The interest rate effect on private saving: Alternative perspectives

Aizenman, Joshua; Cheung, Yin-Wong; Ito, Hiro Working Paper The interest rate effect on private saving: Alternative perspectives ADBI Working Paper, No. 715 Provided in Cooperation with: Asian Development Bank Institute (ADBI), Tokyo Suggested Citation: Aizenman, Joshua; Cheung, Yin-Wong; Ito, Hiro (2017) : The interest rate effect on private saving: Alternative perspectives, ADBI Working Paper, No. 715, Asian Development Bank Institute (ADBI), Tokyo This Version is available at: https://hdl.handle.net/10419/163206 Standard-Nutzungsbedingungen: Die Dokumente auf EconStor dürfen zu eigenen wissenschaftlichen Zwecken und zum Privatgebrauch gespeichert und kopiert werden. Sie dürfen die Dokumente nicht für öffentliche oder kommerzielle Zwecke vervielfältigen, öffentlich ausstellen, öffentlich zugänglich machen, vertreiben oder anderweitig nutzen. Sofern die Verfasser die Dokumente unter Open-Content-Lizenzen (insbesondere CC-Lizenzen) zur Verfügung gestellt haben sollten, gelten abweichend von diesen Nutzungsbedingungen die in der dort genannten Lizenz gewährten Nutzungsrechte. Terms of use: Documents in EconStor may be saved and copied for your personal and scholarly purposes. You are not to copy documents for public or commercial purposes, to exhibit the documents publicly, to make them publicly available on the internet, or to distribute or otherwise use the documents in public. If the documents have been made available under an Open Content Licence (especially Creative Commons Licences), you may exercise further usage rights as specified in the indicated licence. https://creativecommons.org/licenses/by-nc-nd/3.0/igo/ ADBI Working Paper Series THE INTEREST RATE EFFECT ON PRIVATE SAVING: ALTERNATIVE PERSPECTIVES Joshua Aizenman, Yin-Wong Cheung, and Hiro Ito No. 715 April 2017 Asian Development Bank Institute The Working Paper series is a continuation of the formerly named Discussion Paper series; the numbering of the papers continued without interruption or change. ADBI’s working papers reflect initial ideas on a topic and are posted online for discussion. ADBI encourages readers to post their comments on the main page for each working paper (given in the citation below). Some working papers may develop into other forms of publication. Unless otherwise stated, boxes, figures, and tables without explicit sources were prepared by the authors. ADB recognizes “China” as the People’s Republic of China. Suggested citation: Aizenman, J., Y.-W. Cheung, and H. Ito. 2017. The Interest Rate Effect on Private Saving: Alternative Perspectives. ADBI Working Paper 715. Tokyo: Asian Development Bank Institute. Available: https://www.adb.org/publications/interest-rate-effect-private-saving- alternative-perspectives Please contact the authors for information about this paper. Email: [email protected], [email protected], [email protected] Aizenman and Ito gratefully acknowledge the financial support of faculty research funds of University of Southern California and Portland State University. Cheung gratefully acknowledges the Hung Hing Ying and Leung Hau Ling Charitable Foundation (孔慶熒及 梁巧玲慈善基金) for their support through the Hung Hing Ying Chair Professorship of International Economics (孔慶熒講座教授(國際經濟)). Joshua Aizenman is Dockson Chair in Economics and International Relations, University of Southern California. Yin- Wong Cheung is Hung Hing Ying Chair and professor of international economics, City University of Hong Kong. Hiro Ito is professor of economics, Portland State University. The views expressed in this paper are the views of the author and do not necessarily reflect the views or policies of ADBI, ADB, its Board of Directors, or the governments they represent. ADBI does not guarantee the accuracy of the data included in this paper and accepts no responsibility for any consequences of their use. Terminology used may not necessarily be consistent with ADB official terms. Working papers are subject to formal revision and correction before they are finalized and considered published. Asian Development Bank Institute Kasumigaseki Building, 8th Floor 3-2-5 Kasumigaseki, Chiyoda-ku Tokyo 100-6008, Japan Tel: +81-3-3593-5500 Fax: +81-3-3593-5571 URL: www.adbi.org E-mail: [email protected] © 2017 Asian Development Bank Institute ADBI Working Paper 715 Aizenman, Cheung, and Ito Abstract Conventional logic suggests that lowering the policy interest rate will stimulate consumption and investment while discouraging people from saving, but low interest rates may also prompt people to increase their saving to compensate for the low rate of return. Using data on 135 countries from 1995 to 2014, this paper shows that a low-interest rate environment can yield different effects on private saving across country groups under different economic environments. A well-developed financial market, an aging population, and output volatility can all contribute towards turning the relationship between interest rates and saving negative. Among developing countries, when the nominal interest rate is not too low, we detect the substitution effect of the real interest rate on private saving. However, among industrial and emerging economies, the substitution effect is detected only when the nominal interest rate is lower than 2.5%. In contrast, emerging-market Asian countries are found to have the income effect when the nominal interest rate is below 2.5%. When we examine the interactive effects between the real interest rate and the variables for economic conditions and policies, we find that the real interest rate has a negative impact—i.e., income effect—on private saving if any output volatility, old dependency, or financial development is above a certain threshold. Further, when the real interest rate is below 1.5%, greater output volatility would lead to higher private saving in developing countries. JEL Classification: F3, F31, F32, F36 ADBI Working Paper 715 Aizenman, Cheung, and Ito Contents 1. INTRODUCTION ....................................................................................................... 1 2. THEORY AND EVIDENCE ABOUT PRIVATE SAVING............................................. 3 2.1 What Kind of Saving Do We Focus On? ........................................................ 3 2.2 Theoretical Predictions of the Determinants of Private Saving ...................... 4 2.3 Stylized Facts ................................................................................................ 6 3. BASELINE ESTIMATION ........................................................................................ 11 3.1 Estimation Model ......................................................................................... 11 3.2 Estimation Results ....................................................................................... 11 4. INTERACTIVE EFFECTS ........................................................................................ 19 4.1 Empirical Findings ........................................................................................ 19 4.2 Implications for the World and Asia .............................................................. 25 5. CONCLUSION ......................................................................................................... 31 REFERENCES ................................................................................................................... 33 APPENDIX 1: SAMPLE COUNTRY LIST ............................................................................ 36 APPENDIX 2: DATA DESCRIPTIONS ................................................................................ 37 APPENDIX 3: ADDITIONAL ESTIMATION RESULTS ........................................................ 38 ADBI Working Paper 715 Aizenman, Cheung, and Ito 1. INTRODUCTION In the summer of 2014, when the European Central Bank changed its interest rate on excess bank reserves to –0.1%—a negative policy interest rate for the first time in not only its own history but also in the history of major central banks—advanced economies implementing unconventional monetary policies entered a new phase. 