Abstract
This paper explores the factors that have led to a Canada-U.S. productivity gap using a sectoral growth accounting approach. Both fiscal and monetary policies have had significant effects on the saving rate. The Canadian dollar’s appreciation was followed by a protracted period of exchange rate weakness. This paper reviews the institutional aspects of Canada’s real return bond program. The Canadian system provides a successful model for pension reform. Free trade has helped promote the integration of U.S. and Canadian economies, but significant differences remain.
II. Canadian Household Saving—Developments and Risks1
1. The Canadian household saving rate has recently fallen to unusual lows. The saving rate remained in the range of around 5 percent during the 1997-2001 period, after having fallen steadily from its peak of just over 21 percent of disposable income in early 1982 (Figure 1). However, the saving rate plunged in late 2002 and 2003, dropping to a record-low of 1¼ percent in 2003Q2, and now stands below even the U.S. personal saving rate.
Saving, Interest Rates, and Inflation
Long-term gov’t bond rates and 5-year avg. CPI inflation
Citation: IMF Staff Country Reports 2004, 060; 10.5089/9781451806908.002.A002
Source: Haver Analytics.2. The drop in the saving rate raises question about the prospects for household spending. The suddenness of the decline in the saving rate, especially over the past two quarters, suggests that future consumption could slow as households seek to return the saving rate to higher levels.
3. In order to address this question, this chapter examines the extent to which saving has fallen below levels consistent with long-run “fundamentals.” An econometric model of the long-run relationship between the saving rate and household net worth, inflation, interest rates, and government spending is estimated. The results suggest that the secular decline in the saving rate since the early 1980s has reflected a response to the strength of macroeconomic policies in Canada, namely the improvement in the fiscal balance and the success in reducing inflation. However, the model is less successful in explaining the more recent decline and suggests that a saving rate consistent with economic fundamentals would be around 4½ percent, well above the present rate of 1¼ percent.
A. Trends in the Determinants of Household Saving2
4. The household saving rate is closely correlated with the expected inflation rate (Figure 1). From the mid-1960s to the late 1970s, the saving rate rose from around 5 percent to 15 percent, a period in which expected inflation rose from just over 2 percent to nearly 10 percent. The decline in expected inflation rate from the early 1980s to the very low levels achieved in the 1990s was also concurrent with a similar drop in the saving rate. However, as noted by Bérubé and Côté (2000), explanations for this correlation center on the fact that inflation tends to impart an upward bias to measured saving rates since incomes, as measured in the national accounts, include interest income, which tends to increase in nominal terms with inflation.” In addition, risk-averse households may tend to boost their saving during periods of high inflation, in order to shield themselves from further inflationary shocks.3
5. The Canadian personal saving rate also moves closely with the growth in gross government debt. The Canadian government ran persistent fiscal surpluses, averaging around 2¼ percent of GDP on a general government basis from 1962 to 1975. During the late 1970s and early 1980s, however, the budget swung to persistent deficits, averaging 3 percent of GDP. This deterioration in the fiscal position led to rapid debt growth, which was matched by a rise in the household saving rate (Figure 2). Similarly, the decline in the saving rate from its 1982 peak was matched by a marked improvement in the fiscal position, with the general government balance shifting from a deficit of 7 percent of GDP in 1985Q1 to a surplus of 3¾ percent in 2000Q3. The apparent inverse relationship between personal and public saving rates suggests that Canadian households are relatively “Ricardian” and increase their saving in periods of high fiscal deficits in order to prepare for expected future tax hikes.
Personal and Government Finances
Personal saving and growth in general government debt
Citation: IMF Staff Country Reports 2004, 060; 10.5089/9781451806908.002.A002
Source: Haver Analytics.6. A strengthening of household balance sheets also appeared to have contributed to the drop in the saving rate during the 1990s (Figure 3). The ratio of household net worth to disposable income remained roughly stable from the late 1960s to the late 1980s averaging around 3½. Starting in the 1990s, this ratio rose sharply, peaking at around 5¾ during the equity market bubble in 2000Q1, before dropping back to an average of just under 5 in 2001Q2 to 2003Q2.
