On the U.S. budget deficit dilemma: has television contributed?

Darrat, Ali F. ; Bowers, Bill P.

I. Introduction

According to many writers and observers, television has become one of the most powerful vehicles ever invented to influence human behavior, [e.g., LaHaye (1979), Potter (1990), Morrow (1992), and Andersen (1993)]. They contend that television has largely shaped the young generation's values, affected people's performance in the workplace, and has been closely linked to many social problems like violence and crime. Increasingly, people turn to television to learn what society considers appropriate (or inappropriate) behavior in all walks of life.

In this light, it seems reasonable to speculate that television exposure may be at the center of many contemporary public policy problems. The basic aim of this paper is to examine the nature of the relationship (if any) between growth in television and one of the most important economic issues of our time; namely, the federal budget deficit delimma. More precisely, we attempt to address the following question: Has the influence of television on the American public opinion contributed significantly to the recent escalation in the federal budget deficit? Section II of the paper provides some theoretical arguments linking television to the federal budgetary process. Section III discusses the data and empirical methodology used. Section IV presents the empirical results. Section V concludes.

II. Some Theoretical Issues

The power of television to form, shape, mold and even direct public opinion is well documented and is often a generally accepted contention. Once aroused, public opinion then becomes a primary determinant of public policy, or more precisely, the taxing and spending policies of the federal government [see Erickson (1976), Monroe (1979), Page and Shapiro (1983), and Williams (1990)].

In their recent examination of foreign policymaking, Jordan and Page (1992) report that the statements and actions of television commentators and the allegedly nonpartisan experts have had the largest impacts upon the public policy preferences of the public. Similar findings linking television to domestic policy actions have also been documented by Page, Shapiro and Dempsey (1987). To the extent that television influences public opinion, and given the reasonable assumption that policymakers are somewhat sensitive to public choice, it follows that television viewership could play a significant role in determining the federal budget deficit.

That television has contributed to the budget deficit dilemma is also suggested by the claim of some analysts that the television (news) industry generally presents a "liberal" political bias favoring government intervention in the economy [LaHaye (1979), Henry (1992), and Zoglin (1992)]. This liberal bias generates increased support for government funded programs and thus increased government spending. Personal experience might also suggest that television has influenced public opinion which ultimately lead to increased government spending and thus resulted in the huge deficit. In this regard, we might highlight the substantial role played by the media, particularly the TV, to direct the attention of the public to demand government action about child abuse, the homeless, the elderly, AIDS victims, and the like. More recently, the almost daily images from places such as Somalia, Bosnia-Herzegovina, Rwanda, Haiti and the former Soviet Union have brought mounting pressures on the federal government to act, and that action has almost always resulted in higher government spending [Goodman (1993), and Neuman (1993)].(1) Others might hypothesize that TV encourages military build ups and thus the deficit.

Preliminary evidence on the television/budgetary process linkage is indicated by the positive and highly significant correlation (r = 0.79, t = 8.25) between the growth in television, as proxied for example by TV real advertising expenditures and the growth of the real national debt over the past four decades (1950-1992). As seen in Figure 1, television viewership (TV advertising) and national debt (both in real terms) tend to be closely associated since the mid 1960s. A similar picture (not shown here to conserve space) emerges for the close link between growth in television and federal budget deficit (or government spending). Of course, one cannot look at the figure and directly discern whether television growth is behind the escalation in the federal deficit (debt). To do that, a more formal test of "causality" is required, a task performed in this paper.

The preceding discussion appears to suggest the presence of a linkage from television-to-budget deficit which we subject below to empirical testing. Both economic theory and empirical evidence have shown that a number of other macroeconomic variables appear to have influenced the size of the federal budget deficit. Therefore, confining the causality test to the restrictive bivariate model (with only television growth and the federal budget deficit) may yield biased causality inferences due to the "omission of variables" phenomenon [see Sims (1980), and Lutkepohl (1982)]. To avoid such a bias, we expand our testing model and allow for the possible influence of other key determinants of the federal budget deficit.

In particular, the U.S. Employment Act of 1946, and its reaffirmation by the Humphrey-Hawkins' Balanced Growth Act of 1978, stipulate that the federal government is responsible for achieving and maintaining low unemployment. Economic theory also suggests [see Naylor (1985) and Tokheim (1993)] that fluctuations in private domestic investment impact the federal budget deficit through changes in corporate income taxes, investment tax credits, and depreciation schedules. Moreover, several studies have found evidence of a significant effect of interest rates on the size of the budget deficit [e.g., Laumas and McMillin (1984), and Darrat (1991)]. Indeed, prominent political figures (like Congressman and former Presidential candidate Jack Kemp) have repeatedly argued that high interest rates, and the resulting increases in debt servicing, are largely responsible for the recent escalation in federal deficits. Another potentially important variable that could play a significant role in shaping the deficit process is the trade deficit in what is now known as the "twin deficits" phenomenon [see Hutchison and Pigott (1984), and Darrat (1988)]. Darrat (1988), in particular, theorizes that the falling off of net exports appears to have imposed increasing pressures upon the federal government to aid domestic industries and the farming sectors that are hard hit by declining foreign market shares. The adverse economic and financial consequences of trade deficits are often viewed with much concern by the U.S. business community, by labor unions and, perhaps as a result, also by government authorities. Besides these possible determinants, we have also considered a number of other candidates like the monetary base, real GNP and the inflation rate. However, none of these variables proved empirically important for determining the budgetary process, and were thus deleted.(2)

III. Data and Methodology

We use U.S. annual data spanning the period 1950 through 1992 to test the effect of television growth on the size of the federal budget deficit in the context of a multivariate Graner-causality model. Lack of quarterly or monthly data on the television proxies necessitated the use of annual time series.

One main proxy to measure growth in television viewership could be total television advertising expenditures or revenues (TVA). It can be reasonably hypothesized that television viewership and (TVA) are strongly and positively correlated, and that (TVA) react to variations in viewership rather quickly (within one year horizon). These contentions have received some empirical support in the recent marketing literature [e.g., Friedman (1989), and Schmuckler (1992)]. When the rating (viewership) falls for any given TV program, rational behavior dictates that advertisers switch their expenditures towards alternative TV and other non-TV programs (like sales promotions and radio advertising). Indeed, advertisers may even elect to reduce the total amount of funds allocated to advertising when TV viewership declines [see Schmalensee (1983), and Rust et al (1992)]. Besides TVA, growth in television may also be measured by changes in the number of TV stations (TVS), or perhaps also by changes in the number of cable subscribers (TVC). Of course, these three alternative proxies do not exhaust all possible ways of measuring television viewership. Indeed, three more choices seem particularly promising; namely, the total number of television sets, the average number of television sets per household, and the average number of hours of television watched per year.(3) Although data on these latter measures is relatively scanty, we use the available data on these alternative proxies of television viewership to check the robustness of our empirical results.

Based on the preceding discussion, other explanatory variables in the budget deficit model include the unemployment rate (UN), gross private domestic investment (DI), short-term interest rates (RS), and the trade deficit (TD). As to the dependent variable, it is measured by the national income account version of the federal budget deficit (BD). Since disagreement exists regarding the proper measure of the budget deficit, it seems advisable to use other measures in order to check the sensitivity of the results to the choice of variables. Therefore, we alternatively used the federal debt held by the public (PD) to represent the dependent variable.(4)

Data sources are the Economic Report of the President, and the Survey of Current Business for all variables except for the three television proxies. The latter series are compiled from various issues of the Historical Statistics of the U.S., the Statistical Abstract of the U.S., the Advertising Age, and several publications of the New York based McCann-Erickson, Inc.

The focus of the paper is on the causal linkage between television growth and the federal budget deficit. As is common in applied econometrics, we employ the concept of causality in the sense of Granger (1969).(5) A given time series (x) is said to Granger-cause another time series (y) if y can be better predicted by using past values of y and x than by just using past values of y alone. The Granger-causality tests require the use of stationary time series to generate reliable inferences. Therefore, before inclusion in the empirical analysis, each variable is transformed to achieve stationarity. This is done by checking for the presence of unit roots using the Dickey-Fuller (DF) and the Augmented Dickey-Fuller (ADF) tests. As Stock and Watson (1989) point out, F and [[Chi].sup.2] statistics will not have standard distributions if the variables under examination exhibit unit roots and are thus nonstationary. Table 1 reports the test results from applying the DF and the ADF tests (the latter test with one and two lags for the dependent variable).(6) These results suggest that most variables are stationary in first differences except for the three TV proxies (TVA, TVS, and TVC), and also for (BD) which appear rather stationary in levels.

Granger-causality tests further require the selection of the appropriate lag profiles for the variables. Some studies, e.g., Mishkin (1982), use a common lag on each variable. However, as Ahking and Miller (1985) note, the assumption of an equal lag length on each variable is too restrictive and could potentially bias the results. Therefore, we use the Akaike final prediction error (FPE) criterion to determine the proper lag profile for each variable in the testing model. This procedure has recently gained popularity in applied econometrics and thus its full details are not reproduced here to conserve space.(7) Essentially, the FPE procedure uses a stepwise construction of univariate, then, bivariate, trivariate . . . etc., regressions such that the FPE value is minimized as each variable (lag) is added sequentially. This FPE procedure can be thought of as a sequential F-test where the appropriate lag is that which balances the decline in the residual variance due to the added lag against the reduced efficiency of the estimation as the lag length increases.

Since the tests performed here are multivariate in nature, another issue arises; namely, the order by which the variables are included in the testing equations. This aspect of the methodology is important in light of the fact that causality inferences may not be invariant with respect to the entry order. Following recent literature, we rank the variables for inclusion based on the "specific gravity" criterion of Caines et al (1981). The resulting equations are also examined for the presence of residual autocorrelation and for structural instability of the parameters. [TABULAR DATA FOR TABLE 1 OMITTED] In an important, but surprisingly neglected work, Lutkepohl (1989) demonstrates that structural stability of the estimated equation is a key prerequisite for reliable Granger-causality tests. The final step in the procedure is to test for the joint significance of the lagged coefficients and then derive the Granger-causality inferences.

IV. Empirical Results

Based on the procedures outlined in the previous section, Table 2 reports the empirical results from estimating alternative models of the U.S. federal budget deficit over the annual period 1950-1992.(8) Panel (A) in the table displays the results using the television advertising expenditures proxy (TVA), Panel (B) displays the results using instead the number of TV cable subscribes (TVC), while panel (C) reports the results using the number of TV stations (TVS) as the measure of TV viewership.

As the table shows, the proposed model fits the data quite well, judged by the high and significant values of adjusted R-squares (ranging from 0.72 to 0.89). Several autocorrelation tests could not reject the hypothesis of independent errors. This implies, among other things, that the t-values reported in the table are reliable measures of the significance of the coefficients, and that the estimated equations are unlikely to suffer from a serious omission of variables. Another evidence for the adequacy of the estimated equations is the apparent absence of significant structural instability using alternative testing procedures. For several breaking dates, the Chow test could not reject the null hypothesis of parameter stability, a finding corroborated by the Farley and Hinich test.(9)

The most striking finding in the results of Table 2 is that television viewership, however measured, exerts a significant causal and positive impact upon the U.S. budget deficit.(10) Specifically, the null hypothesis that television viewership (TVA, TVC or TVS) does not Granger-cause higher federal budget deficits is soundly rejected at better than the 5 percent level (the corresponding t-statistics are 7.71, 7.69, and 3.61 respectively). Of course, the Granger-causality inferences are usually tested by means of the F-statistics. These F-statistics are reported in Table 3 and they too corroborate the evidence that television viewership exerts a significant causal effect on the federal budget deficit (the associated F-values are 49.13 for TVA, 48.73 for TVC, and 12.19 for TVS). Interestingly, the FPE procedure reveals that the appropriate lag length for each of the TV viewership proxies is rather short (one year). [TABULAR DATA FOR TABLE 2 OMITTED] Therefore, it appears that the impact of television viewership on the budgetary process is both significant and relatively swift as well. To verify the sensitivity of these results to using alternative measures of television viewership, we use the total number of television sets, the [TABULAR DATA FOR TABLE 3 OMITTED] average number of television sets per household, or the average number of hours of television watched per year to represent television viewership in our equations. For the available data (from the same sources as for other viewership proxies) over the annual series 1975-1993, the estimated coefficients are all positive as hypothesized and are generally statistically significant at least at the 10 percent level (the t-statistics are 1.95, 2.33, and 1.12 for the coefficients on TVA, TVC, and TVS respectively).

Still, one might also argue that it is the size of the government, rather than the deficit itself, that matters in most public policy discussions. Thus, we perform another set of regressions using the deficit as a percent of GDP to represent the dependent variable. The estimated coefficients on all three measures of television viewership are positive as expected and are highly statistically significant at better than the 5 percent level (t = 4.04 on the TVA proxy, t = 3.39 on the TVC proxy, and t = 4.09 on the TVS proxy).

All in all, then, it appears that the increase in television viewership (however measured) has contributed significantly to the escalation in government budget deficit (measured either in absolute values or as a percent of GDP).

Before concluding, we should point out that these multivariate Granger-causality tests are also useful for detecting other potential causal factors for the federal budget deficit. For example, Tables 2 and 3 suggest that domestic private investment, the trade deficit, unemployment and interest rates are all important determinants of the U.S. federal budget deficit. It also bears emphasis that the coefficients on the interest rate variable are statistically significant and consistently positive across equations. That is, rising interest rates have Granger-caused high federal budget deficits. This finding supports recent studies on fiscal policy reaction functions [e.g., Laumas and McMillin (1984), Bradley and Potter (1986), and Darrat (1991)]. It further suggests that the observed positive association between federal budget deficits and interest rates in the U.S. is, at least partly, the outcome of interest rates causing budget deficits as some prominent analysts and observers have recently contended [e.g., Jack Kemp and Robert Eisner]. If valid, these results imply that budget deficits cannot be treated as an exogenous variable in interest rate single-equation models as some researchers have done [for example, DeLeeuw and Holloway (1985), Hoelscher (1986), Evans (1987), Plosser (1987), and Zahid (1988)]. Failure to recognize the endogeneity of budget deficits to interest rates may cast some doubts on the reliability of these studies due to the simultaneity bias problem. This also suggests that a more fruitful inquiry into the relationship between budget deficits and interests rates in the U.S. should be performed in the context of a simultaneous-equation model.

V. Concluding Remarks

Based on several theoretical arguments, this paper examines empirically whether growth in television viewership has significantly contributed to the recent escalation in the U.S. budget deficits over the last four decades. We specify a model for the federal budget that includes alternative measures of the importance of television viewing along with a number of other potentially important macro determinants of the deficit. In the context of such a multivariate model, we focus on the Granger-concept of causality and use the FPE procedure in conjunction with the specific gravity criterion to derive our inferences. Before implementing the empirical procedure, we check the stationarity requirements for all series by means of alternative unit root tests.

The empirical results suggest that recent growth in the U.S. television viewership (however measured) has significantly contributed to the size of the federal budget deficit, over and above the effects of several other possible determinants (most notably interest rates, private domestic investment, and the unemployment rate). This finding implies some support to the view that there exists a "liberal" bias within the media and within television in particular that undermines fiscal conservatism.

Of course, given the relative brevity of our sample and in light of possible estimation or measurement problems, the results of this paper are only suggestive and should be interpreted with caution. At the very least, though, these results do suggest that the recent growth in the U.S. television viewership is an important contributing factor behind the escalating federal budget deficit. Policy-makers, therefore, should not ignore the role of television if they hope to understand and ultimately control the U.S. budget deficit dilemma.

Notes

1. For further discussion, interested readers are referred to the recent survey of some relevant issues in Alesina and Perotti (1994).

2. The growth in government budget deficits since the 1970s may also be related to the dramatic increase of entitlement expenditures. No attempt, however, is made here to examine every possible reason for the deficit explosion. Our primary objective in this paper is to examine whether television viewership in particular has played any important role in the recent escalation of the U.S. budget deficit.

3. We thank the anonymous referee for suggesting these proxies to us.

4. Of course, other measures may also be used for the dependent variable like federal government spending, for some related disaggregates (e.g., spending or social programs).

5. Note that controversy surrounds the Granger-causality concept and the implied tests. For a concise account of these difficulties, see Jacobs et al (1979), and Zellner (1979). Many researchers view the Granger-causality procedure as merely a test of the "predictive content" of the time series.

6. A linear time trend was also included in all testing equations and proved statistically significant.

7. For a useful account, see Ahking and Miller (1985). Note that Thornton and Batten (1985) report evidence supporting the use of the FPE criterion over several other lag-length selection procedures, including the Akaike Information (AIC) and the Schwartz criterion.

8. To ensure statistical efficiency of the estimations, we used the Beach-Mckinnon maximum-likelihood procedure to estimate all equations.

9. The Chow test yielded the following F-statistics for the TVA, TVC and TVS equations respectively: 0.51, 0.59, and 2.85 for the mid-point; 1.35, 2.77, and 1.30 for the pre- and post-Reagan periods; 0.11, 0.91, and 0.11 for the pre- and post- Johnson periods; and 0.11, 0.91, and 0.11 for the pre- and post-Vietnam years. The Farely and Hinich test of a gradual structural shift produced the following F-statistics: 1.57, 2.70, and 1.71 or the three equations respectively.

10. The use of the federal debt held by the public as the dependent variable produced similar conclusions. These results are not reported here to conserve space but are available from the authors upon request.

References

Ahking, F. W., and S. M. Miller, "The Relationship Between Government Deficits, Money Growth, and Inflation." Journal of Macroeconomics 7 (Fall 1985), 447-467.

Alesina, A., and R. Perotti, "The Political Economy of Budget Deficits." NBER Working Paper No. 4637 (February 1994).

Andersen, K. E., "The Politics of Ethics and the Ethics of Politics." American Behavioral Scientist 32 (March/April 1989), 479-492.

Bradley, M. D., and S. M. Potter, "The State of the Federal Budget and the State of the Economy: Further Evidence." Economic Inquiry 24 (January 1986), 143-153.

Caines, P. E., C. W. Kang, an S. P. Sethi, "Causality Analysis and Multivariate Autoregressive Modelling with an Application to Supermarket Sales Analysis." Journal of Economic Dynamics and Control 3 (August 1981), 267-298.

Darrat, A. F., "Have Large Budget Deficits Caused Rising Trade Deficits?" Southern Economic Journal 54 (April 1988), 879-887.

-----, "Policy Impacts Under Rational Expectations." Journal of Banking and Finance 15 (April 1991), 257-271.

DeLeeuw, G., and T. M. Holloway, "The Measurement and Significance of the Cyclically Adjusted Federal Budget and Debt." Journal of Money, Credit and Banking 17 (May 1985), 232-242.

Erickson, R., "The Relationship Between Public Opinion and State Policy: A New Look at Some Forgotten Data." American Journal of Political Science 20 (February 1976), 25-36.

Evans, P., "Interest Rates and Expected Future Budget Deficits in the United States." Journal of Political Economy 95 (February 1987), 34-58.

Friedman, L., "How Good is the Seven-Day TV Diary Now?" Journal of Advertising Research 29 (August/September 1989), RC1-RC3.

Goodman, W., "Silent Partner Emerging in Policy Councils: TV." The New York Times (March 6, 1993), section L, 47.

Granger, C. W. J., "Investigating Causal Relations by Econometric Models and Cross-Spectral Methods." Econometrica 37 (July 1969), 424-438.

Henry, W.A., "Are the Media Too Liberal?" Time (October 19, 1992), 46-47.

Hoelscher, G. P., "New Evidence on Deficits and Interest Rates." Journal of Money, Credit and Banking 18 (February 1986), 1-17.

Hutchison, M. M., and C. Pigott, "Budget Deficits, Exchange Rates and the Current Account: Theory and U.S. Evidence." Federal reserve Bank of San Francisco, Economic Review (Fall 1984), 5-25.

Jordan, D., and B. Page, "Shaping Foreign Policy Opinions: The Role of TV News." Journal of Conflict Resolution 36 (June 1992), 227-241.

Jacobs, R. L., E. E. Leamer, and M. P. Ward, "Difficulties with Testing for Causation." Economic Inquiry 17 (July 1979), 401-413.

LaHaye, T. The Battle for the Family (Old Tappan, New Jersey: Fleming H. Revell Company, 1979).

Laumas, G. S., and W. D. McMillin, "Anticipated Fiscal Policy and Real Output." Review of Economics and Statistics 66 (August 1984), 468- 471.

Lutkepohl, H., "Non-Causality Due to Omitted Variables." Journal of Econometrics 19 (August 1982), 367-378.

-----, "The Stability Assumption in Tests of Causality Between Money and Income," in Econometrics of Structural Changes: Studies in Empirical Economics, edited by Watter Kramer (Gelmany: physica-Verlag Heidelberg, 1989).

Mishkin, F. S., "Does Anticipated Aggregate Demand Policy Matter? Further Econometric Results." American Economic Review 72 (September 1982), 688-802.

Monroe, A. D., "Consistency Between Public Preferences and National Policy Decisions." American Politics Quarterly (March 1979), 3-19.

Morrow, L., "Folklore in a Box." Time 140 (September 21, 1992), 591-602.

Naylor, J. A., "A Survey of Post-1970 Empirical Studies of Investment Expenditures," in Modern Concepts in Macroeconomics, edited by T. Haverlesky, and J. T. Boorman (Arlington Heights, ILL: Harlan Davidson, Inc., 1988).

Neuman, J., "Few Bosnia Options for Clinton." USA Today (April 19, 1993), p. 4A.

Page, B., and R. Shapiro, "Effects of Public Opinion on Policy." American Political Science Review 77 (March 1983), 175-190.

-----. and G. Dempsey, "What Moves Public Opinion?" American Political Science Review 81 (March 1987), 23-44.

Potter, W. J., "Adolescents' Perceptions of the Primary Values of Television Programming." Journalism Quarterly 67 (Winter 1990), 843-851.

Rust, R., W. Kamakura, and M. Alpert, "Viewer Preference Segmentation and Viewing Choice Models for Network Television." Journal of Advertising 21 (March 1992), 1-16.

Schmalensee, D. H., "Today's Top Priority Advertising Research Questions." Journal of Advertising Research 23 (April/May 1983), 49-60.

Schmuckler, E., "Buyers are Salivating Over Increases in Net GRP Pool." Mediaweek (June 1, 1992), 4.

Sims, C. A., "Macroeconomics and Reality." Econometrica 48 (January 1980), 1-48.

Thornton, D. L., and D. S. Batten, "Lag-Length Selection and Tests of Granger-Causality Between Money and Income." Journal of Money, Credit and Banking 17 (May 1985), 164-178.

Tokheim, A. B. Legislative and Policy Update (Washington, D.C.: Council on Competitiveness, March 29, 1993).

Williams, W. Mismanaging America (Lawrence, Kansas: The University of Kansas Press, 1990).

Zahid, K. H., "Government Budget Deficits and Interest Rates: The Evidence Since 1971 Using Alternative Deficit Measures." Southern Economic Journal 54 (January 1988), 725-731.

Zellner, A., "Causality and Econometrics," in Three Aspects of Policy and Policymaking: Knowledge, Data and Institutions, edited by K. Brunner and A. H. Meltzer (Amsterdam, North Holland: North-Holland Publishing Company, 1979).

Zoglin, R., "Sitcom Politics." Time (September 21, 1992), 44-47.

Ali F. Darrat and Bill P. Bowers

Professor of Economics, Department of Economics and Finance, Louisiana Tech University, Ruston, LA 71272

Assistant Professor of Business Administration, Department of Business Administration and Economics, Charleston Southern University, Charleston, SC 29423

We wish to thank, without implicating, G. Brown, M. El-Baghdadi, and an anonymous referee for several helpful comments and suggestions.