Journal of Political & Military Sociology

The Civil War’s Demographic Impact on White Males in the Eleven Confederate States: An Analysis by State and Selected Age Groups1

David A. Swanson

University of California, Riverside

Richard R.Verdugo

US National Education Association (retired)

Recent research has put the likely number of Civil War male deaths on both sides of the conflict at 750,000, challenging the conventional number of approximately 620,000. Using similar data and a different methodology, this study supports the newer results in terms of the Confederate states, where we find approximately 346,000 male deaths, a number that far exceeds a long-accepted estimate of 126,000. In fact, our number is about 88,000 more deaths than estimated by even more recent research, putting the number at 258,000. In constructing our number, we estimate the demographic impact of the Civil War on white males who were aged 20–54 in 1870 in the eleven Confederate States. Our approach takes into account both mortality and migration, but it excludes fertility because the youngest age we examine is 20 years. We apply an impact analysis approach using extracts from the micro-level data from the 1850, 1860, and 1870 census counts assembled by the Minnesota Population Center. Using the 1850 and 1860 census counts, we constructed ten-year Cohort Change Ratios (CCRs) for white males by five-year age groups and applied them to 1860 white males aged 10–14, 15–19 . . . 40–44 by state to project the expected number of white males by five-year age groups, 20–24, 25–29 . . . 50–54 for 1870. The results were aggregated into a single age group of 20–54 and compared with the 1870 census numbers for this same age group by state and for all eleven states combined. The Civil War’s demographic impact on white males in all eleven states of the Confederacy due to the combined effects of mortality and migration is estimated by subtracting the 1870 expected number (1,393,125) for age group 20–54 generated by the CCR method from the 1870 actual (census) number (1,047,323). The difference is –345,802 (–24.8%). We also find substantial absolute and relative variation across the eleven states in regard to the war’s demographic impact and discuss these results. Our estimate of Confederate deaths brings the total number of dead on both sides to nearly 850,000, which exceeds the total number of US military deaths resulting from every war and military action in which the United States has participated since the Civil War’s end.

 

As Hacker (2011a) observes, the estimated number of male deaths attributed to the Civil War on both sides became fixed at approximately 620,000 around the start of the twentieth century despite a higher estimate of at least 850,000 by the Director of the 1870 Census, Francis A. Walker, as well as suggestions of higher estimates by others. The estimate of approximately 620,000 remained the conventional figure for over 100 years until it was challenged by Hacker (2011b, 348), who dramatically reopened the discussion on Civil War deaths with an empirically based estimate that suggested the number was higher (in the range of 650,000 to 850,000), with the most likely figure being 750,000. While we will never know the exact number, a reasonable estimate is important, not the least because of the trauma caused by the war and its long-lingering after-effects (Faust 2006; Faust 2008; McPherson 1982; Neely 2007). Hacker (2011b) provides a detailed analysis of the shortcomings of casualty estimates made by others. Thus, we do not repeat the critique here.

Even using the conventional number of 620,000, Civil War deaths are higher than total American fatalities in the Revolution, the War of 1812, the Spanish-American War, World War I, World War II, and the Korean War combined (Faust 2008, xi). If we accept Hacker’s most likely number of 750,000, the Civil War death toll would exceed not only these numbers but also those from the Vietnam War (about 60,000), as well as the deaths by all causes to active duty service members from 1980 to 2010, which number 48,834 (Defense Manpower Data Center 2011). As we aim to illustrate in this article, the number of Civil War deaths may exceed 750,000 and even approach 850,000.

Although there are exceptions (Holmes and Vinovskis 1992; Neely 2007), the usual approach to estimating the demographic impact of the Civil War is to examine mortality among soldiers (Faust 2006; Vinovskis 1989). However, in this article we follow Hacker’s (2011b) lead and take a broader perspective by estimating the full demographic impact of the Civil War, rather than confining our interest to the war’s mortality effects on soldiers. Our approach, the Cohort Change Ratio (CCR) method (Baker et al. 2017), takes into account the complete set of demographic components of change: fertility, mortality, and migration. This approach has already been used by Swanson (2008) and Swanson et al. (2009) in assessing the demographic impact of Hurricane Katrina on the Mississippi gulf coast and New Orleans, respectively.

In this study, we use this same approach by examining the impact of the Civil War on the number of white males who were between the ages of 10 and 44 in 1860 in each of the eleven Confederate states.2 This population is of interest for at least two reasons: (1) it provided the bulk of soldiers who served in the Confederate army; and (2) deaths among Confederate forces are believed to be underestimated even more than those of Union forces (Faust 2006, 2008; Hacker 2011a, 2011b). By 1866, the US government was engaged in an organized attempt to record every Union soldier’s grave and all Union soldiers’ deaths, something not done for the Confederate dead (Faust 2008, 219–220, 241). In addition, the record suggests that Robert E. Lee systematically and intentionally undercounted Confederate deaths in order to conceal his actual troop strength from Union forces (Faust 2006, 253).