Personal Saving and Net Worth
Percent (or share) of disposable income
Citation: IMF Staff Country Reports 2004, 060; 10.5089/9781451806908.002.A002
Source: Haver Analytics.7. The rise in household net wealth resulted from a rapid increases in the value of household holdings of equities and land, which more than offset rising debt. Since 1990, real household net worth has grown by 58 percent, with most of the growth occurring during the first half of the 1990s (Table 1). Increases in the holdings of equities and land accounted for around two-thirds of the growth in real wealth, despite accounting for only around one third of the total value of household wealth. Although household liabilities have also increased significantly since 1990, largely owing to the accumulation of mortgage debt during the early 1990s, real debt rose by only 50 percent during 1990-2003, much slower than the rise in assets (Table 1).
Growth in Selected Components of the Household Balance Sheet
Growth in Selected Components of the Household Balance Sheet
| 1990-1995 | 1995-2003 | 1990-2003 | 1990 | 2003 | |||
|---|---|---|---|---|---|---|---|
| (percent growth) | |||||||
| Net worth | 80.1 | 25.1 | 58.2 | ||||
| Assets | 80.3 | 26.4 | 56.9 | ||||
| (contribution to growth) | (as share of total) | ||||||
| Real estate | 33.9 | 13.3 | 22.6 | 50.0 | 46.5 | ||
| Residential structures | 15.0 | 5.0 | 9.4 | 23.5 | 21.2 | ||
| Land and other real estate | 14.3 | 6.2 | 10.6 | 16.6 | 17.2 | ||
| Other non-financial | 4.6 | 2.1 | 2.6 | 10.0 | 8.1 | ||
| Financial | 46.3 | 13.0 | 34.2 | 50.0 | 53.5 | ||
| Currency | 11.7 | 3.9 | 8.5 | 10.3 | 12.3 | ||
| Bonds and other holdings | 9.6 | -0.5 | 4.2 | 29.6 | 21.7 | ||
| Equity | 25.0 | 9.6 | 21.5 | 10.1 | 19.4 | ||
| (percent growth) | |||||||
| Liabilities | 81.0 | 31.8 | 51.4 | ||||
| (contribution to growth) | (as share of total) | ||||||
| Mortgages | 46.9 | 13.9 | 28.7 | 63.6 | 60.9 | ||
| Other | 34.1 | 17.9 | 22.6 | 36.4 | 39.1 | ||
Growth in Selected Components of the Household Balance Sheet
| 1990-1995 | 1995-2003 | 1990-2003 | 1990 | 2003 | |||
|---|---|---|---|---|---|---|---|
| (percent growth) | |||||||
| Net worth | 80.1 | 25.1 | 58.2 | ||||
| Assets | 80.3 | 26.4 | 56.9 | ||||
| (contribution to growth) | (as share of total) | ||||||
| Real estate | 33.9 | 13.3 | 22.6 | 50.0 | 46.5 | ||
| Residential structures | 15.0 | 5.0 | 9.4 | 23.5 | 21.2 | ||
| Land and other real estate | 14.3 | 6.2 | 10.6 | 16.6 | 17.2 | ||
| Other non-financial | 4.6 | 2.1 | 2.6 | 10.0 | 8.1 | ||
| Financial | 46.3 | 13.0 | 34.2 | 50.0 | 53.5 | ||
| Currency | 11.7 | 3.9 | 8.5 | 10.3 | 12.3 | ||
| Bonds and other holdings | 9.6 | -0.5 | 4.2 | 29.6 | 21.7 | ||
| Equity | 25.0 | 9.6 | 21.5 | 10.1 | 19.4 | ||
| (percent growth) | |||||||
| Liabilities | 81.0 | 31.8 | 51.4 | ||||
| (contribution to growth) | (as share of total) | ||||||
| Mortgages | 46.9 | 13.9 | 28.7 | 63.6 | 60.9 | ||
| Other | 34.1 | 17.9 | 22.6 | 36.4 | 39.1 | ||
B. Determinants of the Canadian Personal Saving Rate
8. The household saving rate is typically viewed as being tied in the long run to household wealth and income. For example, the Permanent Income Hypothesis posits that household consumption and saving decisions reflect expected levels of permanent income and wealth.4 Since saving (st) is income (yt) less consumption, the saving rate should reflect the share of household net worth (nwt) to income. This suggests the following long-run relationship:
where yt is measured as personal disposable income.5
9. The apparent close relationship between the saving and inflation, interest rates, and government fiscal policy suggests the following augmented equation:
where xt is a vector of additional, explanatory variables, including the real interest rate, the government fiscal balance as a percent of GDP, and the expected inflation rate. The inclusion of these additional variables can be justified by arguing that measured household net worth and income are imperfect proxies for household permanent income. The equation is estimated using the Phillips-Hanson Fully Modified Least Squares estimator, and a chi-squared test of the joint significance of the variables suggests that a broad model, which includes all the additional variables, fits the data well (Table 3, Column 5).6