1 Eighteen months later, this action was followed by the Bank of Japan’s decision to adopt negative interest rates. As of the fall 2016, 19 euro countries, plus Japan, Denmark, Sweden, and Switzerland, have adopted negative policy interest rates. As unconventional actions often face opposition in general, negative interest-rate policies have also faced challenges against their effectiveness. Conventionally speaking, lower interest-rate monetary policy is supposed to encourage present-day consumption (as opposed to future consumption), by lowering the rewards for postponing consumption. More simply, lowering the policy interest rate is expected to stimulate consumption and investment while discouraging people from saving. Expected as a further drastic action, negative interest rates would not just discourage, but also penalize people if they postpone consumption. Hence, conceptually, negative interest rates should lead people to spend now rather than later and therefore discourage saving. Recently, debates have proliferated regarding the effectiveness of negative interestrate policy. Some people have argued that negative interest rates may not work as central bankers expect. As for the link between the interest rate and saving, the argument is as follows: lower or negative interest rates may contribute to higher, not lower, saving rates because the rate of return per financial instrument is so low that people may try to compensate by increasing their aggregate amount of saving. This scenario can be especially true in an economy with an aging population, as people might want to target their saving to be better prepared for retirement. Such a tendency can also be strong in an economy in which sufficient social protections such as social securities and unemployment benefits are not available. Generally, people may want to increase their aggregate amount of saving in response to lower interest rates if they face a gloomy and more volatile economic outlook. Thus, the behavior of precautionary saving may change depending on economic or policy conditions. This is not just an issue for advanced economies with low or negative interest rates, but for developing economies as well. In fact, in a developing economy with financial repression, nominal interest rates tend to be artificially repressed and therefore the real rates of return tend to be low. This situation can exacerbate if the economy of concern experiences high inflation. If such an economy is also coupled with underdeveloped public social-protection programs, people have reason to increase the aggregate amount of saving for precautionary purposes. While the interest rate effect on private saving is commonly perceived to be positive, Nabar (2011) notes that the People’s Republic of China (PRC) experienced a combination of rising household saving and declining real interest rates during the 2000s. Using province-level data over the 1996–2009 period, Nabar empirically shows that when the return to saving declines, household saving rises. 1 As an exception, Denmark had lowered its benchmark rate to a negative figure in mid-2012. Another exception is Switzerland, which levied negative interest rates on CHF deposits from non-residents in 1972 to curb rapid capital inflows. This policy lasted until 1978. 1 ADBI Working Paper 715 Aizenman, Cheung, and Ito Is the PRC’s documented interest-rate-saving link an isolated instance or an example of the negative income effect of the interest rate? To shed some light on this question, we employ a panel of countries to conduct an extensive empirical study on the link between interest rates and private saving. At the outset, we recognize that the interest rate effect on private saving can be ambiguous. As noted earlier, low interest rates can discourage saving because of the substitution effect, or conversely, encourage saving via the income effect to achieve, say, a targeted saving goal. Because of the conflicting channels, the observed or final effect of the interest rate on saving can depend on the level of the interest rate itself as well as on other contributing factors. In an environment in which the interest rate is extremely low, the income effect may, for example, outweigh the substitution effect. In other words, in such an environment, agents may be worried about the possibility of not meeting financial investment objectives such as retirement, and therefore try to overcome the low return by increasing the aggregate volume of saving. In this case, lower interest-rate levels would lead to higher levels of saving. Or, the effect of the interest rate on saving may differ depending on macroeconomic or demographical conditions or policy environment. Examining the link between the interest rate and saving is important. In the short term, whether policy interest rates and saving rates have a positive or negative relationship also refers to the kind of impact a monetary policy would have on consumption and is therefore related to the question of stabilization measures. Furthermore, this issue is also important in the context of the global imbalance debate. In the years leading up to the Global Financial Crisis of 2008 (GFC), many emerging market economies in East Asia (most notably the PRC) and oil exporters persistently ran current-account surpluses during the global trend of lower real interest rates. Some economists argue that high savings in rapidly growing emerging markets are responsible for such current account surpluses and thus contributing to global economic instability (Greenspan 2005a, b; and Bernanke 2005). Hence, investigating how an ultra-low-interest rate environment would contribute to saving on a global scale is important. In the long term, the impact of the interest rate on saving is related to the question of capital accumulation, which would determine future income level and thereby presentday consumption and saving. Thus, the nature of the interest-rate–saving link can be an important determinant for the sustainability of long-term economic development. Therefore, we investigate whether the interest rate has the income (i.e., negative) effect or the substitution (i.e., positive) effect on private saving by using panel data of 135 countries over the 1995–2014 period while controlling for other factors that can affect the behavior of private saving. Furthermore, we will empirically examine whether and how the impact of the interest rate on saving can be affected by economic, demographical, and policy conditions. Throughout the paper, we pay special attention to Asian emerging market economies. This is because, first, the Asian region has been identified as one of the most dynamic regions in terms of its robust economic growth and development, and second and more importantly, the region receives much attention and sometimes criticisms for its excess saving allegedly contributing to global current account imbalances. In the next section, we introduce potential determinants of private saving and discuss their impacts. In the same section, we present some stylized facts of private saving and the real interest rates to show general trends of these variables. In Section 3, we introduce our estimation model and discuss the results from the baseline estimations. 2 ADBI Working Paper 715 Aizenman, Cheung, and Ito We extend our analysis and examine whether any interactive effects exist between the real interest rate and other macroeconomic and structural conditions in Section 4. In this section, we also discuss the implications of our estimation results for several major emerging market economies. In Section 5, we offer concluding remarks. 2. THEORY AND EVIDENCE ABOUT PRIVATE SAVING 2.1 What Kind of Saving Do We Focus On? A large number of studies have investigated the determinants of saving; a sample of these studies include Masson et al. (1998), Loayza et al. (2000a, 2000b), Aizenman et al. (2015), and Aizenman and Noy (2013). Since these studies have provided comprehensive reviews on theory and empirical evidence pertaining to the determinants of saving, we focus on the theoretical predictions of the factors relevant to our empirical analysis. Before introducing potential determinants of saving, we need to clarify the kind of saving we are referring to. In this paper, we consider private saving, which we define as the difference between domestic saving and public saving. Considering that our interest is to assess the relative importance of income and substitution effects on shaping the interest-rate impact on saving, it would have been ideal if we had been able to focus on household saving. However, we have two reasons for avoiding using household saving data—one practical and the other conceptual. First, in a practical sense, household saving data are extremely limited. One reason for this scarcity is that household saving data are typically derived from government surveys that could be based on a wide variety of methods across countries (and over time). Even if we had a uniform survey method, disagreements could arise over what to include in consumption, saving, or disposable income when calculating the saving rate. For example, the question exists whether capital gains from financial investments should be included in saving or disposal income, or both. Similar concerns arise for social security payments, or depreciation of household assets in saving or income. Depending on the methodologies of data construction, there can be a wide variety of household saving. 