This paper discusses the intermediation of financial saving in India and the implications for growth. Recent studies linking financial sector development and growth in India are reviewed. The following statistical data are also provided: employment and labor statistics, agricultural production and yields, index of industrial production, saving and investment, price developments, balance of payments, official reserves, reserve money, monetary survey, central and state government operations, indicators of financial system soundness, financial performance of Indian commercial banks, and selected monetary and exchange rate indicators.


This paper discusses the intermediation of financial saving in India and the implications for growth. Recent studies linking financial sector development and growth in India are reviewed. The following statistical data are also provided: employment and labor statistics, agricultural production and yields, index of industrial production, saving and investment, price developments, balance of payments, official reserves, reserve money, monetary survey, central and state government operations, indicators of financial system soundness, financial performance of Indian commercial banks, and selected monetary and exchange rate indicators.

III. Nonresident Deposits in India: Trends and Determinants1

A. Introduction

1. Since the 1980s, nonresident Indians (NRIs) have placed large amounts of funds on deposit with the Indian banking system. Such inflows have provided India with a valuable source of foreign exchange. In March 2003, NRI deposits totaled $28.4 billion, which is about one sixth of the size of deposits of residents with commercial banks. Most of the NRI deposits are repatriable, and if account is taken of two foreign-currency bonds issued exclusively to NRIs—the Resurgent India Bond (RIB) in 1998, and the India Millennium Deposit (IMD) in 2000—NRIs hold 60 percent of India’s external debt that is owed to private creditors.

2. NRI deposit-taking gained momentum in the 1980s in conjunction with the increasing number of Indians going to work overseas, particularly in the Gulf countries. So as to draw their savings back to India, the government formulated NRI deposits schemes that offered attractive interest rates and were fully repatriable. The deposits were made subject to concessionary reserve ratios and liquidity requirements, and the Reserve Bank of India (RBI) assumed the exchange rate risk on foreign currency denominated accounts. However, the schemes proved to be vulnerable during the 1991 balance of payments crisis, when outflows of deposits compounded the pressure on the external accounts (Acharya, 2001). As a result, steps were taken to enhance the stability of the deposits by switching the composition towards rupee denominated accounts, and by reducing the repatriable component (RBI, 2000). The exchange risk on foreign-currency deposits was shifted back to the banks.

Table III. 1.

India: Capital Flows

(In billions of U.S. dollars)

article image
Source: RBI.

April 2000–December 2002.

3. Since the 1991 crisis, NRI deposit inflows have continued to be substantial, but their relative importance in the external accounts has declined. The past decade has witnessed rapid growth in I-T exports and inward foreign investment and India’s foreign reserve position has become increasingly comfortable.2 The authorities have responded by linking the interest rates offered on NRI foreign currency deposits more closely with Libor; by giving the banks flexibility to set interest rates on rupee deposits; and, from April 2002, by making all new deposits fully repatriable.

4. This chapter analyzes trends in the accumulation of NRI deposits and investigates the determinants of these inflows. It finds that monthly deposit flows have been quite stable since the 1991 crisis, with standard deviations comparable to portfolio equity inflows into India, which are themselves of fairly low volatility. Nevertheless, there have been occasions when monthly flows of deposits turned negative. Such incidents have coincided with clearly identifiable adverse domestic or external events, but the effects have tended to be short-lived.

5. The determinants of NRI deposits are analyzed in a simple OLS framework. Econometric analysis shows that the differences between the interest rates on NRI deposits and Libor are determinants of deposit inflows. This implies that the generally higher level of Indian interest rates has been a factor driving these flows. Evidence is found of withdrawals from foreign currency deposits at the time of the two extraordinary bond issues—the RIB and the IMD—both of which offered high interest rates. This supports the finding that NRI deposits are interest sensitive. The regression results also confirm that NRI deposits are affected by political and geopolitical uncertainties, such as the government resigning mid-term, or tensions on India’s borders. In addition, deposits seem to be vulnerable to contagion from regional turbulence such as that which occurred during the Asian crisis.

6. The chapter is organized as follows: Section B discusses trends in NRI deposits. Section C describes the data and methodology used in the econometric exercise, while Section D presents the econometric results. Section E concludes.