The basis of our examination consists of projections of the 1860 white male population of the Confederacy by state (in five-year age groups from 10–14 to 40–44) to 1870. We compare the “expected” results (in five-year age groups from 20–24 to 50–54) to the “actual” results among white males in five-year age groups (from 20–24 to 50–54) as reported in the 1870 census for each of the states that were in the Confederacy. The projection method we employ (the Cohort Change Ratio) encompasses all three of the demographic components of change: fertility, mortality, and migration (Swanson 2008; Swanson and Tayman 2012; Swanson et al. 2009; Baker et al. 2017). Note, however, that fertility does not come into play in this analysis because the population we examine (white males aged 10 to 44 in 1860) would not have been affected by fertility during the period 1860 to 1870 because all of the men would have been born prior to 1851.

The Data: 1850, 1860, and 1870 Census Counts

The data used in this study were provided by the Minnesota Population Center (MPC), the organization used by Hacker (2011b) to construct his census-based estimates. We draw on data collected under the auspices of the “North Atlantic Population Project,” one of several highly regarded data sources established by the MPC (Ruggles et al. 2010). As did Hacker (2011b), we use the “Public Use Microdata Set” (PUMS) developed by MPC for census years bracketing the Civil War. In our case, we use PUMS data extracted from the 1850, 1860, and 1870 census counts of the US. These data sets are samples of the original census records, where the sampling ratio is approximately 1 of 100 records. There are several advantages to using these data in lieu of hardcopy census reports. First, they are machine-readable and, once the data are extracted, they can be easily imported into Excel or a statistical software package, such as SPSS, SAS, STATA, or NCSS. Second, the PUMS files have been cleaned, edited, and assembled using high levels of quality control. Third, because they are individual-level data, the extracted sample data can be aggregated in different ways to suit a given analysis and automatically weighted in order to reproduce the census counts.

In the initial extraction, we selected weighted white males by age and state for each of the three census counts (1850, 1860, and 1870) along with the weights. It is important to note that, throughout our analysis, the use of the term “data” will refer to white males. We used the recode, frequency, and filter procedures provided by the MPC to generate output that could be imported directly into Excel. This allowed us to have an Excel file with the aggregated census counts for the selected five-year age groups appropriately weighted: (1) from 10–14 to 40–44 for the 1850 and 1860 census years; and (2) from 20–24 to 50–54 for the 1860 and 1870 census years.

Once the state-specific data were in Excel, we set up the CCR calculation for each cohort of interest by dividing the number of white males in a given five-year age group (e.g., 20–24) in 1860 by the number in the age group that was ten years younger (e.g., 10–14) in 1850. This was done for each of the five-year age groups that fell within the range from 20–24 to 50–54 as found in the 1860 census and the corresponding age groups ten years younger found the age range 10–14 to 40–44 in the 1850 census such that the following seven Cohort Change Ratios were calculated:

10CCR10–14 = (white males aged 20–24)1860/(white males 10–14)1850,

10CCR15–19 = (white males aged 25–29)1860/(white males 15–19)1850,

10CCR20–24 = (white males aged 30–34)1860/(white males 20–24)1850,

10CCR25–29 = (white males aged 35–39)1860/(white males 25–29)1850 ,

10CCR30–34 = (white males aged 40–44)1860/(white males 30–34)1850 ,

10CCR35–39 = (white males aged 45–49)1860/(white males 35–39)1850,

10CCR40–44 = (white males aged 50–54)1860/(white males 40–44)1850.

These seven CCRs were multiplied by each of the seven five-year age groups from 10–14 to 40–44 in the 1860 census, which yielded a projection of the expected number of white males in each of the corresponding five-year age groups from 20–24 to 50–54 for 1870:

(expected number of white males aged 20–24)1870 = (white males aged 10–14)1860 *[(white males aged 20–24)1860/(white males 10–14)1850],

(expected number of white males aged 25–29)1870 = (white males aged 15–19)1860 *[(white males aged 25–29)1860/(white males 15–19)1850],

(expected number of white males aged 30–34)1870 = (white males aged 20–24)1860 *[(white males aged 30–34)1860/(white males 20–24)1850],

(expected number of white males aged 35–39)1870 = (white males aged 25–29)1860 *[(white males aged 35–39)1860/(white males 25–29)1850],

(expected number of white males aged 40–44)1870 = (white males aged 30–34)1860 *[(white males aged 40–44)1860/(white males 30–34)1850],

(expected number of white males aged 45–49)1870 = (white males aged 35–39)1860 *[(white males aged 45–49)1860/(white males 35–39)1850],

(expected number of white males aged 50–54)1870 = (white males aged 40–44)1860 *[(white males aged 50–54)1860/(white males 40–44)1850],

