Draft 2

Why Didn’t the North Let the South Secede? Evidence of Economic Motivations from Presidential Election Data 1 Zachary D. Liscow University of California, Berkeley June 2007

Abstract This paper helps answer the question of why the North chose to fight the South in the American Civil War. In particular, it tests a prediction of theories that economic motivations were important, using district-level presidential election data. That prediction is borne out by the data, and explanations other than Northern economic concerns about southern secession appear unable to explain the results. Thus, this paper offers support to theories that economic motivations were important to the North’s decision to fight the South.

1

Contact e-mail: [email protected]. I would like to thank Brad DeLong, Barry Eichengreen, Jim Fearon, Jeffrey Hummel, David Laitin, Barry Weingast, Dan Wewers, William Woolston, and Gavin Wright for useful comments and the National Science Foundation for financial support. All errors are my own. Comments are welcome.

2

I.

Introduction Three days after Lincoln was elected president in 1860, the influential New York Tribune

wrote, “If the cotton states shall decide that they can do better out of the Union than in it, we insist on letting them go in peace.” 2 War was not a foregone conclusion. Yet, in Lincoln’s immortal words from his Second Inaugural Address, “the war came.” This paper seeks to help answer the question of why the North chose to fight the South after it seceded and offer some economic explanations. Curiously, many important studies of the Civil War barely discuss why the North decided to fight the South after it seceded, instead focusing almost entirely on the decision-making of the South (i.e. Potter 1942; McPherson 1988). History textbooks often note that the North seemed to oppose the war, then the South attacked the Union at Fort Sumter in April 1861, and then the North supported war. To simplify just somewhat, the textbooks say Lincoln—for complex and somewhat unclear reasons 3 —decided to hold onto Fort Sumter and send it foodstuffs, the South attacked it, and then war was a foregone conclusion (Bailey, et al 1998). The federal government was merely enforcing the laws, Lincoln and the Republicans said (McPherson 1998). There have been mentions of the North’s motivations in the economic literature. Gunderson (1974) argues that the North may have wanted to prevent the spread of slavery. Ransom (2001) notes that most historians think that the North’s decision was non-economic. Wright (1978) argues that a desire to maintain the Union among the leadership of the country was the cause.

New York Tribune, November 9, 1860, quoted from Bailey, et al. 1998. Potter (1942) suggests that Lincoln’s reasoning could either have been to begin the war under favorable terms, with an attack by the South, or to merely to send foodstuffs, hoping to avoid conflict (p. 374).

2 3

3 The basic problem with the literature to date is that it lacks empirical substantiation. Explanations of the North’s decision-making are entirely anecdotal (e.g. McPherson 1988), a laundry list of possible causes (e.g. Stampp, 1970), only mentioned in passing (e.g. Wright 1978), or ignored entirely (e.g., Potter 1942; Bailey, et al 1998). The result is that historians, economic or otherwise, have little idea about why the North made the decision that it did. To my knowledge, this study is the first attempt in the economics literature to systematically try to answer the question of why the North chose to fight the South. It is also the first attempt anywhere to test a hypothesis about the causes of the North’s decision and to quantitatively offer a sense of the importance of some of the North’s motivations. To do this, the study uses controlled regressions with large and relatively disaggregated datasets on the county level. Specifically, using voting patterns as representations of the Northern population’s preferences, this paper empirically tests if the economic motivations of its manufacturing, trading, and commercial sectors might have been an important component of the North’s decision to fight. The hypothesis that the North had economic motivations for keeping the South in the Union yields a specific prediction: counties with relatively large amounts of these economic interests should shift their votes from Democrats to Republicans between 1856 and 1864. The reason is the following: the best way to keep the South in the Union before the Civil War was to vote for the Democrats, reducing the likelihood of secession by voting for the party more accommodating to Southern slavery interests. However, the best way to keep the South in the Union during the War was to vote for the Republicans, who were more likely to pursue the war until victory.

4 Using county-level voting data from the 1856 and 1864 presidential elections 4 and the 1840 and 1850 censuses, I find that, indeed, there is a statistically significant shift toward the Republicans associated with several economic variables. This shift toward the Republicans associated with these economic variables together amounts to 8 percent of the electorate in Northern states. The remainder of the paper proceeds as follows. Section II reviews the arguments the causes of the North’s decision. Section III explains the empirical methodology, while Section IV describes the paper’s data. Section V presents the results and argues that other explanations are unlikely to account for the reported shifts in voting patterns. Section VI concludes.

II.

Theory: Explanations of the North’s Decision to Fight

The debate on the economic motivations of the Civil War’s parties has been raging for many decades. The classic argument for the importance of economic motivations in the Civil War was Charles and Mary Beard’s work in the 1920s, who argued that structural economic differences were the primary cause of the Civil War (Beard and Beard, 1927). The North wanted a strong national government which could promote economic development, and the South preferred strong states and did not want to spend the money for that expansionism. While this does not go far to explain why the North did not allow the South to secede, these arguments helped spawn a literature that focused on the economic differences between the North and the South. 4

County-level voting data has been used to discern shifting interests by Egnal (2001), but he did not use controlled regressions and was asking a different question, about regional electoral realignments and its implications for the Civil War, rather than why the North chose to fight. He argues that the rise of the Great Lakes economy in the 1850s realigned the parties from the Second Party System of the Whigs and Democrats to the Third Party System of the Republicans and the Democrats.

5 There are three broad areas of economic motivations that might have been important to the North’s decision to fight the South: manufacturing interests, commercial interests, and trading interests, which I will address in turn. The first reason is manufacturing interests. Stampp (1970), who provides perhaps the most comprehensive treatment to date on the North’s decision to go to war, provides anecdotal evidence that, indeed, manufacturing interests anticipated harms from the secession of the South. 5 For example, “between January 1860 and January 1861, the average price of a share of stock in fourty-four cotton and woolen mills declined from $518.34 to $304.22.” 6 Likewise, as just one example of Northern manufacturing interests leading to a desire to avoid war before session, John Parsons Foote, the abolitionist uncle of Harriet Beecher Stowe, wrote the following to his industrialist brother Samuel Edmond Foote: “You Connecticut folks support the slave powers because you can send your clocks and buttons among the slave drivers” (Foote 1854). More specifically, there are three economic reasons why northern manufacturers might have wanted to keep the South as part of the Union. First, they may have wanted to maintain the Southern market for tariff-protected manufactured goods priced above the foreign competition. This is especially the case after Northeastern interests gained more power to raise the tariff in the wake of the break-up of the Second Party System in the 1850s, which generally had given the South a veto over tariff policy (Weingast 1998). Second, Northern manufacturing feared a long, porous Southern border, through which un-tariff-protected goods could flow, undercutting the price of their more expensive goods (Stampp 1970; Foner 1941). For example, the London Times said in March 1861, after some southern states seceded, that European businessmen had perceived this opportunity to invade the 5

Stampp analyzes archival materials, especially letters and newspapers, from Lincoln’s election in 1860 through the attack on Fort Sumter in 1861. 6 Stampp, p. 124. Also, this was not a time of deflation.

