (2004). “Distribution of gains from access to

stocks”, mimeo, University of Cyprus.

4. Central Bank of Cyprus: Quantification Study conducted by the Cyprus College

(Pampos Papageorghiou) (2001). “Cyprus Households and the Stock Exchange”.

5. Couch Robert B. (2001). “The Stock-Home Decision and household Savings”

Graduate School of Industrial Administration, Carnegie Mellon University, Pittsburgh,

PA.

6. Carroll Christopher D. (2002). “Portfolios of the Rich” in “Household Portfolios”,

Cambridge, MA: MIT Press, pp. 389-429.

7. Fatih Guvenen (2003). “Does Stockholding Provide Perfect Risk Sharing”, University

of Rochester, Rochester, NY.

8. Gong-meng Chen, Michael Firth, Oliver Meng Rui (2000). “Stock market linkages:

Evidence from Latin America “, Journal of Banking and Finance 26 (2002) pp 1113-

1141.

9. Guiso Luigi, Michael Haliassos, Tullio Jappelli (eds) (2002). “Household Portfolios”,

Cambridge, MA: MIT Press.

10. Guiso Luigi, Michael Haliassos, Tullio Jappelli (2002). “Household Portfolios: An

International Comparison” in “Household Portfolios”, Cambridge, MA: MIT Press, pp.

1-24.

11. Guiso Luigi, Michael Haliassos, Tullio Jappelli, (April 2003). “Household Stockholding

in Europe: Where do we stand and where do we go?” Economic Policy, pp 117-164.

12. Gujarati D. (1988). “Basic Econometrics”, McGraw-Hill.

13. Gujarati D. (1992). “Essentials of Econometrics”, McGraw-Hill.

Page 65: FINANCIAL ASSET PARTICIPATION BY CYPRIOT HOUSEHOLDS-UCY-M.SC. THESIS-2004

65

14. Mishkin F. S. (2001). “The economics of money, banking and financial markets”,

Longman.

15. Haliassos Michael and Carol Bertaut (1995). “Why do so few hold stocks?”, The

Economic Journal , 105, pp. 1110-29.

16. Haliassos Michael., C. Hassapis, A. Karagrigoriou, G. Kyriakou, M. Michael and G.

Syrichas (2001). “Assets of Cyprus Households: Lessons from the First Cyprus

Survey of Consumer Finances “, mimeo, University of Cyprus.

17. Haliassos Michael and Christis Hassapis (2001). “Non-expected Utility, Saving and

Portfolios”, The Economic Journal, 111, pp. 69-102.

18. Haliassos Michael (2002). “Stockholding: Recent lessons from theory and

computations” in Guiso Luigi, Michael Haliassos, Tullio Jappelli (eds), Stockholding

in Europe, Houndmills and New York: Palgrave Macmillan Publishers.

19. Haliassos Michael, Alexander Michaelides (2002). “Calibration and computation of

Household portfolio models”, in Household Portfolios, Cambridge, MA: MIT Press,

pp. 55-102.

20. Haliassos Michael and Christis Hassapis “2002”. “Equity culture and household

behaviour “.Oxford Economic Papers, 54, pp. 719-745.

21. Haliassos Michael., C. Hassapis, A. Karagrigoriou, G. Kyriakou, M. Michael and G.

Syrichas (2003) ”Debts of Cyprus Households: Lessons from the first Cyprus Survey

of Consumer Finances”, mimeo, University of Cyprus.

22. Pindyck R. S. and D. L. Rubinfeld (1998). Econometric models and economic

forecasts, McGraw-Hill.

23. Qiu Jiaping (2002). “The Determinants of Households’ Direct and Indirect Stock “

University of Toronto.

24. Romer D. (2001). “Advanced Macroeconomics”, McGraw-Hill.

Page 66: FINANCIAL ASSET PARTICIPATION BY CYPRIOT HOUSEHOLDS-UCY-M.SC. THESIS-2004

66

Table 1: Descriptive Statistics of Population Sample

1999 2002

Mean, % Mean, %

Age, Years 46,5 48,7

Less than High School 35,4 37,3

High School 37,7 32,6

College 26,9 30,1

Married 88,4 87,8

Single 4,3 3,1

Male 65,0 60,9

Small HH=2-4 Members 66,7 68,2

Big Household >4 Members 27,9 24,6

One Income Household 34,6 29,9

Two Income Household 48,6 53,0

Three Income Household 9,5 8,9

Wage Earner 60,4 60,3

Self-Employed 17,6 24,3

Unemployed 2,0 1,5

Pension Receivers 13,8 11,3

Income, in C£ 14194 15109

Financial Assets, in C£ 14512 18254

Valid N (Weighted) 1362 1195

Actual Number of Households 1097 897