Fear of nuclear war and intercountry differences in the rate of saving.

Slemrod, Joel

FEAR OF NUCLEAR WAR AND INTERCOUNTRY DIFFERENCES IN THE RATE OF SAVING

JOEL SLEMROD [*]

I. INTRODUCTION AND REVIEW OF PREVIOUS LITERATURE

Can differences in the perceived likelihood of nuclear war explain intercountry differences in saving behavior? The analysis presented in this paper, based on recent international surveys of attitudes concerning the likelihood of world war, suggests that the answer to this question is yes. Holding other determinants of saving constant, a country's saving rate is lower the greater is the fraction of its population that believes a world war is imminent.

The hypothesis that changes in public perceptions of the likelihood of a nuclear war influence the rate of saving, first proposed in Slemrod [1982], is consistent with the behavior of an expected utility maximizing consumer faced with a changing probability of the survival of oneself and one's heirs. Whether the relationship is of quantitative significance is another question. The evidence examined to date, while not conclusive, suggests that it may be. Slemrod [1986] shows that, holding other determinants of saving constant, two separate indices of fear of nuclear war have a statistically significant negative correlation with the U.S. net private saving rate between 1948 and 1984. The analysis suggests that a diminution in the fear of war from its highest postwar level to its lowest postwar level is associated with an increase of 1.7 percentage points in the net private saving rate, compared to an average value of 10.6 over this period. Note that during this period as many as 53 percent of Americans believed that the chances of a nuclear war occurring within ten years were at least "50-50". Hendershott and Peek [1985, 1989a], using several alternative definitions of savings and investigating other influences on saving, also find that increased fear has tended to reduce saving in the postwar U.S. economy. They also (Hendershott and Peek [1989b]) contend that changes in nuclear fear are an important factor in the movement of real interest rates in the 1970s and 80s. While Russett and Lackey [1987] find no consistent relationship between savings fear of war at the aggregate level for several advanced countries, their analysis of individual data from the U.S. National Election Survey indicates that, for individuals who did some saving, actual saving is negatively related to fear of nuclear war. They conclude that the most appropriate data set is consistent with the hypothesis that fear of nuclear war reduces saving. [1]

II. METHODOLOGY, DATA AND RESULTS

In recent years the Gallup International Research Institutes have annually conducted a poll in as many as thirty-three countries concerning attitudes about the likelihood of world war. [2] The poll in each country asks a random sample of individuals to assess on a zero to ten scale the likelihood of a world war breaking out in the next ten years. [3]

Table I ranks the countries surveyed in 1986 according to the fraction or respondents who felt that the chance of world war was 50 percent or greater within the next ten years. Most striking to a student of saving behavior is that the U.S. stands at the top of the list, with 49 percent of the respondents indicating at least a 50-50 chance of a world war occurring within ten years. Also of special interest is that Japan, which has had exceptionally high saving rates, is near the bottom of the list with only 15 percent of those interviewed professing to a high fear of world war.

Table : TABLE I Fraction of Respondents Saying in 1986 that the Chance of World War

 Within Ten Years is 50 Percent or Greater, by Country
United States 49 Italy 22
South Africa (blacks) 49 Italy 22
Ecuador 45 Norway 21
Chile 43 Austria 20
Colombia 42 Great Britain 20
Australia 38 South Korea 19
Uruguay 38 Denmark 18
Brazil 34 Greece 18
Canada 34 Luxembourg 18
South Africa (whites) 33 Switzerland 18
India 30 West Germany 18
Argentina 30 Finland 17
Philippines 27 Hongkong 17
Ireland 25 Japan 15
Portugal 25 Sweden 15
Belgium 24 Turkey 15
France 24 Netherlands 14

Source: The Gallup Poll, released January 11, 1987.

The analysis that follows compares the survey responses concerning the likelihood of a world war to average saving rates in nineteen of the thirty-three countries listed in Table I over the period 1981 to 1984. These nineteen are the OECD (Organization of Economic Cooperation and Development) member countries both covered by the Gallup survey and for which data on net private savings are available. [4] The sample was restricted to OECD member countries because standardized and relatively reliable national income data are available, and because of the difficulty in comparing the characteristics of countries in radically different stages of development. The sample period of 1981 to 1984 was chosen because it was the most recent period for which both saving and survey data are available. It is appropriate to consider data averaged over several years because it minimizes the importance of cyclical factors irrelevant to the hypothesis being studied.