B. Trends in NRI Deposit Flows


7. NRI deposits were first introduced in February 1970. The initial scheme was a rupee-denominated account, the Nonresident (External) Rupee Account (NRERA), with repatriable principal and interest. In November 1975, a foreign currency denominated deposit facility, the Foreign Currency Nonresident Account (FCNRA) was added. This deposit was also repatriable and was made attractive to the banks through the RBI assuming the exchange rate risk. Nevertheless, the two schemes got off to a slow start (Figure III.1). In the 1970s, India was a closed economy that made little use of foreign capital other than donor assistance, and there was not much need to mobilize NRI deposits. By March 1980, NRERA deposits had risen to only $850 million, while the FCNRA scheme had attracted less than $200 million.

Figure III.1.
Figure III.1.

NRI Deposits

(In millions of US dollars)

Citation: IMF Staff Country Reports 2003, 261; 10.5089/9781451818574.002.A003

8. In the 1980s, inflows into NRI deposits accelerated, with the FCNRA schemes enjoying particularly rapid growth. These inflows coincided with the widening of the current account deficit, and the need for increased borrowings on commercial terms. About half of the nonconcessional debt inflows during the decade was external commercial borrowing (ECB); the other half came from NRI deposits. By March 1990, the stock of NRI deposits had grown to $12.4 billion, of which 70 percent was foreign-currency denominated. The preferences of depositors for foreign currency deposits reflected the favorable interest rates on such deposits, together with the lack of exchange rate risk. Banks found FCNRA accounts to be an attractive source of funding because the RBI assumed the exchange rate risk, and because they tended to have a less onerous Cash Reserve Ratio (CRR) and Statutory Liquidity Requirement (SLR) than deposits raised domestically.


9. The balance of payments crisis in the early 1990s was associated with some outflow of NRI deposits. The amount outstanding under the NRERA and the FCNRA schemes fell by $904 million during fiscal year 1991–92.3 The crisis revealed a number of fault lines in the NRI deposit system. To begin with, the deposits were fully repatriable and thus free to leave upon maturity. Moreover, while about three-quarters of deposits at that time were of longer than one year maturity (RBI, 2000), deposits could be closed prematurely subject to an interest rate penalty and early encashment may have intensified the pressures on the external accounts.4 Moreover, as the rupee depreciated, the RBI began to sustain large losses on account of the exchange rate guarantee.

10. In the wake of the crisis, the authorities made a number of changes to the deposit schemes. In June 1992, a new rupee-denominated scheme, the NRNR account was introduced, which allowed the repatriation of interest income only. Deposits under this scheme were exempted from SLR and CRR requirements (Figure III.2). In May 1993, a new repatriable foreign-currency scheme, the FCNRB, was introduced, which differed from the FCNRA in that the banks were made to bear the exchange rate risk themselves. The FCNRA scheme was closed to new deposits with effect from August 1994, and by 1997, all remaining balances had been repaid.

Figure III.2.
Figure III.2.

Post-Crisis Cash Reserve Ratios

(In percent)

Citation: IMF Staff Country Reports 2003, 261; 10.5089/9781451818574.002.A003

11. In the period since the crisis, NRI deposits have continued to accumulate. During the 1990s, inflows under the FCNRB scheme were almost sufficient to offset the FCNRA repayments, but there were net inflows of $10 billion during the decade under the rupee schemes. Over the last three years, the pace of NRI deposit accumulation has accelerated, and average inflows have exceeded $2 billion per annum. At the same time, foreign reserves have grown rapidly. With the external position increasingly comfortable, efforts have been made to reduce the interest rates on NRI deposits, and to subject NRI deposits to the same CRR and SLR as resident deposits. From April 1, 2002, the NRNR scheme was discontinued and all new NRI deposits were made fully repatriable.

Properties of NRI Deposits

  • Currency composition. As a result of the post-crisis changes, the share of foreign currency denominated deposits in total NRI deposits has fallen from 70 percent in 1990 to 36 percent in March 2003. The net inflow of foreign-currency deposits since 1994—the sample period considered in Section C below—has been negligible.

