Abstract
The relationship between regional saving and investment is examined to measure the extent of capital mobility. The relationship between total regional saving and investment is significantly negative in Canada and the United Kingdom, in contrast to the significant positive relationship found across countries. The difference is attributed to government subsidies to poor regions. The relationship between personal saving and private investment is insignificant in the U.K. and Germany and is negative in Canada which suggests that capital is mobile for individuals. The relationship between retained earnings and private investment is significantly positive in the U.K. and Canada suggesting capital immobility for firms but a test for the presence of regional corporate liquidity constraints yields no effects.
I. Introduction
Ever since the publication of Feldstein and Horioka (1980), there has been continuous debate on the reasons behind high saving and investment correlations across countries. Many papers have documented the influential role played by government policy in the post war period in limiting current account imbalances, hence producing high saving and investment correlations (Summers (1988), Bayoumi (1990)). This is partly seen in splitting up total saving and investment into their private and government components. Regressing private investment on private saving gives substantially lower coefficients than regressing the total quantities on each other (Bayoumi (1990))
This paper considers the relationship between saving and investment within regions of individual countries where there is no targeting of the regional current account by government. It follows that the relationship between total saving and investment is expected to be less strong across regions than across nations, In fact, given that the financial and goods markets within individual developed countries are likely to be integrated, no relationship is expected between private saving and investment on a regional basis. This paper examines whether this is the case by considering the relationship between saving and investment across the regions of Canada, Great Britain, and Germany. 1/ To determine whether regions are constrained in their investment decisions, the paper considers first the aggregate relationship between total regional saving and investment. It then breaks down both aggregates into their private and government components and finally breaks down private saving into its personal and corporate components. The distinction between personal and corporate saving takes into account the recent literature on imperfections in capital markets which stresses the importance of the availability of credit in explaining business cycle fluctuations. Hubbard, Fazzari, and Petersen (1988) document that variables which proxy for the availability of liquidity (e.g., cash flow) have significant effects in standard investment equations and the importance of these variables depends on the magnitude of the liquidity constraint. 2/ The distinction between personal and corporate saving on a regional level attempts to discover whether some regions are isolated from the capital market centers and hence must place greater reliance on generating their own sources of finance.
The paper concentrates on the saving-investment relationship across regions rather than through time 3/ and therefore analyzes each relationship net of business cycle effects. Surprisingly, the relationship between total saving and investment is significantly negative in Great Britain and in Canada. 4/ By separating the private and government components of both variables, the negative relationship between total saving and investment is found to be heavily influenced by the government component. There is a strong negative relationship between government saving and investment in both countries which outweighs the significantly positive relationship found between private saving and investment in the U.K. and the insignificant relationship in Canada. This negative influence by the government on the regional saving-investment relationship is in complete contrast to its positive influence on the national relationship between saving and investment documented earlier. The private saving-investment relationship is heavily influenced by the corporate saving component because when both personal and corporate saving are analyzed separately, the corporate component has a significant positive influence on investment, whereas the effect of the personal saving component is insignificant in Britain and negative in Canada. For west Germany, both personal and corporate saving rates are insignificantly correlated with investment.
II. Data Sources
The lack of data on saving and investment across states in the United States has thwarted previous attempts to analyze correlations between saving and investment across regions. The most recent analysis (which dates back to 1965) aggregated the different individual components of saving and investment using many data sources. 5/ However, data do exist for Canada, Great Britain, and Germany and therefore this paper focuses on the relationships within these three countries. A list of all regions is given in Appendix I and detailed variable definitions and sources are given in Appendix II. 6/
The data are annual and cover the time period 1961–89 for Canada and 1971–87 for Britain. The data for Germany are only available for even years between 1970 and 1978 and each year from 1980 to 1987.
There are a few difficulties with the data. The first is that the personal saving value is calculated as the difference between personal disposable income and consumption and is therefore less precise than its two components individually. Secondly, the government expenditure and revenue items for the U.K. are not complete and therefore it must be assumed that the excluded categories are regionally distributed in the same proportion as the included categories. Finally, corporate saving is calculated on the basis of regional profits and is therefore perfectly correlated with profits. This leads to a difficulty in isolating the effects of liquidity on investment. The first two issues are not considered further in this paper but an attempt is made to identify liquidity effects.