Unit Root Tests: Saving and Wealth
Unit Root Tests: Saving and Wealth
| ADF | Phillips- Perron |
|
|---|---|---|
| Null hypothesis: data contain a unit root. | ||
| (p-values in parentheses) | ||
| Saving rate | -0.156 | -0.164 |
| (0.939) | (0.938) | |
| lag=3 | ||
| Net worth: share of | -0.765 | -0.912 |
| disposable income | (0.824) | (0.781) |
| lag=7 | ||
Unit Root Tests: Saving and Wealth
| ADF | Phillips- Perron |
|
|---|---|---|
| Null hypothesis: data contain a unit root. | ||
| (p-values in parentheses) | ||
| Saving rate | -0.156 | -0.164 |
| (0.939) | (0.938) | |
| lag=3 | ||
| Net worth: share of | -0.765 | -0.912 |
| disposable income | (0.824) | (0.781) |
| lag=7 | ||
Fully-Modified OLS Regressions
As a share of personal disposable income
As a share of GDP
Joint significance test of all of the variables in the regression
Fully-Modified OLS Regressions
| (1) | (2) | (3) | (4) | (5) | |
|---|---|---|---|---|---|
| Constant | 28.78 | 30.56 | 24.82 | 3.52 | 6.66 |
| (1.94) | (2.79) | (1.61) | (2.34) | (1.98) | |
| Net worth 1/ | -0.05 | -0.03 | -0.04 | -0.01 | -0.01 |
| (0.01) | (0.01) | (0.00) | (0.00) | (0.00) | |
| Government | -0.66 | -0.36 | |||
| saving 2/ | (0.08) | (0.06) | |||
| Real long-term | 1.17 | 0.77 | |||
| interest rates | (0.16) | (0.14) | |||
| Expected inflation | 1.53 (0.12) |
1.18 (0.11) |
|||
| Access to credit 1/ | -0.13 (0.08) |
||||
| Chi-squared 3/ | 87.99 | 63.59 | 194.42 | 411.29 | 660.92 |
| Sample period: 1967Q1-2003Q2 | |||||
As a share of personal disposable income
As a share of GDP
Joint significance test of all of the variables in the regression
Fully-Modified OLS Regressions
| (1) | (2) | (3) | (4) | (5) | |
|---|---|---|---|---|---|
| Constant | 28.78 | 30.56 | 24.82 | 3.52 | 6.66 |
| (1.94) | (2.79) | (1.61) | (2.34) | (1.98) | |
| Net worth 1/ | -0.05 | -0.03 | -0.04 | -0.01 | -0.01 |
| (0.01) | (0.01) | (0.00) | (0.00) | (0.00) | |
| Government | -0.66 | -0.36 | |||
| saving 2/ | (0.08) | (0.06) | |||
| Real long-term | 1.17 | 0.77 | |||
| interest rates | (0.16) | (0.14) | |||
| Expected inflation | 1.53 (0.12) |
1.18 (0.11) |
|||
| Access to credit 1/ | -0.13 (0.08) |
||||
| Chi-squared 3/ | 87.99 | 63.59 | 194.42 | 411.29 | 660.92 |
| Sample period: 1967Q1-2003Q2 | |||||
As a share of personal disposable income
As a share of GDP
Joint significance test of all of the variables in the regression
10. The estimates confirm the important role of household wealth (Table 3). Household net worth is highly statistically significant, regardless of which of the additional explanatory variables are included. However, the parameter estimates are not stable across regressions, suggesting that other variables beyond wealth are important determinants of the saving rate—especially monetary variables, such as expected inflation and interest rates.7
11. The fiscal deficit and inflation also are highly significant. In the broadest version of the equation, the coefficient on the deficit/GDP ratio is -0.36, suggesting that a 1 percentage point increase in the deficit ratio lowers the saving rate by this amount. This confirms the earlier conjecture that Canadian households have to some extent been Ricardian in their response to fiscal policy, insofar as they have tended to view an improvement in the general government fiscal position as having increased household net wealth. The expected inflation rate also exhibited a strong, positive correlation with the saving rate, entering with a coefficient that appears close to unity. Furthermore, the coefficients on the real interest rate and expected inflation are statistically different from one another, suggesting that each plays an important, independent role.8
12. The results suggest that the current saving rate is well below its equilibrium level. The results from the model in Column 5 confirm the tight fit of a model in which household net worth, government saving, interest rates, and inflation expectations all influence saving behavior (Figure 4). However, the results also indicate that the recent decline in saving rate has been anomalous. The estimated equilibrium saving rate is 4.6 percent in 2003Q2—almost 2¼ percentage points higher than the actual saving rate in that quarter.