  • Repatriable component. The post-crisis measures led to the repatriable component of NRI deposits declining from 100 percent in March 1990 to 69 percent by March 2000. However, the discontinuation of the NRNR scheme from April 2002 has caused the repatriable component of deposits to rise back to 88 percent by March 2003. This proportion will eventually increase to 100 percent once the outstanding NRNR deposits mature.

  • Stability. In terms of standard deviations, there is similar volatility in monthly inflows of rupee deposits, foreign currency deposits and foreign portfolio investment into India (FII flows). However, as noted above, the post-1994 series for monthly inflows into foreign currency deposits has a very small mean, and thus a much higher coefficient of variation than either rupee deposits or FII equity flows (Table III.2).

  • Effects of economic and political events. In Table III.3, the effects of selected major economic and political events on net deposit flows are calculated. These include the government resigning mid-term, tensions on the border with Pakistan, credit rating downgrades, the Asian crisis, and substantial increases in the prices of oil. Net monthly deposit inflows are found to be affected by these events, but the magnitude and duration of the effects appear to be small.

  • Effects of alternative bond schemes. To counter pressure on foreign reserves following the imposition of sanctions after the nuclear tests, a Resurgent India Bond was issued in August 1998. The RIB, which raised $4.2 billion, was targeted exclusively at the NRI community and paid an interest rate of 7¾ percent on U.S. dollar deposits. A second scheme, the Indian Millennium Deposit, raised $5.5 billion in October and November 2000. The IMD offered an interest rate on U.S. dollar deposits of 8½ percent.5 There were negative monthly flows of NRI deposits at the time of both bond issues. The RIB seems to have had a particularly pronounced impact on foreign currency deposits, which fell by $600 million during the month it was issued (Table III.3).

Table III.2.

Volatility of NRI Deposits and FII Equity Flows

article image
Data for foreign-currency deposits and Rupee deposits are from 1994:4–2002:12, and the data for FII equity flows are from 1994:4–2001:10.
Table III.3.

Monthly NRI Deposits Flows

(Monthly averages; in millions of U.S. dollars)

article image
Data for foreign-currency deposits and Rupee deposits are from 1994:4–2002:12. Number of months for which the averages have been taken are given in parentheses.

An oil shock is identified as a 10 percent increase in the price of oil over previous month.

C. Data and Methodology

12. Monthly data on foreign currency deposits are consistently available from April 1980–March 1991, and from March 1994–December 2002, and monthly data on rupee deposits are available for the period April 1980–December 2002. However, since only limited data are available for some of the possible explanatory variables for the earlier period, the analysis is restricted to the period March 1994–December 2002.

13. Previous research on NRI deposits has been conducted by Mohanty, Kapur and Sahoo (2000). In comparison to that study, this chapter considers a more comprehensive set of possible determinants of NRI deposits over a slightly longer data period. Moreover, Mohanty et al (2000) find a strong link between NRI deposits and foreign exchange reserves, which they ascribe to the comfort factor that depositors derive from the level of reserves. However, as they recognize, there is a potential for simultaneity in this relationship, and the regressions below instead use instruments such as sovereign ratings and dummies for political and geopolitical events to capture credit risk.

14. The dependent variable is net inflow of deposits measured in U.S. dollars. Separate regressions are estimated for two different variants of NRI deposit flows:

  • (i) Change in foreign currency denominated deposits: FCDEP = FCNRA + FCNRB

  • (ii) Change in rupee denominated deposits: REDEP = NRE + NRNR.

15. In choosing explanatory variables, the hypothesis made is that deposit flows result from a portfolio choice by NRIs. The interest rates on the various deposit schemes are therefore included as explanatory variables in the regressions. To capture relative returns, the returns on alternative investment opportunities open to NRIs such as equities are added. Since the riskiness of holding deposits should affect inflows, variables to proxy for Indian sovereign and geopolitical risk are also included. The wealth levels of NRIs are likely to influence their savings and should be an additional determinant of NRI deposits. However, NRI wealth is difficult to measure. As an imperfect measure, an oil price variable is used to capture the wealth of NRIs based in Gulf countries.