6 Northern market, claiming “The smuggler will redress the errors of the statesman.” 7 The New York Evening Post described the horrifying prospect of all imports coming through the South and “the whole country . .. given up to an immense system of smuggling.” 8 Some at the time even suggested that the agrarian West might choose to join the free-trade South to avoid large tariffs 9 , further harming both industrial and trading prospects. Indeed, Cincinnati began receiving goods tariff-free from New Orleans in 1861 (Stampp 1970). Third, manufacturers may have wanted continued access to southern primary goods, particularly cotton. Primary goods might have been more expensive from abroad for a couple of reasons. First, to maintain some political support for the tariff system among farmers, agricultural tariffs may have remained high even if the South seceded, raising the price of primary goods which would have to be imported without the South in the Union. For example, with the South in the Union, cotton was priced at the world price, but would be priced at the world price plus the tariff if it needed to be imported with a tariff. A second reason that primary goods would have been more expensive is that the South may have punitively refused to sell its cotton to the North; though the Transportation Revolution was in full swing, commodities were still bulky and somewhat costly to ship from far away, so greater transportation costs would have raised the price of cotton The second economic sector, trade, could have been harmed by the secession of the South in two ways: through losing control of the Mississippi River specifically and through harming international trade broadly. Westerners, in particular, wanted to maintain access to the lower Mississippi (McPherson 1988; Hummel 1996; Stampp 1970). Although the West’s trade was increasingly oriented toward the east, due to the construction of railroads, trade along the London Times, quoted in New York Herald, March 17, 1961, from Stampp (1970). New York Evening Post, March 12, 1961, from Stampp (1970). 9 Westerners were a minority, so they might have been unable to block the imposition of tariffs. 7 8

7 Mississippi was still substantial. As the Chicago Tribune wrote in 1861, free navigation along the Mississippi was a right to the people of Northwest “and they will assert it to the extremity of blotting Louisiana out of the map.” 10 Likewise, in December 1860, The Detroit Free Press reported that wheat, pork, hogs, wool, and flour could not be sold at that time because of restrictions in Mississippi River trade introduced by the South. 11 Though the South claimed to be committed to maintaining free access, there was no way for them to credibly commit, and, in fact, there were disruptions (Stampp 1970). International trade more broadly could also have been disrupted. An independent South could seize other territories in the Western hemisphere and set up preferential trade relationships with them which excluded the North. In any case, the North would be deprived for naval bases on the Caribbean and Gulf of Mexico, which could be used to protect against pirates and privateering (Stampp 1970). Even trade with the Pacific could be jeopardized, since the South might have come to control trade through the Isthmus of Panama. Perhaps more importantly, international trade through the North could have been disrupted for the same reason that northern manufacturers feared—a porous Southern border, through which un-tariff-protected goods could flow, encouraging traders to import through the South rather than the North (Foner 1941). For example, the same New York Evening Post article quoted above fears that “not an ounce more would be imported at New York,” all instead flowing through Southern ports. 12 Finally, the commercial sector would have been harmed by the secession of the South. Bolton and Roland (1997) develop a model of nation break-up in which the gains to union are primarily economic, such as the greater amount of free trade and specialization and economies of Chicago Tribune, Feb 25, 1961, from McPherson (1988). Detroit Free Press, December 2, 1860, from Stampp (1970) 12 New York Evening Post, March 12, 1861, from Stampp (1970). 10 11

8 scale in public goods provision available to larger nations (North 1961; Fogel 1989). 13 Since the commercial sector is more affected than average by the state of the domestic economy 14 , and secession would likely have made the economy less efficient, secession would have harmed commerce. Even at the time of the Civil War, some made similar arguments; for example, the title of an 1860 book by Kettell says it all: Southern Wealth and Northern Profits, as Exhibited in Statistical Facts and Official Figures: Showing the Necessity of Union to the Future Prosperity and Welfare of the Republic. The political economy analysis, when applied to the context of the American Civil War, suggests that these economic incentives may have been important to the Northern commercial sector. Note that the preceding explanations make the assumption that war was not inevitable; if it were, then this would not be a very interesting question. For example, if the South had chosen to try to take Western territories or had sent militias into the North to try to capture escaped slaves, then a war may have been inevitable. 15 However, it is common in the literature to take war as not being inevitable (Stampp 1970). As this paper noted at the beginning, powerful interests opposed the war, at least before the Fort Sumter attack. Indeed, the public itself did not consider war inevitable, even after the election of Lincoln in 1860 and the secession of several Southern states (Stampp 1970). Historians, economic or otherwise, have also offered a variety of non-economic explanations for the North’s decision to fight the South. Since they are not tested in this paper, I

13

Although the size of the federal government was a small fraction of what it is today, the federal government did provide services like national defense. Furthermore, one of the platforms of the Whig and Republican parties was to expand the role of the federal government in, for example, expanding development in the West. 14 The commercial sector is influenced by domestic demand shocks, but not foreign ones, removing the possible diversification across economies. They were more influenced by the overall economy than either subsistence or export-oriented farmers, the former largely unaffected by business fluctuations and the latter diversified internationally. Still today, we look at the commercial sector performance for clues on the direction the economy is taking. 15 Thanks to Barry Weingast for making this point.

9 will only mention the explanations only briefly here (please see Appendix A for more detailed explanations). In addition to the explanation above that the Fort Sumter attack spurred the North to fight, explanations include: a nationalistic commitment to preserving the Union (Nagel 1964; Ransom 1989), the leadership’s commitment to the Union (Wright 1978), a moral commitment to containing or eliminating slavery (Gunderson 1974), sectional animosity (Owsley 1941; Ransom 1989), the Republicans’ desire not to lose the Union under their watch (Stampp 1970), and a country’s desire to invest in a reputation for preventing secession (Walter 2006).