Table II presents the average net private saving rate and an index of the average perceived likelihood of world war over the period 1981 to 1984 for these nineteen countries, [5] and Figure 1 graphs these data. The points corresponding to the six largest countries in the sample are highlighted. The data for these six countries, in particular, seem to indicate a negative association between the net private saving rate and the index of fear of war. [6]

Table : TABLE I Average Net Private Saving Rate and the Index of Perceived Likelihood of Nuclear War for Nineteen OECD Countries, 1981-1984

 Average Net Index of Perceived
Country Private Saving Rate Likelihood of Wood War(*)
Australia 5.7 52.3
Austria 13.7 28.0
Belgium 16.0 34.2
Canada 17.5 46.0
Denmark 13.1 29.7
Finland 11.1 26.8
France 11.5 39.9
Germany, F.R. of 11.7 33.8
Greece 18.5 18.2
Ireland 17.9 39.0
Italy 23.4 28.4
Japan* 19.3 (1) 38.0 (2) 29.4
Netherlands 17.8 33.2
Norway 12.7 30.7
Spain 11.0 41.8
Sweden 9.7 30.3
Switzerland 18.9 33.9
United Kingdom 10.5 31.8
United States 10.4 54.5

This apparent association is confirmed in a weighted least squares regression explaining the average net private saving rate expressed as a percentage, S/Y, with a constant and the index of the likelihood of war, denoted WAR. The observations are weighted by the square root of the country's population. [7] This procedure yields [Mathematical Expression Omitted] where standard errors are in parentheses below the estimated coefficients, and the value of [R.sup.2] is based on population-weighted residuals.

As Figure 1 suggests, there is a statistically significant negative correlation between the net private saving rate and the likelihood of war +index. The regression line corresponding to equation (1) is shown in Figure 1. The magnitude of the estimated coefficient indicates that a decline of ten points in the likelihood of war index would increase a country's net private saving rate by slightly more than 2.5 percentage points.

The analysis is next expanded to include other determinants of a country's saving rate. Several studies of intercountry differences in saving behavior exist, most notably Houthakker [1961; 1965], Modigliani [1970], Feldstein [1977; 1980], Barro and MacDonald [1979], Kopits and Gotur [1980], Modigliani and Sterling [1983], and Horioka [1986]. Based on the life-cycle model of saving, these authors examine the effect on saving of both demographic factors and government policies such as the social security program. These studies differ in variable definitions, data sources, the sample of countries, time period, and specification, and the results are not entirely consistent across studies.

The basis for studying the effect on saving of fear of nuclear war is an updated version of the savings function estimated by Feldstein [1980], which is representative of the other studies' methodology. I use this specification to investigate the effect of introducing into the equation the index of the perceived likelihood of war. Feldstein estimated

(*) Professor of Economics, Professor of Business Economics and Public Policy, and Director of the Office of Tax Policy Research, The University of Michigan. Helpful comments on an earlier draft were provided by Patric Hendershott, Jonathan Skinner, an anonymous referee, and participants at the University of Minnesota Applied Microeconomics Workshop. I owe special thanks to Charles Horioka for generously providing data used in this paper, and to Jennifer Sumbler and Michael Orszag for able research assistance.

(*) Data for 1981 is unavailable for Denmark, Finland and Norway, and is unavailable for 1982 and 1984 for Austria. The 1981-1984 averages for these countries are calculated using for the missing values extrapolated figures based on overall annual trends.

(**) See text for explanation of two war indices for Japan. the following equation on a sample of twelve OECD countries, using data from the 1950s:

[(S/Y).sub.t] = [Beta.sub.0] + [Beta.sub.1] [G.sub.t] + [Beta.sub.2] [AGE.sub.t] + [Beta.sub.3] [DEP.sub.t] + [Beta.sub.4] [(B/E).sub.t] + [B.sub.5] [LPAGED.sub.t] + [u.sub.t], (2)

where G is the growth rate of total real private national income, AGE is the ratio of the number of retirees aged sixty-five or over to the population aged twenty to sixty-five, DEP is the ratio of the number of persons under twenty to the working age population, B/E is the benefit replacement ratio of the social security program, and LPAGED is the labor force participation rate of men sixty-five or older. There principal changes were made to Feldstein's analysis. First, the data are updated to reflect more recent conditions. Second, due to data contstraints, the measure of the generosity of social security is the ratio of public pension benefits per person over sixty-five to per capita private national income (denoted SOCSEC), instead of the benefit replacement ratio. Finally, the sample of countries studied is expanded from twelve to nineteen.(8)

