THÔNG TIN TÀI LIỆU
Author's personal copy Journal of Asian Economics 29 (2013) 80–90 Contents lists available at ScienceDirect Journal of Asian Economics The determinants of Australian household debt: A macro level study Xianming Meng 1, Nam T Hoang *, Mahinda Siriwardana UNE Business School, University of New England, Armidale, NSW 2351, Australia A R T I C L E I N F O A B S T R A C T Article history: Received 13 August 2012 Received in revised form 22 August 2013 Accepted 30 August 2013 Available online 12 September 2013 This paper employs a cointegrated Vector Autoregression (CVAR) model to explore the determinants of Australian household debt The results show that housing prices, GDP and the population in the economy have a positive effect on household borrowing Meanwhile, interest rates, the unemployment rate, the number of new dwellings and inflation are found to have a negative effect on Australian household debt Of these, interest rates are the most significant Based on these results, it is judicious to rein in household debt during economic booms through monitoring and intervening in the assets market and using monetary policy in a timely, comprehensive, and careful manner ß 2013 Elsevier Inc All rights reserved JEL classification: E24 C32 H60 Keywords: Australian household debt Cointegrated VAR modelling Housing market Housing prices Interest rates Introduction The rapid increase in household debt in the last twenty years has been an international phenomenon, which has also occurred in Australia As Table shows, Australia’s accumulated household debt level increased from A$187 billion in 1990 to A$905 billion in 2005 The debt-income ratio jumped from 70.6% in 1990 to 162.8% in 2005, even though the gearing ratio (total liability as percentage of total assets) was only 18.6% in due to a rapid increase in total assets To put this into perspective, the average Australian household would have to work for more than one and a half years just to pay off their debt This accelerated growth of Australian household debt has generated serious concerns The Economist (2003, p 70) reported: ‘The profligacy of American and British households is legendary, but Australians have been even more reckless, pushing their borrowing to around 125 per cent of disposable income .there are now concerns that unsustainable rates of borrowing will sooner or later end in tears.’ * Corresponding author Tel.: +61 6773 2682; fax: +61 6773 3596 E-mail address: nam.hoang@une.edu.au (N.T Hoang) Tel.: +61 6773 2046; fax: +61 6773 3596 Tel.: +61 6773 2501; fax: +61 6773 3596 1049-0078/$ – see front matter ß 2013 Elsevier Inc All rights reserved http://dx.doi.org/10.1016/j.asieco.2013.08.008 Author's personal copy X Meng et al / Journal of Asian Economics 29 (2013) 80–90 81 Table Household liabilities, GDP and disposable income, A$ billion Year Total liabilities Total assets Disposable income Liabilities as % of total assets Liabilities as % of disposable income Jun-90 Jun-95 Jun-00 Jun-01 Jun-02 Jun-03 Jun-04 Jun-05 187 267 473 515 596 685 800 905 1495 1926 2919 3180 3554 3932 4505 4875 265 324 413 451 471 486 518 556 12.5 13.9 16.2 16.2 16.8 17.4 17.8 18.6 70.6 82.4 114.5 114.2 126.5 140.9 154.4 162.8 Source: Reserve Bank of Australia, Table B20 http://www.rba.gov.au/statistics/tables/index.html Australian Bureau of Statistics, cat, no 5204.0 http:// www.abs.gov.au/AusStats/ABS@.nsf/MF/5204.0 Fig Composition of Australian household debt.3 On the other hand, others are more optimistic, due to the low gearing ratio in Australia For example, the ANZ bank (2005) claimed that the financial position of Australian households was within a serviceable range However, this perspective may be over-optimistic because a low gearing ratio does not guarantee the safety of Australian households that have large amounts of housing mortgage debt Fig disaggregates Australian household debt into three categories: owner-occupied housing debt; investment housing debt; and other personal loans What is immediately noticeable is that owner-occupied housing accounts for more than half of total household debt in the period under consideration Since it is not practical to pay back this debt by selling owneroccupied housing (household members must live in this house), the gearing ratio is less relevant than the debt-income ratio when the household repayment burden is considered The other distinguishing feature is that debt associated with investment housing increases dramatically during the period, from less than 5% of disposable income in 1990 to more than 40% in 2008 This increase in investment housing debt is largely driven by rising housing prices since the late 1990s In turn, it is also a significant factor in inflating housing prices This debt breakdown explains the high debt-income ratio and the low gearing ratio in the Australian household sector Large amounts of debt lead to a high debt-income ratio, but the high percentage of mortgage debt secured on housing assets reduces the gearing ratio Since housing assets, as the dominant assets in Australian household balance sheets, have been highly inflated in recent years, a decline in housing price will significantly decrease household asset value, and thus increase the gearing ratio Consequently, the low gearing ratio does not guarantee that Australian household debt is of low risk Thanks to the Global Financial Crisis (GFC), the growth of Australian household debt is now slowing down and appears less of a concern However, in order to avoid an event similar to the United States (US) sub-prime crisis, it is necessary to identify the factors that affect Australian household debt The remainder of the paper is organised as follows Section reviews previous studies on household debt Section discusses the construction of the dataset Section is aimed at testing and estimating the empirical model Section interprets and discusses the main findings from the empirical model Section summarises the main conclusions and provides some policy suggestions Household debt in this graph excludes debt owed by unincorporated enterprises, so the debt-income ratios are lower than those shown in Table Author's personal copy 82 X Meng et al / Journal of Asian Economics 29 (2013) 80–90 Previous studies The rapid rise of household debt is a relatively new phenomenon, and thus it is rare to find studies on household debt before the 1990s However, booming household debt of the 1990s triggered considerable academic interest, and consequently a substantial amount of literature has been published As the purpose of this paper is to explore the determinants of Australian household debt, only papers concerning the factors affecting household debt are reviewed here A number of empirical studies on the effects of household debt have been completed for countries other than Australia In South Africa, Aron and Muellbauer (2000) utilised adjusted South African Reserve Bank data to estimate the impact of financial liberalisation on household consumption and household debt They concluded that financial liberalisation and fluctuation in asset values have important implications for consumer spending and increasing household debt in South Africa In the US, Barnes and Young (2003) employed a calibrated partial