The private investment values do not include all sectors in all three countries and therefore they are normalized by their respective sectoral GDP aggregates.
III. Total Saving and Investment Analysis
To obtain a rough graphical indication of the relationship between total regional savings and regional gross investment consider first Figures 1 through 7, which represent time averages of total, government and private saving and investment rates among Canadian and British regions.
Canadian Total Saving and Investment Rates
Citation: IMF Working Papers 1993, 062; 10.5089/9781451848212.001.A001
British Total Saving and Investment Rates
Citation: IMF Working Papers 1993, 062; 10.5089/9781451848212.001.A001
Canadian Government Saving and Investment Rates
Citation: IMF Working Papers 1993, 062; 10.5089/9781451848212.001.A001
British Government Saving and Investment Rates
Citation: IMF Working Papers 1993, 062; 10.5089/9781451848212.001.A001
Canadian Private Saving and Investment Rates
Citation: IMF Working Papers 1993, 062; 10.5089/9781451848212.001.A001
British Private Saving and Investment Rates
Citation: IMF Working Papers 1993, 062; 10.5089/9781451848212.001.A001
Canadian Private Saving and Investment Rates
Citation: IMF Working Papers 1993, 062; 10.5089/9781451848212.001.A001
Figures 1 and 2 present total saving and investment rates for Canadian and British regions. These figures indicate that the regions of both countries can be split up into two groups. The Atlantic provinces of Canada have higher investment rates and substantially lower saving rates than the other Canadian regions suggesting a negative relationship between total regional saving and investment in this country. Similarly for Great Britain, the regions of Wales, Scotland, and the North of England have shortfalls of saving which are financed by the surpluses of the other regions. These charts are striking because they are in complete contrast to the charts representing total saving and investment rates across countries. Cross country graphs which depict total national saving and investment usually lie along a 45° line (see Feldstein and Horioka (1980) or Bayoumi and Rose (1993)).
To ascertain the reason underlying the negative relationship between total regional saving and investment, total values were split into their government and private components. These relationships are presented in Figures 3 to 6.
Figures 3 and 4 suggest that the government saving and investment relationships have two distinct clusters in both countries. The poorer regions of both countries have high levels of government investment which are financed by the government surpluses of the richer regions. In Figure 3 the isolated Atlantic regions of Newfoundland, Prince Edward Island, Nova Scotia, and New Brunswick have substantially higher rates of government investment than the other regions and these regions have experienced consistent government deficits over the period. Not surprisingly the richer regions of Alberta, Ontario, and British Columbia have the lowest government investment rates and these high per capita regions have the largest government surpluses.
Similarly, in Britain government investment rates for Wales, Scotland and the North are much higher than the corresponding rates in other regions suggesting cross region subsidization. Figure 4 indicates that the presence of these three regions yields a strong negative relationship between regional government saving and investment rates. These three regions are among the lowest per capita income areas of Great Britain which explains their low government saving rates. In contrast, the South East stands out as a region with one of the lowest government investment rates but with a government saving rate substantially larger than any other region.
There is a close resemblance between Figures 1 and 2 and 3 and 4, which suggests that the relationship between total regional saving and investment is heavily influenced by government behavior. This is clarified when we consider Figures 5 and 6, which represent the relationship between private saving and investment across Canadian and British regions. In contrast to the negative relationship between regional government saving and investment, the relationship between regional private saving and investment is positive in both countries. Neither positive relationship is strong however.
With broad discussion of data as background, we can now turn to a more structured approach to the analysis and consider a regression relationship between total regional investment and total regional saving. Following the 1980 Feldstein-Horioka article, many papers have reinterpreted their findings by adding extra explanatory variables to proxy government influences, productivity shocks, etc. These analyses have used averages of national saving and investment rates and regressed the pooled saving and investment rates on each other. This averaging over time results in a considerable loss of degrees of freedom. Consequently, all the available information is used in this paper with a time dummy included for each year to take out aggregate business cycle effects.
Recently, a new literature has developed arguing that the correlation of saving and investment is an inappropriate indicator of capital mobility (e.g., Mendoza (1992)). The argument is that over the business cycle there is a high correlation between saving and investment in models which assume perfect capital mobility. However, this literature has ignored the fact that the original issue was posed using long-term averages of saving and investment to net out business cycle effects. Many models produce positive saving-investment correlation over the business cycle because of the motive of consumption smoothing, and this is why this paper attempts to filter out the cycle by including time dummies and output growth variables.