16. NRI deposit inflows are also influenced by government policy. For a large part of the sample period, interest rates on NRI deposits were controlled and thus were a policy variable that could be altered in response to the strength of NRI deposit flows. This creates a potential endogeneity problem in the econometrics that is discussed below. The CRR and SLR are other policy variables available to increase the attractiveness of NRI deposit collection to banks. The CRR, in particular, was changed quite frequently during the sample period.6

17. Interest differentials seem more appropriate than interest rate levels as explanatory variables. The interest rate on dollar deposits (DOLINT) and Libor are highly correlated. The same holds to a lesser extent for the interest rates on rupee deposits (NREINT) and Libor.7 Thus, instead of including the interest rate series separately in the regressions, their difference is considered. The equation for foreign-currency deposits includes the difference between DOLINT and Libor, while the rupee deposit equation includes the difference between NREINT and Libor.

18. As noted, it can be argued that the interest rate on NRI deposits is a policy variable. It is thus potentially endogenous to the deposit flows, and using the contemporaneous interest rate variable may give biased results. However, the lag in collecting data on NRI deposits makes it likely that any policy response to deposit flows also occurs with a lag, rather than contemporaneously. It thus seems legitimate to include contemporaneous interest rates in the regressions.

19. The expected return on rupee deposits depends on the expected path of the exchange rate. However, NREINT is calculated as an ex post return.6 To capture expectations, the current month-on-month exchange rate is therefore also included as an explanatory variable. Here there is another potential endogeneity issue, since the exchange rate is in turn likely to be affected by deposit flows. The exchange rate variable is therefore lagged by one month (LEXCC).

20. Since equities are an alternative place for the NRI community to put their savings, higher stock market returns may imply a smaller flow of deposits. The monthly return on the Dow Jones Industrial Average is therefore a possible explanatory variable (DOWC), with the expectation that the sign of the coefficient will be negative. However, there may be an offsetting effect to the extent that changes in the Dow imply a wealth effect for NRIs. The return on the Bombay Stock Exchange is also included (BSEC). Again, however, it is difficult a priori to predict whether the sign of the coefficient on this variable should be positive (since local stock market returns are likely to vary inversely with perceived country risk) or negative (since like the Dow, the BSE is an alternative investment opportunity).

21. Before committing funds, NRIs are likely to look for possible signs that the banks taking their deposits may experience repayment difficulties, particularly with regard to deposits denominated in foreign currency. A dummy variable is thus adopted for downgrades in India’s sovereign rating (RATING), as well as for political events such as the government resigning midterm (GOVT), and for geopolitical events such as the nuclear tests, the Kargil war, and the Indo-Pakistan stand off in mid-2002 when travel advisories were issued by several countries (GEOPOLT).

22. Evidence shows that the capital flows to emerging markets are susceptible to financial crises in other emerging markets. In the case of India, portfolio inflows slowed during the Asian crisis, but were not affected by crises in other emerging markets (Gordon and Gupta, 2002). In order to test for similar effects on the flow of NRI deposits, dummy variables are introduced for the Asian crisis (ASIA), and for crises in emerging markets outside Asia.

23. Additional dummy variables are also introduced. One set of dummies takes account of the effects of the Resurgent India Bond (DUMRIB), and India Millennium Deposit (DUMIMD) on deposit flows. To reflect possible seasonality in the data for NRI deposits, further dummies are added to capture beginning and end of year effects.

24. An attempt is made to capture the wealth of NRIs using an oil price variable. Since many expatriate Indians are located in the Gulf, the prosperity of that region will affect the flow of their funds back to India.8 Oil price changes are therefore included in the regressions (OIL). However, it is possible that the deposit flows respond only to large swings in oil prices, so an oil-shock dummy is experimented with which takes a value one if the oil price increase exceeds ten percent over the previous month (OIL 10). The robustness of the results to an oil price increase of five percent is also checked.

25. Tests for unit roots and autocorrelation were conducted. The time series properties of the dependent and independent variables were analyzed by estimating the following equation for each variable:


and testing for the null hypothesis, ρ=1, against the alternative hypothesis ρ≠1. The results are presented in Table III.4. Since most of the series are in percentage terms, the series are found to be I(0) and the null hypothesis of a unit root is rejected in favor of the alternative hypothesis.9

Table III.4.