Once we obtained the expected 1870 number of white males for each of the seven five-year age groups, we subtracted from it the number of white males counted in the 1870 census for the corresponding age group, which yielded an estimate of the effect of the Civil War on each of the seven year age groups as of 1860 as follows:

(Impact of the Civil War on white males aged 10–14 years)1860 = (expected number of white males aged 20–24)1870 – (census count of white males aged 20–24)1870

(Impact of the Civil War on white males aged 15–19 years)1860 = (expected number of white males aged 25–29)1870 – (census count of white males aged 25–29)1870

(Impact of the Civil War on white males aged 20–24 years)1860 = (expected number of white males aged 30–34)1870 – (census count of white males aged 30–34)1870

(Impact of the Civil War on white males aged 25–29 years)1860 = (expected number of white males aged 35–39)1870 – (census count of white males aged 35–39)1870

(Impact of the Civil War on white males aged 30–34 years)1860 = (expected number of white males aged 40–44)1870 – (census count of white males aged 40–44)1870

(Impact of the Civil War on white males aged 35–39 years)1860 = (expected number of white males aged 45–49)1870 – (census count of white males aged 45–49)1870

(Impact of the Civil War on white males aged 40–44 years)1860 = (expected number of white males aged 50–54)1870 – (census count of white males aged 50–54)1870

The aggregated and weighted raw data counts on which our analysis is based are found in tables 1 through 3 for the years 1850, 1860, and 1870, respectively. The data show that in 1850, the population of white males aged 10 to 44 in the eleven states of interest stood at 1.2 million. The largest age cohort was among those 10–14, with a subpopulation size of 293,000 white males. We see that Tennessee had the largest male population aged 10–44 in 1850 (217,438), and Florida had the smallest (15,689). By 1860, the male population in Tennessee aged 10–44 had grown to 242,352 and remained the largest among the eleven states of the Confederacy. Florida’s white male population aged 10–44 also grew (to 23,224), but it remained the smallest among the eleven states.

Table 1.  Non-Hispanic White Males by Selected Five-year Age Groups & State: 1850 Census

State/Age

10–14

15–19

20–24

25–29

30–34

35–39

40–44

Total

Alabama

30,660

22,562

17,593

14,329

13,033

11,323

8,913

118,413

Arkansas

11,949

9,082

7,409

6,985

7,596

4,960

3,580

51,560

Florida

2,997

3,125

3,103

2,066

1,593

1,567

1,238

15,689

Georgia

36,629

26,838

25,691

18,475

14,437

12,974

10,829

145,872

Louisiana

17,162

10,707

13,038

14,786

14,215

12,983

8,493

91,383

Mississippi

23,528

14,921

15,577

11,193

10,615

9,780

6,859

92,472

North Carolina

38,664

29,212

26,265

18,656

16,045

15,569

10,938

155,348

South Carolina

20,774

12,954

12,256

9,428

9,760

7,462

6,059

78,693

Tennessee

53,197

43,469

40,007

26,220

21,990

18,586

13,969

217,438

Texas

14,227

8,198

7,637

5,746

6,312

5,287

3,701

51,107

Virginia

42,809

38,923

29,648

23,219

21,051

16,452

12,758

184,860

Total

292,594

219,991

198,222

151,102

136,647

116,941

87,338

1,202,835

Source: Minnesota Population Center (2008). See text for details.

Table 2.   Non-Hispanic White Males by Selected Five-year Age Groups & State: 1860 Census

State/Age Group

10–14

15–19

20–24

25–29

30–34

35–39

40–44

45–49

50–54

Total 10–44

Total 20–54

Alabama

36,004

28,242

27,226

22,546

17,488

12,277

9,710

8,147

9,013

153,493

106,407

Arkansas

22,880

17,532

17,052

15,139

10,377

8,380

8,027

6,866

3,183

99,386

69,023

Florida

5,083

5,207

3,485

3,897

1,512

1,631

2,410

1,603

1,141

23,224

15,678

Georgia

41,796

29,141

29,700

23,032

18,019

16,161

12,340

10,577

6,470

170,188

116,300

Louisiana

19,981

15,440

16,220

20,911

17,860

14,778

11,990

6,640

5,259

117,180

93,658

Mississippi

24,466

20,084

20,427

12,214

11,394

9,446

7,629

6,131

5,471

105,660

72,712

North Carolina

40,012

35,841

30,973

22,599

19,353

17,141

11,777

12,439

9,212

177,695

123,494

South Carolina

20,051

14,717

12,027

11,366

8,277

9,681

6,399

5,423

3,466

82,516

56,637

Tennessee

57,626

44,361

45,003

33,459

26,442

21,005

14,456

12,267

11,646

242,352

164,278

Texas

23,271

19,899

19,624

19,768

15,093

12,557

9,521

8,952

5,609

119,734

91,124

Virginia

42,456

37,996

34,316

26,087

21,832

18,098

14,788

15,182

12,685

195,573

142,988

Total

333,626

268,460

256,051

211,017

167,647

141,155

109,047

94,227

73,155

1,247,604

1,052,299

Source: Minnesota Population Center (2008). See text for details.