Obstfeld (1986) suggests that the reason why saving and investment are highly correlated across countries is that both variables have common influences which have been ignored in previous empirical analyses of the relationship. He suggests that output growth is a strong candidate as a common influence based on the life cycle theory of saving and the accelerator theory of investment. However, the empirical validity of the life cycle hypothesis has recently been questioned by Carroll and Summers (1989). They argue that saving rates fell after the productivity slowdown in 1973, in contrast to the standard implication of life cycle theory that savings should have risen to maintain a lower constant consumption level into the future. Deaton (1989) has developed a model which emphasizes the buffer stock motive for savings. He assumes that individuals are liquidity constrained and therefore when positive shocks (whether transitory or permanent) increase income, this results in an increase in the saving rate to buffer individuals against corresponding negative shocks in future. Therefore although the theoretical justification for a positive relationship between income growth and the saving rate is subject to dispute, the empirical evidence is strong. The same can also be said for the relationship between income growth and investment. Regional income growth also approximates regional productivity growth which is expected to be related to retained profits. For the sake of completeness, results are therefore presented from regressions which include and exclude lags of output growth. The aggregate regression equation between total saving and investment excluding output is as follows:
where Ip is private investment, Ig is government investment, TSp is total private saving, TSg is total government saving, Y is GDP and tt is an annual time dummy. The alternative regression equation also includes real output growth terms (YR is real GDP). 7/
One basic difficulty in estimating saving and investment relationships is simultaneity bias. Saving and investment are plausibly influenced by many common exogenous shocks, proxies for all of which cannot be found. This suggests using instruments and including in the regression all variables which are expected to have common effects. The instruments used for the total savings variable are the regional participation rate, the proportion of individuals between 15 and 44 in each region, the proportion of retirees, the regional ratio of social security benefits to total income and real wage innovations. 8/ It seems reasonable to suppose that the first three variables influence saving and have no obvious effect on investment. The social security ratio is introduced as an instrument for government saving and wage innovations represent an instrument for corporate saving. 9/ These instruments are used to estimate each equation by 2SLS.
Table 1 confirms the visual impression gained from Figures 1 and 2 in that the total regional saving coefficient is significantly negative in both countries. The inclusion of output growth terms has only minor effects on the results. This is not inconsistent with the buffer stock theory because it postulates a positive correlation between output growth and personal saving and there is a noticeable change in coefficients when personal saving is used.
The Relationship Between Total Regional Investment and Total Regional Savings
The Relationship Between Total Regional Investment and Total Regional Savings
| Great Britain | Canada | |||
|---|---|---|---|---|
| 1) β | −0.5596 | (−4.2952) | −0.1015 | (−4.2472) |
| 2) β | −0.5689 | (−4.3628) | −0.1044 | (−4.6574) |
The Relationship Between Total Regional Investment and Total Regional Savings
| Great Britain | Canada | |||
|---|---|---|---|---|
| 1) β | −0.5596 | (−4.2952) | −0.1015 | (−4.2472) |
| 2) β | −0.5689 | (−4.3628) | −0.1044 | (−4.6574) |
This section proceeds to separate the influences of the government and private sector to examine whether either sector has a dominant effect on the total saving-investment relationship. Figures 3 and 4 suggest that there is a similar negative relationship between government saving and investment as between total saving and investment. This relationship is considered below in regressions which again include and exclude real output growth. Total private saving is added as an additional explanatory variable to discover whether shortfalls in government saving are compensated by corresponding increases in private saving (a form of regional Ricardian equivalence). Since a considerable amount of regional government expenditure is not regionally financed and individuals are mobile across regions, the null hypothesis is that no relationship holds.
The above table confirms the visual impression of Figures 3 and 4, namely that regions with high government saving rates have low government investment rates, ceteris paribus. The insignificant coefficient on private saving indicates that private saving does not adjust to compensate for changes in government saving across regions.
The relationship between private saving and investment is now considered to examine whether the slight positive relationships apparent in Figures 5 and 6 appear in regression analysis. The estimated equations are as follows:
and the results are shown below.