2 Different types of household saving data exist for different countries. Also, the type of items that should be included in saving and income to compute the saving rate depends on the aspect of saving behavior a researcher chooses to study. Hence, a data set of household saving rate that is consistently compiled is hard to obtain. Although the Organisation for Economic Co-operation and Development (OECD) publishes consistent household saving data for 33 countries, the data are mostly composed from advanced economies. There is also a conceptual reason that makes it difficult to use household saving data. The line between household and corporate saving, which sum up to define private saving, can be blurry. This issue is prevalent among developing countries because of the existence of vast informal labor markets that make it difficult to separate corporate income from household income and vice versa. To a certain extent, there are also difficulties in disentangling household, corporate income, and consumption in advanced economies. 2 There can also be gross or net household saving. See Audenis et al. (2004) for details. 3 ADBI Working Paper 715 Aizenman, Cheung, and Ito Hence, we focus on private saving as a share of gross domestic product (GDP), in which we obtain the amount of private saving by subtracting the general government– budget balance from domestic saving while assuming the latter equals the sum of household, corporate, and public savings. 2.2 Theoretical Predictions of the Determinants of Private Saving We now discuss the theories underlying the determinants of private saving and, hence, the expected signs of estimated coefficients in the following empirical analysis. Persistence: Considering that economic agents usually try to smooth their consumption, private saving should also be smoothed out, and therefore, it tends to be serially correlated. Also, the determinants of private saving can have impact with some time lags; thus, private saving tends to show inertia. A number of empirical studies include the lagged dependent variable as one of the explanatory variables, and the lagged dependent variable tends to be highly significant with relatively large magnitudes. Public saving: The theory of Ricardian equivalence predicts that, in a world where tax policy creates no distortion, any change in public saving can be offset exactly by the same but opposite change in private saving, which makes its estimate negative with a magnitude of one. However, empirical studies usually show that a full offset is not existent, but that a partial offset is often prevalent, with the average absolute estimate ranging 0.25–0.60.3 Credit growth: If credit constraint is mitigated by credit growth, agents would increase their consumption, and hence, decrease saving (Loayza et al. 2000a, 2000b). Therefore, we can expect the estimate on credit growth to be negative. We include the growth rate of private credit creation (as a share of GDP) as a proxy for credit growth or credit availability. Financial development: Further financial development or deepening could induce more saving through increased depth and sophistication of the financial system. As a contrasting view, more developed financial markets lessen the need for precautionary saving and thereby lower the saving rate. Thus, the predicted sign of the estimate for the financial development variable is ambiguous. We use private credit creation (as a share of GDP) as a proxy for financial development. Financial openness: The impact of financial openness on saving behavior can also be explained similarly to that of financial development. To measure the extent of financial openness, we use the Chinn–Ito index (2006, 2008) of capital account openness.4 Both financial development and financial openness could affect the level of private saving through the price channel. That is, financial development and liberalization usually mitigates financial repression, in which the interest rate tends to be artificially depressed due to regulatory controls and lack of competition. Once financial repression is mitigated, higher interest rates can prevail and affect private saving, although the effect of interest rates on saving can be ambiguous. We can expect, at the very least, to see interactive effects between financial development or openness and the interest rate. 3 See de Mello et al. (2004). 4 For both financial development and financial openness, Chinn et al. (2014) find negative effects on national saving. 4 ADBI Working Paper 715 Aizenman, Cheung, and Ito 3. BASELINE ESTIMATION 3.1 Estimation Model With the above theoretical discussions and stylized facts in mind, we estimate the determinants of private saving using the empirical specification: 01 1 , it it it it it i t it y y rX Z u β β µε − ′′ = + + Γ+ Φ+ + + (1) where yit is private saving (normalized by GDP); X is a vector of endogenous variables; Z is a vector of exogenous variables; and rit is the real interest rate. ui refers to unobserved, time-invariant, country-specific effects, whereas µ t is a time-specific effect variable. ε it is the i.i.d. error term. Equation (1) entails a few possible technical issues. First, as we have already discussed, private saving can involve inertia. To allow for persistency in private saving data, we need to estimate a dynamic specification that can address both short- and long-term effects of explanatory variables. Second, some of the explanatory variables can be jointly determined with the saving rate. Hence, we have to account for joint endogeneity of the explanatory variables. Last, we need to control for unobserved country-specific effects correlated with the regressors. The system generalized method of moments (GMM) estimation method, which can consistently estimate a dynamic panel while allowing for joint endogeneity and controlling for potential biases arising from country-specific effects, is therefore adopted for our empirical exercise (Arellano and Bond 1991; Arellano and Bover 1995; Blundell and Bond 1998). In the vector X of endogenous variables, we include public saving (i.e., the general government budget balance normalized by GDP); financial development that is measured by private credit creation as a share of GDP; credit growth that is measured by the growth rate of private credit creation; and per capita income. These variables are treated as “internal instruments” in the GMM estimation. As exogenous variables, vector Z includes young and old dependency ratios, public healthcare expenditure (as a share of GDP), financial openness, output volatility, and per capita income growth. The variable of our focus is the real interest rate r. If the substitution effect outweighs the income effect, the estimate of β1 is expected to be positive. That is, the higher the interest rate, the more the country would save. On the other hand, if the income effect outweighs the substitution effect, β1 would be negative; that is, the higher the interest rate, the less private saving. 3.2 Estimation Results Table 1 reports the results of the estimations for the full sample and the subsamples of IDC, LDC, EMG, Latin American, Asian economies, (Asia), and the emerging market countries in Asia (Asian–EMG).12 Before discussing the system GMM estimates, we conduct diagnostic tests for the validity of the instruments and serial correlation in estimated residuals. For the former, we conduct the Hansen-J test against the null hypothesis that the instrumental variables are uncorrelated with the residuals. If the test fails to reject the null hypothesis, the specification is free of the issue of over-identification. As for serial 12 The sample period becomes 1995–2014 due to data limitations for the year 2015. 11 ADBI Working Paper 715 Aizenman, Cheung, and Ito correlation, we conduct an AR(2) test with the null hypothesis that the errors in the differenced equation exhibit no second-order correlation. This is because the system GMM method involves a first-difference transformation of the original estimation model to eliminate the unobserved country-specific effect. The estimated system GMM model specification is supported if no evidence exists of second-order autocorrelation (even there is first-order autocorrelation) and the over-identifying restrictions are not rejected at conventional levels of confidence. In Table 1 and the other tables, the reported diagnostic test results—both the Hansen-J and AR(2) test results—support the use of the system GMM model specification for all of these samples. That is, the Hansen test fails to reject the null hypothesis of over-identifying restrictions, and the AR(2) test confirms that the estimated errors in the differenced equation exhibit no second-order correlation.13 Generally, the estimation results are consistent with our theoretical discussions. First, the real interest rate, the variable of our focus, enters the estimation significantly for the full sample and the subsample of Asian economies group with a positive sign. This means we detect that the substitution effect outweighs the income