The results of estimating this saving equation with and without WAR as an explanatory variable are displayed in Table III. The second column of Table III shows that, without WAR, the explanatory variables are not particularly successful in explaining intercountry differences in net private saving rates. The rate of income growth, G, is positively associated with saving, in agreement with most previous studies, although its estimated coefficient is significantly different than zero only at the 90 percent confidence level. Neither of the estimated coefficients on the variables reflecting the age structure of the population, AGE and DEP, is significantly different than zero. Most previous studies found both to be negatively associated with the saving rate. The estimated coefficient on the social security variable, SOCSEC, is positive but not significantly different from zero. This finding is at odds with the negative coefficient found by Feldstein [1977;1980], although neither Barro and MacDonald [1979] nor Modigliani and Sterling [1983] corroborated this result. The estimated coefficient on LPAGED is positive (though not significant), failing to support the prediction of the life-cycle theory and in contrast with earlier empirical results. In sum, while Feldstein [1980] found all five coefficients to be statistically significant and consistent with the qualitative predictions of the extended life cycle model, these same conclusions do not follow from this updated and slightly revised version of the same model.(9)

Table : TABLE III Weighted Ordinary Least Squares Regressions Explaining the Net Private Saving Rate (S/Y) With and Without a Measure of the Perceived Likelihood of World War Within Ten Years

Independent Variable

WAR -0.416
 (0.087)
G 1.76 2.79
 (1.24) (1.99)
AGE -1.06 -0.239
 (0.35) (0.506)
DEP (0.0712) 0.0883
 (0.221) (0.362)
SOCSEC -0.00320 0.0744
 (0.0514) (0.0799)
LPAGED -0.128 0.133
 (0.119) (0.173)
CONSTANT 50.3 2.32
 (21.3) (30.79)
[R.sup.-2] 0.893 0.714

Standard errors are in parentheses. Observations are weighted by the square root of the country's population in 1980. Definition of variables(*):

S/Y: Average net private saving rate, 1981-1984.

WAR: Average index of perceived likelihood of world war, 1981-1984.

G: Growth in real private income per capita, 1976-1984.

AGE: Ratio of population aged 65 or over to population aged 20-64, 1980.

SOCSEC: Public pension benefits per person over 65 as a ratio of per capita private national income, 1980.

LPAGED: Labor-force participation rate of males aged 65 or over, 1975.

(*) All variables are defined in percentage points. A detailed data appendix specifying definations and sources in available fromthe author.

When WAR is included in the regression equation, its estimated coefficient is negative and statistically significant and is therefore consistent with the hypothesis that increased fear of war reduces a country's rate of saving. The magnitude of the estimated effect is large, indicating that an increase of 10 percent in the fraction of the population that believes a world war is likely is associated with a decline of 4.16 percentage points in the net private saving rate. Including this variable also changes the sign of the estimated coefficients on LPAGED and SOCSEC from a positive to the negative coefficient found by Feldstein. With WAR included, all of the estimated coefficients except for DEP have the same sign as estimated by Feldstein, although only the coefficients on G and AGE are significantly different than zero at a 90 percent level of confidence.

The effect on saving of fear of war estimated here is apparently much larger than the effect estimated in Slemrod [1986] using U.S. time-series data. In the earlier paper, the setting of the "doomsday clock" of the Bulletin of the Atomic Scientists was the principal index of nuclear fear. Since 1984, the clock has been moved from three minutes to midnight back to ten minutes to midnight in 1990. According to Slemrod [1986], this change would be associated with an increase in the net private saving rate of 1.2 percentage points. During the same period, the fear index for the U.S. based on the Gallup polls has fallen from its 1981-84 average of 54.5 to 30.2 in December of 1989. According to this paper, such a decline in the fear of nuclear war would be associated with an increase in the net private saving rate of 10.1 percentage points. I am inclined to believe the latter is a substantial overestimate of the true impact that has arisen because, in the cross-country regression analysis, the variation in fear of war is picking up cross-country differences in the propensity to save that are not well captured by the other explanatory variables included in the regression equation.

