A. Differences Based on Measures of Fiscal Sustainability

An economy with a sustainable long-run fiscal position comes closest to satisfying Ricardian features because increases in taxes and expenditures are unlikely to affect the long-run intergenerational balance unless they are sustained for a very long time. Rational individuals would therefore respond to an increase in taxes (expenditures) by reducing (raising) their saving to maintain their consumption profile. Calculations of the long-run fiscal position have been motivated by the future aging of the population and the prospective decline in the participation of working-age men in work because these developments are likely to put a heavy strain on the public finances of the industrialized countries. Auerbach et al. (1999) have used generational accounting to estimate the likely cost of these developments, measured as the unfunded liability of the government. They document that the pension liability is most pronounced for Germany, Italy and Japan while Belgium, France, and the United States have moderately large liabilities, and Canada and Sweden are in generational balance (Table 4). Raffelhüschen et al. (1999) have conducted a similar study on other European countries and they find that the intergenerational liability is large for Finland, Germany, Italy, and Spain, moderate in France and the Netherlands, and is absent in Ireland. Auerbach et al. assume that benefits grow in line with GDP per capita whereas Raffelhüschen et al. assume that benefits grow at the rate of productivity growth. Over the medium term these assumptions are very similar because the employment/population ratio is broadly constant. Cardarelli, Sefton, and Kotlikoff (1999) have estimated generational accounts for the United Kingdom based on the assumption that benefits are indexed by the CPI and health care expenditures per inhabitant decline over time (this is a proxy of the government’s current intentions).³ Under these assumptions they document the existence of a small intergenerational liability for the United Kingdom, Benediktsson et al. (2000) have estimated generational accounts for Iceland based on the assumption that benefits are indexed by labor productivity growth. In contrast to most other studies which categorize all government investments that do not provide revenue as consumption, they only include the depreciation of these assets as consumption. Under these assumptions they find that the intergenerational liability is small in Iceland.

Table 4.

Generational Accounting Liability

(in percent of GDP)

^1/Benediktsson et al (2000)

^2/ Cardarelli, Sefton, and Kotlikoff (1999)

Table 4.

Generational Accounting Liability

(in percent of GDP)

	Kotlikoff et al. estimates	Raffeluschen et al. estimates
Australia	12.6
Belgium	20.2	18.8
Canada	0.1
Finland		253.2
France	26.5	81.3
Germany	51.6	136
Iceland 1/	8.6
Ireland		-4.3
Italy	77.4	107.3
Japan	59.3
Netherlands		75.9
Spain		151.9
Sweden	-14.5
UK 2/	33.0
US	16.9

^1/Benediktsson et al (2000)

^2/ Cardarelli, Sefton, and Kotlikoff (1999)

View Table

Table 4.

Generational Accounting Liability

(in percent of GDP)

	Kotlikoff et al. estimates	Raffeluschen et al. estimates
Australia	12.6
Belgium	20.2	18.8
Canada	0.1
Finland		253.2
France	26.5	81.3
Germany	51.6	136
Iceland 1/	8.6
Ireland		-4.3
Italy	77.4	107.3
Japan	59.3
Netherlands		75.9
Spain		151.9
Sweden	-14.5
UK 2/	33.0
US	16.9

^1/Benediktsson et al (2000)

^2/ Cardarelli, Sefton, and Kotlikoff (1999)