Table 3.  Non-Hispanic White Males by Selected Five-year Age Groups & State: 1870 Census

State/Age Group

20–24

25–29

30–34

35–39

40–44

45–49

50–54

Total

Alabama

23,306

16,594

11,651

14,353

8,301

8,059

9,529

91,792

Arkansas

20,347

14,618

9,627

10,719

6,993

7,502

6,260

76,064

Florida

5,861

3,382

2,211

2,454

2,094

2,145

1,321

19,468

Georgia

33,633

19,308

14,233

15,277

11,797

11,300

10,399

115,947

Louisiana

12,685

14,962

13,432

11,266

8,946

8,706

5,925

75,923

Mississippi

21,129

11,382

11,875

9,632

6,548

8,109

4,936

73,610

North Carolina

31,119

22,074

15,297

16,839

12,480

10,804

11,117

119,729

South Carolina

13,897

8,325

8,387

7,665

4,943

4,695

5,410

53,322

Tennessee

40,390

30,901

25,214

23,687

17,900

15,271

14,972

168,334

Texas

28,961

22,230

20,400

13,839

12,330

10,598

9,277

117,636

Virginia

33,028

23,178

17,168

17,945

13,159

17,705

13,316

135,499

Total

264,355

186,954

149,494

143,677

105,490

104,892

92,462

1,047,323

Source: Minnesota Population Center (2008). See text for details

By 1860, the number of 20 to 54 year olds (the cohort of 10–44 year olds in 1850) declined to 1.052 million white males, a drop of approximately 150,500. Between 1860 and 1870, there was also a decline in the cohort aged 10–44 as it aged to become the 20–54 age group in 1870. In 1860, there were approximately 1.248 million white males in this cohort (10–44). By 1870, as this cohort aged to become those 20–54, it declined to number 1.047 million white males, a drop of approximately 200,300.

The differences just referenced would lead to inaccurate estimates if one were to simply take differences between the same age groups across two census counts. That is, if one found the difference between the number of 20–54 year olds in 1860 and 1870, the result would not take into account the demographic determinants of population change (as shown in Equation 1 and discussed below). A better alternative would be to employ a procedure that, conceptually, controls for confounding issues, such as out- and in-migration. In this article we propose and use such a procedure.

Specifically, our approach is similar to Hacker’s (2011a) in that, like him, we use 1850, 1860, and 1870 census information by age, race, and gender, which means that our work (like Hacker’s) is subject to census errors (particularly differential error across census counts over time). Like Hacker, we also employed the 1870 census as a benchmark against which the Civil War’s effects could be estimated using an “expected” 1870 population and finding the difference between the expected 1870 population (by age) and the 1870 population counted in the census (by age). However, our approach differs from his in the construction of the “expected” 1870 population. We employ the CCRs described earlier to move the 1860 population forward to 1870; Hacker constructed 10-year age specific survival rates using massaged and averaged data from the 1850, 1860, 1870, and 1880 census counts. His aim was to develop survival rates that excluded the mortality effects of the Civil War. Once he had found what he judged to be suitable rates, he multiplied each rate by its corresponding age group in 1860 to project the expected number who survived to become ten years older in 1870.

The construction of these Census Survival Rates (CSRs) by Hacker was a laborious task, one based on native-born whites (which generally eliminated the effect of immigration by the foreign born but not the effect of emigration by the native born, points discussed by Hacker, along with detailed discussions of census count errors and how to mitigate them). The CCR approach that we use is not only age-, race-, and gender-specific, but state-specific as well. While Hacker’s CSR approach also was age-, race-, and gender-specific, it was, not state- or region-specific. Thus, the expected populations based on national level CSRs are likely to contain more “hidden heterogeneity” than the CCRs we developed, which are state-specific. This concept basically means that there is, or could be, substantial variation among the subgroups of a population of interest. For example, “hidden” within an estimate of the United State’s life expectancy at birth, there is a great deal of variation in such life expectancy by sex, race, ethnicity, location, and socio-economic class (for a discussion of this issue and why it is prudent to reduce it, see Vaupel and Yashin 1985). As such, our approach allows us to construct state-specific estimates using state-specific information. In turn, such state-specific estimates can be aggregated to the region of interest—here, the states of the Confederacy.

Methods: The Cohort Change Ratio and Population Projection

As mentioned previously, Tables 1 through 3 provide the data used in this study. The 1850 and 1860 data are used to form the ten year CCRs for five-year age groups for white males by state, from 186020–24/185010–14 to 186050–54/185040–44. As described in detail above, once calculated, the CCRs are applied to the 1860 data to project the number of white males by state among the five-year age groups for 1870: 187020–24 = (186020–24/185010–14)*(186010–14) . . . 187050–54 = (186050–54/185040–44)*(186040–44). The Cohort Change Ratios are found in Table 4.