equilibrium overlapping generation (OLG) model to explain household debt in terms of consumption-income and housing-finance motivations They found that the substantial rise of household debt in the 1990s could be explained by real interest rate, income growth expectations, demographic changes, and the removal of credit constraint Tudela and Young (2005) also used the OLG model to analyse household debt in the UK and claims that changes in interest rates, house prices, preferences, and retirement income affect household debt Jacobsen (2004) employed a flexible dynamic model and the Norwegian quarterly data from 1994 Q1 to 2004 Q1 to estimate the effects of various factors on household debt, Many factors were found to influence household debt, including housing stock, interest rates, the number of house sales, wage income, housing prices, unemployment rate, and the number of students Martins and Villanueva (2003) constructed a data set combining household survey data and administrative record of debt to estimate the responsiveness of long-term household debt to the interest rate change in Portugal They found that the elasticity of the probability of mortgage borrowing to a change in the interest rate was large and negative Magri (2002) analysed the determinants of Italian household participation in the debt market, using data from the Bank of Italy’s Survey of Household Income and Wealth The results suggested that age, income, living area, and the enforcement cost of banks, have important effects on household debt Kearns (2003) employed household-level data to explore why households fell into mortgage arrears during the 1990s in Ireland His study suggested that a modest rise of interest rates would result in a substantial repayment burden for a significant number of newly mortgaged households He concluded that the continuing strong growth of mortgage lending, caused by relaxed lending criteria and by household willingness to accept higher repayment burdens, could lead to a higher rate of mortgage arrears among households Thaicharoen, Ariyapruchya, and Chucherd (2004) claimed that low interest rates, demographics, and declining borrowing constraints, contributed to debt in Thai households They suggested that current debt levels in Thailand did not pose a threat to financial stability and the macro-economy Some studies have integrated data from multiple countries Debelle (2004) analysed the possible determinants and the macroeconomic implications of rising household debt He argued that the rise of household debt reflected the response of households to lower interest rates and an easing of liquidity constraints Increased household debt itself was not likely to be the source of negative shocks to the economy, however he suggested it could amplify shocks from other sources Crook (2003) compared studies of the effects of household debt across a number of countries He found that the debt holding by age followed the life cycle pattern in all countries observed, that there were considerable variations in the determinants of desired levels of debt, and that there was intra- and inter-national variation in the marginal effects of household debt In Australia, prompted by rising household debt levels, some institutions have included household debt information in their surveys, or initiated surveys on this topic The Australian Bureau of Statistics (ABS) has conducted several wave surveys on Household Expenditure (HES) Melbourne University commenced the Household, Income and Labour Dynamics in Australia (HILDA) Survey in 2001, funded by the Federal government, and had completed waves at the time of writing The Reserve Bank of Australia (RBA) completed a survey in 2006 on Household Behaviour Around Housing Equity A number of studies at the household sector level have been based on these surveys, and provide information at the micro level Bray (2001) studied financial stress in Australia using the 1998–99 HES survey La Cava and Simon (2003) employed a logit model using the data from the HES and HILDA surveys to study the factors affecting the financial constraints on Australian households Schwartz, Hampton, Lewis, and Norman (2006) used bivariate and logit models to analyse the data obtained by RBA surveys on Household Behaviour Around Housing Equity At the macro level, the RBA and other institutions have published a number of papers and presentations on household debt Stevens (1997) emphasised the positive effect of low inflation on household borrowings The RBA (1999) attributed the rapid growth of personal credit to: innovations in products offered by banks; the increasing household preference towards the use of credit cards; and continuing economic expansion with low inflation and low interest rates The RBA (2003) also illustrated the composition and distribution of household debt, suggesting that low interest rates, low inflation rates, and financial deregulation may have contributed to a rise in household debt A number of studies support this view The Treasury (2005) suggested that the increase in household debt partly reflected increased house prices due to sustained low inflation and interest rates, and partly reflected the improved product choice and reductions in borrowing costs caused by deregulation of the financial sector in the 1980s and 1990s The ANZ bank (2005) claimed that rising household debt level was caused by a sustained boom in house prices, and that the sustainability of household credit depended on the growth of household disposable income and employment Author's personal copy X Meng et al / Journal of Asian Economics 29 (2013) 80–90 83 All of these studies on Australian household debt are instructive However, they present only financial analyses of Australian household debt To provide a more comprehensive picture of Australian household debt, this study identifies the determinants of Australian household debt by employing empirical models The models use data from household accounts, microeconomic data from surveys, and macroeconomic data Dataset The household debt level is jointly determined by supply and demand That is, the availability of funding, and the household’s decision to take on debt However, the macroeconomic environment ultimately determines both supply and demand Consequently, the determinants of household debt must lie in these macroeconomic factors By analysing factors affecting borrowing and/or lending, we can therefore identify potential determinants of household debt With regard to demand, household desire for borrowing is subject to the level of household disposable income as well as the purpose/s for borrowing Household disposable income is comprised of household gross income plus social transfer, less income tax payable and other outlays We focus here on household gross income, since income tax rates in Australia have undergone minimal change in recent decades, while social transfer and other outlays are small relative to household gross income Household gross income includes wage income and gross mixed income, as well as domestic and overseas investment income At the macro level, these factors can be approximated by Gross Domestic Product (GDP) The purposes of household borrowing include smoothing consumption and investing The level of