Not only can countries in intergenerational balance demonstrate Ricardian features, but any country can manifest such features over isolated periods of time if fiscal measures adopted by a particular country lead to a significant revaluation of its long run fiscal position. Following the recent experiences of Ireland and Denmark in which large fiscal consolidations resulted in booming economies, a number of researchers have begun to emphasize the role of expectations in understanding the relationship between movements in the government budgetary position and total saving. In an intertemporal model of consumption behavior, Bertola and Drazen (1993) argue that the effects of current fiscal policy depend on the expectations it generates for the course of future fiscal policy. If a sizeable fiscal contraction signifies the intent for future fiscal consolidations, the private sector may take this into account and offset a large fraction of it. Similarly, Giavazzi and Pagano (1996) argue that large fiscal consolidations can be expansionary precisely because they signal a permanent and decisive change in the stance of fiscal policy, while small adjustments may have the opposite effect for the opposite reason. Blanchard (1990) emphasizes the non-Keynesian effects of fiscal policy contractions at high levels of debt by postulating the existence of a trigger point beyond which tax increases result in declines in output because of tax distortions. Countries undertaking fiscal consolidation with high debt levels will experience an extra output boost if the debt level falls below the trigger point because of the accompanying reduction in tax distortions. The private sector anticipates this prospective output boost and offsets the contractionary effects of fiscal policy through declines in private saving. Perotti (1999) endogenizes this effect and shows that aggregate consumption is negatively influenced by positive shocks to government expenditure in countries with large intergenerational liabilities. Individuals expect tax increases following increases in expenditure and this expectation causes a bigger fall in their wealth and consumption the larger the initial distortions. Since the tax distortion is a positive function of the intergenerational balance, the effect carries through for the level of the intergenerational balance.

Alesina and Perotti (1997) have considered the relationship between fiscal policy and economic activity in identifying successful fiscal adjustments among OECD countries and comparing macroeconomic developments following successful and unsuccessful fiscal adjustments. They identify a successful fiscal adjustment if one of the two following conditions holds three years after the adoption of a tight fiscal stance: (1) the ratio of the cyclically adjusted primary deficit to GDP is at least 2 percentage points below the level observed in the last year of the fiscal contraction or (2) the ratio of debt to GDP is 5 percent of GDP below the level observed in the last year of the fiscal contraction. They find that following the adoption of a successful fiscal adjustment, a country’s growth profile is significantly higher than the industrial country average, fuelled by strong investment.

Tests were conducted to determine the effects of the intergenerational budget position and sizeable fiscal adjustments as follows. The importance of the intergenerational budget balance was identified by dropping the country specific revenue and expenditure coefficients from equation (1) and adding to it the interaction between the change in government revenues and expenditures and a dummy variable with the value of unity for countries with no estimated intergenerational fiscal liability (Canada, Iceland, Ireland, and Sweden). The long-run level relationship between revenues, expenditures and national saving was also assumed to differ between countries in and out of intertemporal fiscal balance.

The temporary effects of large fiscal adjustments were captured by interacting changes in government revenues and expenditures with a dummy representing periods when fiscal contractions/expansions were large (defined as periods when the structural fiscal balance changed by more than 1 percentage point in either direction, see appendix 2) in countries which are not in intergenerational balance. Separate variables were added for countries with sizeable intergenerational imbalances to capture non-linear effects resulting from the distortionary effects of taxation.

The ranking of countries based on the size of the intergenerational liability is comparable between the Auerbach et al. and Raffelhüschen et al. studies although the actual estimates differ considerably. Germany and Italy are estimated to have large liabilities in both studies and therefore these countries are included in the high debt category in our analysis. Japan and Spain are also included in this category because their liabilities are high relative to those of other countries in their respective sample groups. While the findings of the Raffelhüschen et al. study would suggest that Finland should also be in this category, Finland is currently running fiscal surpluses at about 4 percent of GDP and the decision to set aside these surpluses to pay future pensions has improved its intergenerational fiscal position immensely. For this reason Finland is excluded from the sample of high debt countries in the analysis which follows.

The estimated equation is as follows:

where the subscript ld/hd identifies countries with modest and large intergenerational imbalances and subscript ib identifies countries in intergenerational balance.