Long-Run Estimates of Saving
Actual and estimated values for the personal saving rate
Citation: IMF Staff Country Reports 2004, 060; 10.5089/9781451806908.002.A002
Sources: Haver Analytics; Fund staff estimates.13. The estimates suggest that the decline in inflation expectations has had the largest role in the longer-run decline in the household saving rate. Figure 5 displays the effects of each of the variables on the saving rate. Inflation expectations fell from a peak of 10¼ percent in the early 1980s to 2¼ percent in 2003H1; this decline helped to lower saving rates by around 9½ percentage points. The swing in the fiscal balance from around -9 percent of GDP to around 1½ percent of GDP accounted for a 3¾ percentage point decline in the saving rate. Despite the significant correlation between the household net worth and the saving rate, net worth played a relatively minor role in explaining changes in the saving rate over the past 25 years, and in the past 10 years subtracted only around 1¼ percentage points from the saving rate.
Determinants of the Saving Rate
Contributions to the equilibrium saving rate, ex. constant term
Citation: IMF Staff Country Reports 2004, 060; 10.5089/9781451806908.002.A002
Source: Fund staff estimates.14. Surprisingly, improvements in household access to credit are not found to have had a significant impact on the saving rate. Similar to the results reported by Berube and Cote (2000), a simple model that includes just wealth and a measure of household access to credit (Table 3, Column 2) finds that credit measure is statistically insignificant. The decline in the chi-squared test statistic (from Column 1) also suggests that we could drop the variable from the regression. Moreover, regressions similar to those in Columns 3-5, which include the household access to credit term, also find that the credit term is statistically insignificant. However, regressions over the shorter sample starting in the early 1980s find that the household access to credit term is statistically significant, suggesting that financial innovation may have been an important factor in explaining the drop in the saving rate in the more recent period.
C. Conclusions
15. The results above suggest the following two conclusions:
The current saving rate appears well below its long-run equilibrium. The broadest model fits the data well although the errors—of almost 2¼ percentage points—in the past two quarters have been the largest in almost thirty years. The historical experience also suggests that the saving rate could return to its “equilibrium” relatively quickly.
Both fiscal and monetary policies appear to have had significant effects on the saving rate in the long-run. Increases in the expected inflation rate—to the extent that it is not reflected in nominal interest rates—has been associated with higher saving and lower consumption. Similarly, past episodes of budget deficits had the effect of raising saving and dampening household demand.
APPENDIX Data Sources
Expected inflation is measured using a five-year rolling average of CPI inflation.
Government saving is the ratio of general government saving (in the National Income and Expenditure Accounts) to GDP.
Household saving rates are taken from the National Income and Expenditure Accounts, which calculate saving as the difference between personal income and consumption.
Household access to credit is proxied by the total value of consumer credit outstanding, excluding mortgages (as a share of personal disposable income). Consumer credit (as opposed to mortgages) is typically not backed by a real asset, and increases in credit provisioning implies improved monitoring of default risk and credit pricing.
Real interest rates are defined as the nominal interest rate on 10-year government bonds less expected inflation.
Household net wealth: The National Balance Sheet (NBS) provides the basis for a detailed breakdown of the net wealth position of persons and unincorporated businesses. This chapter uses the NBS definitions for nonfinancial assets, non-equity financial assets, and total liabilities. All of the data are interpolated to a quarterly basis (prior to 1990) using the quarterly flow data to create a “synthetic” stock value; the difference between the synthetic stock value at the end of the year, and the actual EOP value of the data represents the valuation change. This valuation change is distributed, proportionate to the synthetic stocks, throughout the year. Finally, the data are seasonally adjusted using the additive X-12 routine in Eviews 4.1.