Table 3 demonstrates the presence of a significant positive relationship between private saving and investment in Great Britain and an insignificant relationship elsewhere. This indicates that the negative relationship between total aggregate saving and investment is due to the dominant behavior of the government sector over its private counterpart. This behavior of government among regions contrasts sharply with its behavior emphasized in relationships across countries (Bayoumi (1990)). In cross country analysis the presence of a government sector tightens the positive relationship between saving and investment because of concerns of financing large current account deficits. In this cross-regional analysis, the presence of regional government saving and investment differences weakens the relationship between private saving and investment. The positive relationship between private saving and investment in Britain is also contrary to conventional wisdom if one believes that there is high capital mobility across regions of individual countries and that saving coefficients in a regression such as (5) indicate the presence or absence of capital mobility.
The Relationship Between Regional Government Investment and Private and Government Saving
The Relationship Between Regional Government Investment and Private and Government Saving
| Great Britain | Canada | ||||
|---|---|---|---|---|---|
| 3) | β0 | -0.0250 | (-0.5383) | 0.0485 | (1.1097) |
| β1 | -0.2578 | (-7.4199) | -0.1281 | (-8.1275) | |
| 4) | β0 | -0.0229 | (-0.4895) | 0.0582 | (1.3338) |
| β1 | -0.2593 | (-7.3847) | -0.1319 | (-8.4060) | |
The Relationship Between Regional Government Investment and Private and Government Saving
| Great Britain | Canada | ||||
|---|---|---|---|---|---|
| 3) | β0 | -0.0250 | (-0.5383) | 0.0485 | (1.1097) |
| β1 | -0.2578 | (-7.4199) | -0.1281 | (-8.1275) | |
| 4) | β0 | -0.0229 | (-0.4895) | 0.0582 | (1.3338) |
| β1 | -0.2593 | (-7.3847) | -0.1319 | (-8.4060) | |
The Relationship Between Regional Private Investment and Regional Private Saving
The Relationship Between Regional Private Investment and Regional Private Saving
| Great Britain | Canada | West Germany | |||||
|---|---|---|---|---|---|---|---|
| 5) | β | 0.3299 | (2.6047) | -0.0416 | (-0.6164) | -0.0695 | (-0.6522) |
| 6) | β | 0.3795 | (2.9361) | -0.0387 | (-0.6038) | -0.0687 | (-0.6558) |
The Relationship Between Regional Private Investment and Regional Private Saving
| Great Britain | Canada | West Germany | |||||
|---|---|---|---|---|---|---|---|
| 5) | β | 0.3299 | (2.6047) | -0.0416 | (-0.6164) | -0.0695 | (-0.6522) |
| 6) | β | 0.3795 | (2.9361) | -0.0387 | (-0.6038) | -0.0687 | (-0.6558) |
IV. Private Saving and Investment Analysis
To clarify the extent of capital mobility across regions, private saving is decomposed into personal and corporate saving. The relationships between these variables and private investment are presented in Figures 7 through 12.
Figure 7 indicates a negative relationship between personal savings and the investment rate for Canadian regions. The isolated Atlantic provinces (Newfoundland, Nova Scotia, New Brunswick, and Prince Edward Island) have negligible savings whereas the regions with relatively high personal saving rates and low private investment rates are the populated areas of Quebec, Ontario, and to a lesser extent Manitoba. In Figure 8 a positive relationship between the corporate saving rate and the investment rate is evident when Newfoundland and Alberta are included.
Canadian Private Saving and Investment Rates
Citation: IMF Working Papers 1993, 062; 10.5089/9781451848212.001.A001
For Great Britain (Figure 9) there is no obvious relationship between personal savings and investment. The significant outlier is the South East with a low savings rate relative to the other regions. However, the relationship between corporate savings and investment appears stronger (Figure 10), once again influenced by specific regions. In this case Wales, Scotland, and the North of England influence the correlation, having high investment rates and high corporate saving rates. Mason (1989) notes that these regions have had difficulty in attracting venture capital investors from elsewhere and this issue is considered in more detail in the final section. Figures 11 and 12 suggest no relationship between personal and corporate savings and investment in west Germany.
British Private Saving and Investment Rates
Citation: IMF Working Papers 1993, 062; 10.5089/9781451848212.001.A001
British Private Saving and Investment Rates
Citation: IMF Working Papers 1993, 062; 10.5089/9781451848212.001.A001
German Private Saving and Investment Rates
Citation: IMF Working Papers 1993, 062; 10.5089/9781451848212.001.A001
German Private Saving and Investment Rates
Citation: IMF Working Papers 1993, 062; 10.5089/9781451848212.001.A001
Turning to the regression analysis, estimates of equations (7) and (8) are presented in Table 4.