consumption is closely related to the Australian population and the price level (Consumer Price Index – CPI) It may also be related also to the macros affecting consumer confidence, such as unemployment rate and GDP Investment decisions are typically related to interest rates Moreover, from the disaggregation of Australian household debt, we found housing to be the main investment vehicle for Australian households In considering the large amount of housing debt, housing prices and the number of new houses entering the market are important factors With regard to supply, the availability of funding and ease of obtaining finance are largely indicated by interest rates However, to reduce credit risk, lenders may take into account household income level and other macroeconomic variables, including unemployment rate, inflation rate and GDP Among these factors, the household income level can be approximated by GDP and the inflation rate is a monotonic transformation of CPI Including all possible variables affecting Australian household debt yields the following dataset: X ẳ DEBT; GDP; NDWELL; HPI; R; U; CPI; POPị (1) where DEBT – accumulated household debt; GDP – gross domestic product; NDWELL – number of new dwellings approved; HPI – housing price index; R – interest rate; U – unemployment rate; CPI – consumer price index; POP – population It may be argued that real per capita data are preferable because they can reduce heteroscedasticity in the model Nonetheless, this study uses nominal values for a number of reasons First, the White tests show the heteroscedasticity problem in the model is not serious when nominal values are used Second, measurement errors are often a problem in macro time series, and transforming nominal value to real and/or per capita value (divided by CPI or POP) may magnify these errors Finally, some international studies show that population and inflation have a significant influence on household debt Thus, a nominal value is required to test if this holds true for Australia as well The quarterly time series data from 1988Q2 to 2011Q2 for the variables in the information set were collected from the Australian Bureau of Statistics (ABS, 2011), the Reserve Bank of Australia (RBA, 2011) and other institutions The majority of data were obtained from the ABS, including population, unemployment rates, housing prices, GDP, and the CPI The data for new dwellings commencements came from the Housing Industry Association (HIA, 2011) Data for household debt and official interest rates were provided by the RBA Due to different sources and different measurement of data, some data sequences were adjusted before use Specifically, the dataset in this study is described as follows: Household debt (DEBT): seasonally adjusted quarterly data, measured in billion Australian dollars (A$ billion), at the end of quarter GDP: seasonally adjusted quarterly data, measured in billion Australian dollars (A$ billion) Consumer price index (CPI): base year 1989/90 = 100 Housing price index (HPI): CPI on housing 1989/90 = 100 Interest rate (R): official interest rate, quarterly averaged monthly data Unemployment rate (U): quarterly averaged monthly data Population (POP): measured in thousand persons, ABS estimated quarterly data Since the data were available only from June of 1989 onwards, the annual population in 1988 provided by ABS has been adopted as the data in June of 1988 The data from September 1988 to March 1989 have been calculated assuming that population growth rate was stable in this period Number of New dwelling approvals (NDWELL): measured in thousand dwellings, including all types of housing such as units and houses Author's personal copy X Meng et al / Journal of Asian Economics 29 (2013) 80–90 84 The empirical model Since macroeconomic variables are notorious for non-stationarity, cointegrated VAR analysis is used to study the relationship between Australian household debt and other macroeconomic factors The likelihood estimation of a cointegrated VAR model for time series data integrated of order one I(1) was developed by Johansen (1988) and Johansen and Juselius (1990) The cointegrated VAR(k) model of an I(1) time series of p-dimension, Xt, kÀ1 X G i DX ti ỵ et ; t ¼ 1; ; T a; b are P  r matrices can be expressed as follows:(2)DX t ẳ ab X t1 ỵ iẳ1 By the Granger representation theorem, the moving average (MA) form of the VECM model (2) is given by: t X ei ỵ C Lịet ỵ X Xt ẳ C (3) iẳ1 where C ẳ b ? a0? G b ? ị G ¼IÀ kÀ1 X À1 a0? Gi i¼1 C*(L) is an infinite order polynomial lag operator which results C à ðLÞet is the stationary part of the process a ? ; b ? are orthogonal complement matrices of a; b which satisfies a0 a ? ¼ 0; b b ? ¼ X contains the initial values of the process X t Premultiplying (2) with a non-singular p  p matrix B, we have the structural model with contemporaneous effects: kÀ1 X (4) BDX t ¼ Bab X t1 ỵ B G i DX ti ỵ ut ; t ¼ 1; ; T i¼1 ut ¼ Bet , which is the structural shock The MA representation of (4) can be obtained by substituting et ¼ BÀ1 ut in (3): X t ¼ CBÀ1 t X ui ỵ C LịB1 ut ỵ X (5) i¼1 À1 Matrix C ¼ b ? ða0? G b ? Þ a0? has rank (p À r), implying that matrix C˜ ¼ CBÀ1 has, at most, r columns of zero To recover the structural shocks, we have to decide the matrix B so that the residuals et can be transformed into permanent and transitory shocks ut ¼ ðuPt ; uTt ị by the transformation ut ẳ Bet where ut $ INð0; IÞ We follow the approach by Jueslius (2006) and Dennis, Hansen, Johansen, and Juselius (2005) to define the matrix À1 B ¼ HÀ1 KG; G ¼ a V À1 a? where V ¼ Eðe e H is the Cholesky decomposition of ðKGVG0 K Þ, and K is a ð p  pÞ block diagonal matrix to impose à restrictions on matrix ut ¼ Bet and matrix C0 ẳ C 0ịB1 , which are needed for identification of the structural model The restriction matrix K is constructed so that it imposes ð1=2Þð p À rÞð p À r À 1Þ zero restrictions in the last (p À r) non-zero à columns of matrix C˜ t also imposes a ð1=2Þrðr À 1Þ dimensional triangular matrix of zeros in the first r columns of matrix C˜0 In the above model, b X tÀ1 are the cointegration vectors, including all variables in the dataset, namely, DEBT, GDP, NDWELL, HPI, R, U, CPI and POP a is the adjustment coefficients matrix DX t is the first difference of X Gi is the coefficient vector associated with DXtÀi Since Gi represents the dynamic effects (how lagged change can affect present change), it can take any sign depending on how the variables will behave in the short run However, the cointegration vector XtÀ1 indicates the long run relationship, so the sign of coefficients on them should be consistent with economic theory In Table we list the expected signs of elements in the cointegration vector based on economic reasoning GDP should have a positive sign for two reasons First, GDP indicates the size of the economy and aggregate household income and thus the size of debt, given the unchanged debt-income ratio Second, rapid GDP growth indicates good economic times which will encourage both borrowing and lending The impact of NDWELL (the number of new dwellings) on household debt is ambiguous More new dwellings means more housing assets, if the ratio of mortgage debt to housing assets is unchanged, and implies more housing debt (a