In addition, this chapter imputes the value of equity assets indirectly because of statistical problems in the NBS. The equity data (along with other securities) are currently estimated in the NBS as a residual from total assets, and as such, are close in definition to the book value. Because equity wealth rose sharply in the late 1990s, and subsequently fell, this chapter indirectly estimates the total value of equity assets held by households.
Household holdings of equity wealth are assumed to be a fraction of the overall market capitalization of the Toronto Stock Exchange (TSE). Using the NBS data, the fraction of household ownership of equities (at market value) is assumed to be the ratio of personal holdings of equities (in the NBS) to the total, economy-wide holdings of equities. This fraction is then multiplied by the total stock market capitalization (using data from 1965–2002) to arrive at a market-value measure of current household holdings of equities. For 2003, the quarterly growth in household holdings is assumed to match the growth of the TSE300 since end-2002.
Despite shortcomings, this imputation produces estimates of total net worth that more closely capture the trends during the late 1990s, when the saving rate began to decline (Figure 6). The NBS measure of net worth has been relatively smoother. In particular, the NBS data miss the sharp spike in equity values in 2000, which is captured in this new data.
Household Net Worth
From the National Balance Sheet and from equity-imputed data
Citation: IMF Staff Country Reports 2004, 060; 10.5089/9781451806908.002.A002
Sources: Haver Analytics; Toronto Stock Exchange.References
Attanasio, O.P., 1999, “Consumption Demand”, in Handbook of Macroeconomics edited by J.B. Taylor and M. Woodford, Vol.1b (New York: Elsevier).
Bérubé, G. and D. Côté, 2000, “Long-term Determinants of the Personal Saving Rate: Literature Review and Some Empirical Results for Canada”, Bank of Canada Working Paper, No.2000-3 (Ottawa: Bank of Canada).
Cerisola, M. and P. De Masi, 1999, “Determinants of the U.S. Personal Saving Rate”, United States—Selected Issues, SM99/164 (Washington: International Monetary Fund).
MacKinnon, J. G., 1996, “Numerical Distribution Functions for Unit Root and Cointegration Tests”, Journal of Applied Econometrics, Vol.11, pp.601– 18.
Phillips, P.C.B. and B. Hansen, 1990, “Statistical Inference in Instrumental Variables Regression with Processes”, Review of Economic Studies, Vol.58, pp.99– 126.
Phillips, P.C.B. and P. Perron, 1988, “Testing for a Unit Root in a Time Series Regression”, Biometrica, Vol.75, pp.335– 46.
Pichette, L and D. Tremblay, 2003, “Are Wealth Effects Important for Canada?”, Bank of Canada Working Paper, 2003-30 (Ottawa: Bank of Canada).
Said, S. and D. Dickey, 1984, “Testing for Unit Roots in Autoregressive-Moving Average Models of Unknown Order”, Biometrika, Vol.71, pp.599– 607.
Prepared by Christopher M. Faulkner-MacDonagh (WHD).
See the data appendix for the definitions of the variables.
Revisions to the national accounts have also adjusted upward the estimate of income and saving, suggesting that the saving rate may be higher than currently reported. For example, staff at Statistics Canada has noted that the initial estimate of the 2000 saving rate was 3.2 percent; it has subsequently been revised to 4.6 percent.
The literature on consumption behavior, including with regard to the PIH, is reviewed in Attanasio (1999).
Table 2 shows that the saving rate and net worth-income ratios both have a unit root (that is, they are I(1)), using either the Phillips-Perron test (Phillips and Perron, 1988) or the Augmented Dickey Fuller test (Said and Dickey, 1984). Both tests fail to reject the null hypothesis of a unit root at the 5 percent level. Because they are I(1), the Fully-Modified OLS (FM-OLS) estimation procedure in Microfit 4.1 is used to obtain the parameter estimates, similar to estimation used for the U.S. saving rate in Cerisola and De Masi (1999). The FM-OLS parameter estimates and standard errors are corrected to fix statistical problems (Phillips and Hansen, 1990).
An advantage of the FM-OLS approach is that it does not require knowledge of which, if any, of the variables in xt are I(1) or I(0), and whether there are any cointegrating relationships.
The low personal saving rate may also reflect a high level of expected wealth. Pichette and Tremblay (2003) construct such a measure and shows that it is important in determining consumption and saving patterns.
If the coefficients were not statistically different, then they could be combined into a single term that would merely reflect the impact of nominal interest rates on saving, with no independent role for expected inflation.