Table 4.  Ten-Year Cohort Change Ratios for Five-Year Age Groups: (1860 Age AX+10)/(1850 AgeX)

State/CCR

186020-24/185010-14

186025-29/185015-19

186030-34/185020-24

186035-39/185025-29

186040-44/185030-34

186045-49/185035-39

186050-54/185040-44

Alabama

0.88800

0.99930

0.99407

0.85678

0.74500

0.71954

1.01115

Arkansas

1.42701

1.66693

1.40065

1.19977

1.05669

1.38443

0.88912

Florida

1.16302

1.24672

0.48716

0.78939

1.51287

1.02292

0.92173

Georgia

0.81083

0.85816

0.70139

0.87476

0.85477

0.81530

0.59752

Louisiana

0.94511

1.95307

1.36983

0.99950

0.84352

0.51140

0.61918

Mississippi

0.86819

0.81863

0.73148

0.84393

0.71866

0.62695

0.79765

North Carolina

0.80108

0.77361

0.73684

0.91879

0.73402

0.79897

0.84222

South Carolina

0.57895

0.87737

0.67530

1.02679

0.65560

0.72674

0.57198

Tennessee

0.84597

0.76972

0.66094

0.80112

0.65737

0.66000

0.83371

Texas

1.37933

2.41141

1.97634

2.18550

1.50841

1.69332

1.51548

Virginia

0.80161

0.67021

0.73639

0.77945

0.70248

0.92280

0.99425

Total

0.87511

0.95921

0.84575

0.93417

0.79801

0.80577

0.83761

Note: See text for details on calculations and data.

1. Quantitative Population Projection

Before describing the CCR method in concise mathematical notation, it is important to note here that any quantitative approach to projecting a population is constrained to satisfy various mathematical identities (Land 1986). In regard to population projections, an approach should ideally satisfy demographic accounting identities, which is summarized in the fundamental demographic equation:

Pt = P0 + B – D + I – O     [1]

That is, the population at some time in the future, Pt, must be equal to the population at an earlier time, P0, plus the births (B) and in-migrants (I) and less the deaths (D) and out-migrants (O) between time 0 and time t. The most commonly used approach to population projection, cohort-component method, satisfies the fundamental equation, but it is data-intensive (Baker et al. 2017; George et al. 2004; Smith et al. 2013; Swanson and Tayman 2012, 201) and the data from the Civil War period are insufficient to support it.

2. The Cohort Change Ratio

The CCR Method also satisfies the fundamental demographic equation. However, it has far less intensive input data requirements than does the cohort-component method. Instead of mortality, fertility, migration, and total population data, the CCR method requires data only by age from two successive censuses (Baker et al. 2017; Hamilton and Perry 1962; Smith et al. 2013, 176–181; Swanson 2008; Swanson et al. 2009; Swanson et al. 2010; Swanson and Tayman 2012, 201–205; Swanson and Tedrow 2012).

The CCR projects a population by age (and sex) from time t to time t+k using data from two successive censuses. Two steps are needed to accomplish this. The first uses data from two successive censuses to develop CCRs, and the second applies these CCRs to the cohorts of the more recent of the two censuses in order to project them into the future.3 The CCR formula is (see Swanson et al. 2010):

nCCRx,i,t = nPx,i,t / nPx–k,i,t–k     [2]

where,

nPx,i,t is the population aged x to x+n in area i at the most recent census (t),

nPx–k,i,t–k is the population aged x – k to x–k+n in area i at the 2nd most recent census (t – k), and k is the number of years between the most recent census at time t and the one preceding it at time t – k.

The basic formula for the second step, which moves the cohorts of a population into the future is:

nPx+k,i,t+k = nCCRx,i,t / nPx,i,t     [3]

where,

nPx+k,i,t+k is the population aged x+k to x+k+n in area i at time t+k, and

nCCRx,i,t and nPx,i,t are as defined in equation [2].

Given the nature of the CCRs, 10–14 is the youngest five-year age group for which projections can be made if there are 10 years between censuses. To project the population aged 0–4 and 5–9, one can use the Child Woman Ratio (CWR) or, more generally, a “Child Adult Ratio” (CAR). In our case, it is not necessary to use this step because fertility is not a factor for the age groups we examine during the period from 1860 to 1870. Details for dealing with fertility when using the CCR method are found in Smith et al. (2013, 178).

If there is a “terminal, open-ended age group” being analyzed, projections of it also differ slightly from the projections for the “closed” age groups. However, we do not have such an age group in our analysis (e.g., those aged 55 and over) so this is not needed. Details for dealing with an open-ended age group are also discussed in Smith et al. (2013, 178).