major part of household debt) However, increased housing supply may lead to lower housing prices and thus less housing debt The effect of HPI should be positive because of its direct positive impact on housing debt The impact of interest rates (R) depends on the perspectives of different agents in the household debt market In the view of borrowers, a higher interest rate means higher borrowing cost and thus less borrowing In the t t Þ; Table Expected signs of coefficients (DEBT as a dependent variable) Variables GDP NDWELL HPI R U CPI POP Expected signs of coefficient + Ỉ + À Ỉ Ỉ + Source: authors’ projection Author's personal copy X Meng et al / Journal of Asian Economics 29 (2013) 80–90 85 Table The results of cointegration tests with structural changes No of CE(s) Max-Eigen statistic Critical value (0.05) No of CE(s) Trace statistic Critical value (0.05) Critical value with breaks (0.05) Adjusted critical value with breaks (0.05) Nonec At most At most At most At most At most At most At most At most At most 88.82414 86.08198 77.40195 63.64656 51.79813 29.37749 23.22869 19.69299 15.18727 4.820856 68.81206 62.75215 56.70519 50.59985 44.49720 38.33101 32.11832 25.82321 19.38704 12.51798 Nonec,a,b At most 1c,a At most 2c,a At most 3c,a At most 4c At most 5c At most At most At most At most 460.0601 371.2359 285.1539 207.7520 144.1054 92.30730 62.92981 39.70112 20.00813 4.820856 273.1889 228.2979 187.4701 150.5585 117.7082 88.80380 63.87610 42.91525 25.87211 12.51798 357.1008 301.2805 251.3317 206.1153 165.1130 128.2082 95.4744 66.9586 42.4707 21.9103 459.1296 387.3606 323.1408 265.0054 212.2881 164.8391 122.7528 86.08963 54.60519 28.17039 c 2c 3c 4c Source: authors’ tests Note: lags are chosen to minimise Schwartz criterion Minimising AIC will lead to more lags and more cointegration vectors when standard critical values are used But when the adjusted critical value is used, the test also indicates only one cointegration vector a Denotes rejection of the hypothesis at the 0.05 level according to the critical value with breaks b Denotes rejection of the hypothesis at the 0.05 level based on the adjusted critical value with breaks c Denotes rejection of the hypothesis at the 0.05 level according to the standard critical value view of the lenders, a higher interest rate means higher return and thus will stimulate lending However, the dominant factor on the lending side is credit risk, so the impact of interest rates on supply side is expected to be small (the unemployed capital has to find an outlay even if interest rates are low) and thus the overall effect should be positive The impact of unemployment (U) and inflation (indicated by CPI) can be either positive or negative A higher unemployment rate may indicate higher demand for financial assistance (e.g borrowing) But, on the other hand, it creates additional financial constraint for households An unemployed person is less likely to repay their debt in time, and thus is more unlikely to obtain a loan High inflation can stimulate debt-financed purchase of physical assets (e.g housing) but it will discourage lending because of diminishing (or even negative) real returns The impact of population (POP) should be positive because it is positively related to the number of households with debt, other things being equal Since most macroeconomic time series data are non-stationary, unit root tests have to be performed Before formal testing procedures are undertaken, the time series data are plotted to allow for visual inspection Most time series show apparent trends (except NDWELL) The Dickey–Fuller test with GLS detrending (DF-GLS) is performed (the results of these tests being available upon request) The tests suggest first order integration for all variables at a 5% level of significance Perron (1989) argued that structural change may be mistaken for a unit root It appears that there were structural changes for GDP at 2008Q4, R at 1993Q1 and U at 1993Q1 The procedure developed by Perron (1989) and extended by Zivot and Andrews (1992) was used to test for unit root with structural change in each of these series The test results in Appendix 1A shows that existence of structural change does not alter the DF-GLS tests results The t-values for a1 are much higher than the values given by Zivot and Andrews (1992) at 0.05 level Thus, we conclude that all time series in the model are I(1) Since we are working with non-stationary data, the co-integration test must be performed Both Trace and Maxeigenvalue tests are used In implementing the tests, a deterministic linear trend and intercept are included in the cointegration equation The structural changes in the model were confirmed using the recursive test for stable cointegration vector b developed by Hansen and Johansen (1999) The forward and backward recursive tests indicate that there are breaks at 1993Q1 and 2008Q4, so the trend changes at 1993Q1 and 2008Q4 are also included in the cointegration test through two dummy variables DM93Q1 and DM08Q4 The trace test suggests cointegration equations, while the max-eigenvalue test suggests only cointegration equations (see Table 3) However, the presence of intervention dummies (DM93Q1 and DM08Q4) may affect the distribution of cointegration tests (Johansen, Mosconi, & Nielsen, 2000 and Joyeux, 2001) Using the method and the simulation results provided by Johansen et al (2000), the authors calculate the critical values with breaks, shown in the second last column in Table According to these critical values, there are only cointegration relationships in the model Moreover, based on the studies by Reimers (1992) and Cheung and Lai (1993), the small sample size in this study may bias the tests – the asymptotic property of the test is not applicable Following the suggestion by Cheung and Lai (1993), the authors use T/(T À nk) (where T is the sample size, n is the number of variables in the model and k is the number of lags) to scale up the critical value with break so as to obtain the adjusted critical values with breaks This is shown in the last column of Table According to these adjusted critical values with breaks, there is only one cointegrating relationship at the 0.05 level of significance Next, the vector error correction model (VECM) is estimated A constant and a general deterministic trend are included in the cointegration equation Two lags are chosen to minimise Schwartz criterion (SC) The estimated cointegration relationship and adjustment coefficients are shown in Table The first panel of Table shows the estimated cointegration vector The estimated results suggest that, with exception of CPI and DM93Q1, all variables have a significant influence on household debt in the long run The second panel shows the adjustment coefficients from the VECM estimation results The adjustment coefficient for DEBT has the correct sign (negative) and is significant, which means that the error correction is quite effective Author's personal copy X Meng et al / Journal of Asian Economics 29 (2013) 80–90 86 Table The results of VECM estimation and diagnostic tests.a Variables DEBT GDP R POP U CPI HPI NDWELL 93Q1*t 08Q4*t Coefficient S.E t-Value 1.000 À13.4760 (0.0039) [À3.4318] 257.4072 (41.6646) [6.1781] À1.4928 (0.4491) [À3.3236] 119.4678 (31.4057) [3.8040] 24.9666 (18.2657) [1.3669] À40.1032 (9.8366) [À4.0769] 24.2793 (3.2325) [7.5111] À3.8336 (3.5016) [À1.0948] 16.5573 (2.4254) [6.8265] Adjustment coefficient S.E t-Value À0.0348 4.6367 0.0007 À6.51EÀ05 À0.00025 0.00082 0.00017 À0.0072 0.0018 À0.0197 (0.0071) [À4.9011] (1.7230) [2.6910] (0.00038) [1.8392] (1.3EÀ05) [À4.9057] (0.00013) [À1.8985] (0.00068) [1.1985] (0.0013) [0.1348] (0.0026) [À2.7382] (0.0021) [0.8519] (0.0065) [À3.0527] Diagnostic tests: R-squared: 0.6820, adj R-squared: 0.6377, Schwartz (for ECM): 7.2343, Schwartz (for VECM): 42.9581 VEC residual serial correlation LM tests LM-Stat = 104.5274 10th order J.B.