INTRODUCTION
Because of the limited amount of capital accumulation from domestic saving,
developing countries rely on the foreign source of financing. Mody
and Murshid (2005) explained the importance of financial integration and
domestic policy to optimize the benefit of capital inflows in the form of foreign
direct investment on the domestic investment. The potential benefit of capital
inflows is indisputable, especially in the case of limited domestic capital
accumulation. Countries that successfully attract foreign direct investment
perform better than countries that deter foreign direct investment (Baharumshah
and Thanoon, 2006). Therefore, the common policy recommendation is to attract
capital inflows to augment capital accumulation for economic growth.
On the other hand, Calvo et al. (1996) point
out several risk of the financial integration for developing countries, including
volatility of commodity price and world interest rates. Because of its size,
the small open economy may have limited influence on the international factors
such as international interest rate and capital flows. Therefore, these factors
may not be directly controllable by the domestic policies. The slow down or
sudden stop of capital inflows may cause financial and balance of payment crisis
(Calvo, 1998). Furthermore, Bosworth
and Collins (1999) argue that capital inflows may not be fully employed
for investment. Some portion might be used for current consumption. In addition,
capital inflows may not be fully transformed into resource flows because some
parts might be offset by capital outflows, reserve accumulation, or errors and
omissions.
Furthermore, capital inflows may affect developing countries in a different
way from developed country. Capital inflows can be used to finance current account
deficits in developed countries, but may cause current account imbalance in
developing countries (Yan, 2007). The incorrect structure
of the capital inflows may create detrimental effect for economic growth and
financial stability (Van Zyl, 2002). Instead of raising
investment, substantial increase of external debt might have negative effects
on investment (Javed and Sahinoz, 2005).
This study contributes to the analysis of capital flows from the perspective
of developing countries. Literatures on the impact of capital flows on the economy
differ in their methodology and approach. However, the ultimate goal is identical,
i.e., to achieve high economic growth. One strand of literature estimates economic
growth directly from capital flows data (Gruben and McLeod,
1998; Baharumshah and Thanoon, 2006). Other strand
estimates the impact indirectly through investment because investment would
induce capital accumulation needed to accelerate economic growth (Bosworth
and Collins, 1999; Mileva, 2008). The last strand
relates the capital flows with current accounts and financial integration (Blanchard
and Giavazzi, 2002).
Similar with this last strand of literature, this study assesses the impact of capital flows through saving and investment relationship by reviewing the contribution of capital accumulation on economic growth, confirming the capital market openness and estimating the relationship between saving and investment. We analyzed the period of 19841995 in Indonesian economy as a representative of developing countries because of the typical characteristics of a developing country with open capital account. Several common characteristics are large capital inflows, managed exchange rate regime, large investment and high economic growth. Other reasons to focus on this period are the interest rate liberalization in the mid 1983 and the fact that capital account openness in Indonesia was at the highest level during this period.
MATERIALS AND METHODS
Economic Growth and Investment
First, we review the importance of investment or capital accumulation for
economic growth. The average GDP growth during the period 19841995 was 6.4%.
Collins and Bosworth (1996) assert that from 3.7% average
growth of output per worker over the period 19841994 in Indonesia, physical
capital accumulation contribution was 2.3%, while human capital and total factor
productivity contributed for 0.5 and 0.9%, respectively. At this period, the
capital account openness in Indonesia was at the highest level according to
KAOPEN index of Chinn and Ito (2008). Ito
(2000) assesses that capital flows to Asia before Asian crisis contributed
to high economic growth which in turn attracted more capital flows into that
region. Indonesia was one of the country receiving large capital inflows during
this period.
Before this period, Indonesian economy relied on the abundance of the commodity and high international price of oil and gas. With declining oil price, the contribution of crude petroleum and its products to exports was decreasing (Fig. 1). Crude petroleum and its products comprised more than 60% of Indonesian export before 1984, but decreasing until it only had less than 20% share of the total export in the 1990s. High economic growth until early 1980 was mostly attributed to the oil revenue. Therefore, we exclude the period of high oil price to emphasis the role of the capital inflows on investment and economic growth.
To maintain export competitiveness and exchange rate stability, managed exchange
rate regime was adopted. The domestic currency was stabilized based on US dollar.
The development of nominal and real exchange rate is presented in Fig.
2. Nominal exchange rate was stable and gradually depreciated over the period
of analysis, with a sudden depreciation in 1986.