The estimated coefficients of this regression are presented in column 1 of Table 5. The coefficient estimate on the change in government revenues is significantly positive but small for the sample as a whole (β5) whereas the coefficient on the change in government expenditures is significantly negative and considerably larger in magnitude (β6). This confirms the earlier result that Keynesian features are present in these economies. While the small revenue coefficient in the whole sample (β5) does not get offset by the revenue coefficient of the countries with sustainable long-run fiscal positions(β11), the expenditure coefficient (β6) gets completely offset (β12), evident in the insignificant Chi squared statistic. These results are partially supported by the long-run coefficients. In the long-run the relationship between government revenues and national saving is insignificant for all countries in the sample whereas, a significantly negative long-run relationship is identified between expenditures and national saving with a coefficient of about a half. Surprisingly, the expenditure coefficient is comparable for countries in or out of intergenerational balance whereas one would have thought that the expenditure coefficient would be insignificant for countries in intergenerational balance. On the other hand, eliminating the long-run revenue and expenditure variables for countries in intergenerational balance strengthened the short-run offset effects with the coefficient on revenues significantly negative and partially offsetting the positive effect for the whole sample (equation [2]). These results suggest that, in general, countries in intergenerational balance exhibit Ricardian features.

Table 5.

Determinants of Gross National Savings

Table 5.

Determinants of Gross National Savings

Variable	[1]		[2]		[3]		[4]		[5]
β5	0.12	*	0.17	*	0.13	*	0.18	*	-0.04
β6	-0.7	*	-0.72	*	-0.64	*	-0.7	*	-0.63	*
β7	0.36	*	0.3	*	0.25	*	0.19	*	0.06
β8	-0.42	*	-0.41	*	-0.67	*	-0.61	*	-0.48	*
β9	-0.06		-0.12	*	-0.04		-0.07		-0.05	*
β10	0.16	*	0.15	*	0.28	*	0.31	*	0.27	*
β11	-0.07		-0.09	*	-0.09		-0.1		0.15
β12	0.65	*	0.75	*	0.56	*	0.72		0.66	*
β13									0.45	*
β14									-0.3	*
β15	-0.41	*	-0.36	*	-0.44	*	-0.39	*	-0.44	*
β16	0.1		0.09		0.05		0.06		-0.08
β17	-0.59	*	-0.59	*	-0.56	*	-0.55	*	-0.35	*
β18									0.33	*
β19									-0.51	*
β20	0.06				0.03
β21	-0.37	*			-0.4	*
Diagnostic Statistics:
R squared	0.59		0.59		0.62		0.61		0.61
Q stat(l)	3.4		3.4		4.6		4.9		4.6
Q stat(2)	3.8		4		6		6.2		5.6
Hypothesis Tests
	Ho:β12=-β6 Ho:β1l=-β5							Ho:β12=-β6
	Chi Sq(1)=1.6 Chi Sq(1)=8.8*							Chi Sq(1)=0.7
	Ho:β17=β21 Ho:β12=-β6
	Chi Sq(1)=0.7 Chi Sq(1)=0.7

View Table

Table 5.

Determinants of Gross National Savings

Variable	[1]		[2]		[3]		[4]		[5]
β5	0.12	*	0.17	*	0.13	*	0.18	*	-0.04
β6	-0.7	*	-0.72	*	-0.64	*	-0.7	*	-0.63	*
β7	0.36	*	0.3	*	0.25	*	0.19	*	0.06
β8	-0.42	*	-0.41	*	-0.67	*	-0.61	*	-0.48	*
β9	-0.06		-0.12	*	-0.04		-0.07		-0.05	*
β10	0.16	*	0.15	*	0.28	*	0.31	*	0.27	*
β11	-0.07		-0.09	*	-0.09		-0.1		0.15
β12	0.65	*	0.75	*	0.56	*	0.72		0.66	*
β13									0.45	*
β14									-0.3	*
β15	-0.41	*	-0.36	*	-0.44	*	-0.39	*	-0.44	*
β16	0.1		0.09		0.05		0.06		-0.08
β17	-0.59	*	-0.59	*	-0.56	*	-0.55	*	-0.35	*
β18									0.33	*
β19									-0.51	*
β20	0.06				0.03
β21	-0.37	*			-0.4	*
Diagnostic Statistics:
R squared	0.59		0.59		0.62		0.61		0.61
Q stat(l)	3.4		3.4		4.6		4.9		4.6
Q stat(2)	3.8		4		6		6.2		5.6
Hypothesis Tests
	Ho:β12=-β6 Ho:β1l=-β5							Ho:β12=-β6
	Chi Sq(1)=1.6 Chi Sq(1)=8.8*							Chi Sq(1)=0.7
	Ho:β17=β21 Ho:β12=-β6
	Chi Sq(1)=0.7 Chi Sq(1)=0.7

