Abstract National savings play an i

Abstract
National savings play an important role in the economic development of
many developing countries, especially if capital markets are weak. In this
paper we investigate the determinants of national savings for Fiji. We use
the recently developed autoregressive distributed lag modelling approach,
shown to provide robust estimates in small samples, to model Fiji's
savings behaviour. Our results indicate that the life cycle hypothesis helps
explain Fiji's savings behaviour. The key finding is that economic growth
has the biggest impact on savings rate, suggesting that savings will
increase with an increase in economic growth.
Introduction
National savings play an important role in the economic development of
many developing countries with limited avenues for financing investment
projects. These countries, characterised by weak capital markets, are
forced to rely heavily on domestic savings to finance such development
projects. It is the shortage of national savings that is one of the most
critical constraints on economic growth in most developing countries
(Krieckhaaus, 2002). High national savings is likely to stimulate national
investment, which in turn will provide the basis for more rapid economic
growth (Nurkse, 1953).
Fiji Islands is a small isolated country with a weak capital market and is
dependent mainly on national savings for financing investment in the
country. It has a population of around 800,000 and is regarded as a lower-
middle income country with per capita income of US$2110. Fiji's savings
rate, however, has been declining throughout the 1980s and most of the
1990s and started to rise in the last four years (Figure 1). Total national
savings as a percentage of gross national product has averaged only 17%
in the 1968 to 2000 period. Given the importance of savings and the
apparent low and downward trend over the last two decades, this study
attempts to examine the factors that have affected savings in Fiji.
Figure l: Savings Rate in Fiji, 1968 to 2000
National Savings rate
15
10 /
5 l
I Year
l. .- . -. _ _ ..
. _ - .- ----- . ---.. - .. - . -- .- . . .. . - - .- .- - -- . . . -. - . . . .- .. .. . - - - -, .
Source: International Financial Statistics, 200 1.
There are two other reasons for undertaking the present study. First, with
the exception of a few recent studies (Agrawal, 2001; Sarantis and
Stewart, 2001) on savings behaviour in developing countries, many
studies (Khan et al. 1994; Cook, 1995; Cardenas and Escobar, 1998;
Hussain and Brookins, 2001, amongst others) are open to criticism on
econometric grounds for they ignore the fact that variables used in
modelling saving rates are likely to be non-stationary and potentially
integrated. It follows then that inferences made concerning long-run
elasticities are potentially flawed and misleading, as noted by Yule
(1926) and Granger and Newbold (1974)~' Given this limitation in the
literature on the determinants of savings rate, this study attempts to
1 Yule (1926) suggested that regression based on non-stationary series are known as 'nonsense' regression
while Granger and Newfold (1974) termed this problem 'spurious regression'.
correct for spurious regression by using cointegration analysis within the
recently developed autoregressive distributed lag (ARDL) framework
(Pesaran and Shin 1998 and Pesaran, et al. 2001), allowing estimates to
be made of both short-run and long-run relationships.
Second, although studies on savings for groups of countries and regions
have proliferated (see, inter alia, Giovannini, 1983; Edward, 1996; Callen
and Thimann, 1997, Dayal-Gulhati and Thimaun, 1997; Hussain and
Brookins, 2001), the focus has seldom been on a specific country. While
this lacuna may be due to the lack of appropriate time series data on
developing countries, no such problem exists for Fiji. The other issue is
that given the use of cross-sectional data analysis in many studies, one
cannot properly infer from observations across countries at a point in time
what might happen in a country over time. A time series study overcomes
this problem, providing a stronger basis for informed policy making.
The aim of this paper is to delineate the short- and long-run relationships
between savings, real interest rate, income, current account deficits and
age dependency ratio in Fiji using cointegration and error correction
models over the period 1968-2000.
The balance of the article is as follows: the next section discusses the
theoretical framework and provides a brief review of the literature; this is
followed by a discussion of the econometric methodology to be applied,
while the penultimate section presents the empirical results followed by
the conclusion.