III.

Empirical Method

In this paper, I test the three economic reasons for why the North chose to fight the South listed above, along with desegregations of the broader economic reasons. I test if manufacturing interests were important and also the specific reason that Southern primary goods for manufacturing were important. I test separately if Mississippi River trade and if international trade were important. Finally, I test if commercial interests were important. I test these hypotheses by testing a prediction in presidential voting data common to all these hypotheses: voters with an economic interest in keeping the South in the Union should vote as such. Particularly, counties with relatively large sectors in these economic areas should shift their votes from Democrats to Republicans between 1856 and 1864. 16 The reasoning is the following: the best way to keep the South in the Union before the Civil War was to vote for the

16

A note on a methodology not used: I do not use the 1860 voting data for a couple reasons. First, opinions at that point in time were rapidly in flux. Potter (1942) notes that the Republican share of the vote dropped substantially between the November presidential elections and the December local elections. While one might attribute that to differences in political preferences at different levels of government, Potter interprets this as rapidly oscillating preferences for the Republican party. Second, the 1860 election is difficult to compare to the 1856 and 1864 elections because there were four parties competing in the North, with over twelve percent of the vote in the counties under consideration going to neither the Northern Democrats nor the Republicans. These votes were likely peeled off from the two main Northern parties in ways which would make it difficult to analyze the voting share components.

10 Democrats, reducing the likelihood of secession by voting for the party more accommodating to Southern slavery interests. However, the best way to keep the South in the Union during the war was to vote for the Republicans, who were more likely to pursue the war until victory. In the 1856 presidential election, before the Civil War, the best way to keep the South in the Union was to vote for the Democrats. The loss of the Republican candidate John C. Fremont to the Democratic candidate James Buchanan in the 1856 election between was widely seen as a result of a desire to keep the Union together. As Bailey, et al. put it, ‘Perhaps more damaging [to Fremont than even his personal foibles] were the violent threats of the southern “fire-eaters” that the election of a sectional “Black Republican” would be a declaration of war on them, forcing them to secede. Many northerners, anxious to save both the Union and their profitable business connections with the South, were thus intimidated into voting for Buchanan.’ 17 As the New York Times wrote, “The Southern press, of every political shade of opinion, with hardly an exception, threatens disunion in the event of defeat in the present contest for the presidency.”18 Thus, counties with relatively large sectors of the economic interests described above should, all else equal, vote for the Democrats. On the other hand, the best way to keep the South in the Union in 1864, during the Civil War, was to vote for the Republicans. In the 1864 election between Union/Republican 19 candidate Abraham Lincoln and Democratic candidate George McClellan, the parties differed substantially on their opinion of the war, since the platform of the Democrats was officially declared the war a failure and the Union party strongly supported continued fighting. 20 Although

17

Bailey, at al. 1998. See also O’Connor (1968), p. 122. New York Times, August 29, 1856, quoted by Potter (1942). 19 The Republicans cleverly joined with pro-War Democrats to temporarily form the Union party. 20 The war was, indeed, going rather poorly for the Union until shortly before the election, when the North scored several victories, like Admiral Farragut’s capture of Mobile and General Sherman’s seizure of Atlanta, Alabama (Bailey, et al. 1998). 18

11 Democratic candidate McClellan largely ignored the platform and ran on a platform of supporting the war, it was clear that the South interpreted the Lincoln victory as a strike for the Union. For example, Southern soldiers shouted, “Hurrah for McClellan!” Additionally, after Lincoln’s victory, desertions from the Southern army sharply rose, suggesting that they viewed Lincoln’s victory as a sign of the Union’s commitment to the war. 21 Thus, economic interests yield a clear prediction: in counties with a given economic interest, ceteris paribus, there should be an increase in the Republican vote share between 1856 and 1864. 22 Using county-level data, I test this hypothesis by running a controlled regression of the difference between the 1864 and 1856 Republican vote shares on the economic variables and covariates, using state fixed effects. The key, of course, is the ‘ceteris paribus’; to help ensure that the shift in votes is due to the economic variables of interest and not others, I use several techniques. First, since I am measuring a shift in votes, I am differencing the dependent variable. Thus, any unobservable which does not change its effect on the likelihood of voting for the Republicans between 1856 and 1864 is differenced out and does not matter. This also eliminates the need to control for observable but constant characteristics, like a county’s presence in the pro-South Butternut region of southern Ohio, Indiana, and Illinois. Second, I add state fixed effects to control for any characteristic common to states which might shift its influence between 1856 and 1864, thus 21

Bailey, et al., 479. Note that, aside from an absence of the economic interests described above, a belief in 1856 that a war would both be low-cost and have a high likelihood of regaining the South would also yield a failure to reject the null of no economic movtivations. In that case, while those with economic interest in the South would still vote for Republicans in 1864, they would have no tendency toward the Democrats in 1856, since a war would essentially not matter, neither costing them taxes nor risking the loss of the South. While we cannot know what expectations of a war were in 1856, we do know that “the Treasury and War departments and Congress all readily agreed that the war should last less than one year, require 300,000 troops and cost about $265,000,000” (Gunderson 1970, p. 930), hardly a low price tag. Furthermore, from the actual costs of the war and this estimate, Gunderson extrapolates that the perceived likelihood of Northern victory over the South of around 14.3 percent. In addition, the North was only becoming stronger relative to the South from 1856 to 1860, gaining relatively more population and industrialization, only increasing these numbers (Bailey et al. 1998), so in 1856, the perceived cost would have been higher and the likelihood of Northern victory lower. Thus, a failure to reject for this reason is unlikely. 22

12 meaning that my identification is off of intra-state differences. Finally, as described below, I add several covariates. First, I control for the number of foreign-born people, a bias which could go either way. Given the large immigration rates in the mid-19th century and the likelihood that immigrants will settle where other immigrants already are—and in areas with lots of manufacturing, it is reasonable to expect that the fraction of the population that is immigrant is increasing faster in counties with already higher immigrant populations. Thus, if the dominant effect of more immigration is anti-immigrant sentiment among the American-born, one would expect an increase in Republican votes, since one of the reasons for the rise of the Republican Party was anti-immigrant sentiment (Silbey 1985). On the other hand, if the dominant effect is having more foreign-born voters, then one would expect fewer Republican votes. Second, I have controlled for population density, since factors associated with cities (like cultural attitudes) are correlated with manufacturing, commercial, and trade activity and thus could be an important omitted variable. These cultural attitudes could potentially change their effect over time as ideas changed disproportionately fast in urban areas. Finally, I have controlled for the per capita employment in the armament industry 23 , without which the coefficient on manufacturing may be biased downward, since those producing or potentially producing arms would be expected to vote for the Republicans in 1856 and 1864, hoping for war. Thus, for the whole paper, when I refer to manufacturing, I refer to nonarmament manufacturing. The following is the preferred specification of the hypotheses that I test: (1)

23

yi = α + xi*β + ci*γ + ds*ξ + εi

The variable is the addition of those producing weapons and gunpowder, in per capita terms.

13 where (1) y is the change in vote share for the Republican candidate between 1864 and 1856 24 ; (2) x is a vector of explanatory variables of interest consisting of per capita manufacturing employment; per capita employment in manufacturing with Southern agricultural goods 25 , per capita employment in ocean navigation, per capita employment in navigation in states with access to the Mississippi River 26 , and per capita employment in commerce 27 ; (3) c is a vector of controls described above; and (4) d is a vector of dummy variables for the states. The presence of economic motivations predicts a positive coefficient on each element of the β vector. I weight the regression by population size, since what really matters to the question here is the total population, not an average of county-level attitudes among counties which have very different populations. Thus, weighting by population gives a more true sense of the opinion in the North at the time. 28 Note that this methodology does not consider the possible importance of pivotal districts, where a vote is particularly important, or strategic voting, in which a vote does not necessarily reflect a voter’s preferred candidate, but rather an attempts to bring about a particular outcome through voting for another candidate. Note also that this methodology assumes that popular opinion mattered to the decision to fight the South. For example, if Lincoln made the decision to fight the South without any regard for Northern opinion, then that opinion would be meaningless. For example, Wright (1978) 24

Estimating this regression less parametrically by putting the 1856 voting share on the right-hand side and either assuming a linear dependence (but with a coefficient other than 1) or separating it into several categories, along with a variable for linear dependence, may be worthwhile for future versions of the paper. 25 Cotton, sugar, and tobacco. 26 These states are Ohio, Michigan, Illinois, and Indiana. Iowa and Wisconsin are not included because they were territories at the time with little trade. 27 Commercial employment includes employment in commercial houses in foreign trade; commission houses; retail, dry goods, grocery and other stores; lumber yards and trade; internal transport; and butchers, packers, and similar industries. The effect of international trade should be taken up by the ocean navigation variable. 28 A second reason is that, since the observations represent averages within counties, smaller counties will have higher variances, assuming that voting data represents a sample not the whole population. Weighting by population size solves this problem of heteroskedasticity.

14 argues that the decisions regarding whether to fight the South “must be studied at the level of small-group behavior, albeit behavior influenced by perceptions of a social background.” 29 However, others argue that Lincoln was quite “sensitive to the drift of public opinion” (Stampp 1970). 30 Thus, nearly all agree that public opinion played a role in the decision to go to war; the question is how much. That will not be answered here, but the question of from whence that public opinion came remains a useful one. The interpretation of the results using this methodology depends upon whether one assumes that the workers for manufacturing, trade, and commercial enterprises have interests aligned with the owners, insofar as Southern secession was concerned. This assumption is quite reasonable, since harm to all of manufacturing would lower wages for laborers, work in ocean and inland navigation was specialized enough that workers might not easily find as well-paying work elsewhere 31 , and, likewise, for commercial work affected by the overall economy, workers are worse off with a worse economy. Indeed, it was a longstanding Whig and Republican belief that the interests of owners and workers were aligned, as demonstrated particularly by the notion “that the protective tariff was designed primarily to advance the interests of labor” (Foner 1970). 32 Even without this assumption, the per capita sectoral employment variables could be seen as proxies for the economic interests in an area. However, given the reasonableness of the assumption, I will interpret the results as such.

29

Wright, p. 156. Stampp, p. 189. 31 This assumes the absence of costless migration to the South, where workers could again be employed in the untariff-protected trade there. 32 Foner, p. 20. 30

15 IV.

Data

The explanatory variables on total manufacturing, total population, and foreign population come from the 1860 U.S. Census of Population and Housing, from at the Geospatial and Statistical Data Center at the University of Virginia Library. The explanatory variables on commercial activity and on the disaggregated industrial activity come from the 1840 U.S. Census of Population and Housing, Census of Manufactures, accessed from the Inter-University Consortium for Political and Social Research (IPCSR). Unfortunately, this data does not exist in an accessible form from the 1860 census. 33 These variables refer only to the men (not the women) involved in various commercial and industrial activities, but this difference is unimportant since almost no women were involved in these activities and did not vote in any case. 34 There are no obvious reasons why the census data would be biased in any way. The use of census data is standard in the literature, and respondents have no reason to be dishonest in their responses. As a test of how relevant the 1840 data is to the realities in the 1850s and 1860s, I can see how high the correlation is for the one variable that I have in both the 1840 and 1860 censuses, the fraction of the population in manufacturing. The correlation between the two variables is 0.745, weighted by population, and 0.661, unweighted, which are quite high, suggesting that the 1840 data may be a good proxy for the later data. To the extent that the variables are mismeasured, the mismeasurement should only attenuate results toward zero, unless the error (the change between 1840 and 1860) is correlated with the 1840 level. For the one variable for

33

In future versions of this paper, I hope to incorporate data from the 1860 census. There are, of course, the famous Lowell loom women, but, even in 1860, they formed a small portion of the population and, in any case, this would not add bias, but rather just make the results less precise.

34

16 which I have data in both years, the fraction of employment in manufacturing, a simple bivariate regression of the change on the 1840 levels fails to yield statistical significance. While this is just one variable, this is encouraging for the assumption that the errors for the rest of the variables are uncorrelated with the change in levels. County areas from the 1950 City and County Data Book, also accessed from ICPSR, were used to compute population densities, along with historical research to determine the size of the half-dozen counties which changed sizes or no longer existed in 1950. The dependant variable, the percentage of the presidential vote received by the Republican candidate received at the county level in 1856 and 1864 from Union states, I obtained from the ICPSR. Votes from soldiers are not included, since they are tabulated at the state level. Their precise magnitude is difficult to concern because data is given in percentages, not numbers of votes. However, with approximately 2,100,000 men participating in the Union army over the course of the war and a Union population of approximately 20,000,000, if half of the total was serving in 1864, then approximately 10 percent of the male population would not be included in the analysis (McPherson 1988). The dataset is limited to 508 counties for several reasons. First, I include only states which appeared in the 1840 census as either states or territories which became states by the 1856 elections. Second, only non-Border state members of the Union during the Civil War are included. Border states cannot be included, since they had no votes recorded for the Republican candidate in 1856, presumably since there were no local Republican organizations to distribute ballots, making shifts in voting impossible to discern. 35 Finally, I include only counties which

35

Thanks to Jeffrey Hummel for helping me understand that parties produced their own ballots at the time. The one border state which did have some votes for Republicans is Delaware (where the Republican received between 0 and 5 percent); however, for consistency, I decided not to include Delaware. A further confounding factor is that many

17 were in existence in both 1840 and 1860 censuses, resulting in 53 dropped counties from states which were states in 1840 and 82 dropped counties from states which were territories at the time. These omitted observations are more rural and less industrialized than the rest of the sample, but their omission is relatively unimportant because the weighting would make their impact minimal. 36 Summary statistics are tabulated in Table 1. I use presidential election results, rather than Congressional voting records on important votes leading up to the Civil War, for several reasons. 37 First, using Presidential elections allows me to have data closer to the time of Southern secession and the Northern declaration of war; whereas many of the important Congressional votes were in the early 1850s, presidential elections which are useful for testing predictions happened right before and after these events. Second, since there are more counties than Congressional districts, using counties yields a greater number of more precise observations. Third, Congressional districts and counties do not match up and would have required somewhat arbitrary divisions of both, which would have led to less precise measurements of the variables. Finally, Congressmen may vote strategically—for example voting in favor of pro-slavery policies in return for votes for pro-development economic policy. 38 It is likely that, due to uncertainties over electoral outcomes and coordination problems, voters are less likely to be able to vote strategically.

counties in Missouri had extremely high vote shares—some as high as 100 percent—for the Republican in 1864, perhaps a result of the Union military presence. 36 I plan to include more rigorous statistical analysis using data I do have on the dropped observations in future versions of the paper. 37 I plan to look into exactly how far suffrage had extended by the 1850s and 1860s in each of the states, since this could affect the interpretation of the regressions. 38 Weingast (1998), p. 58.

18

Table 1: Summary Statistics, Weighted by Population Size Variable Obs Mean Std. Dev. Min Change in Fraction Republican Vote Share Between 1856 and 1864 508 0.095147 0.134535 -0.303 Fraction of Population Employed in 508 0.060874 0.062564 0.000579 Manufacturing Fraction of Population in States with Mississippi Access Employed in Navigation 508 0.000671 0.003116 0 Fraction of Population Employed in 508 0.00499 0.016743 0 Ocean Navigation (International Trade) Fraction of Population Employed Manufacturing with Southern Agricultural Goods 508 0.007568 0.019564 0 Fraction of Population Employed Manufacturing Weapons and Gunpowder 508 0.000132 0.00055 0 Fraction of Population Employed in 508 0.009978 0.012624 0 Commerce Fraction of Population of Foreign Origin 508 0.185357 0.132978 0.00471 Butternut State Dummy 508 0.30115 0.45921 0 Population Density 508 2131.386 7944.713 3.303306

V.

Max 0.762 0.280036

0.043269 0.179095

0.213025

0.008189 0.168109 0.590783 1 36984.95

Results & Interpretation

Table 2 presents the results of the preferred specification. With high significance, the fraction of employees employed in manufacturing predicts a shift in votes toward the Republicans, with a coefficient of 0.856. Likewise, the regression shows high significance for the fraction of the population employed in ocean navigation, the variable for international trade and the fraction of the population employed in commerce. The variable for Mississippi River trade, that for manufacturing using Southern agricultural goods, and that for employment making armaments are not significant. Using the summary statistics weighted by population of the county, the interpretation of the results is as follows. If a one percentage point increase in fraction of the population in manufacturing increases the voting share for Republicans by 0.865 percentage points, then, given

19 an average employment rate in manufacturing of 6.09 percent, then (multiplying the two together) 5.27 percent of the total population changed its votes to Republicans out of concern for manufacturing.

Table 2: Preferred Specification of Regression on Change in Fraction Republican Vote Share Between 1856 and 1864, with State Fixed Effects Fraction of Population Employed in Manufacturing Fraction of Population Employed Manufacturing with Southern Agricultural Goods Fraction of Population Employed in Ocean Navigation (International Trade) Fraction of Population in States with Mississippi Access Employed in Navigation Fraction of Population Employed in Commerce Fraction of Population Employed Manufacturing Weapons and Gunpowder Fraction of Population of Foreign Origin Population Density Constant Observations R-squared Absolute value of t statistics in parentheses * significant at 5%; ** significant at 1%

0.856 (5.69)** 0.107 (0.25) 1.527 (4.08)** 2.352 (1.32) 2.263 (4.42)** 3.464 (0.37) -0.189 (3.10)** -2.60E-09 (8.05E-07) -0.147 (2.20)* 508 0.42

Similar analyses can be done for the other significant variables. An increase in Republican vote share by 1.527 percentage points for each percentage point increase in the fraction of the population involved in ocean navigation, given a mean fraction of 0.499 percentage points, means a shift toward for the Republicans of 0.76 percentage points. Likewise, an increase in Republican vote share of 2.263 percentage points with each percentage point increase in commercial employment and a 0.998 percentage share of the population in commerce imply an increase in the Republican vote share of 2.258 percentage points associated with commerce.

20 Since these two coefficients are greater than one, these results imply that those outside of the profession are also affected by considerations of ocean navigation and commerce. This result is reasonable, since others, like family members and services for participants in these sectors, also have a stake in them. Adding these three economic components together—manufacturing, international trade, and commerce—respectively multiplied by their sample means, leads to an increase in the Republican vote share of 8.027 percentage points (p-value 0.000; t-statistic 8.08, for the linear hypothesis). Given that the overall shift was only 9.515 percentage points and the Republican candidate received only 44.79 and 54.31 percent of the vote in 1856 and 1864, respectively, this is quite substantial. The insignificant variables here also yield interesting results. Returning to the theory of Section II, there were three reasons that manufacturing interests might matter—access to Southern markets, fear of a porous Southern border, and access to Southern primary goods. The insignificant coefficient on industries using Southern primary goods, with a modest standard error (0.434), suggests that the first two reasons may be more important than the final one. Likewise, there are two reasons that trade might matter—access to the Mississippi River specifically and international trade more broadly. The insignificant coefficient on Mississippi River trade suggests that the latter may have been more important than the former, though the standard error on the coefficient is large, meaning that this result could be driven by noisy data rather than a lack of an effect.

21 A. Robustness checks

These results are robust to several specification checks. Hederoskedastic-robust standard errors leave the significant variables as significant, except for putting the commerce variable slightly below significance at the five percent level (p = 0.058). The results are broadly robust to the removal of the fixed effects, except that the coefficient on ocean navigation becomes insignificant and the coefficient on Mississippi trade becomes significant. Controlling for per capita agricultural revenue and per capita church accommodation (which might matter because of the religious connection to abolitionism) does not substantively change the results. Also, using different variables for the reliance on Southern agricultural goods (just cotton; cotton and mixed linens; cotton, mixed linens, and sugar; cotton, mixed linens, sugar, and tobacco) does not make a difference.

B. Interpretation

At worst, these results lead to a failure to reject the hypothesis that economic concerns were an important motivation for the North’s decision to fight the South, since the prediction has been satisfied. However, I argue that the interpretation goes further than that, to offering support for the hypothesis. There are two main problems to be addressed in further interpreting these results. The first is the conventional problem of omitted variable bias: the possibility that variables other than those in the regression explain the shift in voting patterns. Even if there is no problem of omitted variable bias, the second is interpreting the coefficients as resulting to economic concerns about holding onto the South, rather than some other concern.

22 Of course, it cannot be proven definitively that there is no problem of omitted variable bias. However, the state fixed effects should go a long way toward controlling for many unobserved characteristics—those common to any given state in the sample. Also, as mentioned above, the coefficients are robust to controlling for per capita church accommodation and per capita agricultural profits. Perhaps more importantly the differencing methodology of the study goes a long way toward removing omitted variable problems. Due to the differencing in the 1856 and 1864 electoral outcomes, the only omitted variables that would matter are those which change their effect between 1856 and 1864, not those which affect preferences for Republicans, per se. The second issue in interpretation is that of attributing this shift in voting patterns toward concerns about keeping the South as part of the Union. As described above, there are good economic reasons to believe that these economic interests would be important. However, to believe that these shifts in voting are attributable to concern about the South, I must argue that there were not other concerns that could have caused this shift. One concern which might easily have affected economic interests is the tariff. 39 However, with the increasing tariffs that came with the passage of the Morrill Tariff in early 1861 and subsequent increase to pay for the war, the tariff became a less important issue to manufacturers, meaning that, all else equal, they would be more likely to vote for pro-low tariff Democrats in 1864 (Taussig 1910). Likewise, commercial interests and international traders who would have benefited from lower tariffs would have become more likely to vote for Democrats in 1864 as a result of the increasing tariffs, which were harming them. Thus, the effect of concerns for the tariff would, if anything, decrease the magnitude of the results.

39

Thanks to Jeffrey Hummel for his help understanding the history of the tariff at this time.

23 A second issue could be that it was not concern for keeping the South in the Union, but rather profiting off of the war itself that drove manufacturing and commercial interests. However, I controlled for the armament industry. And, in any case, if profiting from the war were an important motivation, one would predict voting patterns that would be more likely to lead to more war in both years—that is, voting Republican and both years and a prediction of a low vote change, which cannot explain the results reported above. A third question might be what was happening to interests other than those I test for, like agriculture, over the time between 1856 and 1864: perhaps they were trending away from the Republicans in ways that make it appear that the economic interests whose effects I am testing are increasing their support for the Republicans. Indeed, as Weingast (1998) argues, there was a shifting of coalitions around this time: partly because of increased immigration, Northern business interests became more willing to accept Western expansion through the provision of free land. So perhaps farmers in the east or west may have been more willing to vote for Republicans instead of the pro-agrarian Democrats across time. However, that merely strengthens these results, since the effects of the agricultural interest are trending in the same direction as the economic interests for which I am testing. In any case, though, as noted above, controlling for per capita agricultural profits did not affect the results. A fourth issue is that, as discussed in the data section, the population voting in 1856 and 1864 are somewhat different, since soldiers votes were counted at the state level and could not be included here. Soldiers did vote differently than the rest of the population, but they voted more for Republicans. So, this change in the sample biases result towards zero, since removing soldiers removes more Republicans than Democrats from the 1864 sample.

24 Finally, for only the international trade variable, it might be argued that their primary concern was not so much keeping the Union intact as avoiding war, which tends to impede international trade. In that case, though, one would predict a consistently high voting share for the Democrats, the decision less likely to lead to war, which cannot explain the shift in areas with large amounts of ocean navigation. 40 Note also how well these economic explanations fit with the standard story of there being resistance to war up until the Fort Sumter attack and then support afterwards (Foner 1941). The Fort Sumter attack was precisely the point at which the best way to keep the South in the Union flipped from accommodation to war; thus the standard story might be mistaking a change in information for a change in preferences. The typical argument is that Northern preferences changed: the North was ‘galvanized’ by the attack on Fort Sumter and were then willing to be led into war. However, this galvanization also coincided with new information: the South was, indeed, intent on seceding; the first-best option of peacable reunionization was no longer an option, so it was time to go to the second-best option. Stampp (1970) notes that, because of their commanding political and economic positions the compromise drive was always spearheaded by eastern merchants and manufacturers.” 41 For example, “practically every prominent New York capitalist” met in December, 1860, to appeal to the South for compromise. 42 Bostonian businessmen acted similarly. However, once it was clear

40

A separate issue from either omitted variables bias or attribution of vote shifts to concern about Southern secession is that of regression toward the mean. (If, in 1856, these three economic variables were associated with voting for the Democrats and then, in 1864, they were still associated with voting for the Democrats, just less so, that could be attributable to regression toward the mean.) However, regression toward the mean is not an issue here, since I am using means of county-level aggregates, which are unlikely to exhibit the random deviations than would observations of individuals. To the extent that there are small-sample problems for the counties means, I am weighting away from those toward those with larger samples underlying the means. 41 Stampp (1970), p. 126. 42 Ibid. Stampp notes, though, that these attempts at compromise were superficial at best and fraudulent, at worst. For example, New York merchants refused to come out in favor of a law to liberalize the American shipping

25 that it would take war to keep the South in the Union, areas that had been in favor of reconciliation turned quickly toward war. “Union rallies were most numerous in the large cities of the East and in the border areas of the West where the people favored conciliation overwhelmingly. There the specter of financial ruin made other issues shrink into insignificance.” 43 Thus, these economic hypotheses also agree with anecdotal evidence on the timing of changes in support for the war.

VI.

Conclusion: Next Steps and Extensions

I have argued that an approximately 8 point shift in voting for Republicans between 1856 and 1864 is largely attributable to the economic interests of manufacturing, international trade, and commerce. Causes other than a desire to keep the South in the Union seem able to explain this shift, suggesting that a concern for keeping the South as part of the Union motivated this 8 percent of the electorate. If 8 percent of the population were on the margin between voting for the Democrats and Republicans and changed their votes to keep the South in the Union, then the actual fraction of the population with these economic interests was, no doubt, much larger, since many with this interest either kept voting Republican or were motivated enough by other concerns to keep voting Democrat. Thus, given that Lincoln could not have taken any action that did not reflect the will of at least a substantial part of the population, this paper offers strong evidence that economic motivations were important in the North’s decision to fight the South after it seceded. 44

monopoly, which increased shipping in the costs in the South, despite the great irritation that this caused to the South. 43 Stampp (1970), p. 126. 44 However, these results should not be taken to mean that economic motivations were the only or even the most important cause of the North’s decision to fight the South.

26 Furthermore, I show that the relevant economic interests seem to be of several types. First, manufacturing interests are most important, but I find no evidence that the import of Southern primary goods is important, suggesting that it was access for tariff-protected goods to the Southern market and keeping un-tariff-protected goods from flowing in from the South that manufacturers cared most about. I show that international trade, endangered especially in the Gulf of Mexico and from the possibility of more imports flowing through the South, was also likely an important interest, but find no evidence that trade access to the Mississippi River was important, perhaps because of the reorientation of trade to railroad-driven east-west routes. Finally, I show that commercial interests were also likely important, again suggesting that trade with the South was likely an important motivation. More broadly, this paper is a case study on why states tend to try to fight secessionist movements, even in democracies where leaders are less able to extract rents from seceding regions (e.g., Bolton and Roland 1997; Walter 2006). 45 These results suggest that powerful groups may benefit economically from having a union and that their influence may encourage countries to resist secession, even of relatively poor regions like the South.

45

Resistance to secession makes sense in dictatorships, in which a leader can extract rents from his subjects, but that is not the case in democracies. There are examples of a democracy breaking up willingly , like Czechoslovakia, Belgium and Netherlands, Norway and Sweden, and Singapore and Malaysia. However, Yugoslavia, Ethiopia/Eritrea, and Ireland/United Kingdom, for example, were bloody. Even in cases, like Spain or Canada, where portions wish to secede, the national government tries hard to keep them part of the country, through tax transfers or partial autonomy.

27

Appendix A: Non-Economic Explanations for the North’s Decision to Fight the South Several non-economic explanations have been offered for the North’s decision to fight the South. As Ransom (2001), one of the leading economic historians of the Civil War notes, “Most writers argue that the decision for war on Lincoln’s part was not based primarily on economic grounds.” 46 The first is that the nature of American nationalism meant that Americans were ideologically committed to preserving a Union (Nagel 1964; Ransom 1989). As Nagel (1964) argues, “The Union became a supreme ideal and image, as it increasingly answered [the] desire to elude personal and national peril. It proved a reassuring symbol of the American dream in the nation’s first age of insecurity.” 47 This explanation is perhaps most related to the standard textbook explanation that strong nationalistic identities spurred by the attack on Fort Sumter, both for the Union and the South, was an important cause of the war (Bailey et al. 1998). Gunderson (1974) argues that a moral desire to contain slavery was the main motive for the North. 48 He argues that the cost of paying for the emancipation of the slaves—2.5 to 3.7 billion dollars—is on the same order of magnitude as the cost of the war (Goldin 1973; Goldin and Lewis 1975). Thus, if the North were interested in freeing the slaves, a war may have been a relatively cost-effective way of doing it. Furthermore, if the North’s interest was in containing slavery not only in the United States but also in the rest of the Western hemisphere, it could only accomplish its goal if it did not have a sovereign state south of it promoting the spread of slavery as well (Ransom 1989). 49 As Ransom (1989) notes, “the South of the mid-nineteenth century was an expansionist system that coveted land to the west and to the south.” Weingast (1998) responds, though, that this expansionism was not motivated by a desire for land per se, but rather 46

Ransom, 2001. Nagel, p. iv. 48 Stopping the spread of slavery had several components to it. First, the morality-based ideology of abolitionism, which was spurred especially by the Second Great Awakening and later by activists like Harriet Beecher Stowe, wanted slavery to end (McPherson 1988). However, much of anti-slavery sentiment was due to the Free Soilers, who wanted to keep slaves out of the territories, both to avoid low-wage competition and possibly for reasons of racism; there is little reason that Free Soilers would care about the spread of slavery elsewhere in the western hemisphere, as long as it was not spreading into the territories (Foner 1970). 49 For example, the Constitution of the Confederate States of America said, “In all such territory the institution of negro slavery, as it now exists in the Confederate States, shall be recognized and protected be Congress and by the Territorial government” (http://www.yale.edu/lawweb/avalon/csa/csa.htm). Furthermore, the Confederacy was interested in annexing Cuba and could have used their power to thwart Northern attempts to eliminate slavery in the hemisphere. 47

28 a desire for political balance in the Senate. Nevertheless, the one common denominator of the Republican party was the opposition to the expansion of slavery—its free soil platform—and any threat of this would have galvanized much of the Republican base (Stampp 1970). 50 Another potential contributor, suggested by Ransom (1989) is that there was pent-up animosity between the North and the South, and the North was motivated in part by the “irritation, frustration, and tension of almost continual political crises” 51 so strong feelings were easier to rile up. Relatedly, some argue that strong sectional identities, particularly vis a vis the other part of the country, were important (Owsley 1941). Another contributor is the political economy of the political parties—that secession would decimate the Republicans (Stampp 1970). The mere association of the Republicans with the destruction of the Union, along with the undermining of the Republicans’ credibility due to their prior commitment to maintaining the Union, would have ended existence of the new party. Additionally, Walter (2006) argues, using data on all separatist movements between 1956 and 2002, that governments invest in reputation; they are less likely to accommodate separatist challenges if: (1) the value of land that could come under dispute in the future is high or (2) the government is likely to face a second or third challenger in the future. As reviewed by Walter (2006), existing literature focuses on present stakes and capabilities: governments balance the economic, strategic, or psychological value of land against the costs of defending against secession, determined by the relative strengths of the government and challenger. As Fogel (1989) and Schlesinger (1922) point out, there were substantial parts of the county that in the previous decades had considered secession, so it is possible that Walter’s explanation may matter for the American experience. Indeed, some in 1861 suggested that New York City secede so that it could have lower tariffs and compete with Southern ports (Foner, P. 1941).

50

As Stampp notes, “Their hostility to slavery and the southern way of life was a partial expression of the broader nineteenth-century middle-class liberalism which they espoused. They put their faith in individual freedom, in the right of each man to the fruits of his enterprise.” (p.148). 51 Ransom p. 167.

29

Table 3: By-State Distribution of Counties in Sample State CONNECTICUT ILLINOIS INDIANA IOWA MAINE MASSACHUSETTS MICHIGAN NEW HAMPSHIRE NEW JERSEY NEW YORK OHIO PENNSYLVANIA RHODE ISLAND VERMONT WISCONSIN Total

Number of Counties 8 83 87 19 13 14 29 8 18 56 79 54 5 13 22 508

Percentage of Counties 1.57 16.34 17.13 3.74 2.56 2.76 5.71 1.57 3.54 11.02 15.55 10.63 0.98 2.56 4.33 100

Percentage of Population 2.75 9.22 7.88 2.02 3.25 7.36 3.98 1.72 3.58 22.73 13.01 16.42 1.04 1.86 3.18 100

30 References Bailey TA, Kennedy DM, Cohen L. 1998. The American Pageant Boston: Houghton Mifflin. 11th ed.

Beard CA, Beard MR. 1927. The Rise of American Civilization, Vol. 1. New York: Macmillan.

--- 1930. The Rise of American Civilization, Vol. 2. New York: Macmillan. 3-121pp.

Bolton P, Roland G. 1997. The Breakup of Nations: A Political Economy Analysis. QJE. 112 : 1057-1090.

Clubb JM, Flanigan WH, and Zingale, Nancy H. 2006. Electoral Data for Counties in the United States: Presidential and Congressional Races, 1840-1972 [Computer File] ICPSR08611-v1.

Egnal M. 2001. The Beards were Right: Parties in the North, 1840-1860. Civil War History. 47 : 30-56.

Fogel RW. 1989. Without Consent Or Contract:: The Rise and Fall of American Slavery, New York: Norton.

Foner E. 1970. Free Soil, Free Labor, Free Men: The Ideology of the Republican Party before the Civil War, London: Oxford University Press.

Foner PS. 1941. Business and Slavery: The New York Merchants & the Irrepressible Conflict, Chapel Hill: University of North Carolina Press.

31 Foner PS, Shapiro H, eds. 1994. Northern Labor and Antislavery: A Documentary History, Westport, CT: Greenwood Press.

Foote JP. March 14, 1854. Letter to Brother Samuel E. Foote. http://cgi.ebay.com/ws/eBayISAPI.dll?ViewItem&ih=014&sspagename=STRK%3AMEW A%3AIT&viewitem=&item=330064672118&rd=1&rd=1.

Geospatial and Statistical Data Center, University of Virginia Library. 2004. 1860 U.S. Census of Population and Housing. http://fisher.lib.virginia.edu/collections/stats/histcensus/.

Goldin C. 1973. The Economics of Emancipation. Journal of Economic History. 33 : 66-85.

Goldin C, Lewis F. 1975. The Economic Costs of the American Civil War: Estimates and Implications. Journal of Economic History. 35 : 299-326.

Gunderson G. 1974. The Origin of the American Civil War. Journal of Economic History. 34 : 915-950.

Hummel, J. 1996. Emancipating Slaves, Enslaving Free Men: A History of the American Civil War, Chicago: Open Court.

Inter-university Consortium for Political and Social Research. 1999. United States Historical Election Returns, 1824-1968, Ann Arbor, MI: Inter-university Consortium for Political and Social Research. 2nd ICPSR ed ed.

32 Kettell TP. 1860. Southern Wealth and Northern Profits, as Exhibited in Statistical Facts and Official Figures: Showing the Necessity of Union to the Future Prosperity and Welfare of the Republic, New York: G. W. & J. A. Wood.

McPherson JB. 1988. Battle Cry of Freedom, New York: Oxford University Press.

Nagel PC. 1964. One Nation Indivisible: The Union in American Thought, 1776-1861, New York: Oxford University Press.

North D. 1961. The Economic Growth of the United States, 1790-1860, Englewood Cliffs: Prentice Hall.

O'Connor TH. 1968. Lords of the Loom, New York: Charles Scribner's Sons.

Owsley F. 1941. The Fundamental Cause of the Civil War: Egocentric Sectionalism. Journal of Southern History. 7 : 3-18.

Potter DM. 1942. Lincoln and His Party in the Secession Crisis, New Haven: Yale UP.

Ransom RL. 1989. Conflict and Compromise: The Political Economy of Slavery, Emancipation and the American Civil War, Cambridge: Cambridge University Press.

---2001. The Economics of the Civil War. http://eh.net/encyclopedia/article/ransom.civil.war.us.

Schlesinger AM. 1922. New Viewpoints in American History, New York: Macmillan.

33 Silbey JH. 1985. The Partisan Imperative : The Dynamics of American Politics before the Civil War, New York: Oxford University Press.

Taussig FW. 1910. The Tariff History of the United States, New York: The Knickerbocker Press.

Walter BF. 2006. Building Reputation:Why Governments Fight some Separatists but Not Others. American Journal of Political Science. 50 : 313-330.

Weingast BR. 1995. The Economic Role of Political Institutions: Market-Preserving Federalism and Economic Development. Journal of Law, Economics, & Organization. 11 : 1-31.

--- 1998. Political Stability and Civil War: Institutions, Commitment, and American Democracy. In Analytic Narratives, ed. Bates, Robert et al, Princeton: Princeton University Press.

Woodworth SE. 1996. the American Civil War: A Handbook of Literature and Research, Westport, Conn.: Greenwood Press.

Wright G. 1978. The Political Economy of the Cotton South: Households, Markets, and Wealth in the Nineteenth Century, New York: W. W. Norton.