The Relationship Between Private Investment and Personal and Corporate Saving
The Relationship Between Private Investment and Personal and Corporate Saving
| Great Britain | Canada | West Germany | |||||
|---|---|---|---|---|---|---|---|
| 7) | β0 | 0.2821 | (1.2431) | -0.5560 | (-5.2645) | -0.2280 | (-1.2779) |
| β1 | 0.5721 | (1.8067) | 0.7328 | (6.2299) | 0.4047 | (0.8501) | |
| 8) | β0 | 0.1826 | (0.6959) | -0.5473 | (-4.9300) | -0.1724 | (-1.0061) |
| β1 | 0.9154 | (2.3301) | 0.7141 | (5.7266) | 0.2236 | (0.4827) | |
The Relationship Between Private Investment and Personal and Corporate Saving
| Great Britain | Canada | West Germany | |||||
|---|---|---|---|---|---|---|---|
| 7) | β0 | 0.2821 | (1.2431) | -0.5560 | (-5.2645) | -0.2280 | (-1.2779) |
| β1 | 0.5721 | (1.8067) | 0.7328 | (6.2299) | 0.4047 | (0.8501) | |
| 8) | β0 | 0.1826 | (0.6959) | -0.5473 | (-4.9300) | -0.1724 | (-1.0061) |
| β1 | 0.9154 | (2.3301) | 0.7141 | (5.7266) | 0.2236 | (0.4827) | |
where PS/Y refers to the personal savings rate and CS/Y refers to the corporate savings rate.
The above table yields insight on the issue of capital mobility which arose from Table 3. The fact that the personal savings coefficient is insignificant in Britain and Germany although negative in Canada suggests that the savings of individuals is mobile across regions. 10/ The existence of nationwide branching of banks in these countries obviously contributes to the mobility of savings. The opposite signs of the personal saving and corporate saving coefficients for Canada and Germany also suggests that there is no veil between households and firms in these countries. When output growth terms are introduced the only noticeable effect is that the corporate saving coefficient for Britain increases from 0.5721 to 0.9154.
An alternative test of capital mobility was conducted using the Euler equation approach. According to this approach, differences in consumption growth across regions should approximate a random walk if capital is mobile. Therefore a test was conducted to discover whether the deviation of consumption growth in each region from its national average could be forecast using past values of explanatory variables. The explanatory variables used in the analysis were two period lags of the consumption growth differential and the disposable income differential. 11/ For Canada and Germany, the null hypothesis that the explanatory variables were insignificant was accepted, but this was not the case for the United Kingdom. This test therefore strengthens the mobility results documented above for Canada and Germany but leaves the extent of capital mobility in the United Kingdom an open question.
The significance of the corporate saving coefficient requires mention. The significance of the coefficient is evident in Canada and Britain whether output growth terms are included or not (although not in Germany). The issue to consider is how to interpret its significance. In the introduction it was mentioned that part of the paper was motivated to examine whether there are significant liquidity effects across regions. Liquidity effects are linked with significant corporate saving coefficients because if equities, bonds and retained earnings are perfect substitutes, one would not expect to find a relationship between retained earnings and private investment. There is a major difficulty, however, in ascribing liquidity effects to the significant corporate saving coefficient. This is because the positive correlation between retained earnings and private investment may represent endogenous responses of both of these variables to shocks which have not been controlled for in the regressions analyzed. I have attempted to counter this critique by including real wage innovations as an instrument assuming that real wage innovations affect profits and are not immediately affected by investment.
An alternative strategy to capture liquidity effects across regions is to split the data into two groups along regional lines. As long as the problem of endogeneity is of a similar magnitude across the two groups, the estimated difference in the coefficients across groups is an unbiased estimate of the true difference. This method of bypassing potential endogeneity problems has frequently been applied to the relationship between corporate structure and liquidity (e.g., Hubbard, Fazzari, and Petersen (1988)).