effect for these groups of countries. For the other samples, the estimates are positive, except in the cases of the Latin American (LATAM) and Asian EMG, which are not significant. The behavior of private saving is found to be somewhat persistent. The degree of persistency is 0.390 for the full sample, although this varies across different subsamples. The groups of Asian economies and Asian EMG have higher degrees of persistency, 0.70 and 0.67 respectively, which is consistent with the prevailing observation that Asian economies’ saving rates are consistently high. We can observe evidence for the partial Ricardian offset in the estimated coefficient for public saving. The results of the full sample indicate that about 44% of an increase in public saving would be offset by a worsening of private saving. The size of the offset is much larger among industrialized countries than in developing economies, which may be because the tax system in the former is less distortive than in the latter. While the level of financial development only matters for industrialized economies, credit growth is found to be a negative contributor for developing economies. Once credit conditions improve, a developing country tends to experience growth in its consumption—that is, a fall in its saving rate. Financial openness, in contrast, is a positive contributor, although only for the IDC and the LATAM. For these economies, financial openness helps increase private saving through increasing investment opportunities. Although both the level and the growth of per capita income are found to positively contribute to private saving, output volatility has opposing effects for developed countries and the group of Asian–EMG economies. 13 However, Roodman (2006) argues that including too many instruments can not only overly fit endogenous variables, but also weaken the power of the Hansen test to detect over-identification. He suggests that high p-values (such as “1.00”) for the Hansen test may signal that the test wrongly failed to detect over-identification. In fact, in Table 1 and others, we see that the smaller the sample is (such as IDC, EMG, and regional country groups), the more tendency there is for the Hansen test’s p-value to take the value of “1.00.” This can be related to the fact that in a smaller sample, N (= the number of countries) tends to be small relatively to T (= the number of years)—the GMM estimation is more suitable for a data set with the dimension of large N and small T. However, when we apply the random effect model (not reported), the estimation results are qualitatively intact; in fact, they tend to become more robust. Hence, we focus on discussing the results from the GMM estimations. 12 ADBI Working Paper 715 Aizenman, Cheung, and Ito Table 1: Determinants of Private saving – System–GMM, 1995–2014 FULL IDC LDC EMG LATAM Asia Asia EMG (1) (2) (3) (4) (5) (6) (7) Private saving (t–1) 0.390 0.250 0.360 0.484 0.366 0.704 0.672 (0.080)*** (0.078)*** (0.088)*** (0.094)*** (0.076)*** (0.088)*** (0.064)*** Public saving –0.443 –0.715 –0.317 –0.651 –0.634 –0.458 –0.364 (0.150)*** (0.130)*** (0.167)* (0.102)*** (0.125)*** (0.140)*** (0.137)*** Credit growth –0.041 –0.020 –0.034 –0.026 –0.012 –0.005 0.015 (0.012)*** (0.024) (0.014)** (0.017) (0.024) (0.017) (0.024) Fin. development, HP-filtered –0.040 –0.022 –0.013 0.016 –0.085 0.014 0.140 (0.023)* (0.013)* (0.038) (0.047) (0.055) (0.033) (0.032)*** Income/capita level (log, PPP) 0.091 0.205 0.103 0.041 0.066 0.024 0.006 (0.028)*** (0.042)*** (0.031)*** (0.026) (0.036)* (0.018) (0.015) Real interest rate 0.075 0.048 0.070 0.020 –0.054 0.080 –0.002 (0.045)* (0.193) (0.044) (0.052) (0.047) (0.040)** (0.058) Old dependency –0.172 –0.206 –0.156 –0.268 –0.538 –0.259 0.136 (0.130) (0.191) (0.182) (0.199) (0.159)*** (0.122)** (0.334) Young dependency 0.099 –0.348 0.147 –0.147 0.023 –0.092 –0.057 (0.098) (0.235) (0.119) (0.136) (0.177) (0.100) (0.076) Health expenditure –1.321 –0.400 –1.748 –1.749 –0.206 –0.904 –3.592 (% of GDP) (0.486)*** (0.482) (0.488)*** (0.534)*** (0.493) (0.289)*** (0.903)*** Financial openness –0.009 0.017 –0.020 –0.007 0.059 0.012 –0.008 (0.021) (0.034) (0.022) (0.020) (0.016)*** (0.029) (0.024) Output volatility –0.009 0.870 0.001 0.271 0.376 –0.150 –0.357 (0.109) (0.519)* (0.119) (0.198) (0.311) (0.294) (0.165)** Income/capita growth 0.173 0.326 0.198 0.192 0.125 0.209 0.001 (0.058)*** (0.137)** (0.062)*** (0.081)** (0.068)* (0.091)** (0.081) N 2,313 431 1,882 755 436 364 218 # of countries 135 23 112 42 24 21 11 Hansen test (p-value) 0.08 1.00 0.58 1.00 1.00 1.00 1.00 AR(1) test (p-value) 0.00 0.03 0.00 0.01 0.06 0.00 0.02 AR(2) test (p-value) 0.47 0.80 0.40 0.85 0.33 0.93 0.87 IDC = industrialized countries; LDC = least developed countries; EMG = emerging market countries; LATAM = Latin America. Notes: * p<0.1; ** p<0.05; *** p<0.01. The dependent variable is private saving as a share of GDP. The system GMM estimation method is employed. Although the constant term is estimated, it is omitted from presentation. The subsample “Asia” includes Japan and East and South Asian economies. The higher the country’s level of old dependency, the lower the rate of private saving it tends to experience. Although the estimate on the old dependency variable is not significant for the LDC or EMG group, the estimates for the subgroups of LATAM, Asia, and Asian–EMG are significant and their magnitudes tend to be large. The fact that smaller numbers of countries are included in each of the estimations indicates that demographical change happened rather drastically in the sample period and had significant impact on private saving for the countries in these subsamples. Healthcare expenditure, which we measure by public health expenditure as a share of GDP, has a negative impact on private saving. That is, if healthcare is more readily available with the support of the public sector, people would reduce saving because they would not have to save for precautionary reasons. The estimate is robust across the different country groups based on income levels (i.e., full, IDC, LDC, and EMG). 13 ADBI Working Paper 715 Aizenman, Cheung, and Ito Also, when we use social expenditure as a share of GDP that is available in the OECD database, the results are essentially unchanged.14 Although we carefully chose explanatory variables, this sort of exercise can still be subject to missing variable bias. Here, we test two variables as potential determinants of private saving. The first one we suspect as a potential determinant is net investment position. Depending on time preferences and endowments, some become net lenders (i.e., current account surplus countries) at the present time while others become net borrowers. Hence, net investment positions, whose incremental changes are comparable to current account balances, can be related to private saving. From a different angle, foreign saving may crowd out or complement domestic private saving. Developing countries often try to mitigate credit constraint in their own domestic markets by importing foreign saving, though they also have to face external borrowing constraints such as difficulties in borrowing in their own currencies or for long-terms (i.e., the “original sin” argument).15 We test whether net investment positions affect the private saving rate by including a dummy for country-years in which the net position is negative.16 The estimation results (Table 3A) show that the saving rate tends to be lower for net debt countries, indicating that the saving rate tends to be lower for net debt countries. That means that foreign saving complements domestic saving. Another variable of our suspect is property prices. A rise in house prices could create a “wealth effect” on consumption while simultaneously mitigating credit constraint. Either way, we expect property prices to have a negative impact on saving. When we include real property price index in the estimation, we find such a negative impact only for EMG (Appendix 3, Table A1). For that group, as (real) property prices rise, the saving rate tends to fall.17 However, when we test the growth-rate impact of property prices, we find that its estimate is significantly negative for the full sample and the LDC subgroup (Table A2). In these samples, what matters is not so much the level of property prices as its growth rate. A rapid rise in property prices may signal an increase in future or permanent income flows. 1.1.1 Level Impacts of the Interest Rates The weak evidence of the real interest-rate effect in Table 1 is likely to be attributable to its dependency on other economic conditions affecting the saving decision. Our sample period, for instance, includes the GFC and consequential implementations of unconventional monetary policy by advanced economies, such as quantitative easing and negative interest-rate policies. These unconventional monetary policies were implemented primarily in response to financial instabilities experienced by the US and several euro member countries. However, these policies also created repercussions among emerging market economies through surges of capital flows triggered by 14 The data are available only for OECD countries as well as for 1980, 1985, 1990, 1995, 2000, 2005, and 2009–2014. 