Theoretical Framework
Following on fiom various studies on determinants of savings
(Modigliani, 1963; Giovannini, 1983; Doshi, 1994; Edwards, 1996; and
Agrawal, 200 1) the Modigliani-Brumberg-Ando 'Life Cycle' hypothesis
is applied to examine the savings pattern in Fiji. The 'Life cycle'
hypothesis contends that current savings decisions of households are a
consequence of an act to distribute their lifetime consumption evenly over
their lives so as to maintain the same lifestyle during retirement
(Modigliani 1970). This hypothesis implies that savings rate depends on
growth of income andlor the age structure of the population.
With respect to the growth rate of income, the life cycle theory postulates
a positive relationship between savings rate and income growth rate,
since higher income growth makes the young richer than the old; hence
the young will be saving more than the old will be dis-saving. Recent
empirical studies show this relationship to be true for some developing
countries (Agrawal 2001). However, it is also possible that the effect of
the rate of economic growth will be positive only to the extent that
households, on average, accumulate wealth when they are young in order
to dispose of this wealth when they are old. It follows then that growth
rates may adversely affect savings rate in countries where households
spend before they earn by borrowing. Furthermore, the rate of growth
will have little or no effect on national savings if the age dependency
ratio - defined as the population younger than 15 years of age plus
population over 65 years old as a proportion of working age population -
is high. Studies by Carroll (1 992) and Carroll and Weil (1 994) justifiably
argue that a negative relationship between the two variables is also
possible. Their argument is based on the notion that, other things being
equal, an exogenous increase in growth level may make consumers feel
wealthier and thus increase their consumption levels, leading to a lower
savings ratio. Against this background, the expected sign of the savings
growth rate is ambiguous.
The age structure of the population is also seen as an important
determinant of savings in the life cycle model since a declining share of
dependent population to working population is believed to correlate with
higher savings rates. However, the existing empirical evidence on this
relationship is mixed. For instance, Leff (1969) showed that this is true
for both developed and the developing countries, while Ram (1982),
Doshi (1994) and Kelly (1988), found no or very weak evidence of this
relationship between savings rate and the age dependency ratio in
developing countries. Agrawal (2001) found a significant negative
relationship between savings and the age dependency ratio for three of
the seven Asian countries.
Apart kom the life cycle hypothesis, following Agarwal (2001), Hussain
and Brookins (2001) and others, we employ current account deficits and
real interest rates as additional explanatory variables. The relationship
between national savings and current account deficits as a share of gross
national product is clear cut - high foreign savings leads to a fall in
national savings, since they are substitutes.
The real interest rate, defined as the difference between the nominal
interest rate and the expected inflation rate, is the reward for postponing
present consumption into the fbture. The net iinpact of the real interest
rate on savings is unclear both theoretically and empirically, though the
usual presumption is that the total effect is positive. One of the main
reasons for the ambiguity follows from the distinction between the
resulting income and substitution effects from an interest rate change (see
Agrawal, 200 1 ; Giovannini, 1983, 1985). The overall iinpact of interest
rates on savings could be positive or negative depending on which effect
is larger. Hence, the direction of the relationship is an empirical issue
depending upon the relative strength of income and substitution effects. If
the substitution effect outweighs the income effect savings would rise
with an increase in real interest rate. Fry (1980, 1995) found that interest
rates have a positive but small impact on savings while Giovannini (1 983,
1985) found no statistically significant relationship between savings and
interest rates. Agrawal (2001) found mixed results on this relationship
for four Asian co~ntries.~
Model specification
Based on the theory reviewed in the previous section, we posit the
following specification to deteimine the national savings rate for Fiji.
where S, is the savings rate which is calculated as a ratio of gross national
savings to gross national product; GR,is the growth rate of per capita
Taiwan's savings were significantly related to interest rate elasticities (-
0.190), whereas savings from Malaysia, Korea and India were unaffected
by real interest rates.
