«Pablo Querubin Harvard Academy for International and Area Studies Harvard University James M. Snyder, Jr. Department of Government Harvard University ...»
Wealth Accumulation by U.S. Congressmen, 1845-1875:
Were the Civil War Years Exceptional(ly Good)?∗
Harvard Academy for International and Area Studies
James M. Snyder, Jr.
Department of Government
We thank Andrea Camacho, Tewﬁk Cassis, Katharine Lauderdale, Kaitlin Lebbad, Jessica Lee and
Luis Felipe Martinez for their excellent research assistance. We also thank the participants of seminars at the University of British Columbia, Caltech, UC Berkeley, UCLA, the University of Chicago, Fedesarrollo, Harvard University, the University of Maryland, MIT, the University of Rochester, the University of Virginia, the University of Wisconsin - Madison, for their many helpful comments.
Abstract In this paper we use historical census data from the U.S. to estimate the pecuniary returns to holding a seat in the U.S. House of Representatives during the 1850s and 1860s. We employ a regression discontinuity design (RDD) based on close elections and compare wealth accumulation in the decades between 1850 and 1870 among those who won or lost their ﬁrst congressional race by a small margin. We ﬁnd no evidence of large returns to congressional seats for the 1850s or the second half of the 1860s.
However, we do ﬁnd evidence of signiﬁcant returns for the ﬁrst half of the 1860s, during the Civil War. Those who won their ﬁrst election by a narrow margin and served during the period 1861-1866 (37th-39th Congresses) accumulated, on average, 33-55% more wealth between 1860 and 1870 than candidates who lost the election and did not serve – for the median congressman this corresponds to an additional $700,000-$1,200,000 in present values. We hypothesize that the sudden spike in government spending during the war and the decrease in oversight from government agencies might have made it easier for incumbent congressmen – and probably other politicians – to collect rents.
We ﬁnd evidence that wealth accumulation was particularly large for congressmen who represented states that were home to the major military contractors during the war, and for congressmen who served during the Civil War in committees that were responsible for most military appropriations – the latter accumulated up to 70% more wealth relative to those who never served. These results are robust to the inclusion of state ﬁxed eﬀects, and to the inclusion of a broad set of controls including age, initial wealth and occupation dummies. Placebo regressions reveal that these results are not driven by pre-existing diﬀerences in wealth accumulation or other covariates prior to serving in congress. We also show that all of the main results hold when we use the number of domestic servants – a good proxy for wealth – as the dependent variable.
Finally, we show that a simple “before-and-after” design using only winning candidates yields surprisingly similar estimates to the RDD.
1 Introduction An extensive literature in political economy stresses the importance of conﬂicts of interest between elected representatives and their constituencies. The main concern is that elected representatives, once in oﬃce, may use their political power to redistribute resources to themselves or to favor certain interest groups in return for bribes or campaign contributions.
The models in this literature generally predict ineﬃcient and/or distorted policies.1 Such rents may also be inconsistent with the protection of property rights and a level playing ﬁeld that provide correct incentives for innovation and investment –features at the heart of institutional theories of comparative development. Several papers, including Mauro (1995), Knack and Keefer (1995), Olken (2007) and Reinikka and Svensson (2004), document the detrimental eﬀects of corruption on development, and recently the World Bank identiﬁed corruption as one of the greatest obstacles to economic development. Even if the rents accruing to politicians do not imply any speciﬁc ineﬃciency or distortion, estimates of these rents may help assess arguments about the quality of politicians and the eﬀects of quality on policy, as in Caselli and Morelli (2004), Messner and Polborn (2004), and Mattozzi and Merlo (2006, 2007, 2008).
A major empirical question in this context is understanding the environments under which rent extraction by politicians is more likely to occur. The magnitude of such rents may crucially depend on the nature of the political environment and institutions. Rent extraction may be limited if the political environment is highly competitive and the existing institutions provide an appropriate level of checks and oversight on politicians behavior.
Similarly, a free and independent media may allow voters to monitor their representatives.
On the other hand, in weakly institutionalized environments, politicians may be able to capture the political system and extract rents without punishment from oversight institutions The literature includes Barro (1973), Ferejohn (1986), Banks and Sundaram (1993, 1998), Harrington (1993), Persson and Tabellini (2000), Fearon (1999), Berganza (2000), Hindriks and Belleﬂamme (2001), Le Borgne and Lockwood (2001, 2006), Smart and Sturm (2003, 2004), Besley (2006), and Padro i Miquel (2007), as well as Stigler (1971), Peltzman (1976), Denzau and Munger (1986), Austen-Smith (1987), Baron (1994), Grossman and Helpman (1994, 1996, 2001), and Persson and Tabellini (2000).
or their constituencies.
Unfortunately, the study of rent extraction faces substantial empirical challenges, because it is often diﬃcult to detect or measure the accumulation of rents by politicians in a systematic way. One way to assess the magnitude of political rents is to track the wealth of politicians. To the degree that rents are large, we should observe politicians accumulating substantially more wealth while in oﬃce than they would have otherwise.