15 Aizenman, et al. (2007) estimate that only 10% of the capital stock in developing countries is funded with foreign saving, which means that 90% is self-financed. They also show that countries with higher self-financing ratios grew significantly faster than countries with low self-financing ratios. 16 When we normalize external assets minus liabilities, both from the Lane–Ferretti dataset (2001, 2007, updates), by GDP, we find that the net investment position variable enters the estimation insignificantly for all the samples (not reported). This is not surprising given that the data for financial center countries (e.g., Ireland; Hong Kong, China; Singapore) and heavily indebted countries can be outliers affecting the estimation results. 17 Nabar (2011) and Geerolf and Grjebine (2013) find similar results. 14 ADBI Working Paper 715 Aizenman, Cheung, and Ito extremely low rates of return in advanced economies and now possible retrenchment of such flows due to US monetary contraction, which began in late 2013. Thus, spillovers of the GFC and unconventional monetary policy heightened the level of uncertainty among advanced economies as well as emerging market economies, which may have impacted saving behavior. More specifically, low interest rates may signal future monetary uncertainty or financial condition uncertainty and thereby encourage people toward precautionary saving. Against this backdrop, we examine whether low real or nominal interest rates have any impact on the link between the real interest rate and the private saving rate. The estimation model shown below includes the interaction between the real interest rate and the dummy for a certain threshold of the real or nominal interest rate. In the following regression equation, D takes a value of one when the interest rate of concern is below a certain threshold; that is D = I (interest rate < threshold value), 01 1 2 3 . it it it it it it it it i it it y y r Dr D X Z u v β ββ β ε −′′ = + + ⋅ + + Γ+ Φ+ + + (2) Here, we are interested in examining whether any threshold impact exists regarding the real or nominal interest rates, or both,. Conceptually, it is reasonable to simply focus on the real interest rate as a threshold. However, since the implementation of zero- or negative-interest rate policies, the nominal interest rate has received more general attention. Also, given nominal rigidities that create a money illusion, setting the nominal interest rate at an extremely low level can have more than mere announcement effects. Hence, we investigate whether and how low real and nominal interest rates impact private saving. Table 2 reports the estimation results. The first column of the top of Panel (a) reports only the estimates for the real interest-rate variable ( 1 β ), and for the interaction term ( 2 β ) between the real interest rate and the dummy variable that assumes a value of one when the real interest rate is below –2%. The other estimates are omitted to conserve space. The second column reports the estimates for the real interest rate and its interaction term but the threshold is –1%, with the other columns showing the cases of 0%, 1%, and 2% thresholds, respectively, toward the farthest right.18 The bottom of Panel (a) reports the estimates on the same variables, but the value of the dummy variable is assigned based on the threshold of the nominal interest instead of the real interest rate, taking the values of 0.5%, 1%, 1.5%, 2%, or 2.5%, as seen from the farthest left column to the right. While Panel (a) uses the full sample for the estimation, Panels (b) through (f) report the results for IDC, LDC, EMG, Asia, and Asian EMG, respectively. When the estimate ( 2 β ) is found to be significant, it would mean that the impact of the real interest rate on private saving changes when the real or nominal interest rate is below a certain level. 18 For example, column 1 of the top of Panel (a) shows that the estimate on the real interest rate (0.104) is the response of private saving to the real interest rate when it is above –2%, whereas the response is (0.104–0.048) when the real interest rate is below –2%, although both estimates are statistically insignificant. When the nominal interest rate is used as the threshold, the response would not be different from when the nominal interest rate is above 2.5% because the estimate of the interaction (i.e., 0.045) is statistically insignificant. 15 ADBI Working Paper 715 Aizenman, Cheung, and Ito Table 2: Impacts of Extremely Low Interest Rates (a) Full Sample (1) (2) (3) (4) (5) Threshold: Real Interest Rate –2% –1% 0% 1% 2% β 1: Real interest rate 0.092 0.091 0.096 0.103 0.098 (0.078) (0.078) (0.084) (0.083) (0.084) β 2:Real interest rate x D(real) –0.027 –0.034 –0.041 –0.046 –0.041 (0.078) (0.076) (0.081) (0.079) (0.079) Threshold: Nominal Interest Rate 0.5% 1% 1.5% 2% 2.5% β 1: Real interest rate 0.074 0.077 0.078 0.074 0.079 (0.044)* (0.044)* (0.042)* (0.040)* (0.041)* β 2:Real interest rate x D(nominal) 0.440 0.107 0.054 0.253 0.164 (0.230)* (0.163) (0.149) (0.128)** (0.127) (b) Industrial (IDC) (1) (2) (3) (4) (5) Threshold: Real Interest Rate –2% –1% 0% 1% 2% β 1: Real interest rate 0.084 –0.051 0.071 –0.030 0.115 (0.205) (0.154) (0.201) (0.190) (0.265) β 2:Real interest rate x D(real) 1.392 0.981 0.532 0.390 0.139 (1.082) (0.289)*** (0.349) (0.286) (0.289) Threshold: Nominal Interest Rate 0.5% 1% 1.5% 2% 2.5% β 1: Real interest rate 0.134 0.070 0.073 0.044 0.071 (0.185) (0.225) (0.204) (0.204) (0.208) β 2:Real interest rate x D(nominal) 0.420 0.421 0.400 0.434 0.492 (0.282) (0.234)* (0.221)* (0.231)* (0.237)** (c) Developing (LDC) (1) (2) (3) (4) (5) Threshold: Real Interest Rate –2% –1% 0% 1% 2% β 1: Real interest rate 0.078 0.076 0.076 0.084 0.078 (0.082) (0.081) (0.087) (0.088) (0.089) β 2:Real interest rate x D(real) –0.008 –0.014 –0.018 –0.025 –0.019 (0.081) (0.080) (0.083) (0.082) (0.083) Threshold: Nominal Interest Rate 0.5% 1% 1.5% 2% 2.5% β 1: Real interest rate 0.069 0.073 0.074 0.069 0.074 (0.042) (0.043)* (0.041)* (0.040)* (0.040)* β 2:Real interest rate x D(nominal) 0.503 0.060 0.017 0.205 0.160 (0.294)* (0.204) (0.172) (0.131) (0.123) continued on next page 16 ADBI Working Paper 715 Aizenman, Cheung, and Ito Table 2 continued (d) Emerging (EMG) (1) (2) (3) (4) (5) Threshold: Real Interest Rate –2% –1% 0% 1% 2% β 1: Real interest rate 0.023 0.043 0.046 0.009 –0.041 (0.082) (0.082) (0.087) (0.081) (0.081) β 2:Real interest rate x D(real) 0.017 –0.008 –0.038 –0.013 0.044 (0.097) (0.093) (0.102) (0.091) (0.088) Threshold: Nominal Interest Rate 0.5% 1% 1.5% 2% 2.5% β 1: Real interest rate 0.018 0.022 0.020 0.006 0.014 (0.052) (0.053) (0.053) (0.045) (0.046) β 2:Real interest rate x D(nominal) 0.741 0.236 0.103 0.285 0.250 (0.403)* (0.174) (0.150) (0.120)** (0.130)* (e) Asia (1) (2) (3) (4) (5) Threshold: Real Interest Rate –2% –1% 0% 1% 2% β 1: Real interest rate –0.023 –0.063 –0.050 –0.065 –0.057 (0.101) (0.107) (0.115) (0.111) (0.124) β 2:Real interest rate x D(real) 0.288 0.282 0.238 0.269 0.227 (0.235) (0.237) (0.231) (0.217) (0.213) Threshold: Nominal Interest Rate 0.5% 1% 1.5% 2% 2.5% β 1: Real interest rate 0.064 0.053 0.063 0.066 0.071 (0.038)* (0.037) (0.038)* (0.038)* (0.036)* β 2:Real interest rate x D(nominal) –0.253 0.246 0.154 0.265 0.111 (0.458) (0.342) (0.317) (0.357) (0.285) (f) Asian EMG (1) (2) (3) (4) (5) Threshold: Real Interest Rate –2% –1% 0% 1% 2% β 1: Real interest rate 0.037 0.030 0.021 –0.005 –0.046 (0.091) (0.100) (0.076) (0.079) (0.099) β 2:Real interest rate x D(real) 0.053 –0.090 –0.106 0.034 0.069 (0.247) (0.321) (0.246) (0.156) (0.134) Threshold: Nominal Interest Rate 0.5% 1% 1.5% 2% 2.5% β 1: Real interest rate 0.007 0.031 0.048 0.024 0.034 (0.058) (0.058) (0.053) (0.046) (0.048) β 2: Real interest rate x D(nominal) 0.956 –0.134 –0.297 –0.541 –0.464 (0.247)*** (0.190) (0.159)* (0.123)*** (0.109)*** IDC = industrialized countries; LDC = least developed countries; EMG = emerging market countries. 