Fig. 1: 
Export composition. Source: Author’s calculation 

Fig. 2: 
Nominal and real exchange rate. Source: Author’s calculation 
It is apparent that exchange rate was managed to maintain purchasing power
parity and to sustain export competitiveness.
Figure 3 shows three items of the balance of payment. We can see that capital inflows were always positive with the average value of 3.7% of GDP. At the same time, current accounts were always negative. Another typical characteristic of the developing countries is the negative net factor income from abroad. Because foreign assets in domestic economy are larger than domestic assets abroad, developing countries usually have to pay for the factor income.
Data Source and Empirical FrameworkThe data are mainly from the International
Financial Statistics (IFS) 2009 CDROM of the International Monetary Fund (IMF).
Quarterly GDP data for the period 19841989 are taken from Statistic Indonesia
because quarterly GDP data from IFS start from 1990. Interest rates are also
from IMF, except for domestic interest rate of several months in 1984 and 1986
which are taken from Bank Indonesia. Figures for these months are missing in
IFS. We use money market rate to represent domestic interest rate and U.S. LIBOR
dollar rate as the benchmark rate. Estimations on interest rates are conducted
on a monthly basis, while estimations on saving and investment are on a quarterly
basis.

Fig. 3: 
Balance of payment analysis. Source: Author’s calculation 
The period of analysis is 19841995. The analysis is conducted in two main
parts. The first part reviews the capital flows in the context of free capital
mobility. We employ MundellFleming framework (Fleming,
1962; Mundell, 1963) to get evidence of the capital
account openness. The second part elaborates the relationship between saving
and investment. We estimate the savinginvestment relationship using Feldstein
and Horioka (1980) framework.
In the MundellFleming framework (Fleming, 1962; Mundell,
1963), if the exchange rate is fixed (or managed), free capital mobility
implies zero monetary policy independence. To assess the monetary independence,
we estimate the relationship between domestic and benchmark interest rate. In
addition to OLS estimation on the level and first differences of the variables,
we also employ the generalized error correction model (Davidson
and MacKinnon, 1993). This dynamic approach is suggested by Obstfeld
et al. (2005) and Shambaugh (2004). The specifications
of the model are as follows:
where, rd is domestic interest rate and rb is benchmark interest rate.
The shortrun relation can be seen from β in Eq. 1 and 3, while the longrun relation from β in Eq. 2 and γ in Eq. 3. The coefficient θ represents the speed of adjustment toward the longrun equilibrium. With apparent control on exchange rate during the period of analysis (Fig. 1), we may use monetary independence as evidence of free capital mobility. The null hypothesis of the free capital mobility condition or zero monetary independence is β = 1 or γ = 1.
In the saving and investment analysis, we define gross domestic product (GDP) as the sum of goods produced domestically together with imports (M) produced abroad, are consumed by private sector (C) and government (G), used as investment (I), or sent as exports (X).
Domestic saving is GDP less private and government consumption. Net export (NX) is the gap between domestic saving (S_{d}) and domestic investment (I_{d}).
Net factor income from abroad (NFIA) is the income from international capital flows, usually in the form of interest or dividend. National saving (S_{n}) consists of domestic saving and net factor income from abroad. Net export and net factor income from abroad comprise Current Account (CA) which also represent Net Foreign Investment (NFI). National saving may be invested domestically or abroad as NFI. In the end, national saving equals to national investment.
Equation 8 and 9 imply that investment
is financed by saving. In a closed economy, all domestic investment comes from
domestic saving. The benefit of the open capital market for developing countries
is the ability to attract international capital that could enhance their capital
accumulation in addition to the domestic saving. However, there might be possibilities
