77331 THE WORLD BANK ECONOMIC REVIEW, VOL. 14, NO. 3: 571-94 Private Saving in India Norman Loayza and Rashmi Shankar This article studies the evolution of the private saving rate in India during 1960-9S. Its distinctive feature is that it proposes three new measures of private saving, which are incremental improvements to the (naive) national accounts measure. The improvements consist of accounting for capital losses to private net worth due to inflation, including expenditures on durable goods as a form of saving, and expanding the definition of saving to include human capital expenditures. After examining descriptive trends and reviewing the felated literature, the article tests the hypothesis that households that save in India "pierce the corporate veil. * The evidence shows that, in fact, changes in corpo- rate saving are offset by changes in household saving, indicating that the unit of analysis should be aggregate private saving. The core analytical section of the article studies how the behavior of the private saving rate is related to the real interest rate, per capita income, the dependency ratio, financial depth, the government saving rate, and the share of agriculture in gross domestic product. The empirical analysis is done by estimating error-correction models on aggregate annual data, although most of the discussion cen- ters on long-run effects. There are many reasons for studying private saving in India. India's saving per- formance has surpassed that of other developing countries with comparable per capita income. However, given the Indian government's ambitious growth tar- gets, and the need in the present global environment to generate investable re- sources by and large internally, the design of policies aimed at enhancing saving acquires great significance. With this in mind we assess the extent to which an increase in public saving is offset by a reduction in private saving (effectively testing for Ricardian equivalence), whether an increase in rates of return in finan- cial markets leads to a rise in private saving rates (which would have implications for the effectiveness of tax incentives on saving instruments), and the extent to which an increase in average private disposable income brings about a rise in private saving rates (which has implications for the wide range of policies that raise private income). In addition to studying the policy determinants of private saving, it is also important to take into account the impact that India's remarkable demographic Norman Loayza is with the Central Bank of Chile and the Economic Research Group at the World Bank, and Rashmi Shankar is a doctoral candidate at the University of California, Santa Cruz. Their e- mail addresses are nloayza@condor.bcentral.cl and nhankar@ivorldbank.org. The authors are grateful to Luis Serven, an anonymous referee, and Francois Bourguignon for valuable comments. © 2000 The International Bank for Reconstruction and Development /THE WORLD BANK 571 572 THE WORLD BANK ECONOMIC REVIEW, VOL. 14, NO. 3 transition in the past 30 years has had on current and future saving rates. The age structure of the population in India has changed dramatically. The number of people more than 65 years of age as a proportion of total adults (that is, people between 16 and 65 years of age) rose from 6 to 8 percent. The corresponding figure for children less than the age of 16 has declined from 72 percent in 1965 to 58 percent in 1994. We are interested in determining the impact of this change on saving because in the next few years "demographic inertia" is likely to result in a continuing rise in the share of working adults in the population. The major difference between this study and other macroeconomic studies of saving in India is that here we adjust private saving and income figures to derive measures that are conceptually more correct. The goal of these adjustments is to approach saving figures that correspond to changes in net worth. The first adjustment concerns asset price changes. Although in theory we should correct for capital gains and losses for all types of assets, data availability con- strains us to focusing only on inflation-related capital changes. Since we do not have long enough time series for stock, private bond, and real estate prices, the first adjustment considers only the effect that inflation has on private and public income and saving. Although limited, this correction is important. When infla- tion occurs, the value of government debt (including money balances) held by the private sector decreases, which implies a loss of wealth to the private sector and a concomitant gain to the government. We obtain the inflation-adjusted saving and income figures from the World Saving Database (see Loayza and others 1998 for a complete explanation of the adjustment procedure, input sources, and the corresponding quality controls). In fact, the provision of these adjusted saving figures, as well as the consistent definition of the public and private sector for a large number of countries since the 1960s, is one of the main contributions of this new database. Clearly, correcting for capital gains and losses due to inflation becomes more important as the level of government debt held by the private sector increases or inflation rises. In India the inflation-adjusted figures are significantly different from the unadjusted figures, particularly those in the 1980s. We find significant differences between our results and those emphasized in the received historiogra- phy, particularly regarding Ricardian offset coefficients, and believe this to be largely due to our use of inflation-adjusted figures. We also correct private and public saving figures to include expenditures on consumer durables and investment in human capital. We first present a measure of household saving, and hence private and national saving, that includes private expenditures on consumer durables, including personal transportation.1 This is not only because households perceive expenditures on durables as deferred con- sumption but also because in India most consumer durables and motor vehicles have an extremely high resale value. In the absence of data on gold imports over 1960-94, we also hope to use this definition of saving to correct partly for the 1. We follow the methodology in Jorgenson and Fraumeni (1989). Loayza and Shankar 573 understatement of household physical saving that excludes jewelry, a tradition- ally important form of saving in Indian families. The third measure of saving involves both private and public saving. To ad- justed private saving, we add not only expenditures on consumer durables but also expenditures on education and health. Correspondingly, we augment public adjusted saving with the government's final consumption expenditures on health and education. These adjustments are made to proxy for the accumulation of hu- man wealth. I. MAJOR TRENDS IN INDIAN SAVING RATES The trend of the national saving rate (defined as gross national saving, includ- ing net current transfers, divided by gross national disposable income) in India compares favorably to that of low-income countries and countries of the Organisation for Economic Co-operation and Development (OECD) but falls short of that of China and other East Asian countries (table 1). In 1965 India's na- tional saving rate was similar to the average of all developing countries. It has since followed a rising trend, surpassing the average national saving rate of the OECD countries by the late 1980s and gaining about 4 percentage points on the average of developing countries by 1995. Although strong, the rising trend of India's national saving rate has fallen short of that of East Asian countries, pro- ducing a gap between the two of more than 10 percentage points by 1995. The inflation-adjusted median private saving rate appears to be consistently lower in India than in the East Asian or OECD countries, although it has been growing faster than both (table 2). For India and other countries, except the OECD countries, the definition of the public sector we use includes not only the central government but also state and local governments and public enterprises. In the OECD countries the public sector does not include public enterprises, which are ascribed to the private sector. This overstates the OECD's private saving rate relative to that of India and other countries. We next look at the relationship among saving, investment, and growth across countries. Each country relies to a varying extent on national saving to finance Table 1. Median National Saving Rates, 1960-94 (percentage of gross national disposable income) Period India China Low-income South Asia1 East Asia OECD 1960-73 15 26 11 12 20 25 1974-82 21 33 13 15 29 23 1983-91 21 36 11 17 33 21 1992-94 . 