(2) = 2.3630 (ECM equation for DEBT) Jarque–Bera residual test for normality: J.B.(20) = 22.5482 (VECM system) White heteroscedasticity test: Chi-squared(1210) = 736.416 p = 0.3585 p = 0.3068 p = 0.0673 p = 0.0962 Source: authors’ estimation and tests a There is a general trend and intercept in the model, but they are not displayed in the table 2-lag is chosen to minimise Schwartz criterion Table Structural-contemporaneous long-run impact matrix: C˜ ¼ CBÀ1 uT1;t DEBT NDWELL HPI CPI U POP R GDP 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 uP1;t 5.243 À0.002 0.278 0.279 0.006 19.527 À0.019 1.000 uP2;t À6.296 À0.967 1.473 À0.980 À0.584 À82.637 1.000 0.000 uP3;t 0.206 À0.050 0.049 0.021 0.003 1.000 0.000 0.000 uP4;t 2.701 0.100 2.152 0.624 1.000 0.000 0.000 0.000 uP5;t À1.775 À1.546 0.416 1.000 0.000 0.000 0.000 0.000 uP6;t 13.081 4.633 1.000 0.000 0.000 0.000 0.000 0.000 uP7;t À5.980 1.000 0.000 0.000 0.000 0.000 0.000 0.000 Source: authors’ estimation The estimation passes all diagnostic tests shown in the last panel of Table The ECM equation for DEBT has a reasonable R-squared (0.68) and adjusted R-squared (0.64), indicating that the equation explains the behaviour of change in household debt in the short run well The LM tests up to 10th lag could not find autocorrelation at a significant level: the large p-value indicates that we could not reject the null hypothesis of no serial correlation The Jarque–Bera test confirms the normal distribution of residuals for displayed ECM with a p-value of 0.3068: even for the VECM system, the J.B test could not reject the null of normality at conventional level The White test shows no serious heteroscedasticity problem in the model Finally, we can estimate the structural-contemporaneous long-run impact from the structural model Since there is only one cointegration vector in the model, one transitory shock, and seven permanent shocks The response of household debt is the main concern of this study, and economic reasoning shows that all other variables should have an influence on household debt We therefore view the shock of DEBT as transitory and the other shocks as permanent The order of variables is important in imposing restriction, through which the structural model can be identified The order of other variables is based on the macroeconomic influence of each variable and the purpose of this study For instance, DEBT is included at the end of the list because it is influenced by any other variables On the other hand, GDP has profound impacts on many aspects of the economy, so is positioned first The order of variables other than DEBT is arranged as presented in Table The permanent and transitory decomposition of structural shocks is listed in Table in the next section, and the rotation matrix is listed in Appendix 2A Analysis of estimation results According to the first panel of Table (the sign on coefficients in the table should be read opposite, since the dependent variable DEBT is on the same side of other variables in the cointegration vector), factors negatively affecting household debt are interest rate, unemployment rate, the number of new dwelling approvals, and CPI Those acting positively on household debt include housing prices, GDP, and population In addition, both dummies with trends are significant We discuss these three categories of factors in turn 5.1 Factors negatively affecting household debt First, the model shows that the interest rate has a dramatically negative effect on household debt The interval estimates show that if the interest rate increases by one base point (0.01%), household debt level would decrease by A$2.57 billion over Author's personal copy X Meng et al / Journal of Asian Economics 29 (2013) 80–90 87 time The direct reason for this is that interest rate rises will increase the borrowing cost, which in turn will deter households from borrowing, or at least reduce the amount they are inclined to borrow Moreover, for households that have already incurred debt, the repayment burden will increase if the debt is based on a variable interest rate, as is the case for most Australian housing loans If the repayment burden is unbearable, some households are required to sell their property to pay off their debt As a result, household debt will decrease Increase in interest rates also indirectly affects household debt by discouraging investment, and the reduction in investment will slow down the whole economy Scaling back the economy may reduce household income, increase the unemployment rate and thus reduce household borrowing If households’ newly incurred debt is less than the amount of their scheduled repayments, the household debt level will also decrease From the results we can infer that rising household debt in Australia must be associated with a decrease in interest rates This is exactly the case The rising household debt in Australia has coincided with the low interest rates after financial deregulation During and shortly after financial deregulation in the 1980s, the Australian interest rates were generally more than 10% For example, in the early 1990s, the official interest rate reached 17%, the rate for lending to business was 20.50%, and the rate for personal loans was 23.50% Thanks to the delayed effect of financial deregulation, easy credit at a global level, and the RBA’s adoption of a targeting inflation policy, the official interest rate decreased to about 6% in the 1990s and below 5% in the early 2000s Given the dramatic negative impact of interest rates shown in Table 4, low interest rates in the most recent two decades must have a remarkable role in the rise of household debt Second, another important factor affecting household debt negatively is the unemployment rate This is due to the negative effect of unemployment on household income Generally speaking, a high unemployment rate means there is less income for all households and thus a greater desire for loans to finance consumption From this point of view, high unemployment rates will lead increased household debt However, lower income due to unemployment casts doubt on the future income, with two implications The first implication is that households without regular employment will be discouraged from borrowing because of concerns about the ability to repay the loan, while the second implication is that unemployment increases the possibility of financial constraints These two factors actually limit household demand for financing and depress the growth of household debt More importantly, the rising unemployment rate indicates a deteriorated economic situation, in which investors are very