that the capital inflows are used to raise consumption instead of investment
(Bosworth and Collins, 1999). As a result, saving would
be reduced. Capital inflows would increase investment, decrease saving or both.
Consequently, saving and investment would move independently with the existence
of the capital inflows. The relationship can be described as follows:
Subtracting saving from both side and multiplying with minus one would result the relation between saving and current accounts or net foreign investment.
In an open economy, any increase in saving will cause an increase in investment
in all countries. The smaller the country, the smaller proportion of domestic
saving would be retained domestically. There would be no correlation between
saving and investment (β = 0) in an open economy. On the contrary, increased
domestic saving is retained entirely and transformed into domestic investment
in a closed economy. There would be one to one saving and investment relationship
(β = 1). For this reason, Feldstein and Horioka (1980)
argue that domestic saving and investment relation implies the degree of capital
mobility.
Many literatures employ time series approach for individual country analysis,
such as Jansen and Schulze (1996), Sarno
and Taylor (1998) and Taylor (2002). As in the interest
rate estimation, we employ regression in levels, differences and also generalized
error correction model. The specifications to estimate Eq. 10
are as follows:
The small characters mean that the variables are stated as a ratio to income. This study employs both definition of income (GDP and GNP). Both domestic saving and national saving are regressed on domestic investment.
RESULTS AND DISCUSSION
First, we test the assumption of free capital mobility to confirm the capital account openness. Table 1 shows the estimates of domestic interest rate responsiveness to the benchmark rate. The null hypothesis of unity of the shortrun coefficients in the first model (β in Eq. 1) and third model (β in Eq. 3) could not be rejected. However, the shortrun coefficients are not statistically significant due to large standard deviation. The longrun coefficients from the second model (β in Eq. 2) and third model (γ in Eq. 3) are highly significant and are statistically not different from unity. Moreover, domestic interest rate seems to adjust rapidly, with adjustment coefficient of 0.5, indicating half life of one month. Therefore, we may conclude that domestic interest rate followed the movement of the benchmark rate. Monetary policy was not independent. It also implies that capital was freely mobile during the period 19841995.
The lack of independence of the monetary policy in this period is in support
of Shambaugh (2004) results that countries with pegged
exchange rate would have no monetary independence. Interest rate follows the
benchmark rate if capital is freely mobile. This result also confirms Chinn
and Ito (2008) assessment of the capital mobility in Indonesia with the
highest value of KAOPEN index in this period.
According to FeldsteinHorioka criterion, if capital is freely mobile and the economy is small, saving coefficient would be zero. On contrary, coefficient of one in saving and investment relation would mean closed economy or zero capital mobility because all incremental saving is used to finance domestic investment. The null hypothesis of a closed economy is β = 1 and γ = 1. Because Table 1 implies free capital mobility, we should be able to reject the null hypothesis of the closed economy. We estimate domestic investment relation using two definitions of saving.
Table 1: 
Domestic interest rate responsiveness to benchmark rate 

*,** and *** represent 10, 5 and 1% levels of significance,
respectively. Standard errors reported in parentheses are robust to heteroskedasticity
and autocorrelation. The unitroot hypothesis is rejected for the residuals
from all models. The benchmark rate is U.S. LIBOR dollar rate 
Table 2: 
Domestic saving and domestic investment 

*,** and *** represent 10%, 5% and 1% levels of significance,
respectively. Standard errors reported in parentheses are robust to heteroskedasticity
and autocorrelation. The unitroot hypothesis is rejected for the residuals
from all models 
Table 3: 
National saving and domestic investment 