22 41 12 18 34 19 Note: For country groups, the median is of individual country averages, while for India and China the averages across time are reported. The saving rate is gross national saving, including net current transfers, divided by gross national disposable income. a. Excluding India. s± Source: Authors' calculations "based on data in the World Bank's World Saving Database. 574 THE WORLD BANK ECONOMIC REVIEW, VOL. 14, NO. 3 Table 2. Inflation-Adjusted Median Private Saving Rates, 1960-94 (percentage of inflation-adjusted private disposable income) Period India Low-income South Asia* East Asia OECD 1960-73 12 9 n.a. 20 23 1974-82 16 10 10 27 23 1983-91 17 9 10 29 22 1992-94 18 10 n.a. 29 23 Note: See table 1 for the composition of country groupings. a. Excluding India. Source: Authors' calculations based on data in the World Bank's World Saving Database. the investment necessary for growth. Comparing performances on these fronts allows us to make some inferences about the best use of resources garnered through saving. The saving rate in China, for example, was, on average, 70 percent higher than that in India, investment was approximately 50 percent higher, and growth was 100 percent higher (table 3). East Asia achieved a growth rate twice as high as India's with an investment rate that was 34 percent higher and a saving rate that was about 45 percent higher. India's long-run averages of saving, invest- ment, and growth are similar to those of the OECD countries. Looking at the saving-investment gap, India had to rely on foreign saving more than the East Asian countries, but notably less than other low-income countries. Saving Rates and Other Macroeconomic Aggregates in India We divide the period 1960-95 into four subperiods (table 4). The first two, 1960-1973 and 1974-82, are separated by an oil crisis, as are the second and third. The third and fourth subperiods, 1983-91 and 1992-95, are distinguished by the Indian government's 1991 structural adjustment program. Inflation increased fairly smoothly across these broad subperiods, as did real GDP growth and the rate of investment (given by the ratio of gross domestic in- vestment to gross national disposable income), although the investment rate dipped slightly in the last period. Table 3. Median Saving, Investment, and Gross Domestic Product, 1960-94 Low- South East Indicator India China income Asia1 Asia OECD National saving rate (percent of gross national disposable income) 20 34 11 16 19 22 Gross domestic investment (percent of national disposable income) 21 32 17 17 27 21 GDP growth (percent) 4 8 3 4 8 4 a. Excluding India. Source: Authors' calculations based on data in the World Bank's World Saving Database. Loayxa and Shankar 575 Table 4. Basic Macroeconomic Indicators in India, 1960-98 Gross domestic National saving rate Inflation investment (gross national saving (GDP GDP (percentage of as a percentage deflator; growth gross national of gross national Period percent) (percent) disposable income) disposable income) 1960-73 7.0 3.3 17 15 1974-82 7.7 3.9 21 21 1983-91 8.5 5.2 23 21 1992-95 9.0 4.9 23 22 1996-98 7.0 5.9 23 21 Source: Authors' calculations based on data in the World Bank's World Saving Database. In 1960-82, India achieved its peak saving rate in 1978, when the national saving rate rose to 22 percent. Private saving increased sharply in the second period, 1974-82, by about 5 percentage points. Several factors operated to in- crease the saving rate in the late 1970s, notably, the end of a decade of vigorous bank expansion and foreign remittances from Indians working in the Gulf area. These changes are also indicated by the fall in the average current account deficit (by definition, the difference between investment and saving) during this period. Real GDP growth did not exhibit any marked trend relative to the previous period. The national saving rate remained at about 21 percent in the last three peri- ods, despite the sharp increase in the GDP growth rate between the second and third periods, its slight decline in the fourth, and the post-1996 increase. The investment rate remained closely linked to the saving rate, resulting in a small current account deficit throughout 1960-95. This points to an important styl- ized fact: saving in India depends on national, as opposed to foreign, savings. The slight increase in the investment-saving gap in the last period still leaves the deficit at a low 2 percent of gross national disposable income. Changing Composition of National Saving in India We study the composition of national saving by, first, describing the behavior of the components of private saving and, second, describing the trends in public and private saving, paying special attention to the comparison between adjusted and unadjusted figures. (Recall that the definition of the public sector includes the central government, state and local governments, and public enterprises.) The entire 1960-94 period was marked by an increase in the share of house- hold saving as a proportion of unadjusted private disposable income, whereas the analogous measure for corporate saving remained mostly stagnant, picking up only in the mid-1990s (table 5). It can be argued that the sharp rise in national saving following the nationalization of banks and vigorous branch expansion beginning in 1969 was spurred by a rapid growth in financial saving in the 1970s. The 1970s were also characterized by a jump in remittances from abroad—mostly from the Middle East—which could have contributed some of the increase in household saving between the first and second periods. 576 THE WORLD BANK ECONOMIC REVIEW, VOL. 14, NO. 3 Table 5. The Composition of Private Saving in India, 1960-95 (percentage of unadjusted private disposable income) Household Household Total Corporate Period physical saving financial sewing household saving saving 1960-73 7.61 3.62 11.23 2.46 1974-82 9.83 6.40 16.23 3.03 1983-91 9.82 8.81 18.63 2.64 1992-95 8.67 11.16 19.83 3.46 Note: Figures are not adjusted for capital losses due to inflation, since there it no sound basis on which to apportion the adjustment to household and corporate sectors. Source: Authors' calculations based on data in the World Bank's World Saving Database. The rate of household physical saving also increased in the first three periods; however, in the last period, 1992-95, it fell somewhat, reflecting in part a portfo- lio shift from physical to financial saving. The slight decline in household physi- cal saving generated some debate in India, especially by 1994, when the decline appeared to be more substantial. One view holds that this decline is a statistical illusion (Athukorala and Sen 1995). The Central Statistical Organization (CSO) sets household physical saving exactly equal to household investment, which in turn is determined residually. In national accounts domestic investment is ad- justed to equal the sum of domestic and foreign saving. Then only investment is adjusted for errors and omissions. This asymmetric adjustment is advocated