cautious to lend The estimation results show that the negative effect dominates This result is quite reasonable considering the dominance of the supply side in the Australian household debt market For various reasons the demand for credit is strong in Australia Amongst other things, these reasons include the Australian home ownership dream, rising housing prices, high rental prices, and housing grants provided by the Australian government However, this demand is regulated by the supply side, which is primarily determined by credit risk An apparent situation related to rising unemployment is that, when the lender perceives that the borrower is unlikely to pay back the loan, the high chance of making a loss prohibits him from providing credit, and thus the supply of credit will shrink dramatically This is what happened during the GFC It is noticeable from the estimated results that the unemployment rate is less influential than the interest rate This finding is consistent with previous studies For example, Debelle (2004, p 57) argues that the unemployment rate is less severe than interest rate because ‘unemployment generally affects only a relatively small section of population, and the degree of overlap between those households with a higher risk of unemployment and those with high debt level has historically been low’ Third, the number of new dwelling approvals also has a significant influence The effects of new dwellings are two-fold On one hand, new dwellings increase the total housing assets in the market Given the high demand for housing assets and the popularity of housing mortgage loans in Australia, this implies more housing debt for households On the other hand, new dwellings entering the market means greater housing supply If the housing demand is unchanged, housing prices will drop Decreased housing prices will reduce the market value of housing assets and thus reduce the amount of housing loans The estimated results show that the latter effect dominates the former Finally, CPI shows an insignificant negative effect on household debt Inflation (the percentage increase in the CPI) has different effects on borrowing and lending With regard to borrowing, inflation will devalue the debt, providing a strong stimulus for households to borrow However, on the supply side, inflation will erode the principal and discourage lending The insignificant negative effect of CPI indicates that the supply side dominates That is, in the face of high inflation, fewer funds are lent, and household debt will decrease The dominant effect of the supply side in the household debt market concurs with our earlier analysis of the effect of unemployment rate Moreover, this finding is also consistent with previous research For example, many studies (RBA, 2003; Debelle, 2004; and The Treasury, 2005) suggest that low inflation could be a reason for rising household debt because it may decrease the financial constraints on households (lower inflation leads to lower interest rates and thus less income is needed for the reduced scheduled payment) and encourages lending (lower inflation erodes principal more slowly) 5.2 Factors positively affecting household debt With regard to the positive factors, the empirical model shows that housing prices have a substantial influence on household debt, while both GDP and the population also have a very significant effect First, the significant positive effect of the housing price index can be understood easily, by considering the importance of housing prices in housing assets and the importance of housing assets in Australian household investment portfolios Increasing housing prices will scale up housing asset value From the demand perspective, this has two implications For new Author's personal copy 88 X Meng et al / Journal of Asian Economics 29 (2013) 80–90 home buyers, it means they have to take on substantially more debt to buy housing, other things being equal For those who have already taken housing loans, the increased housing assets provide them with a good opportunity to withdraw housing equity, to obtain more loans against the increased value of their housing assets From the supply perspective, rising housing assets prices imply good economic times This tends to indicate low credit risk and thus stimulates lending As a result, household debt will increase along with the housing prices There are a number of factors contributing to rising housing prices One is financial deregulation and targeted inflation policy, which provide an environment in favour of housing demand Australia’s financial deregulation abolished a large number of restrictions on lending and borrowing, and thus facilitated debt financed housing purchases However to some degree, oversimplified regulation of the mortgage market and/or no regulation on some markets (such as the markets of financial brokers and complex financial products) pushes housing demand to an unsustainable level Meanwhile, interest rates are low due to changes in the RBA’s interest rate policy, targeting the inflation rate Since housing price has very little weighting in the CPI basket, the low inflation of daily consumption items means low official interest rates in Australia even where housing prices skyrocket The low borrowing cost and simplified borrowing procedure facilitate and encourage housing demand Another factor contributing to rising housing prices is the home ownership incentives provided by the Federal government To ease the financial burden on households after the introduction of the GST in 1999, the government offered various associated policies Some of these policies were intended to reduce the cost of home ownership For example, first home buyers were entitled to a government grant of A$7000 A number of other compensation policies were enacted For the owner-occupied property, the imputed rental income and capital gains on sales were exempt of tax For investment housing, expenses such as interest payment and depreciation are tax deductible Finally, capital gains on selling investment housing were taxed at half the taxpayers’ marginal tax rate All these policies tended to stimulate housing demand The state governments’ tight control of land use is also an important contributor of rapidly rising housing prices because it put a rigid constraint on housing supply Due to this tight control, the number of new dwelling approvals has barely changed in recent decades, fluctuating at around 38,000 approvals per quarter in the period 1988–2012 In other words, the housing supply in Australia is very inelastic Given the tremendously increased demand and a very inelastic housing supply, rapid increases in housing price and thus in household debt are logical consequences Second, the positive influence of GDP on household debt may arise in two ways On the one hand, the magnitude of GDP indicates the size of the economy and thus the capacity of household borrowing and lending A higher GDP implies higher income for households and higher gross operating surplus for firms With higher income, households would be less creditconstrained On the other hand, the higher income and profit provide richer funding sources for banks The other channel may come from household confidence The growth rate of GDP is a