To show that the CCR method satisfies the fundamental demographic equation, we restate equation [2] using the terms in equation [1]:

Pi,t+k = Pi,t + Bi – Di + Ii – Oi

where,

Pi,t = Population of area i at time t (the launch year),

Pi,t+k = Population of area i at time t+k (the projection year),

Bi = Births in area i between time t and t+k,

Di = Deaths in area i between time t and t+k,

Ii = In-migrants in area i between time t and t+k, and

Oi = Out-migrants in area i between time t and t+k,

then,

nCCRx,i,t = (nPx–k,i,t–k + Bi – Di + Ii – Oi )/nPx–k,i,t–k     [4]

Since we can also express equation [3] in terms of equation [1]:

nPx+k,i,t+k = ((nPx–k,i,t–k + Bi – Di + Ii – Oi) / (nPx–k,i,t–k)) / (nPx,i,t)     [5]

where x+k >= 10, then,

nCCRx,i,t = (nPx–k,i,t–k – Di + Ii – Oi) / nPx–k,i,t–k, and since

Ni = Ii – Oi, where x+k ≥ 10, we have

nCCRx,i,t = (nPx–k,i,t–k – Di + Ni) /nPx–k,i,t–k     [6]

Equations [4], [5], and [6] show that the CCR Method is not only consistent with the fundamental demographic equation but also closely related to the cohort-component method. Hacker’s (2011b) approach also is related to the fundamental demographic equation but, to be fully consistent with it, one must accept his assumption that migration did not affect the CSRs he constructed.

The CCR Method expresses the individual components of change—births, deaths, and migration—in terms of CCRs. As such, it satisfies the fundamental demographic equation. However, as we will see in the following section, expressing the components of population change in this way can be exploited. An important reason why the population projection method must be consistent with the fundamental demographic equation is so to minimize potential errors associated with hidden heterogeneity (Vaupel and Yashin 1985), the problem we described earlier. For the same reason—to minimize the effect of hidden heterogeneity—we use five-year age groups and conduct our analysis by state.

Results

Table 5 provides the 1870 projection of the male population by state and five-year age groups from 20–24 to 50–54. Tables 6 and 7 provide absolute and relative differences between projected and census counts, respectively. Table 6 displays the absolute differences between the projected 1870 population of white males by state and five-year age groups from 20–24 to 50–54 and the corresponding 1870 census counts. Table 7 provides relative differences: the percentage difference between the actual census counts and our projected numbers.

Table 5.  Expected (Forecasted) Non-Hispanic White Males by Selected Age Group and State: 1870

State/Age Group

20–24

25–29

30–34

35–39

40–44

45–49

50–54

Total

Alabama

31,971

28,222

27,065

19,317

13,029

8,833

9,818

138,256

Arkansas

32,650

29,225

23,883

18,163

10,965

11,601

7,137

133,624

Florida

5,912

6,492

1,698

3,076

2,287

1,668

2,222

23,353

Georgia

33,889

25,007

20,831

20,147

15,402

13,176

7,374

135,827

Louisiana

18,884

30,156

22,218

20,901

15,065

7,558

7,424

122,206

Mississippi

21,241

16,441

14,942

10,308

8,188

5,922

6,085

83,128

North Carolina

32,053

27,727

22,822

20,763

14,206

13,695

9,919

141,185

South Carolina

11,608

12,912

8,122

11,670

5,426

7,036

3,660

60,434

Tennessee

48,750

34,146

29,744

26,804

17,382

13,864

12,052

182,742

Texas

32,099

47,985

38,783

43,203

22,767

21,263

14,430

220,529

Virginia

34,033

25,465

25,270

20,333

15,337

16,701

14,703

151,842

Total

303,091

283,778

235,377

214,687

140,054

121,317

94,822

1,393,125

Note: See text for details on calculations and data.

Table 6.  Demographic Impact of Civil War on Non-Hispanics White Males by Selected Age Groups and State: Difference between Actual (Census) and Expected (Forecasted), 1870

State/Age Group

20–24

25–29

30–34

35–39

40–44

45–49

50–54

Total

Alabama

–8,666

–11,628

–15,413

–4,964

–4,728

–774

–289

–46,463

Arkansas

–12,304

–14,607

–14,256

–7,444

–3,972

–4,100

–877

–57,560

Florida

–50

–3,109

513

–622

–193

476

–900

–3,886

Georgia

–256

–5,699

–6,598

–4,870

–3,605

–1,877

3,026

–19,880

Louisiana

–6,199

–15,193

–8,786

–9,635

–6,119

1,148

–1,499

–46,283

Mississippi

–113

–5,059

–3,067

–676

–1,641

2,186

–1,149

–9,518

North Carolina

–934

–5,653

–7,525

–3,925

–1,726

–2,891

1,198

–21,456

South Carolina

2,288

–4,587

266

–4,005

–483

–2,341

1,750

–7,112

Tennessee

–8,360

–3,245

–4,531

–3,117

517

1,408

2,920

–14,408

Texas

–3,138

–25,755

–18,383

–29,364

–10,437

–10,665

–5,152

–102,894

Virginia

–1,006

–2,287

–8,102

–2,388

–2,178

1,004

–1,387

–16,343

Total

–38,736

–96,824

–85,883

–71,010

–34,564

–16,425

–2,360

–345,802

Note: See text for details on calculations and data.