17 ADBI Working Paper 715 Aizenman, Cheung, and Ito In Panel (a), in the presence of real interest rate regime variables, there is no evidence of significant the real interest rate effect ( 1 β ). However, when we control for low nominal interest rate regimes, the real interest rate effect becomes significantly positive—the estimated substitution effect is in accordance with the full sample result in Table 1. Furthermore, the magnitude of the substitution effect gets much larger when the nominal interest rate is below 2%.19 This result suggests that as far as the full sample is concerned, low nominal interest rates affect the way the real interest rate affects private saving. For the subsample of industrialized countries (Panel (b)), the real interest rate has the substitution effect when the real interest rate is lower than –1% or the nominal interest rate is lower than 2.5%. In fact, when we test the threshold of 3%, the interaction term is still significant, and it becomes insignificant at the 3.5% threshold (not reported). For this group of countries, the substitution effect is dominant but only when the real or nominal interest rate is low. Results in Panel (c) are quite similar to the results of the full sample. According to the panel, when the nominal interest rate is above 0.5%, the real interest rate has the substitution effect on private saving whereas countries with nominal interest rates below 0.5% have much stronger substitution effects. 20 These results indicate that, overall, the positive real interest rate effect is the norm for this group of economies, and the threshold of the nominal interest rate is more relevant than that of the real interest rate. When we look at the EMG group (panel (d)), the real interest rate has the positive effect on private saving when the nominal interest rate is below 2.5%.21 Also, again, the magnitude of the effect is quite large. When we restrict our sample to Asian economies (Panel (e)), we only find that the real interest rate generally has a substitution effect on private saving. The group of Asian EMG economies, however, displays a different pattern of real interest rate effects. In Panel (f), the estimated 2 β is now significantly negative for the threshold of 2.5% while the threshold of 3% is found to be insignificant (not reported).22 That is, when the nominal interest rate is below 2.5%, private saving for Asian EMG would negatively respond to the real interest rate movement. That is, the income effect outweighs the substitution effect.23 19 The dummy for the 0.5% threshold is also significant, but it can be considered as reflecting a “subset” of the dummy for 2.0%. 20 The countries with nominal interest rates below 0.5% in this sample include Panama, The Bahamas, Belize, Trinidad and Tobago, Bahrain, Cyprus, Oman, Qatar, Nepal, Singapore, Algeria, Bulgaria, Czech Republic, Slovak Republic, Estonia, Latvia, Lithuania, Croatia, and Slovenia in years after the GFC. 21 The threshold of 3% is found to be insignificant (not reported). 22 When the nominal interest rate is below 0.5%, the real interest rate effect becomes positive with a large magnitude. In this case, however, the dummy for the nominal interest rate threshold only reflects Singapore from 2011 through 2014. Hence, the positive estimate here is only specific to this country. 23 The countries whose nominal interest rates are below 2.5% in this sample include Republic of Korea, Malaysia, Pakistan, Philippines, Singapore, and Thailand. 18 ADBI Working Paper 715 Aizenman, Cheung, and Ito 4. INTERACTIVE EFFECTS 4.1 Empirical Findings Results in the previous section show that the real interest-rate effect, if significant, tends to be positive; the substitution effect tends to dominate the income effect. The effect varies across different country groups, and its magnitude can be influenced by the level of nominal interest rate. In the case of the Asian EMG group, the real interestrate effect has become negative when the nominal interest rate is lower than 2.5%. Overall, these results suggest that the effect of the real interest rate on private saving can depend on the economic environment at large. In this section, we use interaction variables to explore the real interest-rate effect under alternative economic conditions. For example, when an economy experiences a high level of output volatility, a low interest rate can be interpreted as a sign of economic weakness and thus, can strengthen the saving incentive. Alternatively, for an economy in which old dependency is increasing, a lowering of the interest rate might encourage people to increase their rates of saving to reach pre-determined target levels of retirement saving. In the following, we investigate influences of the economic environment because we suspect that the threshold effect of the nominal interest rate on the real interest rate may be reflecting the economic conditions where it is in. Specifically, we investigate the effect of output volatility, old dependency, healthcare expenditure, financial development, and financial openness on real interest-rate effects. In the estimation, we include the term it it rW⋅ , where it W is the economic environment variable under consideration to examine the interactive effect in the modified saving regression equation: 01 1 2 3 . it it it it it it it i it it y y r rW W X Z u v β ββ β ε − ′′ = + + ⋅ + + Γ+ Φ+ + + (3) Table 3 presents the effect of the real interest rate under alternative output-volatility scenarios.24 The real interest-rate variable has a positive coefficient estimate for the full-country sample and the three subsamples, but it is only statistically significant for the full sample and the subsample of LDC. The output volatility is insignificant in all four samples under consideration. The interaction term between output volatility and real interest rate is positive and statistically significant in the case of the IDC subsample and negative in the other three cases. Most likely, the significant negative effect of the interaction term found in the full sample is driven by the LDC subsample. 24 To ensure a wider variation in the variables, we report results only for the full, IDC, LDC, and EMG samples. Also, because the estimations with the interaction terms between the real interest rate and healthcare expenditure or financial openness turn out to be consistently insignificant, we only discuss the results from the estimations with interaction terms of output volatility, old dependency ratios, and financial development. 19 ADBI Working Paper 715 Aizenman, Cheung, and Ito Table 3: Determinants of Private Saving, Interacting with Output Volatility FULL IDC LDC EMG (1) (2) (3) (4) Private saving (t–1) 0.367 0.264 0.339 0.506 (0.077)*** (0.063)*** (0.085)*** (0.081)*** Public saving –0.466 –0.688 –0.335 –0.625 (0.155)*** (0.125)*** (0.175)* (0.107)*** Credit growth –0.046 –0.019 –0.040 –0.029 (0.016)*** (0.024) (0.017)** (0.016)* Fin. development, HP-filtered –0.041 –0.022 –0.014 0.021 (0.023)* (0.013)* (0.038) (0.047) Income/capita level 0.095 0.200 0.104 0.038 (log, PPP) (0.023)*** (0.039)*** (0.026)*** (0.027) Real interest rate 0.209 –0.373 0.193 0.116 (0.047)*** (0.290) (0.048)*** (0.125) Old dependency –0.158 –0.175 –0.153 –0.241 (0.128) (0.185) (0.170) (0.185) Young dependency 0.105 –0.314 0.145 –0.138 (0.083) (0.220) (0.098) (0.132) Health expenditure –1.397 –0.448 –1.811 –1.677 (% of GDP) (0.432)*** (0.463) (0.471)*** (0.515)*** Financial openness –0.011 0.016 –0.021 –0.006 (0.019) (0.033) (0.020) (0.021) Output volatility 0.021 0.539 0.030 0.269 (0.103) (0.479) (0.113) (0.176) Output volatility x –2.262 21.430 –1.993 –3.012 Real interest rate (0.618)*** (7.094)*** (0.635)*** (3.089) Income/capita growth 0.179 0.274 0.202 0.174 (0.060)*** (0.136)** (0.063)*** (0.081)** N 2,313 431 1,882 755 # of countries 135 23 112 42 Hansen test (p-value) 0.07 1.00 0.60 1.00 AR(1) test (p-value) 0.00 0.02 0.00 0.01 AR(2) test (p-value) 0.46 0.99 0.42 0.83 IDC = industrialized countries; LDC = least developed countries; EMG = emerging market countries. Notes: * p<0.1; ** p<0.05; *** p<0.01. The dependent variable is private saving as a share of GDP. The system GMM estimation method is employed. Although the constant term is estimated, it is omitted from presentation. Results in Table 3 indicate the possibility that, when output volatility increases, the real interest-rate effect can change from positive to negative in the cases of the full-country sample and the LDC sample. For instance, the estimates from the full sample suggest that when the output volatility is less than 9.24%, the marginal real interest-rate effect is positive, and when it is larger than that amount, the marginal effect will be negative.25 25 For the full sample, the estimate of is found to be . Thus, the output volatility threshold of the marginal real interest-rate effect is given . 