Turning to the effects of significant fiscal consolidations, the coefficient on the change in government revenues is significantly positive for countries with small intergenerational imbalances (β7) but insignificant for countries with large intergenerational imbalances (β9). This suggests that in countries with small imbalances changes in government revenues have bigger impacts on total saving when fiscal adjustment is large. Moreover, while the effect on national saving of large expenditure adjustments gets accentuated for countries with modest intergenerational imbalances (β8<0), it gets attenuated for countries with large intertemporal fiscal imbalances (β10 >0). Since the sum of β6 and β8 is greater than one, the private sector accommodates sizable changes in the fiscal position in countries with modest intergenerational imbalances which is surprising. On the other hand the more modest impact on national saving of expenditure changes in countries with large intertemporal fiscal imbalances is consistent with Perotti (1999) and provides support for the presence of expectation effects related to fiscal actions and/or large tax distortions. Moreover, the behavioral difference in countries with large imbalances manifests itself in response to expenditure rather than revenue changes. This is consistent with the fact that fiscal consolidations achieved through changes in revenue are thought to be less durable than those achieved through changes in expenditure because they are more easily reversed.

The empirical literature has focused on current government obligations in defining countries with high debt levels whereas this paper has used the intertemporal definition, which arguably is closer in spirit to the Ricardian hypothesis.⁴ It could be claimed, however, that the findings are sensitive to changes in the definition of countries with high debt. To test for this, high debt countries were defined as those with large ratios of government obligations to GDP. According to this definition, Belgium is categorized as a high debt country and Germany and Spain are categorized as countries with moderate debts (see Table 6). Consistent with our previous finding, in countries with high debt ratios, the private sector offsets government expenditure changes rather than revenue changes (β9~0, β10>0) during periods of large fiscal adjustments. For every 1 percentage point increase in expenditures, national saving declines by 0.36 percentage points in countries with high debt ratios whereas it declines by over 1 percentage point in countries with moderate imbalances. This difference is larger than under the alternative definition of countries with high debt and the explanatory power of the equation is higher.

Table 6.

Average Import and Gross Debt Levels 1980-99

(in percent of GDP)

Table 6.

Average Import and Gross Debt Levels 1980-99

(in percent of GDP)

	Import Ratio	Gross Debt Ratio
BELGIUM	67.8	119.2
CANADA	29.0	78.6
FINLAND	28.0	30.3
FRANCE	22.1	38.2
GERMANY	26.0	46.3
ICELAND	34.7	39.7
IRELAND	57.2	86.0
ITALY	21.2	95.7
JAPAN	9.7	77.4
NETHERLANDS	54.8	70.6
SPAIN	20.9	45.8
SWEDEN	31.8	62.7
UNITED KINGDOM	26.7	46.8
UNITED STATES	11.0	62.0

View Table

Table 6.

Average Import and Gross Debt Levels 1980-99

(in percent of GDP)

	Import Ratio	Gross Debt Ratio
BELGIUM	67.8	119.2
CANADA	29.0	78.6
FINLAND	28.0	30.3
FRANCE	22.1	38.2
GERMANY	26.0	46.3
ICELAND	34.7	39.7
IRELAND	57.2	86.0
ITALY	21.2	95.7
JAPAN	9.7	77.4
NETHERLANDS	54.8	70.6
SPAIN	20.9	45.8
SWEDEN	31.8	62.7
UNITED KINGDOM	26.7	46.8
UNITED STATES	11.0	62.0

This paper’s approach in determining Ricardian effects has a number of similarities with the analysis of Giavazzi and Pagano (GP, 2000). One major difference however is that GP ignore the distinction between countries in and out of intergenerational balance. While they find small revenue and large expenditure offset coefficients, similar to this study, they also find that large fiscal adjustments moderate the impact of changes in government incomes on national saving for all countries, and that the impact of revenue/expenditure changes on national saving is independent of the level of government debt. A potential explanation for the different results is that GP do not isolate periods with large fiscal adjustments when interacting government revenues and expenditures with dummies for countries with high debt ratios. It is quite likely that the impact of the high debt burden is greatest during these periods because they suggest a significant break in the debt profile.