Table 7.  Relative Demographic Impact of Civil War on Non-Hispanic White Males by Selected Age Groups and State: Percent Difference Between Actual (Census) and Expected (Forecasted), 1870

State/Age Group

20–24

25–29

30–34

35–39

40–44

45–49

50–54

Total

Arkansas

–37.68%

–49.98%

–59.69%

–40.99%

–36.23%

–35.34%

–12.28%

–43.08%

Alabama

–27.10%

–41.20%

–56.95%

–25.70%

–36.29%

–8.77%

–2.94%

–33.61%

Florida

–0.85%

–47.90%

30.22%

–20.23%

–8.43%

28.55%

–40.53%

–16.64%

Georgia

–0.76%

–22.79%

–31.68%

–24.17%

–23.41%

–14.24%

41.04%

–14.64%

Louisiana

–32.83%

–50.38%

–39.55%

–46.10%

–40.62%

15.19%

–20.19%

–37.87%

Mississippi

–0.53%

–30.77%

–20.53%

–6.56%

–20.04%

36.92%

–18.88%

–11.45%

North Carolina

–2.91%

–20.39%

–32.97%

–18.90%

–12.15%

–21.11%

12.08%

–15.20%

South Carolina

19.71%

–35.52%

3.27%

–34.32%

–8.90%

–33.27%

47.80%

–11.77%

Tennessee

–17.15%

–9.50%

–15.23%

–11.63%

2.98%

10.15%

24.23%

–7.88%

Texas

–9.77%

–53.67%

–47.40%

–67.97%

–45.84%

–50.16%

–35.71%

–46.66%

Virginia

–2.95%

–8.98%

–32.06%

–11.74%

–14.20%

6.01%

–9.43%

–10.76%

Total

–12.78%

–34.12%

–36.49%

–33.08%

–24.68%

–13.54%

–2.49%

–24.82%

Note: See text for details on calculations and data.

1. Projected Counts by State

Table 5 shows that the total projected size of the 1870 white male age cohort 20–54 is approximately 1.4 million. It also shows a declining monotonic relationship between age and cohort size. For example, the three largest cohorts are the youngest—that is, those aged 20–24 (n = 303,091), 25–29 (n = 283,778), and 30–34 (n = 235,377). The three largest Confederate states, in terms of white males aged 20 to 54, are Texas (n = 220,529), Tennessee (n = 182,742), and North Carolina (n = 141,185). There exists considerable variation across the Confederate states in the size of the white male cohort aged 20–54. The implication is that the demographic impact of the Civil War was not consistent across Confederate states.

2. Absolute Differences

As can be seen in Table 6, there are 345,802 fewer white males aged 20–54 found in 1870 across all eleven states of the confederacy than expected (–24.82%). Not surprisingly, the largest deficits are found in the age groups most likely to have served in the Confederate army, with 36.49 percent fewer (–85,883) in age group 30–34, just over 34 percent fewer (–96,824) in age group 25–29, and over 33 percent fewer (–71,01) in age group 35–39. Additionally, there are 38,736 (–12.78 percent) in age group 20–24 and 34,564 fewer (–24.68 percent) in age group 40–44. Thus, in an absolute sense, the largest deficit for white males across all eleven states is found for age group 25–29, while the largest relative deficit is found for age group 30–34.

Although judged by many to be incomplete (e.g., Hacker 2014), the number of men from the eleven Confederate states who were killed in battle, died of wounds, and died of disease is estimated to be 126,476 (Fox 1889).4 Given that this population was comprised of white males who would have been between the ages of 20–54 in 1870 had they survived, this total can be compared to the sum of the differences between actual and expected for those in 1870 who were aged 20–54. Our estimate is –345,802. Given that little of this difference is due to net out-migration, these comparisons suggest that deaths from all causes to white males who served in the Confederate army is 2.73 times greater than the estimate of 126,476 provided by Fox (1889) and 1.34 times greater than the educated guess of 258,000 provided by Faust (2006), which matches an estimate by the National Park Service (n.d.), suggesting that the Confederate dead from battle deaths (94,000) and disease (164,000) was 258,000, a figure considerably higher than that provided by Fox (1889). However, our analysis suggests that the number of deaths was even higher, likely in excess of 340,000 and that, when suppressed net in-migration is considered, the total effect is over 345,000. We turn to this latter topic in the next section.