12 W ββ + 0.209 2.262 it W− 0.209 / 2.262 0.0924 it W<= 20 ADBI Working Paper 715 Aizenman, Cheung, and Ito Figure 6: Interactive Effects – Real Interest Rate and Financial Development We can see that on average, EMG countries have an average level of financial development above the threshold. However, the two other conditions, i.e., output volatility and the old dependency ratio, are below the threshold. This applies to the group of ex-the PRC Asian EMG, and, to a lesser extent, Latin American EMG, and non-EMG LDC. Both the PRC and Hong Kong, China stand out from the EMG group with their high levels of financial development, which contribute to these two countries facing the negative impact of the real interest rate. Furthermore, Hong Kong, China has an average old dependency ratio above the threshold, providing an example in which the real interest rate can have an income effect on an aging-population economy. Figure 8 illustrates the actual real interest rate effects conditional upon output volatility, old dependency, and financial development for the PRC; Hong Kong, China; Republic of Korea; and the group of Asian emerging market economies excluding the PRC. The fourth bar from the left-hand side of the figure (i.e., the light blue bar) shows the real interest rate effects conditional on output volatility, old dependency, and financial development when each of the three economic conditional variables takes the average over the 1995–1999 period, that is, 12 2 1995 99 1995 99 ˆˆ ˆ __ OV OD Output vol Old dep ββ β −− ++ 1995 99 2 ˆFD FD β − + , whereas the first three bars from the left-hand side of the figure show the effects for each of the three disaggregates, namely, 21995 99 ˆ_ OV Output vol β − , 21995 99 ˆ_ OD Old dep β − , and 1995 99 2 ˆ FD FD β − , respectively. The set of four bars on the right-hand side are comparable to the left four bars, except that the economic conditional variables are averaged as of the 2010–2014 period. These bar figures help us grasp how the real interest rate effect has changed over time. As we saw in Figures 2 and 3, the first 5 years represent the period when the real interest rate was relatively high while the last 5 years is the period with very low real interest rates. 27 ADBI Working Paper 715 Aizenman, Cheung, and Ito Figure 7: Triangle Charts LATAM = Latin America; PRC = People’s Republic of China; EMG = emerging market countries; LDC = least developed countries. 28 ADBI Working Paper 715 Aizenman, Cheung, and Ito Figure 8: The Real Interest Rate Effect Conditional on Economic Conditions PRC = People’s Republic of China; EMG = emerging market countries. We can make several interesting observations from the figure. First, for all the three economies and the ex-PRC Asian EMG, the real interest rate effect is negative for both periods. Second, the magnitude of the negative effect increased between the two periods. The extent of increase in the absolute magnitude is especially bigger for the three individual economies. Based on the estimation results reported in Table 6, the short-term real interest rate effect for the PRC conditional upon the three economic condition variables as of 2010–2014 is –0.381, which means the long-term effect is –0.585(=–0.381/(1–0.349)). These figures are higher compared with the short- and long-term effects of the real interest rate as of 1995–1999 that are –0.210 and –0.323, respectively. A 4 percentage point decline in the real interest rate, which is about the same as one standard deviation for the PRC and also the same as the change that occurred between 1995–1999 and 2010–2014, would lead to a 2.3 percentage point increase in the country’s private saving rate. Given that a 2.3 percentage point increase is equivalent to a 0.567 standard deviation increase in the private saving, the effect is economically significant. Third, when we focus on the disaggregated effects of the real interest rate for each of the three conditional variables, the panels in the figure illustrate that the effect of financial development is the largest, followed by old dependency and output volatility, though the interactive effect of old dependency is found to be statistically insignificant (Table 6). Furthermore, the impact of financial development on the real interest rate effect has increased in the last 2 decades because the economies of our concern all experienced further financial development. Hence, it is safe to conclude that the reason 29 ADBI Working Paper 715 Aizenman, Cheung, and Ito why Asian emerging market economies are experiencing weaker substitution effects or stronger income effects in their real interest rates in recent years is mainly because these economies have experienced financial development. As a last issue, let us look at the impact of low real interest rates on private saving for the economies of our interest. Table 3 and Figure 4 show that when the real interest rate is below 1.5%, greater output volatility would lead to higher private saving. Tables 4 and 5 (and Figures 5 and 6) show that the old dependency ratio and financial development can have negative impacts on private saving, but such negative impacts in absolute values tend to become smaller as the real interest rate falls. Thus, under low real interest rates, output volatility tends to increase private saving, and old dependency ratio and the stage of financial development display a reduced negative impact on private saving. Figure 9 illustrates the ratios of private saving in GDP and the real interest rates, but only for selected Asian economies, EMG, non-EMG LDC, and Latin American EMG. The dotted line depicts the threshold of 1.5% for the impact of output volatility for developing countries. Figure 9: Private Saving and the Real Interest Rate for Asia and Others EMG = emerging markets; LATAM = Latin America; H.K. = Hong Kong, China; LDC = least developed countries; Korea = Republic of Korea. In this figure, we can see that Asian developing economies are distributed at lower levels of the interest rate, with all of them, except for Sri Lanka, below the 1.5% threshold. Thus, these economies tend to respond negatively to output volatility and less negatively to shocks to old dependency, thus, to financial development. 30 ADBI Working Paper 715 Aizenman, Cheung, and Ito 5. CONCLUSION In the aftermath of the GFC, unconventional monetary policies, such as quantitative easing and negative interest-rate policies were implemented by advanced economies. While such policies may have contributed to jumpstarting these economies, their implementation also created uncertainty over the future direction of the economies and the financial systems. In particular, the effectiveness of interest rate policies such as zero or negative interest-rate policies have been questioned, along with implications for the financial sector. One frequently asked question is whether an extremely low or negative interest-rate policy would lead to lower or higher consumption or saving. In this paper, we focus on this question and empirically investigate the link between the interest rate and private saving. Our primary focus is whether the interest rate effect is dominated by the income (i.e., negative) or the substitution (i.e., positive) effect. First, our baseline estimations generally affirm the positive effect of the real interest rate on private saving, although its estimate is significant only for the full sample and marginal for the subsample of Asian economies. Given the weakly positive estimates, we suspect that if the interest rate has any impact on private saving, its effect can be masked by uncertain economic environment. Our motive for this investigation is that recent low interest rates may be coupled with greater uncertainty of future monetary or financial conditions and thereby encourage people to engage in precautionary saving when interest rates become very low. When we investigate whether the real interest rate affects private saving differently depending on whether the real, or nominal, interest rate is below a certain threshold, we find some evidence that the impact of the real interest rate on private saving changes when the nominal interest rate is below a relatively low level. This finding may indicate that certain economic environments affect the way interest rate policy is conducted and can impact interest rate effects. Therefore, we examine the impact of the real interest rate conditional upon economic circumstances such as output volatility, old dependency ratio, and financial development. From this investigation, we find that these conditions matter. Extremely high levels of output volatility could make the interest rate effect negative. In economies with high levels of old dependency, the income effect associated with a low interest rate dominates, and a similar observation applies to countries with well-developed financial markets. We also find that the impacts of such economic factors could also be affected by the real interest rate. The impact of output volatility is found to be conditional upon the real interest rate, especially when it is at a low level. That is, when the real interest rate is below 1.5%, greater output volatility would lead to higher private saving in developing countries. Lastly, we find that an old dependency ratio and financial development have negative impacts on private saving, but that negative impacts in absolute values tend to become smaller as the real interest rate falls. Thus, a low-interest rate environment can yield different effects on private saving across country groups under different economic environments. This means that low-interest rate policies adopted by advanced countries to stimulate their economies can yield contractionary effects on developing countries through encouraging saving and reducing consumption. 