The past half century witnessed a transformation of the finances of American families. Whereas in 1962 only 19% of American households owned corporate stock, by 1992 the rate had increased to 37.4%, and by 2007, fully 65% of American households were stock owners (Poterba and Samwick 1995, McCarthy 2015). Both researchers and political pundits have argued that this fundamental shift in the property interests of American households caused them to identify with the investor class, contributing to increased support for the Republican party (Duca and Saving 2008). This theory has been so influential that it partly inspired the Bush Administration’s 2005 proposal to privatise social security. By further expanding stock ownership, it was hoped that private social security accounts would help create a permanent Republican majority (Conlan 2008).
Yet the political effects of public policy changes that broaden the ownership of financial assets may not be so simple. Even if owning financial assets actually causes voters to identify with the investor class, it also exposes them to financial market fluctuations. In response to such fluctuations, voters who own financial assets may punish incumbents when financial markets perform poorly, or reward them in response to strong returns, in a pattern consistent with models of retrospective voting behaviour. Such models argue that voters’ choices are driven by backwards-looking assessments of how well the government has performed during an incumbent’s tenure (e.g. Achen and Bartels 2016, Healy et al. 2017). Rather than simply supporting a business-friendly political party, owning financial assets may induce voters to focus on the performance of financial markets during incumbents’ time in office.
In a recent paper, we study the electoral consequences of the liberty bond drives of WWI, which induced tens of millions of American households to purchase government bonds (Hilt and Rahn 2018). Prior to those bond drives, relatively few Americans held any financial assets other than bank accounts. But the liberty loan campaigns were conducted on a massive scale and sought to attract the widest possible participation, in an effort to strengthen public support for the war effort. Ordinary citizens responded to the campaigns by subscribing at extraordinary rates, and by 1919 the share of American households that owned a liberty bond was likely greater than the share of modern American households that own stocks.
Although they were marketed as the “safest investment in the world”, the value of the liberty bonds proved to be quite volatile. Beginning in late 1919, in an effort to restrain the growth of credit and prices that had developed during and after the war, the Federal Reserve enacted a series of substantial increases in interest rates. This caused the prices of liberty bonds to fall, and millions of American households suffered capital losses. Then in 1921 the Fed began to lower interest rates, causing liberty bonds to appreciate. These fluctuations present an opportunity to test how voters who have been induced to hold financial assets by government policy respond to fluctuations in those assets’ values.
Volatility induced by monetary policy changes
The Fed’s policy changes, and their consequences for liberty bond owners, are illustrated below. Figure 1 presents the Fed’s discount rate, which was held at 4% up through the first half of 1919, in part to hold down the interest rates that the liberty bonds would need to offer when they were issued. Then in late 1919 and early 1920, the Fed raised interest rates dramatically, increasing its discount rate to 7%, a level that would not be reached again until the 1970s.
Figure 1 Discount rate, Federal Reserve Bank of New York
Figure 2 shows what happened to the prices of the fourth liberty loan, the most widely held liberty bond. In order for their yields to rise to levels commensurate with prevailing rates, the prices of these bonds fell to about 85% of their subscription price, and remained near that level through late 1921.
Figure 2 Prices, fourth liberty loan
The cumulative returns earned by liberty bond holders are shown in Figure 3. The capital losses resulting from the price decreases were far larger than the 4.25% annual coupon payments received by the bond holders, which meant that in mid-1920 the cumulative returns that had been received by liberty bond owners were sharply negative, in both nominal and real terms.
Figure 3 Cumulative returns, fourth liberty loan
In mid-1921, following a sharp contraction in output and a decline in the price level, the Fed began to ease rates. The decline in prevailing interest rates led to a recovery of liberty bond prices, and an increase in the cumulative returns they paid. By 1924, the effect of the losses experienced in 1919–21 had been erased by strong capital gains.