Several variations of the basic estimation strategy were investigated to test the robustness of the finding with respect to changes in specification. Weighting the observations by population (as done by Feldstein [1977; 1980] and Horioka [1986]) rather than the square root of population increased the absolute magnitude of the coefficient on WAR to 0.482 and substantially decreased (to 0.051) its standard error.(10) Estimating the equation with unweighed ordinary least squares does, however, slightly weaken the estimated association between saving and the perceived likelihood of world war. The estimated coefficient on the WAR variable becomes -0.331 with a standard error of 0.152.

The negative relationship between S/Y and WAR survives the addition of several other potential influences on saving. One of special interest is the level of real national income which, in the absence of the WAR variable, has a significant negative association with the saving rate. When WAR is included, this association disappears and WAR retains a significant negative association. Including as an explanatory variable the fraction of gross national product devoted to military spending did not affect the main result, and it attracted a negative sign not significantly different than zero. Finally, the analysis was repeated excluding the U.S. from the sample. The estimated coefficient on WAR increased (in absolute value) to -0.581, although its standard error also increased, to 0.189. The negative coefficient remained significantly different from zero.

III. CONCLUSION

This paper establishes that an index of the perceived likelihood of nuclear war in a country is negatively correlated with the country's rate of net private saving, holding other determinants of saving constant. This finding is thus consistent with other evidence based on U.S. aggregate time series and cross-individual data suggesting that fear of nuclear war decreases saving. That proposition has profound implications for our interpretation of the performance of the post-nuclear world economy to date, including the declining rate of saving in the 1970s and mid-1980s and the relatively low saving rate of the U.S. compared to other developed countries. It also alerts us to a link between the state of international tensions and aggregate economic activity that may prove to be important in the future. Because of its profound implications, this hypothesis deserves further attention and study.

( 1.) Stewart and Venieris [1985] find empirical support for a related hypothesis, that sociopolitical instability reduces the saving rate of developing countries.

( 2.) At least for Americans, the interpretation of the term "world war" as a catastrophic nuclear exchange is not in doubt. According to survey evidence, throughout the postwar period over 60 percent of Americans believed that a "world war" would involvenuclear weapons, and as many as 60 percent viewed their chances of surviving such a nuclear exchange as poor. (See Slemrod [1986] for survey details.) The interpretationof the term "world war" may, however, vary by country, especially considering the difficulty of precise translation into other languages. To the extent that the interpretation varies across countries, the survey results measure with error the perception of a war of constant magnitude.

( 3.) The precise wording of the question was: "I'd like your opinion of the chances of a world war breaking out in the next ten years. If ten means it is absolutely certain that a world war will break out and zero means that there is no chance of a world war breaking out, where on this scale of ten to zero would you rate the chances of world war breaking out in the next ten years?"

( 4.) OECD member countries Luxembourg, New Zealand, Portugal, Turkey, and Yugoslavia were omitted because of the unavailability of data on net private savings. Iceland was not included in the Gallup survey.

( 5.) In Table II and all subsequent regression results the index of likelihood of war is defined as the fraction of respondents who registered an opinion that the chance of war was 50-50 or greater. All the regressions were also run with an alternative index, the average percentage chance of war indicated by those who registered an opinion. None of the conclusions discussed in the text depend on which index is chosen.

( 6.) Two data points for Japan are shown. JAP (1) calculates the average index of fear by averaging the 1981 through 1984 indices. However, the survey results for 1981 (the first year this question was asked in Japan) are highly suspect. In that year 66.0 percent of the respondents said that the chance of a world war was 50-50 or greater, compared to 26.0 percent in 1982 and less than that in subsequent years. The 66.0 percent figure is higher than any figure recorded for any other country in any year. JAP (2) calculates the average fear index using a 1981 figure based on an extrapolation of the 1982-84 Japanese figures and overall trends. All regression results use the average value including the reported 1981 figure.

( 7.) This procedure is appropriate if the variance of the error term is proportional to the reciprocal of population, which would occur if each country's saving rate represented an average of independent units, with a homogeneous variance at the unit level. Barro and MacDonald [1979] found that the error variance in a similar cross-country regression analysis of saving behavior does decline with population, although not quite as rapidly as this weighting scheme implies. The sensitivity of the results to the weighting scheme is reported elsewhere in this paper.