popular indicator of economic development Robust growth of GDP makes people more confident so that they feel safe to borrow and lend With the increased demand (willingness and ability to borrow) and increased supply (willingness to lend), household debt may grow in line with GDP Last, the population in the economy also has a significant positive affect on household debt The reasoning behind the estimation results is that the growth of population is likely to increase the number of households with debt and thus the total household debt level This result concurs with previous studies (e.g Crook, 2003; Thaicharoen et al., 2004; and Tudela & Young, 2005) 5.3 Effects of structural changes The structural change in 1993 coincided with the policy change of the RBA That is, following the gradual deregulation of the financial system in the 1980s, the inflation targeting policy was introduced in 1993 The insignificant positive estimation results indicate that financial deregulation and RBA monetary policy change may have a mild effect of increasing Australian household debt Practically, this mild positive effect can be explained by the easier access to finance thanks to financial deregulation and more confidence in the new interest rate regime The markedly significant structural change in 2008 is apparently influenced by the GFC The estimation results show that the trend of household debt decreases remarkably after 2008Q4 The negative effect of this structural change is consistent with the negative effect of the GFC on the economy As the GFC hit, confidence in both consumption and investment declined sharply Facing an uncertain economic future, households tend to save more and borrow less, so it is reasonable to see household debt growing at a slower pace (although even as the quarterly household borrowing decreases, the debt level keeps rising as it is an accumulated value) However, while the negative effect of the GFC on household debt has been undoubted, caution must be taken in the interpretation of the estimated coefficient of T*DM08Q4 since we only have two and a half years post-GFC data As the time series extends in the future, the degree of significance of T*DM08Q4 may change 5.4 Responses to structural shocks The responses to structural shocks are shown in Table For the purpose of this study, we discuss the results with an emphasis on the response of household debt to shocks on other variables (the first row in Table 5) Since the shock of DEBT (uT1;t in Table 5) is identified as the transitory shock, its long run impact is negligible because the first column of the long-run impact matrix is zero The first permanent shock, uP1;t , is identified as productivity shock or GDP shock The second permanent shock, uP2;t as the part of the disturbances in interest rate which is not explained by uP1;t The third permanent shock, uP3;t can be identified as the part of disturbances in population which is not explained by uP1;t and uP2;t , Author's personal copy X Meng et al / Journal of Asian Economics 29 (2013) 80–90 89 and so forth The response of DEBT to a shock on housing prices is tremendously positive, that is an A$13.08 billion increase in household debt in the long-run in response to a unit increase in housing price index It has also a substantial positive response to a shock on GDP Positive responses in the long-run are also found to shocks on unemployment rate and on population, though they are much milder The responses to shocks on other variables are negative The interest rates have the largest negative influence in the long-run on household debt, followed by new dwelling approvals and CPI The large responses of household debt to shocks on variables like GDP, HPI, R and NDWELL have important policy implications The positive response to an increase in GDP indicates that an increase in household debt is a natural thing In other words, one should not worry too much about a moderate increase in household debt in good economic times The notable influence of housing prices on household debt shows the importance of the housing market on household debt To reduce the pace of rising household debt, the government should intervene in the household market to reduce housing prices One way to intervene is to reduce housing demand by suspending home ownership incentives, while another way is to increase housing supply by easing restrictions on land use The substantial negative influence of interest rates suggests that monetary policy is a useful tool to contain household debt An increase in the official interest rate (or tightening the money supply) can reduce the amount of household borrowing substantially Concluding remarks The empirical results presented in this paper reveal that the rapidly growing Australian household debt is mainly the result of a favourable macroeconomic environment, a booming housing market, and a rising population The robust economic development in Australia (growing GDP with a low interest rate, low unemployment and a low inflation rate) causes optimistic expectations When households are sufficiently optimistic to borrow and investors are confident to lend, household debt surges The housing market has also played a significant role in the rapid growth of Australian household debt Australian households are in favour of housing investment The high demand for housing pushes up house prices and the expectation of rising housing prices in turn encourages investment demand for housing A booming housing market leads to a high level of housing mortgage debt The results also demonstrate that the GFC has had a markedly negative impact on household borrowing, while the financial market deregulation prior to 1993 and the RBA targeting inflation policy introduced in 1993 seem to have encouraged household borrowing Based on the findings of this paper, the authors suggest that it is necessary to rein in rising household debt To this, there are a number of instruments at the government’s disposal First, proper regulation and standardisation of financial markets is necessary Overly unregulated lending and borrowing practices will give rise to unaffordable home loans and cause household debt bubbles Proper levels of regulation can minimise irresponsible and irrational behaviours in financial markets, and thus reduce the risk of financial bubbles Second, the official interest rate is an effective tool but must be used in a timely and careful way Due to the dramatic impact of a change in interest rates and the time lag of this policy, this instrument must be used decisively when it is certain that the economy is heading in an unwanted direction Furthermore, it is better to change interest rates gradually Dramatic changes in interest rates with inappropriate timing may cause large economic fluctuations Moreover, the decision to change an official interest rate should be based on comprehensive analysis Many factors other than inflation rates have to be taken into account, including housing prices, the unemployment rate and the growth of household debt Finally, the government should monitor the assets market closely (especially the housing market in Australia), and intervene when necessary New dwelling approval is a useful tool in managing housing markets Since new housing entering the market can damp housing prices, state governments in Australia must loosen land use restrictions in the face of rapid rising housing prices and household debt The various incentives for home ownership (such as the first home grant), and for investment housing, have played a significant role in encouraging demand for housing in Australia These kinds of stimuli need to be abolished or suspended when a housing bubble is around the corner Appendix A Results of unit root tests with structural change.