31 ADBI Working Paper 715 Aizenman, Cheung, and Ito Such findings are relevant to Asian economies. Many of them are characterized by relatively well-developed financial markets. Some of these economies are also experiencing rapidly aging populations. Our empirical findings suggest that these factors are associated with the dominance of the income effect on private saving. It has been documented that advanced economies’ monetary or financial conditions can have spillover effects on emerging market economies (e.g., Aizenman et al. 2016a and 2016b). This means that, in emerging market economies, unconventional monetary policies can guide interest rates to lower levels. Low interest rates could then contribute to higher private saving. All of these findings suggest that an active low-interest rate policy in advanced economies can contribute to keeping global imbalances perennial. 32 ADBI Working Paper 715 Aizenman, Cheung, and Ito REFERENCES Aizenman, J., E. Cavallo, and I. Noy. 2015. Precautionary Strategies and Household Saving. Open Economies Review 26: 911–939. Aizenman, J., M. Chinn, and H. Ito. 2016a. Monetary Policy Spillovers and the Trilemma in the New Normal: Periphery Country Sensitivity to Core Country Conditions. Journal of International Money and Finance 68(November): 298–330. Also available as NBER Working Paper 21128. Aizenman, J., M. Chinn, and H. Ito. 2016b. Balance Sheet Effects on Monetary and Financial Spillovers: The East Asian Crisis Plus 20. NBER Working Paper 22737 (October). Cambridge, MA: National Bureau of Economic Research. Aizenman, J., and I. Noy. 2013. Public and Private Saving and the Long Shadow of Macroeconomic Shocks. NBER Working Paper 19067. Cambridge, MA: National Bureau of Economic Research. Aizenman J., B. Pinto, and A. Radziwill. 2007. Sources for Financing Domestic Capital: Is Foreign Saving a Viable Option for Developing Countries? Journal of International Money and Finance 26(5): 682–702. Arellano, M., and S. Bond. 1991. Some Tests of Specification for Panel Data: Monte Carlo Evidence and an Application to Employment Equations. Review of Economic Studies 58(2): 277–297. Arellano, M., and O. Bover. 1995. Another Look at the Instrumental Variable Estimation of Error-Components Models. Journal of Econometrics 68(1): 29–51. Ayyagari, M., A. Demirgüç–Kunt, and A. Maksimovic. 2010. Formal versus Informal Finance: Evidence from China. Review of Financial Studies 23(8): 3048–3097. Beck, T. 2007. Financing Constraints of SMEs in Developing Countries: Evidence, Determinants and Solutions. Washington, DC: The World Bank. Mimeographed document. ———. 2011. SME Finance: What Do We Know and What Do We Need to Know? Tilburg, The Netherlands: Tilburg University. Mimeographed document. Bernanke, B. 2005. The Global Saving Glut and the U.S. Current Account. Remarks at the Sandridge Lecture, Virginia Association of Economics, Richmond, VA, 10 March. Blanchard, O., and S. Fischer. 1989. Lectures on Macroeconomics. Cambridge, MA: MIT Press. Buera, F., J. Kaboski, and Y. Shin. 2011. Finance and Development: A Tale of Two Sectors. American Economic Review 101(5): 1964–2002. Caballero R., T. Hoshi, and A. Kashyap. 2008. Zombie Lending and Depressed Restructuring in Japan. American Economic Review 98(5): 1943–1977. Chinn, M., B. Eichengreen, and H. Ito. 2014. Oxford Economic Papers 66 (2): 465–490, also available as NBER Working Paper Series 17513 (October 2011). Feldstein, M., and C. Horioka. 1980. Domestic Saving and International Capital Flows. Economic Journal 90(358): 314–329. Friedman, M. 1957. The Permanent Income Hypothesis: A Theory of the Consumption Function. Princeton, NJ: Princeton University Press. ISBN 0-691-04182-2. 33 ADBI Working Paper 715 Aizenman, Cheung, and Ito Geerolf, F., and T. Grjebine. 2013. House Prices Drive Current Accounts: Evidence from Property Tax Variations. CEPII Working Paper 2013-18. Greenspan, A. 2005a. Current Account. At Advancing Enterprise 2005 Conference, London, England, 4 February. ———. 2005b. Mortgage Banking. At American Bankers Association Annual Convention, Palm Desert, California, 26 September. Hansen, L. P., and T. J. Sargent. 2010. Wanting Robustness in Macroeconomics. Handbook of Monetary Economics 3: 1097–1057. Horioka, C. Y., and J. Wan. 2006. The Determinants of Household Saving in China: A Dynamic Panel Analysis of Provincial Data. NBER Working Paper 12723. Cambridge, MA: National Bureau of Economic Research. Huang, Y. 2011. Can the Precautionary Motive Explain the Chinese Business Savings Puzzle? Evidence for the Liquid Assets Perspective. Washington, DC: International Monetary Fund. Mimeographed document. Lane, P., and G. M. Milesi–Ferretti. 2001. The External Wealth of Nations: Measures of Foreign Assets and Liabilities for Industrial and Developing Countries. Journal of International Economics 55: 263–294. ———. 2007. The External Wealth of Nations Mark II: Revised and Extended Estimates of Foreign Assets and Liabilities, 1970–2004. Journal of International Economics 73(2): 223−250. Loayza, N., K. Schmidt–Hebbel, and L. Servén. 2000a. Saving in Developing Countries: An Overview. The World Bank Economic Review 14(3): 393–414. ———. 2000b. What Drives Private Saving across the World. Review of Economics and Statistics 82(2): 165–181. Masson, P. R., T. Bayoumi, and H. Samiei. 1998. International Evidence on the Determinants of Private Saving. The World Bank Economic Review 12. McKinnon, R. 1973. Money and Capital in Economic Development. Washington, DC: Brookings Institution. de Mello, L., P. M. Kongsrud, and R. Price. 2004. Saving Behavior and the Effectiveness of Fiscal Policy. OECD Economics Department Working Paper 397. Paris: Organisation for Economic Co-operation and Development. Midrigan, V., and D. Y. Xu. 2014. Finance and Misallocation: Evidence from Plant-Level Data. American Economic Review 104(2): 422–458. DOI: 10.1257/aer.104.2.422 . Modigliani, F., and R. Brumberg. 1954. Utility Analysis and the Consumption Function. In Post-Keynesian Economics, edited by K. K. Kurihara. New Brunswick, NJ: Rutgers University Press. Mody, A., F. Ohnsorge, and D. Sandri. 2012. Precautionary Saving in the Great Recession. IMF Working Paper 12/42. Washington, DC: International Monetary Fund. Nabar, M. 2011. Targets, Interest Rates, and Household Saving in Urban China. IMF Working Paper. Washington, DC: International Monetary Fund. Obstfeld, M., and K. Rogoff. 1996. Foundations of International Macroeconomics. Cambridge, MA: MIT Press. 34 ADBI Working Paper 715 Aizenman, Cheung, and Ito Röhn, O. 2010. New Evidence on the Private Saving Offset and Ricardian Equivalence. OECD Economics Department Working Paper 762. Paris: OECD Publishing. http://dx.doi.org/10.1787/5kmft7qb5kq3-en Roodman, D. 2006. How to Do xtabond2: An Introduction to ‘Difference’ and ‘System’ GMM in Stata. Center for Global Development Working Paper 103. Washington, DC: Center for Global Development. Skinner, J. 1988. Risky Income, Life-Cycle Consumption, and Precautionary Savings. Journal of Monetary Economics 22(2): 237–255. Zeldes, S. P. 1989. Consumption and Liquidity Constraints: An Empirical Investigation. Journal of Political Economy 97(2): 305–346. 35 ADBI Working Paper 715 Aizenman, Cheung, and Ito APPENDIX 1: SAMPLE COUNTRY LIST Industrialized Countries Australia Austria Belgium Canada Denmark Finland France Germany Greece Iceland Ireland Italy Japan Malta Netherlands New Zealand Norway Portugal Spain Sweden Switzerland United Kingdom United States Developing Countries Albania Algeria Angola Antigua and Barbuda Argentina Armenia Azerbaijan Bahamas, The Bahrain Bangladesh (AE) Barbados Belarus Belize Benin Bolivia Botswana Brazil (LE) Bulgaria Burkina Faso Burundi Cote d'Ivoire Cameroon Central African Republic Chad Chile Colombia Comoros Congo, Dem. Rep. Congo, Rep. Costa Rica Croatia Cyprus Czech Rep. Dominican Rep. Ecuador Egypt El Salvador Estonia Fiji Gabon Gambia, The Georgia Ghana Grenada Guinea–Bissau Hungary India (AE) Indonesia (AE) Israel Jamaica Jordan Kazakhstan Kenya Korea, Rep. of (AE) Kuwait Kyrgyz Republic Lao PDR Latvia Lebanon Lithuania Madagascar Malawi Malaysia (AE) Maldives Mali Mauritius Mexico Moldova Mongolia Morocco Mozambique Myanmar Namibia Nepal Niger Nigeria Oman Pakistan (AE) Panama Paraguay Peru Philippines (AE) Poland PRC (AE) Qatar Romania Russian Federation Rwanda Senegal Seychelles Sierra Leone Singapore (AE) Slovak Rep. Slovenia South Africa Sri Lanka (AE) St. Lucia St. Vincent and the Grenadine Swaziland Tajikistan Tanzania Thailand (AE) Togo Trinidad & Tobago Tunisia (AE) refers to Asian emerging market economies. 36