Unit Root Tests of Dependent and Independent Variables1/

article image
Source: Authors’ own calculations.

The Phillips-Perron test allows for serial correlation and heteroscedasticity in the error term. The P value is the probability with which the null hypothesis of unit root can be accepted. The tests have been conducted including a constant, but no trend, in the regressions. For variables NRNRC and NREC the time period included in the regressions is 1994:4–2002:4, for all other variables it is 1994:4–2002:12.

D. Econometric Results

26. In the regression analysis, a general specification is first estimated which includes all the variables. Then the variables with the least significant coefficients are dropped. Results reported in Table III.5 are for a subset of the equations estimated using the parsimonious set of regressors. Sensitivity analysis was conducted (not reported) which reveals most of the results to be robust.

Table III.5.

Regression Results for Foreign Currency and Rupee Deposits

article image
t- Statistics in parentheses. *, **, ***, **** indicate significance at 15, 10, 5 and 1 percent respectively.

27. Regression results indicate a positive association between NRI deposit inflows and the difference between NRT deposit rates and Libor. The coefficient of the interest rate differential variable is more significant for rupee deposits than for foreign currency deposits, perhaps reflecting that foreign currency deposit rates have tended to be very similar to Libor.

28. The finding of interest rate sensitivity is supported by the behavior of deposits at the time the extraordinary bonds were issued. In particular, the floating of the RIB seems to have led to some diversion of foreign-currency deposits into these high interest bonds. In the FCDEP regression, the coefficient of the variable DUMRIB is negative, and significant at the 1 percent level. By contrast, the coefficient of the variable DUMIMD is positive, but insignificant. In the REDEP regressions, both dummies have coefficients insignificantly different from zero.

29. Concessionary CRRs do not appear to be a major factor explaining deposit inflows. The coefficient on the CRR differential—defined as the difference between the CRR applying to NRI deposits (CRRFD) and the CRR applying to resident deposits (CRRRD)—is negative, as would be expected, but not significant. This contrasts with Mohanty et al (2000) who find a significant effect of a dummy variable for periods when differential CRRs applied.

30. Domestic political uncertainty is associated with a reduced inflow of foreign currency deposits. The coefficient on GOVT is significant at the 10 percent or 15 percent level. The variable is mostly insignificant in the regressions for rupee deposits. What seems to matter more for rupee deposit flows is the uncertainty associated with geopolitical risks. The coefficient of this variable is negative and significant at 5 percent or higher levels in the regressions for rupee deposits.

31. NRI deposit flows are found to have been affected adversely by the Asian crisis, but not by crises elsewhere in the world. The coefficients on ASIA for both foreign currency and rupee denominated deposits are negative and significant at the usual significance levels. The coefficient of a dummy variable for major crises in Mexico, Russia, Turkey, Brazil, and Argentina that occurred during the sample period is negative, but insignificant (results not reported). As noted, similar findings have been made for inward portfolio investment into India.

32. The period of the Asian crisis was marked by tumultuous political events in India. These included the Gujral Government resigning in November 1997, and the nuclear tests in May 1998. To test for the possibility that the ASIA dummy is picking up the effects of these domestic events, the ASIA, GOVT and GEOPOLT variables are included separately in the equations. The coefficients of the included variables are found to be somewhat more significant when only one variable (rather than all) are included in the regressions, but otherwise the results mostly remain unchanged.10

33. Rupee deposit flows are found to be associated negatively with exchange rate depreciation, but the coefficient is not significant. This is puzzling given that the bivariate correlation between rupee deposits and the lagged exchange rate (LEXCC) is strongly negative. However, rupee depreciation can also be shown to be positively correlated with the Asian crisis (ASIA) and domestic political events (GOVT). Since these variables are both associated negatively with rupee deposit flows, it seems likely that including them in the REDEP regression is taking away the significance of the LEXCC variable.

34. Ratings downgrades do not appear to discourage NRI deposits. The coefficient of the credit rating downgrade variable in the foreign currency deposit equation is positive and marginally significant, and negative, but insignificant, in the rupee deposit equations. This result is at odds with the finding in Gordon and Gupta (2002) that credit downgrades are associated with smaller portfolio flows into India. However, that foreign investors (FIIs) rely more on the views of foreign ratings agencies than do expatriate Indians (NRIs) is perhaps not surprising.