Newfoundland and Alberta appear to influence the positive long-run relationship between private investment and corporate saving in Canada and the peripheral regions of Scotland, Wales, and the North have similar effects in Great Britain. If firms in these regions have greater difficulty in obtaining external funds than those in other regions, variations in corporate savings should have greater effects on investment in these regions than elsewhere.
This method of isolating liquidity effects is applied by estimating equation (9) which is a standard accelerator investment equation except for the exclusion of Tobin's q which is not available across regions. 12/ The dependent variable is the annual change in regional investment and explanatory variables are made up of contemporaneous 13/ and lagged changes in regional gross profit (GP) and regional and national real GDP. 14/ Time dummies are excluded to exploit the time dimension of the data, and separate coefficients on gross profits for peripheral (potentially liquidity constrained) and central regions are estimated. 15/
The major difference in behavior between the peripheral and central regions of Great Britain is that the second lag of gross profits is significant for the former regions and insignificant for the latter. This suggests that changes in regional grossprofits have greater effects on the investment behavior of the peripheral regions. However this difference is insignificant at the 5 percent level. For Canada no variable is significant. This test suggests that liquidity effects are not evident across regions.
The Relationship Through Time Between Private Investment and Corporate Saving in Great Britain and Canada
The Relationship Through Time Between Private Investment and Corporate Saving in Great Britain and Canada
| Great Britain | Canada | |||
|---|---|---|---|---|
| β0 | -0.1586 | (-0.5058) | 1.1092 | (0.3632) |
| βp1 | 0.0678 | (0.3798) | -0.0665 | (-0.2186) |
| βp2 | 0.2201 | (2.1040) | 0.4776 | (0.5547) |
| βc1 | 0.0102 | (0.0575) | -0.4521 | (-1.2779) |
| βc2 | 0.1165 | (1.0901) | 0.2008 | (0.2366) |
| γ0 | 0.6154 | (1.0976) | 0.7709 | (0.4078) |
| γ1 | 0.4896 | (0.7708) | -0.0463 | (-0.0921) |
| γ2 | -0.3568 | (-1.1522) | -0.2828 | (-0.2007) |
| δ0 | 0.5452 | (0.9736) | -1.3485 | (-0.3247) |
| δ1 | 1.0042 | (2.4339) | 1.5773 | (1.1115) |
| δ2 | -0.4155 | (-0.7503) | -1.5528 | (-1.4690) |
The Relationship Through Time Between Private Investment and Corporate Saving in Great Britain and Canada
| Great Britain | Canada | |||
|---|---|---|---|---|
| β0 | -0.1586 | (-0.5058) | 1.1092 | (0.3632) |
| βp1 | 0.0678 | (0.3798) | -0.0665 | (-0.2186) |
| βp2 | 0.2201 | (2.1040) | 0.4776 | (0.5547) |
| βc1 | 0.0102 | (0.0575) | -0.4521 | (-1.2779) |
| βc2 | 0.1165 | (1.0901) | 0.2008 | (0.2366) |
| γ0 | 0.6154 | (1.0976) | 0.7709 | (0.4078) |
| γ1 | 0.4896 | (0.7708) | -0.0463 | (-0.0921) |
| γ2 | -0.3568 | (-1.1522) | -0.2828 | (-0.2007) |
| δ0 | 0.5452 | (0.9736) | -1.3485 | (-0.3247) |
| δ1 | 1.0042 | (2.4339) | 1.5773 | (1.1115) |
| δ2 | -0.4155 | (-0.7503) | -1.5528 | (-1.4690) |
V. Conclusion
This paper has considered the relationship between saving and investment across the regions of Canada, Germany, and Great Britain and has found a number of empirical regularities: (1) no positive relationship exists between regional personal saving and private investment in the three countries studied. This suggests that personal savings are perfectly mobile across regions, although the Euler equation approach rejects this finding for Britain. And (2) there is a positive relationship between regional retained profits and private investment in Britain and Canada, which may be indicative of the presence of capital market imperfections. However when regions are separated on the basis of contributing to the long-run relationship between investment and corporate saving, gross profits are not found to influence investment significantly differently between these regions.
There is a negative relationship between government saving and government investment which is the cause of the negative relationship between total saving and investment across regions. This implies that government behavior in a regional context contrasts sharply with its behavior in a national context. In cross-country analysis the presence of a government sector tightens the positive relationship between saving and investment because of concerns of financing large current account deficits. By contrast, in this cross regional analysis the presence of regional government saving and investment differences weakens the relationship between private saving and investment.