Using county-level data on liberty loan subscription rates, we study the effect of these fluctuations on presidential election outcomes. The 1920s were a period of Republican dominance in presidential politics, with Harding, Coolidge, and Hoover winning substantial majorities of the popular and electoral votes in 1920, 1924, and 1928. We find that relative to their voting patterns from the 1908–1916 elections, counties that subscribed to the liberty loans at higher rates turned sharply against the Democratic party in the 1920 and 1924 elections. This was a reaction to the depreciation of the bonds prior to the 1920 election (when the Democrats held the presidency), and the appreciation of the bonds in the early 1920s (under a Republican president). These results suggest the liberty bond campaigns had unintended political consequences – an effort to increase support for the war contributed to an electoral backlash against the party that created them.
Of course, liberty bond subscriptions may have been influenced by unobservable county attributes not reflected in historical voting patterns, which may, in turn, have influenced voting behaviour in the 1920s. In order to address this possibility, we instrument for liberty bond participation using a measure of the predicted local severity of the fall 1918 influenza epidemic. The most lethal wave of the epidemic, which occurred in October 1918, coincided with the fourth liberty loan drive. Our measure of the predicted severity of the influenza epidemic is based on a county’s distance to large military training camps, which were the most likely source of the epidemic within the civilian population of the US. Greater distance from military camps was strongly correlated with participation in the fourth loan, as the bond drive was hampered by both the influenza epidemic itself, and by efforts to control the spread of the epidemic. Our instrumental variable estimates of the effect of liberty bond ownership on the Democratic Party vote share indicate that a one standard deviation increase in a county’s liberty bond participation rate led to a decrease in the Democratic share in presidential elections of 3.3 percentage points on average over the 1920–32 period.
In order to assess whether the electoral effects of liberty bonds could have been decisive, we estimate the same empirical model using state-level data. We focus on the 1920 presidential election, in which Democrat James Cox won only 12 states and Harding won 37, for electoral vote totals of 127 to 404. Counterfactual estimates of the Democratic Party vote share for the 1920 presidential election by state indicate that in the absence of the liberty bonds, the Democratic Party would have won 12 additional states, but would still have lost the electoral vote. This implies that the effect our analysis attributes to liberty bonds contributed significantly to Republican electoral margins but was unlikely to have been decisive.
Our results are clear evidence that the composition of assets owned by households can influence their political behaviour, but not necessarily in ways consistent with ownership society theories. In our context, bond ownership made the finances of ordinary American households more sensitive to changes in financial markets, which led them to reject incumbents who had presided over asset price decreases, and to support political candidates who claimed that they had brought fiscal stability and higher bond prices. This is consistent with a ‘pocketbook’ view of retrospective voting behaviour.
The results also suggest that the Bush administration’s 2005 proposal to transform social security into a system of private accounts may not have had its intended political effects if it had been implemented. Voters who were induced to finance their social security benefits by contributing to an account invested in financial assets would likely have become quite sensitive to financial market fluctuations. It is not unreasonable to think that the steep decline in asset prices in 2008 associated with the financial crisis would have led to an even stronger backlash against the Republican Party in the presidential elections of that year.
Achen, C H and L Bartels (2016), Democracy for Realists: Why Elections Do Not Produce Responsive Government, Princeton University Press.
Conlan, T J (2008), “Federalism, the Bush Administration, and the evolution of American politics,” in Morgan and Davies (eds), The Federal Nation: Perspectives on American Federalism, Palgrave MacMillan, 11–25.
Duca, J V and J L Saving (2008), “Stock ownership and congressional elections: The political economy of the mutual fund revolution,” Economic Inquiry 46(3): 454–79.
Healy, A J, M Persson and E Snowberg (2017), “Digging into the pocketbook: Evidence on economic voting from income registry data matched to a voter survey,” American Political Science Review 111(4): 771–785.
Hilt, E and W Rahn (2018), “Financial asset ownership and political partisanship: Liberty Bonds and republican electoral success in the 1920s,” NBER Working paper 24719.
McCarthy, J (2015), “Little change in the percentage of Americans who own stocks,” Gallup, 22 April.
Poterba, J and A Samwick (1995), “Stock ownership patterns, stock market fluctuations, and consumption,” Brookings Papers on Economic Activity 2: 295–3.