In this paper we use historical census data from the U.S. to estimate the pecuniary returns to holding a seat in the U.S. House of Representatives during the 1850s and 1860s. We focus on representatives who served during the period 1845 to 1875, as well as individuals who ran for a seat in the U.S. House but lost their election. The U.S. census recorded wealth in 1850, 1860, and 1870, and we have found the individual census records for a large sample of candidates. We also collected information on the number of domestic servants in each candidate’s household, as well as the number of slaves from the census slave schedules, as other proxies for wealth. We study the number of servants because wealth was self-reported, so there could be concerns about misreporting. (We address this in more detail below.) It was diﬃcult to misreport the number of servants, since census enumerators visited each home in person. When studying the 1860 to 1870 period we focus on “free” states where slavery was prohibited, because prior to emancipation slaves were counted as part of personal wealth.
Thus, comparing wealth ﬁgures before and after the abolition of slavery may be misleading for former slave owners. We analyze wealth and slave holding in slave states during the 1850 to 1860 period.
We employ a regression discontinuity design (RDD) based on close elections to estimate the causal eﬀect of serving in Congress on wealth accumulation during this period.2 We compare wealth accumulation in the decades between 1850 and 1870 among those who won or lost their ﬁrst congressional race by a small margin. The outcome of close elections provides us with quasi-random assignment of political power. It therefore allows us to isolate See Hahn, Todd and Van der Klaauw (2001) for a general discussion of regression discontinuity designs and Lee (2008) for a concrete application to close elections.
the eﬀect of serving in Congress from the eﬀect of other characteristics of these individuals – such as talent, connections, or charisma – that are correlated with serving in congress and wealth accumulation.
We ﬁnd no evidence of large returns to congressional seats for the 1850s or the second half of the 1860s. However, we do ﬁnd evidence of signiﬁcant returns for the ﬁrst half of the 1860s, during the Civil War. Those who won their ﬁrst election by a narrow margin and served during the 37th and 38th Congresses (1861-1865) accumulated, on average, about 30% more wealth between 1860 and 1870 than candidates who lost the election and did not serve. For the median congressman this corresponds to about $15,000 in additional wealth – roughly $750,000 in present values. Thus, our results indicate that the returns to a seat in the House were low during “normal” times in the mid-19th century, suggesting that U.S. institutions appear to have been eﬀective at controlling politicians’ behavior. However, the returns to oﬃce increased between 1861 and 1865, when federal government spending expanded sharply to unprecedented levels in order to fund the war.
Figure 1 illustrates the evolution of nominal spending by the Federal government between 1845 and 1880. There was a dramatic spike in nominal government spending during the Civil War years, from about $60 million just before the outbreak of the war to almost $1,200 million at the highest point during the war – an increase of almost 2,000%. This was driven by the need mobilize, equip, feed, and move armies on a scale never before seen in U.S. history.
The sudden spike in government spending might have made it easier for incumbent congressmen (and other politicians) to collect rents. For example, they could channel contracts towards ﬁrms in which they had an interest, or collect side-payments or legal fees from contractors in exchange for favorable treatment. Procurement was especially frantic and disorganized during the ﬁrst half of the war, as an army of almost 700,000 men was built essentially from scratch (see Wilson, 2006). Under severe pressure, and focused on the gloomy military situation in the east, it is unlikely that the agencies of the federal government were capable of carefully overseeing and auditing much of the contracting. In addition, rent extraction would have been more diﬃcult to detect during the Civil War than during the 1850s because the same dollar amount of rents represented a much smaller fraction of total government spending. Using the $15,000 ﬁgure from above, the total amount of rents extracted by all incumbent congressmen serving during the Civil War would have represented less than 0.1% of total federal wartime spending. However, this would have represented almost 2% of the average level of pre-war spending (over four years), and about 0.5% of post-war spending.
Thus, rent extraction comparable in scale to what we estimate for the Civil War years would have been much easier to detect during “normal” times.
We also ﬁnd evidence that wealth accumulation was particularly signiﬁcant by representatives who represented states that played an important role providing supplies during the war. Congressmen from these states accumulated 40-50% more wealth than similar individuals who never served. Moreover, we ﬁnd that congressmen who served during the Civil War in committees that were responsible for most military appropriations became richer than other members of congress and that candidates who ran but never served. These individuals accumulated almost 70% more wealth relative to those who never served. This, together with additional anecdotal evidence give us further conﬁdence in our interpretation.
Our main results are robust to the inclusion of state ﬁxed eﬀects, and to the inclusion of a broad set of controls including age, initial wealth and occupation dummies. Placebo regressions reveal that these results are not driven by pre-existing diﬀerences in wealth accumulation or other covariates prior to serving in congress. We also test whether politicians appear to accumulate additional rents after leaving congress – which would be consistent with the idea that politicians beneﬁt in the long run from the connections and networks established while in oﬃce – but do not ﬁnd signiﬁcant evidence of such returns. We also show that all of the main results hold when we use the number of domestic servants as the dependent variable. In fact, the results for servants are even more robust than those using wealth.