on the argument that public and corporate saving data are more reliable than invest- ment data. This might be true in the case of public saving; however, the estimate of corporate saving (and investment) is based on small, not necessarily represen- tative, samples and relies on voluntary responses from enterprises. At any rate, although there are reasons to believe that the figures for private saving, particu- larly those of household physical saving, are erroneous, it is not at all obvious that measurement error is behind the recent trends.2 Private income and saving figures must be adjusted to take into account the redistribution of wealth from the public to the private sector due to inflation's erosion of the value of public debt held by the. private sector. This correction is sizable. Adjusted public saving as a ratio of gross national disposable income increased by 3 percentage points over 1960-94, while the corresponding unad- justed figure rose by about hjilf as much. There was 3. steady increase in adjusted private saving over the entire period, except between 1974-82 and 1983-91, when it remained stable at about 14 percent (table 6). The unadjusted figures for private saving show a sharper in- crease, from 12 to 21 percent over 1960-94, as opposed to the adjusted figures, which show an increase from 11 to 16 percent. The unadjusted public saving rate declined from 3 to 1 percent over the full period, reaching a low of 0.94 percent 2. Some physical assets, traditionally preferred as saving tools in India, such as jewelry and gold, are not covered in the CSO estimates. This further increases the likelihood of measurement error in this cat- egory of saving. Loayza and Shankar S77 Table 6. Private and Public Saving Rate in India, 1960-94 Adjusted Adjusted Unadjusted Unadjusted private public private public Period saving rate saving rate saving rate saving rate 1960-73 Unaugmented 0.11 0.04 0.12 0.03 Includes consumer durables 0.13 0.15 Includes expenditures on health and education 0.16 0.06 0.18 0.04 1974-82 Unaugmented 0.14 0.07 0.17 0.04 Includes consumer durables 0.18 0.21 Includes expenditures on health and education 0.22 0.10 0.24 0.07 1983-91 Unaugmented 0.14 0.07 0.19 0.02 Includes consumer durables 0.19 0.24 Includes expenditures on health and education 0.23 0.11 0.27 0.06 1992-94 Unaugmented 0.16 0.07 0.21 0.01 Includes consumer durables 0.20 0.25 Includes expenditures on health and education 0.23 0.11 0.28 0.05 Note: Adjusted means adjusted for capital gains or losses due to inflation. Source: Authors' calculations based on data in the World Bank's World Saving Database. in 1984. However, adjusted public saving increased from 4 to 7 percent between 1960 and 1994. What is striking about these data is that the difference between the adjusted and unadjusted figures becomes larger as time goes on. In the first period, 1960- 73, the differences are marginal. However, in the next subperiod, there is a 3 percent difference between the adjusted and unadjusted figures. This difference increased to 5 percent in later years. These increases were due to the gradual rise of inflation and, most important, the increase in public debt held by the private sector. Whereas the unadjusted public saving rate shows a declining trend since the early 1970s, the adjusted public saving rate remained basically flat. Table 6 also presents private saving rates augmented to include expenditures on consumer durables and expenditures on health and education. Saving by any definition rose sharply in the early and mid-1970s and then again in the early 1980s. The flattening observed in common measures of saving since the early 1990s was less pronounced for the augmented measures, which continued to rise, albeit by less than in the 1986-88 period. This is in part due to the continu- ous increase in spending on health and education. S78 THE WORLD BANK ECONOMIC REVIEW, VOL. 14, NO. 3 n. SAVINGS ISSUES IN INDIA Much of the literature on saving in India distinguishes between private and public saving for analytical purposes and treats public saving as exogenous. Pub- lic saving has been conspicuous in the lack of attention it has received. According to Pandit (1991) public saving in India has been largely residual and mostly driven by expansionary fiscal policy and public sector pricing policy. A notable excep- tion to this view is that of Cashin, Olekalns, and Sahay (1998). They find evi- dence that the central government in India exhibits both tax-smoothing and tax- tilting behavior. Thus the government relies on seigniorage and financial repression as revenue sources, whereas taxes rise only in response to permanent changes in expenditures. According to their study, state tax revenues are relatively volatile in the face of even temporary changes in expenditure. Most of the literature, how- ever, focuses, as we do, on aggregate private saving. In diis context several issues have been raised. Dependence on Public Satring The issue here is the extent to which the private sector internalizes the government's budget constraint and hence the extent to which an increase in public saving is offset by an increase in private saving. Muhleisen (1996) presents evidence that Ricardian equivalence is of minor importance in India. He finds that long-run aggregate private saving decreases by 0.25 percent in response to a 1 percent increase in public saving. Estimates for other developing countries range from 0 percent (Haque and Montiel 1989) to 50 percent (Corbo and Schmidt- Hebbel 1991). Income Growth Lahiri (1989) finds that the rate of growth of personal disposable income is a significant determinant of private saving in all the countries in his sample of Asian countries, including India. Lahiri bases bis empirical findings on individual time-series analyses for each country in his sample. He claims that such an ap- proach has an advantage over a panel-based analysis in that the marginal re- sponse of the saving rate to various factors need not be assumed uniform across countries. Muhleisen (1996) uses a vector autoregression (VAR) process in loga- rithms to jointly model the relationship between the private saving rate and growth and between private and public saving (that is, Ricardian equivalence). Using Granger-causality tests, Muhleisen shows that growth leads to both higher pub- lic and private saving rates. Demography Lahiri (1989) finds that the age-dependency ratio (the fraction of the popula- tion under the age of 16 and over the age of 64) is a significant determinant of private saving. Under Lahiri's specification a 1 percentage point increase in the Loayza and Shankar 579 dependency ratio lowers the long-run average propensity to save by 1.6 percent- age points in India, the Republic of Korea, Malaysia, Singapore, and Sri Lanka. Muhleisen (1996) finds that the age-dependency ratio is the most significant de- terminant of private saving, with the usual negative relationship between the two variables. Level of Financial Development According to Muhleisen (1996) public policy could play a role in providing credible and stable rules for financial intermediation and thus, by encouraging saving in financial assets, also further the development of financial markets. He highlights the need for reform in insurance markets, pension schemes, and mutual funds. He calls for policies that would mobilize greater financial saving and favor long-term saving instruments, such as reducing the government's recourse to cap- tive saving by allowing greater flexibility in portfolio allocation to pension funds and the life Insurance Corporation. Muhleisen bases his call for greater financial development on his empirical results, in particular those that find that the ratio of M2 to GDP—a proxy for financial depth—has a positive long-run relationship with private saving. His methodological tool is cointegration analysis of time-series data, following the maximum-likelihood method of Johansen (1991). Urban versus Rural Propensities to Save Pandit (1991) tests single-equation cross-sectional models of aggregate house- hold saving with the purpose of, among others, contrasting the propensities to save in rural and urban areas. He compares cross-sectional data for 1967-68 and 1975-76. He does not find convincing or consistent evidence supporting the view that both the average propensity to save and the marginal propensity to save are higher in urban areas and thus that a worsening of agricultural terms of trade could lead to higher saving. Pandit does not attempt a more disaggregated analy- sis, that is, one that studies the propensities to save at the level of the physical and financial components of household saving. This distinction might be important given that urban and rural households have different access to financial saving instruments. Rate of Interest The presence of imperfect, segmented capital markets with administered rates of interest in India may make estimating the interest elasticity of private saving problematic. The available empirical evidence (see, for example, Muhleisen 1996) does not support the hypothesis that aggregate private saving is increasing in the real rate of interest. Muhleisen argues that, to the extent that the measured inter- est rate reflects only the rate of return on controlled financial instruments and not the rate of return on investment, changes in measured interest rates mostly generate a substitution effect between physical and financial assets and between different kinds of financial assets. 580 THE WORLD BANX ECONOMIC REVIEW, VOL. 14, NO. 3 Decomposition of Private Saving Pandit (1991) further decomposes private saving into household and corpo- rate saving, mostly in an attempt to explain the household's portfolio decision. • Household physical and household financial saving. Pandit finds that house- hold investment (equal to household physical saving) declines when house- hold financial saving increases, suggesting substitutability between the two components of household saving and, therefore, the need to treat them joindy. Further decomposing financial saving into the real demand for money, sav- ing in time deposits, saving in insurance premiums, and saving in provident funds, Pandit finds that the composition of household financial saving is, as expected, driven by the rates of return on each type of financial saving and, to some extent, by bank expansion. • Corporate saving. Corporate saving is defined as the excess of profits over dividends. Pandit estimates a two-equation model using ordinary least squares, one equation for after-tax corporate profits and one equation for corporate dividends. He finds that after-tax profits are positively related to sales and to the nonagricultural terms of trade and negatively correlated with per unit wage costs. Dividends are explained by after-tax profits and the firm's access to external financing. in. Do HOUSEHOLDS IN INDIA "PIERCE THE CORPORATE VEIL"? The household sector, which includes unincorporated enterprises, is the ulti- mate owner of incorporated business. The question we address in this section is the extent to which the household sector takes into account the saving decisions of corporations in formulating its own saving decisions. In order for households to treat corporate saving as their own, they must both understand corporate actions and have a marginal propensity to save out of wealth equal to the mar- ginal propensity to save out of disposable income. As Poterba's (1987) seminal paper explains, if households "pierce the corporate veil," aggregate private sav- ing becomes the variable of interest, and little information is gained from break- ing it down further into household and corporate saving. However, if household and corporate saving decisions are made independently, then they must be stud- ied separately, for it is likely that they have different determinants. We argue here that household saving reacts sufficiently to corporate saving so as to allow a focus on aggregate private saving. We study the relationship between household and corporate saving in the frame- work of time-series cointegration. Augmented Dickey-Fuller tests provide evi- dence that all series involved are integrated of order 1 and, therefore, that cointegration analysis can be relevant (table A-l). However, if household saving is perceived as equivalent to corporate saving, the two variables should be cointegrated, possibly including other variables in the long-run relationship, with the cointegrating vector (1, 1). Loayza and Skankar 581 We first test for cointegration using the Engle and Granger approach and then follow a maximum-likelihood method to estimate the cointegrating vector, al- lowing for short-run dynamics (see table 7). In addition to the ratios of (unad- justed) household and corporate saving to private disposable income, we include in the model unadjusted private disposable income, the dependency ratio, the real interest rate (an-ex post composite of borrowing and lending rates), and the rate of inflation (to serve as an additional measure of the rate of return on non- financial assets). The saving rates and income figures are not adjusted for infla- tionary capital losses, given our inability to apportion the adjustment on private sector saving between the household and corporate sectors. However, we do account for consumer durables as a form of household saving. Thus we present estimation results for two representations of the dependent variable: the house- hold saving ratios obtained excluding and including durable consumption. We reject the null of no cointegration when we include in the model the vari- ables mentioned above (other variables could be excluded from the long-run re- lationship of household saving). Both excluding and including consumer durables in household saving, the coefficients on the corporate saving ratio are signifi- cantly negative and large in magnitude (table 7). This result is consistent with the hypothesis that households allocate their portfolio between corporate and house- Table 7. Empirical Results: Do Indian Households Pierce the "Corporate Veil"? Household saving Household saving ratio not including ratio including Variable consumption durables consumption durables Dependency ratio 12 -0.18 (4.4)* (9.5)* Inflation in GDP deflator -0.35 0.09 (2.7)* (1.94) Real interest rate -0.16 0.24 (1.7) (5.9)* Ratio of corporate saving to unadjusted -0.72 -0.99 private disposable income (5.07)» (11.3)* Log of real unadjusted private 0.49 0.05 disposable income per capita (6.12)* (18.J)* Engel-Granger cointegration test (Ho: no cointegration): Augmented Dickey-Fuller test -5.5M.7* -8.01/-4.7* statistic/5 percent critical value (Phillips-Ouliaris) Test on piercing the corporate veil (Ho: coefficient on corporate saving rate = -1): t-test statistic/critical value 1.97/1.96* 0.114/1.96* * Significant at 5 percent. Note: (-statistics are in parentheses. Source: Authors' calculations based on data in the World Bank's World Saving Database. 582 THE WORLD BANK ECONOMIC REVIEW, VOL. 14, NO. 3 hold saving and that there is a strong substitution between the two. Moreover, in the case of household saving including consumer durables, the estimated coeffi- cient is close to and not statistically different from - 1 , indicating that (correctly measured) household saving offsets a change in corporate saving. This result is consistent with the hypothesis that households pierce the corporate veil and thus indicates that we should focus on the behavior of aggregate private saving. We follow this strategy in the following section, where we study the economic and demographic determinants of private saving rates. IV. DETERMINANTS OF PRIVATE SAVING In this section we examine how the evolution of the private saving rate over 1960-94 has been related to the behavior of other economic and demographic variables. Instead of adhering to a specific consumption/saving model, we esti- mate a reduced-form model with the private saving rate as the dependent vari- able and its most important proposed determinants as explanatory variables. We define me public sector as including not only the central government but also state and local governments and public enterprises. We consider three new mea- sures of the private saving rate, each of which represents an incremental adjust- ment to the naive (unadjusted) measure. Thus we consider the unadjusted saving rate, the saving rate adjusted for inflationary capital gains and losses, the inflation- adjusted saving rate that includes expenditures on consumer durables, and the adjusted saving rate that includes expenditures on education and health in addi- tion to consumer durables. Using the conceptually correct definition of the public sector, adjusting for capital gains, and applying the theoretically correct concept of net worth (which should include consumer durables and human capital assets) lead us to different results. In selecting the explanatory variables, we take into account the literature on Indian saving, the current cross-country studies on private saving (see Loayza, Schmidt-Hebbel, and Serven 2000), and the availability of data for the whole period. The explanatory variables we consider are the ratio of public saving to private income (to evaluate the extent of Ricardian equivalence), the ratio of private domestic credit to GDP (to examine the importance of financial depth), the dependency ratio (to account for life-cycle effects), the share of agriculture in GDP (to gauge the effect of occupational structure and income uncertainty), the real interest rate (to serve as the relative price of current consumption with re- spect to future consumption), and the log of per capita disposable income (to examine income elasticities and the importance of subsistence consumption). We use aggregate annual data and therefore work with a sample of 35 obser- vations. We are interested mostly in the long-run relationship between the pri- vate saving rate and various economic variables. However, working with annual data forces us to consider a model that also takes short-run effects into account. In addition, unit-root tests indicate that the variables in our model are integrated of order 1 (see table A-l). Both the need to account for long-run and short-run Loayza and Shankar 583 effects and the importance of avoiding spurious correlations among integrated variables lead us to estimate the relationship between saving and its determinants using an error-correction model (see Johansen 1991 and Pesaran and Shin 1997, 1999). We performed an Engle-Granger test for cointegration (see table 8), and we could not reject the hypothesis that the variables in the model are cointegrated; that is, at least one linear combination of the variables is stationary. We estimate the error-correction model under the assumption that there is a single long-run relationship between private saving rates and their proposed determinants. Al- though Johansen tests indicated the presence of more than one cointegrating re- lationship, we chose to work under the restriction of a single cointegration vec- tor, given that we did not have appropriate restrictions to differentiate between various long-run saving relationships (see Pesaran 1997). Given that we estimate a dynamic model with only 35 observations, degrees-of-freedom considerations prevent us from either working with a larger set of explanatory variables or ex- amining the importance of nonlinear and interactive effects. Correlation Results Before presenting the estimation results, we consider simple time-series corre- lations between saving rates and potential saving determinants (table A-2). The first point to notice is that the four measures of the private saving rate are highly correlated. The correlation between the unadjusted and inflation-adjusted pri- vate saving rates is 0.87, which means that the two series share basic trends and cyclical fluctuations. Introducing consumer durables produces a larger change in the saving figures (the correlation of the series with and without durables is 0.75), whereas the inclusion of health and education expenditures produces a series that is highly correlated with the series that ignores them (correlation 0.99). Regarding public saving ratios, it appears that the inflation adjustment does produce a major change in the series (correlation 0.39), while the inclusion of health and education expenditures does not (correlation 0.99). The private sav- ing rate that includes all adjustments is negatively correlated with the depen- dency ratio (-0.74) and the share of agriculture in GDP (-0.72) and is positively correlated with per capita private income (0.72), financial depth (0.65), the real interest rate (0.50), and the government saving rate (0.43). We do not dwell on the meaning of these simple correlations because, as we will see, the signs of the corresponding coefficients change when estimated in a multivariate setting. We present the correlation coefficients to serve as a reference when analyzing .the results of the error-correction model. Estimation Results Although we present the estimated long-run and short-term coefficients, in- cluding the adjustment term, of the error-correction model (tables 8 and 9), we emphasize the long-run coefficients for two reasons. First, we are mainly inter- ested in the long-term evolution of the private saving rate, and, second, most of the estimated short-term parameters are not statistically significant. Among the 584 THE WORLD BANK .ECONOMIC REVIEW, VOL. 14, NO. 3 Table 8. Empirical Results: Determinants of the Private Saving Rate, Long- Run Effects Unadjusted Adjusted private Adjusted private private saving/ Adjusted saving including saving including unadjusted private saving/ consumer consumer durables private adjusted private durables/adjusted and human capital/ disposable disposable private disposable adjusted private Variable income income income disposable income Public saving/private -0.166 -0.411* -0.31* -0.454* disposable income* (1.07) (4.76) (2.79) (3.19) Domestic -0.002 -0.002* -0.00003 0.0002 credit/GDP (7.13») (4.97) (0.07) (0.456) Dependency ratio -1.26* -2.04* -0.657* -1.25* (4.83) (8.05) (2.8) (4.9) Real interest rate -0.1087* 0.08 0.28* 0.26* (4.23) (1.56) (5.47) (4.31) Log of real private 0.05 -0.09* -0.16* -0.21* disposable income (1.96) (2.7) (4.44) (5.21) per capita1" Share of agriculture 0.69* 1.14* 0.06 0.47* in GDP (4.17) (6.57) (0.39) (2.8) Engel-Granger cointegration test (Ho: no cointegration): Augmented -5. 96/-4.74 -5.18/-4.74 -5.24M.74 -5.17/-4.74 Dickey-Fuller test statistic/5 percent critical value (Phillips-Ouliaris) 'Significant at 5 percent. Note: ^-statistics are in parentheses. a. Adjusted for inflation if dependent variable is adjusted private saving/private disposable income; adjusted for inflation and inclusive of human capital expenditure if dependent variable is adjusted private saving including expenditure on human capital/adjusted private disposable income. If dependent variable is private saving including consumer durables/adjusted private disposable income, then we use adjusted public saving unaugmented/adjusted private disposable income; if dependent variable ii unadjusted private saving/unadjusted private disposable income, we use unadjusted public saving/unadjusted private disposable income. b. If dependent variable is adjusted private saving/adjusted private disposable income, augmented or unaugmented, we use log of real private disposable income per capita; if dependent variable is unadjusted private saving/unadjusted private disposable income, we use log of real unadjusted private disposable income per capita. Sourci: Authors' calculations based on data in the 'World Bank's World Saving Database. measures of the private saving rate, we give more importance to the measure that we believe is conceptually most correct the private saving rate that adjusts for inflationary capital gains and losses and includes expenditures on consumer du- rable goods and expenditures on health and education. The first explanatory variable is the public saving ratio. We can reject full Ricardian equivalence for all measures of the private saving rate. That is, the coefficient on the public saving ratio is statistically less than - 1 . However, we do find a negative relationship between private and public saving rates. This rela- Loayza and Shankar S8S tionship is significant for all measures except for the unadjusted private saving rate. The "offset" coefficient is lowest for the unadjusted measure of private saving (-0.166) and largest for the most conceptually correct measure (-0.454). The latter result indicates that, beyond internalizing the public budget constraint, the private sector considers some degree of substitution between public and pri- vate expenditure on health and education. It is noteworthy that we find an offset between private and public saving only when we account for inflationary capital g?ins and losses. Assuming that we have made this adjustment correctly, we can interpret the offset as an indication that the private sector internalizes the government budget constraint and that the private sector recognizes that inflation entails a transfer of income to the government. Table 9. Empirical Results: Determinants of the Private Saving Rate, Short- Run Effects Inflation-adjusted Inflation-adjusted private saving private saving Unadjusted Inflation- including including consumer private saving/ adjusted private durables/ durables and unadjusted saving/adjusted adjusted human capital private private private expenditure/adjusted disposable disposable disposable private disposable Variable" income income income income CR -0.39 -0.30 -0.07 -0.015 (1.46) (0.77) (0.17) (0.04) Self(-l) -0.30 -0.09 -0.04 -0.06 (1.24) (0.28) (0.09) (0.04) D[DOMCRGDP(-1)] 0.0004 -0.12 0.0002 0.0002 (0.69) (1.05) (0.16) (0.19) D[RR70_EX(-l)] 1.30* 0.07 -0.14 -0.14 (2.08) (0.68) (1.26) (1.31) D[SHAGRI(-1)] -0.14 0.65 0.29 0.29 (0.49) (1.37) (0.55) (0.55) D[GSR(-l)]b 0.39 -0.36 -0.20 -0.22 (1.31) (1.44) (0.67) (0.74) DfLRPDIMJf 0.04 -0.25 -0.19 -0.20 (0.32) (1.40) (0.88) (0.96) D[DEP(-1)] -2.53 -3.14 -0.47 -0.01 (1.25) (0.90) (0.15) (0.003) Note: f-statistics are in parentheses. a. See table A-2. b. Adjusted for inflation if dependent variable is adjusted private saving/private disposable income; adjusted for inflation and inclusive of human capital expenditure if dependent variable is adjusted private saving including expenditure on human capital/adjusted private disposable income. If dependent variable is private saving including consumer durables/adjusted private disposable income, then we use adjusted public saving unaugmented/adjusted private disposable income; if dependent variable is unadjusted private saving/unadjusted private disposable income, we use unadjusted public saving/unadjusted private disposable income. c If dependent variable is adjusted private saving/adjusted private disposable income, augmented or nnangmented, we use log of real private disposable income per capita; if dependent variable is unadjusted private saving/unadjusted private disposable income, we use log of real unadjusted private disposable income per capita. Source: Authors' calculations based on data in the World Bank's World Saving Database. 586 THE WORLD BANK ECONOMIC REVIEW, VOL. 14, NO. 3 However, another, less amenable, interpretation is also possible. If the inflation- related adjustment introduces measurement error in private and public saving, the negative relationship between the two would be spurious (given that, by construc- tion, the measurement error wduld have the opposite sign for private and public saving). Unfortunately, our methodology does not control for measurement error, and the interpretation of the negative coefficient on public saving has to rely on the quality of the adjustment. We believe that, on balance, the adjustment for inflation is correct and brings the saving figures closer to analytically correct measures. Fur- thermore, the size of the offset coefficient is consistent with that estimated for other countries, in both panel and single-country studies (see Loayza, Schmidt-Hebbel, and Served 2000 and Corbo and Schmidt-Hebbel 1991). The second explanatory variable, the ratio of domestic credit to the private sector to GDP, is an indicator of financial depth. It has a negative and significant relationship with private saying rates, both unadjusted and adjusted for infla- tionary capital gains and losses. This negative relationship is consistent with the notion that financial development allows households and small firms to use col- lateral more widely to reduce down payments on loans for housing and con- sumer durables. This should reduce private saving as individuals are able to fi- nance higher consumption at their current income level. When the measure of private saving includes consumer durables, the negative relationship drops nota- bly (in absolute value) and becomes insignificant. This indicates that financial development induces private agents to change the composition of their assets to favor durable goods but does not affect the total volume of saving once this is correctly measured. This conclusion invites a reinterpretation of the finding that financial development reduces private saving, which is common in several indi- vidual and cross-country studies that ignore consumer durables (see Loayza, Schmidt-Hebbel, and Serven 2000 and Bandiera and others 2000). The third explanatory variable is the dependency ratio, which captures life-cycle effects. As expected, its estimated coefficient is negative and significant for all four measures of the private saving rate. In other words, the private saving rate moves in the same direction as the share of working-age people in the total population. Thus India's demographic transition in the past 30 years must have contributed to expand the aggregate private saving rate. We should expect that as India moves to the next phase of the demographic transition, in which the share of old people in total population expands, private saving rates will decrease accordingly. The effect of the real interest rate on saving differs drastically depending on the measure of saving used. The effect of the real interest rate on unadjusted private saving appears to be negative, meaning that the (negative) income effect of an increase in the real interest rate dominates the (positive) substitution effect. Once we adjust for inflation-related capital losses, however, the effect of interest rate changes sign, although it is not statistically significant. Given that our mea- sure of real interest rates is derived from nominal rates that may not respond fully to price changes, a decrease in real interest rates reflects a rise in inflation. Thus the more positive impact of real interest rates on saving rates once they are Loayza and Shankar 587 inflation-adjusted can be explained by the fact that a change in inflation affects both real interest rates and adjusted saving rates in the same direction. A corol- lary of this explanation is that if we had a measure of real interest rates that were fully independent of inflation, we would not see a different effect of the former on adjusted or unadjusted private saving rates. When we include consumer durables in the measure of private saving, the positive effect of the real interest rate becomes stronger and statistically significant. Taken at face value, this indi- cates that the substitution effect of interest rate changes is larger (or the income effect is smaller) when applied to expenditures on durables with respect to other forms of saving. The effect of per capita private disposable income on saving is also highly dependent on the measure of private saving used. This effect is positive for unad- justed saving, but becomes significantly negative for inflation-adjusted saving and even more negative when saving is augmented to include consumer durables and expenditures on education and health. Most of the literature finds that per- manent rises in income have a positive effect on the private saving rate. This has been regarded as a puzzle, given that a permanent rise in income should increase consumption by the same magnitude, and thus have no effect on the level of saving, but should reduce the saving rate. In India the puzzling positive effect of income does not apply once we correct for inflationary wealth losses and include durables and human capital investment in the measure of private saving. That is, a permanent increase in income is consumed, and this consumption mostly takes the form of nondurables. Given that in a country as poor as India most people have yet to satisfy their vital consumption needs, it is only natural that an in- crease in income translates into an increase in basic consumption, and not into financial saving, durable consumption, or expenditures on human capital. Finally, we consider the long-run effect on saving of an occupation structural variable, the share of agriculture in GDP. This variable is important given that India remains a largely agricultural economy. The dominance of agriculture im- plies that a large share of the population faces an uncertain income. The share of agriculture in GDP appears to have a positive effect on private saving rates (table 8). This finding is consistent with a precautionary saving motive. Taking into account monsoon risk, Indian farmers have a higher propensity to save than people dedicated to other activities. For the sake of completeness, we present the short-term coefficients of the error-correction model for each measure of the private saving rate. These coeffi- cients are, however, uninteresting as they are not statistically significant. V. CONCLUSION: WHAT IS SPECIAL ABOUT SAVING IN INDIA? India's saving rate has been consistently higher than that of most other coun- tries with comparable per capita income. Regarding aggregate national saving as a ratio of gross national disposable income, India has performed as well as the OECD countries. India also traditionally has relied largely on national saving to S88 THE WORLD BANK ECONOMIC REVIEW, VOL. 14, NO. 3 fuel investment needs, relying relatively less on foreign saving compared with other developing countries. Even the 1991 liberalization that led to an increase in foreign capital inflows has not significantly changed this scenario. Given the ambitious growth targets of the government and the current global environment, it is likely that policies oriented toward raising aggregate national saving will play a key role in Indian economic development. Our results on the determinants of private saving rates in India highlight six issues. First, Indian households that save appear to pierce the corporate veil, in that they internalize the saving actions of the corporate sector. This finding supports our strategy of focusing on aggregate private saving, rather than on its components. We find equivalence between household and corporate saving only when we augment the measure of household saving to include expendi- tures on durable goods. As with many other conclusions of the article, the cor- rections we make to measures of private saving are key to uncovering impor- tant relationships. Second, the private saving rate rises with the share of agriculture in GDP. A large share of India's GDP comes from agriculture, and a large proportion of the population lives in rural areas. This suggests that the income of many households is characterized by the uncertainty associated with agriculture in India. Uncer- tainty should introduce a precautionary motive to save that should manifest itself as a positive relationship between aggregate private saving rates and the share of agricultural income in GDP. Our empirical results support this claim. Third, the real interest rate is positively associated with private saving rates once these are adjusted for inflation-related capital losses and augmented to in- clude consumer durables. Taken at face value, this result indicates that in India the substitution effect of interest rate changes is larger (or the income effect is smaller) when applied to consumption of durables with respect to other forms of saving. A controversial implication of this result is that tax incentives that in- crease the net rate of return on saving instruments could increase private saving rates. We are, however, hesitant to recommend tax incentives given their distor- tionary effect on resource allocation and the associated fiscal impact. Fourth, in India financial development has induced private agents to change the composition of their assets in favor of consumer durable goods, but this does not affect the total volume of saving once it is correctly measured. Extending this result to other countries, this conclusion invites a reinterpretation of the finding that financial development reduces private saving (see Loayza, Schmidt-Hebbel, and Serven 2000 and Bandiera and others 2000). Fifth, in India the puzzling positive effect of income on private saving rates found in other studies does not apply once we correct for inflationary wealth losses and include durables and human capital investment in the measure of pri- vate saving. That is, a permanent increase in income is consumed, and this con- sumption mostly takes the form of nondurables. Whether this reflects the behav- ior of forward-looking agents who are not financially restricted or the behavior of people living below subsistence consumption, we cannot tell for sure. Natu- Loayza and Shankar 589 rally, given the extent of poverty in India, we favor the subsistence-consumption interpretation. Last, the dependency ratio has a negative effect on private saving rates, as expected from life-cycle considerations. Thus India's demographic transition in the past 30 years must have contributed to an increase in the aggregate private saving rate. As India moves to the next phase of the demographic transition, in which the share of elderly people in the total population expands, private saving rates should decrease accordingly. Although important, demographic variables such as the dependency ratio do not appear to be the sole determinants of private saving rates, as previous studies had indicated. We contend that the differences between our study's findings and previous findings are at least partially due to our use of inflation-corrected figures for income and saving and our inclusion in saving of private expenditures on consumer durables and accumulation in hu- man capital. Table A-l. Unit-Root Test Results Adjusted Dickey- Fuller test statistic/ Order of Variable critical value* integration DEP (dependency ratio) 0.38/-3.6 1(1)" 1 lag; intercept GSR (inflation-adjusted public saving including human -1.99/-3.6 1(1)" capital expenditure/adjusted private disposable income) 1 lag; intercept GSR_PU (unadjusted public saving/unadjusted private -0.99/-3.6 KD" disposable income) 1 lag; intercept GSR_PA (inflation-adjusted public saving/adjusted private -1.3/-3.6 l(D" disposable income) 1 lag; intercept PSR (inflation-adjusted private saving including human -1.7/-3.6 I(D" capital expenditure/adjusted private disposable income) 1 lag; intercept PSR_PA (inflation-adjusted private saving/adjusted private -2.0/-3.6 1(1)" disposable income) 1 lag; intercept PSRCON (inflation-adjusted private saving including -1.8/-3.6 I(D" durables/adjusted private disposable income) 1 lag; intercept PSR_PU (unadjusted private saving/unadjusted private -1.2/-3.6 KD" disposable income) 1 lag; intercept RR70_EX (real ex post rate of interest) -3J2/-3.6 1(1)" 2 lags; intercept SHAGRI (share of agricultural income in GDP) -0.81/-3.6 KD" 2 lags; intercept LRPDI_PU (natural log of unadjusted private disposable 1.2/-3.6 1(1)" income per capita) 1 lag; intercept LRPDI_PA (natural log of adjusted private disposable 1.39/-3.6 I(D" income per capita) 1 lag; intercept CORPSR (corporate saving/unadjusted private -0.87/-3.6 I(D** disposable income) 1 lag; intercept HHCON (household saving including durables/ -1.28/-3.6 KD" unadjusted private disposable income) 1 lag; intercept DOMCRGDP (domestic credit to the private sector/GDP) -1.1/-3.6 KD" 1 lag; no inter- cept, no trend * Significant at 5 percent. * * Significant at 1 percent; the null hypothesis is that there is a unit root. Source: Authors' calculations based on data in the World Bank's World Saving Database. S90 THE WORLD BANK ECONOMIC REVIEW, VOL. 14, NO. 3 Table A-2. Correlations between Saving Rates and Potential Saving Determinants Inflation-adjusted public saving Corporate including human saving/ Domestic capital unadjusted credit to the expenditure/ private- private adjusted private disposable Dependency sector/ disposable income ratio GDP income Corporate saving/unadjusted private disposable income 1.00 -0.17 0.16 0.02 Dependency ratio -0.17 1.00 -0.88 -0.61 Domestic credit to the private sector/GDP 0.16 -0.88 1.00 0.62 Inflation-adjusted public saving including human capital expenditure/adjusted private disposable income 0.02 -0.61 0.62 1.00 Inflation-adjusted public saving/adjusted private disposable income -0.01 -0.49 0.53 0.99 .Unadjusted public saving/ unadjusted private disposable income 0.03 0.39 -0.25 0.29 Household saving rate -0.13 -0.87 0.74 0.76 Inflation rate -0.05 -0.19 0.16 0.49 Natural log of adjusted private disposable income per capita 0.16 -0.95 0.79 0.47 Natural log of unadjusted private disposable income per capita 0.14 -0.97 0.82 0.54 Inflation-adjusted private saving including human capital expenditure/adjusted private disposable income 0.06 -0.74 0.65 0.51 Inflation-adjusted private saving/adjusted private disposable income 0.42 -0.72 0.57 t).35 Unadjusted private saving/ unadjusted private disposable income 0.29 -0.88 0.74 0.69 Inflation-adjusted private saving including durables/adjusted private disposable income 0.04 -0.73 0.62 0.47 Real ex post interest rate 0.23 -0.36 0.22 -0.07 Share of agricultural income inca>p -0.17 0.98 -0.86 -0.67 Loayza and Shankar 591 Inflation adjusted public Unadjusted sat/ing/ public saving/ adjusted unadjusted Natural log of private private adjusted private disposable disposable Household Inflation disposable income income saving rate rate income per capita -0.01 0.03 -0.13 -0.05 0.16 -0.49 0.39 -0.87 -0.19 -0.95 0.53 -0.25 0.74 0.16 0.79 0.99 0.29 0.76 0.49 0.47 1.00 0.39 0.68 0.51 0.35 0.39 1.00 -0.14 0.00 -0.47 0.68 -0.14 1.00 0.20 0.85 0.51 0.00 0.20 1.00 0.03 0.35 -0.47 0.85 0.03 1.00 0.43 -0.45 0.88 0.10 0.99 0.43 0.07 0.75 0.03 0.72 0.27 -0.03 0.68 -0.22 0.77 0.61 -0.12 0.90 0.14 0.88 0.39 0.02 0.73 0.04 0.72 -0.14 -0.11 0.34 -0.72 0.54 -0.57 0.33 -0.87 -0.27 -0.90 (Table continues on the following page.) 592 THE WORLD BANX ECONOMIC REVIEW, VOL. 14, NO. 3 Table A-2. (continued) Inflation-adjusted private savings Inflation-adjusted Natural log of including human private unadjusted private capital expenditure/ saving/adjusted disposable income adjusted private private disposable per capita disposable income income Corporate saving/unadjusted private disposable income 0.14 0.06 0.42 Dependency ratio -0.97 -0.74 -0.72 Domestic credit to the private sector/GDP 0.82 0.65 0.57 Inflation-adjusted public saving including human capital expenditure/adjusted private disposable income 0.54 0.51 0.35 Inflation-adjusted public saving/adjusted private disposable income 0.43 0.43 0.27 Unadjusted public saving/ unadjusted private disposable income -0.45 0.07 -0.03 Household saving rate 0.88 0.75 0.68 Inflation rate 0.10 0.03 -0.22 Natural log of adjusted private disposable income per capita 0.99 0.72 0.77 Natural log of unadjusted private disposable income per capita 1.00 0.71 0.74 Inflation-adjusted private saving including human capital expenditure/adjusted private disposable income 0.71 1.00 0.77 Inflation-adjusted private saving/adjusted private disposable income 0.74 0.77 1.00 Unadjusted private saving/ unadjusted private disposable income 0.89 0.74 0.87 Inflation-adjusted private saving including durables/adjusted private disposable income 0.71 0.99 0.75 Real ex post interest rate 0.47 0.50 0.72 Share of agricultural income in GDP -0.93 -0.72 -0.69 Source: Authors' calculations bated on data in the World Bank's World Saving Database. Loayza and Shankar 593 Inflation-adjusted private saving Unadjusted private including saving/unadjusted durables/adjusted Share of private disposable private disposable- Real ex post agricultural mcome mcome tnterestrate income in GDP 0.29 0.04 0.23 -0.17 -0.88 -0.73 -0.36 0.98 0.74 0.62 0.22 -0.86 0.69 0.47 -0.07 -0.67 0.61 0.39 -0.14 -0.57 -0.12 0.02 -0.11 0.33 0.90 0.73 0.34 -0.87 0.14 0.04 -0.72 -0.27 0.88 0.72 0.54 -0.90 0.89 0.71 0.47 -0.93 0.74 0.99 0.50 -0.72 0.87 0.75 0.72 -0.69 1.00 0.72 0.47 -0.88 0.72 1.00 0.50 -0.71 0.47 0.50 1.00 -0.28 -0.88 -0.71 -0.28 1.00 594 THE WORLD BANK ECONOMIC REVIEW, VOL. 14, NO. 3 REFERENCES The word "processed" describes informally reproduced works that may not be com- monly available through library systems. Athukorala, Premachandra, and Kunal Sen. 1995. "Economic Reforms and Rate of Sav- ing in India." Economic and Political Weekly 30:2184-90. Bandiera, Oriana, Gerard Caprio, Patrick Honohan, and Fabio Schiantarelli. 2000. "Does Financial Reform Raise or Reduce Private Savings?" Review of Economics and Statis- tics 82(2):239-63 Cashin, Paul, Nilss Olekalns, and Ratna Sahay. 1998. "Tax Smoothing in a Financially Repressed Economy: Evidence from India." IMF Working Paper WP/98/122. Interna- tional Monetary Fund, Washington, D.C. Processed. Corbo, Vittorio, and Klaus Schmidt-Hebbel. 1991. "Public Policies and Saving in Devel- oping Countries." Journal of Development Economics 36(1):89-115. Haque, Nadeem U., and Peter Montiel. 1989. "Consumption in Developing Countries: Tests for Liquidity Constraints and Finite Horizons." Review of Economics and Sta- tistics 71(August):408-15. Johansen, Sorenoren. 1991. "Estimation and Hypothesis Testing of Cointegrating Vec- tors in Gaussian Vector Autorcgressive Models." Econometrica 59:1551-80. Jorgenson, Dale W., and Barbara M. Fraumeni. 1989. "The Accumulation of Human and Nonhuman Capital, 1948-84." In Robert E. Lipsey and Helen Stone Tice, eds., The Measurement of Saving, Investment, and Wealth. Cambridge, Mass.: National Bureau of Economic Research. Lahiri, Ashok. 1989. "Dynamics of Asian Savings: The Role of Growth and Age Struc- ture." IMF Staff Papers 36:228-61. Loayza, Norman, Humberto Lopez, Klaus Schmidt-Hebbel, and Luis Serven. 1998. "The World Saving Data Base." World Bank, Washington, D.C. Processed. Loayza, Norman, Klaus Schmidt-Hebbel, and Luis Serv6n. 2000. "What Drives Private Saving across the World?" Review of Economics and Statistics 82(2):165-81. Muhleisen, Michael. 1996. "India—Policies to Increase Domestic Saving." International Monetary Fund, Washington, D.C. Processed. Pandit, B. L. 1991. The Growth and Structure of Savings in India. Bombay: Oxford University Press. Pesaran, Hashem. 1997. "The Role of Economic Theory in Modelling the Long-Run." Economic Journal 107:178-91. Pesaran, Hashem, and Yongcheol Shin. 1997. "An Autoregressive Distributed Lag Mod- elling Approach to Cointegration Analysis." University of Cambridge, Cambridge, U.K. Processed. . 1999. "Long-Run Structural Modelling." University of Cambridge, Cambridge, U.K. Processed. Poterba, James M. 1987. "Finite Lifetimes and the Effects of Budget Deficits on National Saving." Journal of Monetary Economics 20(September):369—91.