*,** and *** represent 10%, 5% and 1% levels of significance,
respectively. Standard errors reported in parentheses are robust to heteroskedasticity
and autocorrelation. The unitroot hypothesis is rejected for the residuals
from all models 
Table 2 shows the results using domestic saving. Shortrun and longrun coefficients in all models are highly significant and statistically not different from unity. Adjustment speed is high with half life of one month. Moreover, adjusted R^{2} values are very high, more than 90%. Domestic investment moved one to one with domestic saving.
Table 3 shows the results of estimations using national saving. The use of national saving does not change the result. The shortrun and longrung coefficients are highly significant and also not different from unity. National saving contributed for more than 90% of the variation in investment. In other words, domestic investment followed the movement of national saving almost perfectly.
The results indicate one to one relationship between saving and investment
in the shortrun and longrun. This implies that domestic investment was financed
by domestic or national saving, not capital inflows. This fact is against our
intuition, considering the fact of large capital inflows. This condition is
known as the FeldsteinHorioka puzzle. Sinha and Sinha (2004)
and KayaBahce and Ozmen (2008) also found a high correlation
between saving and investment in Indonesia. However, the interpretation is different,
because they conclude that capital mobility was low in Indonesia. We believe
that high correlation between saving and investment coexists with free capital
mobility in Indonesia.
As a robustness check, we estimate the relation between national saving and current accounts (ca), which also represents net foreign investment as in Eq. 12. After finding close relationship between saving and investment, we expect that saving would not be correlated with net foreign investment. There would be no correlation between saving and current account. Saving coefficients would be zero in the shortrun and longrun. Therefore, the null hypotheses are β = 0 and γ = 0. Table 4 shows the results. The null hypothesis could not be rejected. The saving coefficients in the shortrun and longrun from all models are zero. In this sense, saving did not have any relationship with net foreign investment. Excess saving was invested domestically, while lack of saving did not attract foreign investment.
The results indicate that domestic investment was only influenced by saving. All domestic investment was financed by domestic saving. Therefore, the relation between capital inflows and domestic investment may not be as strong as we expected. Considering the amount only, it may imply that capital inflows were used to pay for the net factor income.
Table 4: 
National saving and current account 

*,** and *** represent 10%, 5% and 1% levels of significance,
respectively. Standard errors reported in parentheses are robust to heteroskedasticity
and autocorrelation. The unitroot hypothesis is rejected for the residuals
from all models 

Fig. 4: 
Investment income payment to abroad and financial account.
Source: Author’s calculation 
Alternatively, it may indicate that domestic saving was high because of high burden to pay for the factor income abroad and capital inflows were used to finance domestic investment. Figure 4 compares the annual figures of net investment income payment to abroad with the financial accounts which represents the capital inflows during the periods.
Income payments were larger than financial account in many cases. There was
a large burden to service foreign investment. In other words, resource was transferred
out of Indonesia. Net resource outflows happened at the same time with capital
inflows. The results in this study provide evidence for Bosworth
and Collins (1999) remarks that capital inflows are not always equal to
resource transfer. In Indonesian case, capital inflows have been offset by payments
of the net factor income to abroad. This fact is interesting because economists
believe that physical capital was the main source of economic growth at that
time. In this sense, capital inflows played a crucial role. Naturally, we expect
that capital inflows would mean net resource inflows and thus higher investment.
Further analysis on the composition of the income payment is presented in Fig. 5. We can see that direct investment comprised a significant portion of the income payment until the end of 1980s. This fact creates another interesting issue because the amount of direct investment until this period was not as large as other investment.
The implication of the results is that simplistic approach to attract capital
inflows might not be an optimal solution for developing countries. Development
the domestic infrastructure and capability is necessary to attain the highest
benefit from international capital. Therefore, domestic productivity may improve
in parallel with the increase of foreign capital. The importance of total factor
productivity increase and spillovers are also suggested by Haskel
et al. (2007). The increased domestic productivity is crucial to
service debt and other factor payment to abroad.

Fig. 5: 
Composition of investment income payment to abroad. Source:
Author’s calculation 
CONCLUSIONS
The potential benefit of foreign capital inflows to developing countries is indisputable. Capital inflows raise economic growth by accelerating capital accumulation. However, there might be risks of the capital inflows and financial integration. The most apparent risk would be the slow down and sudden stop of the capital inflows which may cause crisis in developing countries. The impact on resource transfers may also need attention because capital inflows may not automatically mean resource inflows.
In Indonesian case, domestic investment moved one to one with saving. This is inconsistent with the fact of open capital market. Analysis on the balance of payment showed that the amount of capital inflows was smaller than net factor income payment to abroad. Resource outflows happened at the same time with capital inflows.
In short, policies directed toward absorbing foreign capital inflows and discouraging capital outflows might not be sufficient. Domestic policies should be directed towards optimal utilization of the foreign capital inflows to improve domestic productivity.
ACKNOWLEDGMENT
The author would like to thank Prof. Komatsu Masaaki for invaluable comments and support.