35. Large swings in oil prices seem to be important in determining the flow of foreign currency deposits, but not the flow of rupee deposits. This asymmetry is difficult to explain. If higher oil prices do make NRIs better off, it is difficult to see why the wealth effects should apply to foreign-currency and not rupee deposits. Thus it seems likely that the oil price variable is capturing some other effect.

36. Finally, stock market returns do not appear to be important influences on deposit inflows. While foreign currency deposits are found to be associated positively with returns on the BSE, the coefficient in the rupee deposit equation is not significantly different from zero (results not reported). Moreover, the return on foreign stock markets (DOWC) is not found to be associated significantly with either foreign-currency or rupee deposit flows.

E. Conclusion

37. While NRI deposits have provided a substantial and stable source of foreign exchange, they appear to be influenced by standard risk and return variables. India’s ability to tap a large pool of expatriate savings is an option available to few countries. Moreover, in the period since the 1991 crisis, NRI deposits have supported India’s balance of payments during a number of episodes of market turbulence. Nevertheless, NRI deposits are found to be risk and return sensitive, so they do not substitute for the need to maintain stable economic conditions and a sound policy stance.


  • Acharya, Shankar, 2001, “India’s Macroeconomic Management in the Nineties,ICRIER Working Paper, ICRIER, New Delhi.

  • Gordon, James and Poonam Gupta, 2003, “Portfolio Flows into India: Do Domestic Fundamentals Matter?,IMF Working Paper 03/20 (Washington: International Monetary Fund).

    • Search Google Scholar
    • Export Citation
  • Hamilton, James, D., 1994, Time Series Analysis (Princeton, New Jersey: Princeton University Press).

  • Judge, G., W. E. Griffiths, R. C. Hill, H. Lutkepohl, and T. Lee, 1985, The Theory and Practice of Econometrics, (New York: Wiley, 2nd Edition).

    • Search Google Scholar
    • Export Citation
  • Mohanty, Deepak, Muneesh Kapur and Satyananda Sahoo, 2000, “Management of Non-Resident Deposits,” in External Debt Management-Role of Financial Intermediaries, Proceedings of CRISIL Seminar, Goa, September 6–8, pp. 335346.

    • Search Google Scholar
    • Export Citation
  • Pindyck, Robert. S. and Daniel L. Rubinfeld, 1981, Econometric Models and Economic Forecasts, (New York: McGraw Hill).

  • Reserve Bank of India, 2000, “Report of the IDF Project on Non-Resident Deposits” (Mumbai; RBI, Department of Economic Analysis and Policy) (October).

    • Search Google Scholar
    • Export Citation

Prepared by James Gordon and Poonam Gupta.


See Chapter II of this volume.


It is likely that the outflow was even larger during the course of 1991–92, but monthly data are not available for the crisis period.


Mohanty, Kapur, and Sahoo (2000) note that the ability to encash deposits before maturity endows depositors with a put option which makes the deposits effectively short-term debt.


A precursor to these schemes was the India Development Bond issued to NRIs during the crisis in 1991. This bond raised $1.6 billion and was repaid in 1996.


By contrast, the SLR applicable to NRI deposits has not been changed very often. The SLR on NRE deposits was 30 percent in the first half of the 1990s compared to 38 percent for domestic rupee deposits. However, since 1997, the SLR on both has been set at 25 percent (RBI, 2000).


NREINT is a dollar return calculated as the interest rate on NRE deposits of up to one year, less the depreciation of the rupee over the previous 12 months.


In addition to bank deposits, the savings of NRIs also return to India in the form of remittances. It would be interesting to jointly analyze these two inflows since they probably share some common determinants. However, data on transfers are only available quarterly.


See Hamilton (1994). The tests used are Augmented Dicky-Fuller (ADF) and Phillips-Perron (PP).


A seasonal effect is also found in the flow of NRI deposits (Timel dummy): deposits are on average higher in the first four months of the year than during the rest of the year.

India: Selected Issues and Statistical Appendix
Author: International Monetary Fund