( 8.) Two other methodological differences should be noted. The first is that Feldstein used two-stage least squares to account for the endogeneity of LPAGED. The results, though, are not significantly different from those obtained with OLS. Second, Feldstein weighted the observations by the country's population, rather than the square root of population. The motivation and implications of alternative weighting schemes are discussed elsewhere in this paper.

( 9.) It is difficult to pinpoint why the results reported in Table III differ so greatly from the results reported in Feldstein [1980]. Merely updating the saving rates used by Feldstein does not substantially change the results, nor does the difference in the weighting scheme. However, updating G, AGE, and DEP, or changing B/E to an updated SOCSEC does change the results markedly. Furthermore, the results reported in Table III (without the WAR variable) change quite a bit when the sample is restricted to the same twelve countries studied by Feldstein. Thus, any of a number of changes in the analysis is sufficient to cause the results to differ significantly.

(10.) Changing to weighting by population does not, though, rescue the equation without the WAR variable, which has no estimated coefficient other than G significantly different than zero and of the sign found by Feldstein.

REFERENCES

Barro, Robert J. and Glenn M. MacDonald, "Social Security and Consumer Spending in

an International Cross Section." Journal of Public Economics, June 1979, 275-89.

Feldstein, Martin S. "Social Security and Private Savings: International Evidence in an

Extended Life-Cycle Model," in The Economics of Public Services, edited by M. Feldstein

and R. Inman. London: MacMillan Press Ltd., 1977, 174-205.

___. "International Differences in Social Security and Saving." Journal of Public Economics,

October 1980, 225-44.

Hendershott, Patric and Joe Peek, "Household Saving: An Econometric Investigation," in

The Level and Composition of Household Saving, edited by P. Hendershott. Cambridge,

Mass.: Ballinger, 1985, 63-100.

___. "Aggregate Private Saving in the United States: Conceptual Measures and Empirical

Tests," in Measurement of Saving, Investment, and Wealth, edited by R. Lipsey and H.

Tice. Chicago: University of Chicago Press, 1989a.

___. "Treasury Bill Rates in the 1970s and 1980s." National Bureau of Economic Research

Working Paper No. 2900, 1989b.

Horioka, Charles. "Why is Japan's Private Savings Rate So High?" Finance and Development,

December 1986, 22-5.

Houthakker, Hendrick S. "An International Comparison of Personal Saving." Bulletin of

the International Statistical Institute, 38, 1961, 56-9.

___. "On Some Determinants of Saving in Developed and Underdeveloped Countries,"

in Problems of Economic Development, edited by E.A.G. Robinson. London: Macmillan,

1965, 212-24.

Kopits, George and Padma Gotur. "The Influence of Social Security on Household Saving:

A Cross-Country Investigation." International Monetary Fund Staff Papers, 27, 1980,

161-90.

Modigliani, Franco. "The Life Cycle Hypothesis of Saving and Intercountry Differences in

the Saving Ratio," in Induction, Growth, and Trade: Essays in Honor of Sir Roy Harrow,

edited by W.A. Ellis, M. Scott, and J.N. Wolfe. London: Clarendon Press/Oxford,

1970, 197-225.

Modigliani, Franco and Arlie Sterling. "Determinants of Private Saving with Special

Reference to the Role of Social Security - Cross-Country Tests," in The Determinants

of National Saving and Wealth, edited by F. Modigliani and R. Hemming. New York:

St. Martin's Press, 1983, 24-55.

Organization for Economic Cooperation and Development. National Accounts, 1974-1986,

Volume 1: Main Aggregates. Paris: Organization for Economic Cooperation and

Development, 1989.

Russett, Bruce and Miles Lackey. "In the Shadow of the Cloud: If There's No Tomorrow,

Why Save Today?" Political Science Quarterly, Summer 1987, 259-72.

Slemrod, Joel. "Postwar Capital Accumulation and the Threat of Nuclear War." National

Bureau of Economic Research Working Paper No. 887, 1982.

___. "Saving and the Fear of Nuclear War." Journal of Conflict Resolution, September 1986,

403-19.

___. "Further Evidence on the Relationship Between Saving and the Fear of Nuclear

War." Photocopy, The University of Michigan, 1989.

Stewart, Douglas B. and Yiannis P. Venieris. "Sociopolitical Instability and the Behavior of

Savings in Less-Developed Countries." Review of Economics and Statistics, 1985, 557-63.