* Variable Items GDP with break at 08Q4 Coefficient t-Value Coefficient t-Value Coefficient t-Value U with break at 93Q1 R with break at 93Q1 T l Critical value (0.05)** 93 0.88 À5.08 93 0.22 À5.08 93 0.22 À5.08 a0 a1 a2 u1 u2 À1009.017 À0.874563 0.421078 3.009933 3.286257 4.176358 0.020157 1.154451 À0.072596 À3.152300 À0.174409 À4.415341 8.374419 0.200283 0.022707 2.073827 À0.087803 À3.067263 À6521.01 À3.35636 À0.24774 À2.25818 À0.33901 À0.12147 817.2433 3.339856 À0.02759 À2.24019 0.084325 2.984417 Source: authors’ estimation k X * Model for test: Y t ¼ a0 þ a1 à Y tÀ1 þ a2 à t þ u1 DM ỵ u2 DT ỵ bi DY ti ỵ et iẳ1 DT ẳ t T l ift > T l; and otherwise: * Lags are chosen in order to minimise AIC ** Obtained from Table 4A in Zivot and Andrews (1992) Author's personal copy 90 X Meng et al / Journal of Asian Economics 29 (2013) 80–90 Appendix B Rotation matrix B: ut ¼ Bet uT1;t uP1;t uP2;t uP3;t uP4;t uP5;t uP6;t uP7;t e1;t e2;t e3;t e4;t e5;t e6;t e7;t e8;t 0.006 0.077 0.019 À0.017 0.039 0.003 0.069 0.090 À0.169 0.041 0.016 À0.214 0.077 0.037 0.226 À0.294 0.533 À0.049 0.759 0.404 0.339 À0.058 0.610 À0.444 0.175 À0.499 À0.924 À0.593 0.169 1.475 À0.102 À0.073 2.269 À2.300 À3.241 À1.362 5.312 À4.278 1.730 0.879 0.055 0.056 À0.027 0.005 0.001 À0.020 À0.007 À0.000 À2.128 0.927 À0.260 0.195 2.540 0.403 À1.842 À0.518 0.616 À0.093 0.282 À0.343 0.078 À0.085 À0.175 À0.001 Source: authors’ estimation References ABS (Australian Bureau of Statistics) (2011) Various statistic tables http://www.abs.gov.au/ausstats/abs@.nsf/webpages/statistics Accessed 20.08.13 ANZ bank (2005) Submission to the Senate Economic References Committee public inquiry Aron, J., & Muellbauer, J (2000) Financial liberalisation, consumption and debt in South Africa, Centre for the Study of African Economies, Working Paper Series 132 Barnes, S., & Young, G (2003) The rise in US household debt: Assessing its causes and sustainability, Bank of England Working Paper No 206 Bray, J (2001) Hardship in Australia: An analysis of financial stress indicators in the 1998–99 Australian Bureau of Statistics Household Expenditure Survey, Department of Family and Community Services Occasional Paper No Cheung, Y.-W., & Lai, K S (1993) Finite-sample sizes of Johansen’s likelihood ratio tests for cointegration Oxford Bulletin of Economics and Statistics, 55, 313–328 Crook, J (2003) The demand and supply for household debt: A cross country comparison, Credit Research Centre Working Paper No Debelle, G (2004, March) Household debt and the macro-economy BIS Quarterly Review, 51–64 Dennis, J G., Hansen, H., Johansen, S., & Juselius, K (2005) CATS in RATS, Version Evanston, IL: Estima Hansen, H., & Johansen, S (1999) Some tests for parameter constancy in cointegrated VAR model Econometric Journal, 2, 306–333 Housing Industry Association Economics Group (2011) Dwelling unit commencements in Australia http://economics.hia.com.au/publications/state_outlook.aspx Accessed 20.08.13 Jacobsen, D (2004) What influences the growth of household debt? Economic Bulletin, 04, Q3 Johansen, S (1988) Statistical analysis of cointegration vectors Journal of Economic Dynamics and Control, 12, 231–254 Johansen, S., & Jueslius, K (1990) Maximum likelihood estimation and inference on cointegration – with applications to the demand for money Oxford Bulletin of Economics and Statistics, 52, 169–210 Johansen, S., Mosconi, R., & Nielsen, B (2000) Cointegration analysis in the presence of structural breaks in the deterministic trend Econometrics Journal, 3, 216– 249 Joyeux, R (2001) How to deal with structural breaks in practical cointegration analysis, Research Papers 0112 Macquarie University, Department of Economics Jueslius, K (2006) The cointegrated VAR model UK: Oxford University Press Kearns, A (2003) Mortgage arrears in the 1990s: Lessons for today, Central Bank and Financial Services Authority of Ireland Quarterly Bulletin, Autumn, 97–113 La Cava, G., & Simon, J (2003) A tale of two surveys: Household debt and financial constraints in Australia, Reserve Bank of Australia, Research Discussion Paper, 2003– 08, July 2003 Magri, S (2002) Italian households’ debt: Determinants of demand and supply Rome: Bank of Italy: Mimeo Martins, N., & Villanueva, E (2003) The impact of interest-rate subsidies on long-term household debt: Evidence from a large program, economics working papers 713, department of economics and business Universitat Pompeu Fabra Perron, P (1989) The great crash, the oil price shock, and the unit root hypothesis Econometrica, 57, 1361–1401 RBA (Reserve Bank of Australia) (1999) Consumer credit and household finance Reserve Bank of Australia Bulletin, June: 11–17 RBA (Reserve Bank of Australia) (2003) Household debt: What the data show Reserve Bank Bulletin, March: 1–11 RBA (Reserve Bank of Australia) (2011) Various statistics tables http://www.rba.gov.au/Statistics/tables/index.html Accessed 20.08.13 Reimers, H.-E (1992) Comparisons of tests for multivariate cointegration Statistical Papers, 33, 335–359 Schwartz, C., Hampton, T., Lewis, C., & Norman, D (2006) A survey of housing equity withdrawal and injection in Australia, RBA Research Discussion Papers RDP 200608 Stevens, G (1997) Some observations on low inflation and household finances Reserve Bank of Australia Bulletin, October: 38–47 Thaicharoen, Y., Ariyapruchya, K., & Chucherd, T (2004) Rising Thai household debt: Assessing risk and policy implications Bank of Thailand Symposium 2004 The Economist (2003) Finance and economics: Living in never-never land London: The Economist http://www.economist.com/node/1522783 Accessed 20.08.13, 11.01.03 The Treasury, Australian Government (2005) Treasury submission to Senate Economics References Committee public inquiry http://www.treasury.gov.au/ documents/710/HTML/docshell.asp?URL=04.asp Accessed 20.08.13 Tudela, M., & Young, G (2005) The determinants of household debt and balance sheets in the United Kingdom, Bank of England Working Paper No 266 Zivot, E., & Andrews, D (1992) Further evidence on the great crash, the oil-price shock, and the unit-root hypothesis Journal of Business & Economic Statistics, 10(3), 251–270 ... Liabilities as % of total assets Liabilities as % of disposable income Jun-90 Jun-95 Jun-00 Jun-01 Jun-02 Jun-03 Jun-04 Jun-05 187 267 473 515 596 685 800 905 1495 1926 2919 3180 3554 3932 4505... [À3.0527] Diagnostic tests: R-squared: 0.6820, adj R-squared: 0.6377, Schwartz (for ECM): 7.2343, Schwartz (for VECM): 42.9581 VEC residual serial correlation LM tests LM-Stat = 104.5274 10th order... in the model, but they are not displayed in the table 2-lag is chosen to minimise Schwartz criterion Table Structural-contemporaneous long-run impact matrix: C˜ ¼ CBÀ1 uT1;t DEBT NDWELL HPI CPI
Ngày đăng: 09/12/2017, 08:36
Xem thêm: