Evidence

Distributing News and Political Influence 99

The address goes on to take credit for projects such as roads from farms to markets and the provision of relief to specific groups such as homeowners and teachers. Ra-dio was apparently used to remind voters of past favors as well to make new cam-paign promises.

Both these uses of radio probably changed the political benefits of government programs, because information about past favors and campaign promises helped voters to identify and vote for politicians who had furthered their interests in their last term in office or who promised to do so in the coming term. For example, if a governor in the early 1920s would have promised to start building roads from farms to markets, the return in the form of rural votes might have been meager, because many of those people concerned lived in rural areas, did not have a daily newspaper, and would not have been aware of this promise. Ten years later this governor could go on the radio and make this promise directly to an increasing number of such voters. This, of course, increased the political benefits of such projects. Similarly, before the advent of radio many of the people living in rural areas would not have known whom to credit for the roads they benefited from, while 10 years later a gov-ernor could go on the radio and tell an increasing number of these voters directly that the credit was his or hers. This increased the incentive to launch such projects.

The incentives created by radio could also have worked at the local level. Accord-ing to Dunn (1936), a local relief chairman reportedly laments when the FERA is cutting back its activities: “‘This is likely to hinder my chances for re-election,’ since there would undoubtedly be a feeling of bitterness created on the part of a number of people whom he might find it necessary to refuse.” My model suggests that the chair-man would be more reluctant to refuse people who were likely to vote and people who were likely to know that he was responsible for their refusal.

If politicians were affected by the incentives to provide favorable policies to in-formed voters is an empirical question. If this were true, the patterns we would expect in the data would be high spending in areas with high voter turnout where many voters have radios and few voters are illiterate.

Results

Before a more structured investigation of the data, it may be helpful to look at some univariate correlations. Illiteracy, low voter turnout, and low radio ownership are negatively correlated with government spending, whereas unemployment is posi-tively correlated with spending and low bank deposits are only weakly correlated with spending.

In a multivariate regression analysis of the determinants of spending, a similar pattern emerges. Factors related to low socioeconomic status are positively related to spending if they indicate a need for income assistance (high level of unemploy-ment, low bank deposits, low house values) and negatively related to spending if they indicate low levels of political participation and information (low voter turnout,

high levels of illiteracy, low radio use). Thus poor counties are not automatically politically weak. This makes sense, because the votes of the poor may be more easily swayed by economic favors. The weakness comes about because the poor participate less in politics and are not well informed.

The estimated effects of radio use and voter turnout on government spending are sizable. The estimate coefficients from the foregoing regression imply that a 1 per-cent increase in the share of households with radios in a county is associated with an increase in relief spending of 0.52 percent, and a 1 percent increase in voter turnout is associated with a 0.61 percent increase in relief spending.

Next I used a panel of county data from 1920–30 to investigate whether voter turn-out was related to the increasing use of radio. In a fixed effects regression I found an extremely significant positive correlation between increasing radio use and increasing voter turnout. The estimated coefficients from this regression imply that a 1 percent increase in radio use is associated with a 0.12 percent increase in voter turnout.

Figure 6.1 summarizes radio’s total estimated effects on relief spending. A 1 per-cent increase in the share of households with radios in a county is estimated to in-crease relief spending by 0.52 percent directly as people in the county become better at attracting public funds. As every 1 percent increase in voter turnout increases spending by 0.61 percent, radio’s effect on spending via voter turnout is 0.07 percent.

Thus the total increase in relief spending from a 1 percent increase in radio owner-ship is estimated to be 0.59 percent.

A possible concern is that the correlation between the share of households with a radio and relief spending may arise simply because counties where many people have radios have a greater need for relief spending; however, the opposite seems more likely. Counties where many people have radios have characteristics that sug-gest a lower need for relief spending, that is, lower unemployment, higher wages, higher property values, and so on.¹

Discussion

The effects of radio ownership on both relief spending and voter turnout are signifi-cantly higher in rural than in urban counties. This is expected given the cost structure 1. To further investigate whether it really is radio penetration that matters, radio ownership was instrumented using ground conductivity, a geological feature that the Federal Communica-tions Commission uses to predict the propagation of AM signals across the United States, and the sum of the power of all AM antennas in 1934, weighted by the inverse square root of the distance between the county seat and the antenna. The instruments are strongly correlated with the share of households with radios in the expected way. The instrumental variables estimates show significant effects of radio on both spending and turnout. However, while the instrumen-tal variables-estimated effect of radio on turnout remains significant after the inclusion of state effects, this is not true for the spending equation.

Distributing News and Political Influence 101

of radio and newspaper news delivery. It is also in line with contemporary accounts of how radio reduced the informational disadvantage of rural populations. In a sym-posium on radio and rural life organized in 1935 by E. Brunner, a professor of educa-tion at Columbia University, radio’s effectiveness in breaking down rural isolaeduca-tion was repeatedly stressed. For example, R. F. Fricke (1935, p. 26) stated:

I believe the radio to be a very important factor in giving to the farmer a much broader knowledge in public affairs . . . I believe the radio has meant more to the farmer than to the city person, and that he uses it more especially in listen-ing to programs deallisten-ing with public affairs and for that reason, I believe the farmer to be better informed than the average city man, thanks to the radio.

My results (Strömberg 1999, 2001a) indicate that radio not only broke rural isola-tion, but in doing so also increased the political power of rural counties. The effect was considerable. Estimates indicate that radio increased FERA funds allocated to a rural county, relative to an identical urban county, by 20 percent.

Moreover, radio’s effect on voter turnout is interesting in its own right. The ag-gregate effects of increasing radio use on voter turnout are far from negligible. In 1920 fewer than 1 percent of the U.S. population used radios. By 1940 around 80 percent of households had radios. The estimate suggests that this would have led to an increase in votes per capita of around 5.5 percent. Between 1920 and 1940 votes per capita in the United States increased by about 12 percent, from 25 to 37 percent, in both gubernatorial and presidential elections. According to the estimates the in-crease would only have been about half as large without radio. The estimates are based on time series variation using year dummy variables, so they are not merely picking up the time trend in both series.

Figure 6.1. Estimated Effects of Radio

Note: Figures in parentheses are standard errors.

Source: Author.

Radio Government

spending

Voter turnout

1% 0.52

(0.20) 0.12

(0.01)

0.61 (0.12)

0.52 + 0.12 * 0.61 =

0.59%

I estimated the media’s impact on policy in the United States in the early 1930s, that is, for an industrial, democratic country with a free press (Strömberg 1999, 2001a).

This impact may, of course, be different in other settings, though Besley and Burgess (forthcoming) found a similar impact in a developing country. They found that In-dian states with higher newspaper circulation are more effective in attracting public food distribution and calamity relief expenditure in cases of drought. However, the findings of Djankov and others (forthcoming) that state ownership is negatively cor-related with good government and of Besley and Prat (2001) that less press freedom is associated with lower government turnover indicate that in a less democratic coun-try with less free press the effects are likely to be smaller.

Extension: Television 1950–60

I carried out a similar study on the effects of the expanding use of television. Al-though the effects of television on voter turnout are as precisely estimated as those of radio, the estimated effects on spending are cruder, because I did not look at one large program with well-defined goals, but used a 1962 cross-section of all intergov-ernmental transfers from U.S. states to counties. This included spending on educa-tion, highways, public welfare, and for other purposes, and thus carefully controlling for other determinants of spending was much more difficult. Unfortunately, the 1950 census took place before the large-scale expansion of television, when only 9 percent of U.S. homes had television, and the 1960 census took place after most of the expan-sion, when 87 percent of homes had a television. So the cross-sectional variation is not as great as in the radio study.

While radio was important for rural areas, television seemed to have been impor-tant for African-Americans and people with low education. McCombs (1968) found that during the 1952–60 expansion of television, the share of the population that used neither television nor newspapers extensively fell from 71 to 49 percent among African-Americans with less than a high school education. Among whites at the same educa-tional level the share of low media users actually grew by 5 percent. Among whites with a high school education or more, the share of low media users doubled from 16 to 38 percent. Judging from these studies, television seems to have reduced the infor-mational disadvantage of African-Americans and people with a low level of educa-tion. I therefore tested whether television increased the political strength and participation of these groups.

The statistical analysis took the same form as in the study of radio: one regression studied the determinants of intergovernmental transfers and another studied the determinants of changes in voter turnout during 1950–60. The results show that while high voter turnout is correlated with receiving more government transfers, televi-sion ownership is only significantly correlated with government transfers in coun-ties with many African-Americans. Furthermore, an increase in television use is clearly

Distributing News and Political Influence 103

associated with increases in voter turnout, especially in counties with many people with low education levels. The effects of television are weaker than those of radio:

the positive effect of television on voter turnout is only a third of that of radio. Per-haps this indicates increasing media saturation.

The results indicate that television increased the ability of African-Americans and people with a low level of education to attract government funds. For African-Americans the effect seems primarily to be that television directly increased their ability to attract government funds, perhaps by helping them to vote more accu-rately for politicians who furthered their interests. In contrast, television seems to have increased the political power of people with low levels of education by increas-ing their voter turnout more than for the average citizen.

That the increasing use of television significantly increased voter turnout is some-what surprising. Watching television news is often not a significant predictor of po-litical knowledge in cross-sectional studies of survey data (Delli Carpini and Keeter 1996). However, this may indicate that people who watch much television are less knowledgeable to start with, not that they do not learn from television (Price and Zaller 1993). Critics of television further claim that instead of stimulating viewers’

interest and involvement in social action, television news may instead spread a po-litical malaise that discourages popo-litical participation (Putnam 2000; Robinson 1976).

My analysis (Strömberg 1999, 2001a) strongly rejects the idea that the increasing use of television created a political malaise in the 1950s. Instead my findings suggest that television increased political participation.