income; IR, is the real interest rate3; CAD, is the current account deficit as
a proportion of GNP; and DEC is the share of dependents (below 15 and
above 65) to the

Kết quả (Anh) 1:

Abstract National savings play an important role in the economic development of many developing countries, especially if capital markets are weak. In this paper we investigate the determinants of national savings for Fiji. We use the recently developed autoregressive lag distributed modelling approach, shown to provide robust estimates in small samples, to model Fiji's savings behaviour. Our results indicate that the life cycle hypothesis helps explain Fiji's savings behaviour. The key finding is that economic growth has the biggest impact on savings rate, suggesting that savings will increase with an increase in economic growth. Introduction National savings play an important role in the economic development of many developing countries with limited avenues for financing investment projects. These countries, characterised by weak capital markets, are forced to rely heavily on domestic savings to finance such development projects. It is the shortage of national savings that is one of the most critical constraints on economic growth in most developing countries (Krieckhaaus, 2002). High national savings is likely to stimulate national investment, which in turn will provide the basis for more rapid economic growth (Nurkse, 1953). Fiji Islands is a small isolated country with a weak capital market and is dependent mainly on national savings for financing investment in the country. It has a population of around 800.000 and is regarded as a lower- middle income country with per capita income of US $ 2110. Fiji's savings rate, however, has been declining throughout the 1980s and most of the 1990s and started to rise in the last four years (Figure 1). Total national savings as a percentage of gross national product has averaged only 17% in the 1968 to 2000 period. Given the importance of savings and the apparent low and downward trend over the last two decades, this study attempts to examine the factors that have affected savings in Fiji. Figure l: Savings Rate in Fiji, 1968 to 2000 National Savings rate 15 10/ 5 l I Year l.. - . -. _ _ .. . _ - .- ----- . ---.. - .. - . -- .- . . .. . - - .- .- - -- . . . -. - . . . .- .. .. . - - - -, . Source: International Financial Statistics, 200 1. There are two other reasons for undertaking the present study. First, with the exception of a few recent studies (Agrawal, 2001; Sarantis and Stewart, 2001) on savings behaviour in developing countries, many Studies (Khan et al. 1994; Cook, 1995; Cardenas and Escobar, 1998; Hussain and Brookins, 2001, amongst others) are open to criticism on econometric grounds for they ignore the fact that variables used in modelling saving rates are likely to be non-stationary and potentially integrated. It follows then that inferences made concerning long-run elasticities are potentially flawed and misleading, as noted by Yule (1926) and Granger and Newbold (1974) ~ ' Given this limitation in the literature on the determinants of savings rate, this study attempts to 1 Yule (1926) suggested that regression based on non-stationary series are known as 'nonsense' regression while Granger and Newfold (1974) termed this problem 'spurious regression'. correct for spurious regression by using cointegration analysis within the recently developed autoregressive distributed lag (ARDL) framework (Pesaran and Shin 1998 and Pesaran, et al. 2001), allowing estimates to be made of both short-run and long-run relationships. Second, although studies on savings for groups of countries and regions have proliferated (see, inter alia, Giovannini, 1983; Edward, 1996; Callen and Thimann, 1997, Dayal-Gulhati and Thimaun, 1997; Hussain and Brookins, 2001), the focus has seldom been on a specific country. While this lacuna may be due to the lack of appropriate time series data on developing countries, no such problem exists for Fiji. The other issue is that given the use of cross-sectional data analysis in many studies, one cannot properly infer from observations across countries at a point in time what might happen in a country over time. A time series study overcomes this problem, providing a stronger basis for informed policy making. The aim of this paper is to delineate the short- and long-run relationships between savings, real interest rate, income, current account deficits and age dependency ratio in Fiji using cointegration and error correction models over the period 1968-2000. The balance of the article is as follows: the next section discusses the theoretical framework and provides a brief review of the literature; this is followed by a discussion of the econometric methodology to be applied, while the penultimate section presents the empirical results followed by the conclusion. Theoretical Framework Following on fiom various studies on determinants of savings (Modigliani, 1963; Giovannini, 1983; Doshi, 1994; Edwards, 1996; and Agrawal, 200 1) the Modigliani-Brumberg-Ando 'Life Cycle' hypothesis is applied to examine the savings pattern in Fiji. The 'Life cycle' hypothesis contends that current savings decisions of households are a consequence of an act to distribute their lifetime consumption evenly over their lives so as to maintain the same lifestyle during retirement (Modigliani 1970). This hypothesis implies that savings rate depends on growth of income andlor the age structure of the population. With respect to the growth rate of income, the life cycle theory postulates a positive relationship between savings rate and income growth rate, since higher income growth makes the young richer than the old; hence the young will be saving more than the old will be dis-saving. Recent empirical studies show this relationship to be true for some developing countries (Agrawal 2001). However, it is also possible that the effect of the rate of economic growth will be positive only to the extent that households, on average, accumulate wealth when they are young in order to dispose of this wealth when they are old. It follows then that growth rates may adversely affect savings rate in countries where households spend before they earn by borrowing. Furthermore, the rate of growth will have little or no effect on national savings if the age dependency ratio - defined as the population younger than 15 years of age plus population over 65 years old as a proportion of working age population - is high. Studies by Carroll (1 992) and Carroll and Weil (1 994) justifiably argue that a negative relationship between the two variables is also possible. Their argument is based on the notion that, other things being equal, an exogenous increase in growth level may make consumers feel wealthier and thus increase their consumption levels, leading to a lower savings ratio. Against this background, the expected sign of the savings growth rate is ambiguous. The age structure of the population is also seen as an important determinant of savings in the life cycle model since a declining share of dependent population to working population is believed to correlate with higher savings rates. However, the existing empirical evidence on this relationship is mixed. For instance, Leff (1969) showed that this is true for both developed and the developing countries, while Ram (1982), Doshi (1994) and Kelly (1988), found no or very weak evidence of this relationship between savings rate and the age dependency ratio in
developing countries. Agrawal (2001) found a significant negative
relationship between savings and the age dependency ratio for three of
the seven Asian countries.
Apart kom the life cycle hypothesis, following Agarwal (2001), Hussain
and Brookins (2001) and others, we employ current account deficits and
real interest rates as additional explanatory variables. The relationship
between national savings and current account deficits as a share of gross
national product is clear cut - high foreign savings leads to a fall in
national savings, since they are substitutes.
The real interest rate, defined as the difference between the nominal
interest rate and the expected inflation rate, is the reward for postponing
present consumption into the fbture. The net iinpact of the real interest
rate on savings is unclear both theoretically and empirically, though the
usual presumption is that the total effect is positive. One of the main
reasons for the ambiguity follows from the distinction between the
resulting income and substitution effects from an interest rate change (see
Agrawal, 200 1 ; Giovannini, 1983, 1985). The overall iinpact of interest
rates on savings could be positive or negative depending on which effect
is larger. Hence, the direction of the relationship is an empirical issue
depending upon the relative strength of income and substitution effects. If
the substitution effect outweighs the income effect savings would rise
with an increase in real interest rate. Fry (1980, 1995) found that interest
rates have a positive but small impact on savings while Giovannini (1 983,
1985) found no statistically significant relationship between savings and
interest rates. Agrawal (2001) found mixed results on this relationship
for four Asian co~ntries.~
Model specification
Based on the theory reviewed in the previous section, we posit the
following specification to deteimine the national savings rate for Fiji.
where S, is the savings rate which is calculated as a ratio of gross national
savings to gross national product; GR,is the growth rate of per capita
Taiwan's savings were significantly related to interest rate elasticities (-
0.190), whereas savings from Malaysia, Korea and India were unaffected
by real interest rates.
income; IR, is the real interest rate3; CAD, is the current account deficit as
a proportion of GNP; and DEC is the share of dependents (below 15 and
above 65) to the