APPENDIX I
Region List
Region List
| Great Britain | Canada | West Germany |
|---|---|---|
| North | Newfoundland | Schleswig-Holstein |
| Yorks+Humbs | Prince Edward Island | Niedersachen |
| East Midlands | Nova Scotia | Nordrhein-Westfalen |
| East Anglia | New Brunswick | Hessen |
| South East | Quebec | Rheinland-Pfalz |
| South West | Ontario | Baden-Wurtemberg |
| West Midlands | Manitoba | Bayern |
| North West | Saskatchewan | Saarland |
| Wales | Alberta | Hamburg |
| Scotland | British Columbia | Bremen |
| West Berlin |
Region List
| Great Britain | Canada | West Germany |
|---|---|---|
| North | Newfoundland | Schleswig-Holstein |
| Yorks+Humbs | Prince Edward Island | Niedersachen |
| East Midlands | Nova Scotia | Nordrhein-Westfalen |
| East Anglia | New Brunswick | Hessen |
| South East | Quebec | Rheinland-Pfalz |
| South West | Ontario | Baden-Wurtemberg |
| West Midlands | Manitoba | Bayern |
| North West | Saskatchewan | Saarland |
| Wales | Alberta | Hamburg |
| Scotland | British Columbia | Bremen |
| West Berlin |
APPENDIX II Data Sources
Canada
The Canadian regional data source is the Income and Expenditure Accounts Division at Statistics Canada.
Private investment: coverage of all sectors of the economy
Government investment: coverage of all sectors of the economy
Personal income: aggregation of wages, salaries, transfers, dividends and net interest. Dividends and net interest are imputed according to information obtained from individual tax returns. 16/ Information on rent payments is obtained from the Labor Force Survey which, when multiplied by the total provincial stock of dwellings from the Census, gives provincial gross rent payments. The Census is taken every four years with annual updates on the basis of building permits. Imputed rent for owner occupiers is calculated in a similar way, supplemented by information from the Census on the characteristics of houses.
Income Taxation: Regional taxation totals are available from Revenue Canada
Consumption: The Retail Commodity Survey, the Family Expenditure Survey and provincial surveys on specific items provide detailed regional consumption data.
Personal Saving = Personal Income-Income Taxation-Consumption
Gross profits: Gross profits in Canadian manufacturing, construction and mining are first calculated nationally and are allocated among provinces using value added figures from annual provincial surveys on an establishment basis. These surveys rule out the problem of determining which province to allocate the profits of multi-establishment businesses. For the other sectors the national profit totals are divided among provinces in proportion to provincial taxable income, wages, and salaries.
Corporate Saving: This is calculated by using regional gross profits as weights in attributing the national retained profit amount to the separate regions. This manipulation assumes that dividend payments, corporate taxes and interest have the same regional weight as the retained profit total.
Government saving: complete coverage
Regional output: Regional GDP.
Great Britain
The British regional data are obtained from two sources: Regional Trends and Financial and General Rating Statistics.
Private investment: coverage of all categories except construction, shipping and services.
Government investment: coverage of all categories
Personal income: aggregation of wages, salaries, transfers, dividends and net interest. The “net investment income” component of personal income is provided by the Internal Revenue Service, rent paid to landlords is obtained from the Family Expenditure Survey (FES) and imputed rent for owner occupiers makes use of measures of domestic rateable values from the FES and house price changes in the open market.
Income Taxation: Regional tax payments are allocated using estimates of tax payments provided by the Survey of Personal Incomes
Consumption: The FES provides estimates of household expenditure by detailed category based on a regionally balanced sample.
Personal Saving = Personal Income-Income Taxation-Consumption.
Gross profits: Gross profits in the British manufacturing sector are obtained from estimates of net output provided by the Annual Census of Production. Profits for all the other sectors are allocated in proportion to regional employment.
Corporate saving: as for Canada.
Government revenue: The combined contribution of regional personal tax payments and corporate tax payments. The regional corporate tax payments are constructed using regional profits as weights. The construction of this variable assumes that the regional distribution of indirect taxes and other forms of government revenue are equivalent to those of personal and corporate tax payments.