B. Differences Based on the Degree of Openness

The strong relationship between changes in government revenues and expenditures and total saving indicates the presence of Keynesian features in these economies. In such economies the extent to which fiscal policy can impact the economy depends on the degree of openness since expenditures leak out of the economy via imports. This effect is clear from the national income identity:

where S_p is the gross private saving rate, S_g is the gross saving position of the government, Ip is the private investment rate, Ig is the public investment rate, and CA is the current account. In open economies, changes in government saving generate significant movements in the current account and hence in national saving although the output effects are moderated because of the expenditure leakage into imports.

To capture the effects of openness to trade, a proxy variable was introduced, defined as the ratio of imports of goods and services to GDP. Based on this definition, Belgium is the most open economy in the sample with an average import ratio of over 67 percent of GDP over the 1980-99 period (Table 6). The import ratios for Ireland and the Netherlands are also high at over 50 percent. Most of the other import ratios are grouped in two country categories, one between 25 and 35 percent of GDP (Iceland, Sweden, Canada, Finland, the UK, and Germany) and the other between 20 and 22 percent of GDP (France, Italy, and Spain) with Japan and the United States taking up the rear. Countries with a high level of import penetration were defined as those with an import ratio above 25 percent of GDP and extra short and long-run revenue and expenditure coefficients were included for countries with this characteristic.⁵ The revised equation is as follows:

where the subscript o denotes variables which have been multiplied by a dummy associated with countries with large import ratios.

In this specification the coefficients on changes in government revenues and expenditures interacted with dummies for intergenerational balance and significant fiscal adjustments are comparable to the previous specification (column 5 in Table 5) with the exception that government revenue changes are insignificant in the short run.⁶ Of the new variables of interest, the degree of openness affects the spillage of government revenues and expenditures into the current account (significant coefficients on β13 and β14) with stronger effects for revenues than expenditures. This could be associated with the fact that government expenditures are generally spent on home goods so the initial spillage into imports is likely to be less. Turning to the long-run coefficients, a 1 percentage point increase in revenues has no effect on national saving in closed economies but results in a 0.33 percentage point increase in open economies. On expenditures, a 1 percentage point increase in expenditure leads, in the long-run, to a 0.35 percentage point decline in national saving in closed economies and a 0.50 percentage point decline in open economies.

The finding that countries in intergenerational balance demonstrate Ricardian features has considerable intuitive appeal. However, the extent to which this finding is robust is debatable. Apart from the works already cited there are few studies on Ricardian equivalence which focus on Canada, Iceland, Ireland, and Sweden. However, those that do so generally find that Sweden, Iceland and Ireland manifest these features although the Canadian case is more uncertain. Carroll and Summers (1987) compare saving rates in Canada and the United States and, controlling for other determinants, find that the difference in the saving rates can be explained assuming that the Canadian economy has Ricardian features. Evans and Hasan (1994) find that the lagged level of real per capita net worth is unable to explain changes in private consumption in Canada, consistent with the view that the planning horizon for private sector consumption decisions is infinite. However, in a more recent paper Evans and Karras (1996) find that changes in current income have a significant effect on changes in consumption for most countries, including Canada, suggesting that liquidity constraints are present in these economies. For Iceland, Einarrsson (1990) estimates an empirical model of the Icelandic economy and simulates its response to a temporary tax cut. He finds very little change in the current account following this tax cut and argues that this is suggestive of strong Ricardian effects. Evans and Karras’ findings for Ireland and Sweden indicate that changes in current income have no effect on current consumption. This finding on Sweden is supported by Brunila (1997) who estimates Euler equations for a number of European countries and finds that current income has no effect on current consumption in the Netherlands, Finland, and Sweden but has a significant effect in Austria, Belgium, France, Germany, Greece, Italy, and the UK.