Government expenditure: A direct estimate of total regional government expenditure for the British regions is unavailable, but the health, social security, and local authority components are available. These components of expenditure have been used to attribute regional weights to the national government expenditure value. The three categories make up 60 percent of total government expenditure on goods and services; therefore assuming the other categories follow the same regional pattern, it represents a reasonable proxy.
Regional output: Regional GDP.
Germany
The German data were obtained from Volkswirtschaftliche Gesamtrechnungen, Statistisches Bundesamt.
No data are available on regional government saving and investment in Germany and therefore the analysis is restricted to the private component.
Private investment: only covers the manufacturing sector and is only available for even years between 1970 and 1978 and each year from 1980 to 1987.
Personal Saving = Personal Income-Income Taxation-Consumption.
Corporate saving: This is computed by attributing the national value to regions on the basis of regional capital income. It is therefore assumed that rent payments in addition to dividends and net interest have the same regional weight as retained profits.
Regional output: Regional GDP.
References
Bayoumi, T., “Saving Investment Correlations: Immobile Capital, Government Policy or Endogenous Behavior?,” Staff Papers (Washington: International Monetary Fund, 1990).
Bayoumi, T., and A. Rose, “Domestic Savings and Intranational Capital Flows,” forthcoming European Economic Review (1993).
Carroll, C., and L. Summers, “Consumption Growth Parallels Income Growth: Some New Evidence,” NBER Working paper No. 3090 (1989).
Deaton, A., “Saving and Liquidity Constraints,” mimeo (1989).
Feldstein, M., and C. Horioka, “Domestic Saving and International Capital Flows,” Economic Journal, Vol. 90 (1980).
Feldstein, M., and C. Horioka, and P. Bacchetta, “National Saving and International Investment,” NBER Working Paper No. 3164 (November 1989).
Hubbard, Fazzari, and B. Petersen, “Financing Constraints and Corporate Investment,” Brookings Papers on Economic Activity, 1988:1.
Mason, C., “Venture capital in the UK - a geographical perspective,” Nat West Bank Quarterly Review (1989).
Mendoza, Enrique G., “Robustness of Macroeconomic Indicators of Capital Mobility” (Washington, International Monetary Fund), Working Paper No. WP/92/111 (1992).
Obstfeld, M., “Capital Mobility in the World Economy: Theory and Measurement” (Carnegie-Rochester Conference Series on Public Policy), Vol. 24 (Spring 1986).
Romans, J.T., “Capital Exports and Growth Among US Regions” (1965).
Summers, L., “Tax Policy and International Competitiveness,” International Aspects of Fiscal Policies (1988).
I wish to thank Olivier Blanchard, Paul Krugman, Tamim Bayoumi, Paul Masson, Peter Clark, Simon Gilchrist, and Donogh Mcdonald for comments and the MIT International workshop and the Federal Reserve Board for the opportunity of presenting previous versions. The paper represents the final chapter of my dissertation at M.I.T.
The data only extend to 1987, so it is west Germany which is analyzed.
They find different cash flow coefficients across groups of firms classified by their dividend behavior.
This distinction has often been neglected in the literature.
Data are only available on private saving and investment for west Germany.
The area incorporating Yukon and the North West territories in Canada has been left out of the analysis because this area has a private investment rate of 0.55 compared to the regional average of 0.22. Similarly, in this area the government investment rate is 0.22 compared to a regional average of 0.05. If included, its outlying position would dominate the results.
The national GDP deflator is used to convert nominal regional GDP values into real values.
The real wage innovations are the residuals from a regression of the real wage on itself lagged and on regional dummies.
It may be argued that real wage innovations are not a valid instrument because they are correlated with investment. It is extremely difficult to think of any regional variable which is expected to shift corporate saving but not shift investment. In the sample the correlation betweenreal wage innovations and the investment rate is zero and therefore it seems the best instrument available.
Bayoumi and Rose (1993) report similar results for the British personal saving rate.
Two period lags were used because of the potential of first-order autocorrelation in the use of time averaged variables.
This is defined as the ratio of the market value of firms to the total replacement value of capital in each region.
Lagged values of profits and output are used as instruments for the contemporaneous value.
All level terms of variables are expressed in logarithms.
British peripheral regions are the North, Scotland and Wales. The Canadian peripheral regions are the Atlantic provinces and Alberta.
A check on the quality of deposit interest data is made by comparing the total with bank liabilities and trust and mortgage loans across provinces.