3. Relative Differences

Table 7 shows the relative impact of the Civil War on the demography of the Confederate states—that is, the difference in what the population of white males would have been in 1870 as a percent of the actually observed count in 1870. For example, in Arkansas, the actual count of white males aged 20–24 was 20,347 in 1870, and our estimated count using the CCR method for this age cohort was 32,659. The difference is –12,304 white males. Thus, the relative impact of the Civil War in Arkansas was 37.68% (–12,304/32,659)*100. In other words, were it not for the Civil War, the number of white males aged 20–24 in 1870 would be greater by 12,304, or approximately 38 percent of the estimated size of 32,659.

The relative impact of the Civil War varied by age cohort. The three most affected cohorts were 30–34 (–36.49), 25–29 (–34.12), and 35–39 (–33.08). The impact scores indicate that each of these age cohorts would have been larger by 37 percent, 34 percent, and 33 percent, respectively.

In addition to variation among age cohort, relative impact scores also varied considerably by state. The three states exhibiting the greatest impact scores were Texas (–46.66), Arkansas (–43.08), and Louisiana (–37.87). Hence, if it were not for the Civil War, the male population in 1870 in Texas would have been approximately 47 percent greater, 43 percent greater in Arkansas, and 38 percent greater in Louisiana. The Confederate state with the lowest impact score was Tennessee (–7.88).

Overall, the Civil War had a significant impact on the Confederate states. Indeed, the overall impact score for all eleven Confederate states is approximately –25 percent. If it were not for the Civil War, the eleven Confederate states’ white male population aged 20–54 would have been about 25 percent larger than it actually was by 1870.

Discussion

In order to set the table for the discussion of suppressed net in-migration, we need to compare the (expected) 1850–1860 CCRs to the (actual) 1860–70 CCRs. Table 8 lists the (actual) 1860–70 CCRs by state.

Table 8.  Actual Ten-year Cohort Change Ratios for Five-year Age Groups: (1870 AGEX+10)/(1860 AGEX)

State/CCR

187020-24/ 186010-14

187025-29/186015-19

187030-34/186020-24

187035-39/186025-29

187040-44/186030-34

187045-49/186035-39

187050-54/186040-44

Alabama

0.64731

0.58756

0.42795

0.63659

0.47464

0.65647

0.98138

Arkansas

0.88926

0.83376

0.56460

0.70803

0.67389

0.89520

0.77989

Florida

1.15316

0.64957

0.63439

0.62967

1.38531

1.31500

0.54819

Georgia

0.80470

0.66259

0.47922

0.66331

0.65468

0.69917

0.84272

Louisiana

0.63488

0.96904

0.82813

0.53875

0.50091

0.58908

0.49418

Mississippi

0.86358

0.56673

0.58134

0.78861

0.57465

0.85840

0.64703

North Carolina

0.77774

0.61587

0.49388

0.74511

0.64485

0.63032

0.94395

South Carolina

0.69308

0.56569

0.69738

0.67444

0.59726

0.48493

0.84541

Tennessee

0.70089

0.69658

0.56026

0.70795

0.67694

0.72702

1.03570

Texas

1.24451

1.11713

1.03955

0.70009

0.81693

0.84398

0.97434

Virginia

0.77792

0.61001

0.50028

0.68791

0.60273

0.97830

0.90046

Total

0.79237

0.69639

0.58384

0.68088

0.62924

0.74310

0.84791

Source: Minnesota Population Center. Calculations by authors. See text for details

The (actual) CCRs in Table 8 can be compared to the corresponding (expected) CCRs by state and age group as shown in Table 4, above. These data are pertinent to a discussion of suppressed net in-migration. Note that the expected 1850–1860 CCR for the age group 20–24/10–14 in Texas is 1.37933. When a CCR exceeds 1.00, net in–migration is occurring over the period in question. During the period 1850–60 net in-migration affected every age group in Texas because every CCRs exceeds 1.00. Comparing these “expected” CCRs with the corresponding “actual” CCRs shown in Table 8, we see that only among three age groups (20–24/10–14, 25–29/15–19, and 30–34/20–24) does the CCR exceed 1.00. Moreover, each of these three CCRs is less than the corresponding CCRs found in Table 4. The implication here is that the substantial levels of net in-migration found in Texas for 1850–1860 were suppressed by the war.

Although not as dramatic, similar results are found in Arkansas and Louisiana. States, such as Texas, were on the “western edge” of the south and, as such, experienced substantial net in-migration before the war, which the war suppressed. Florida is similar as well, although it appears to have escaped the high levels of suppression of in-migration found in Arkansas, Louisiana, and, particularly, Texas.

On the opposite end of the net in-migration spectrum, one finds North Carolina and Virginia, where none of the expected or actual CCRs exceed 1.00. To some extent these states appear to have been experiencing little in-migration and, perhaps, a small degree of net out-migration. Thus, the demographic impact of the Civil War on these states is largely due to battle deaths and deaths due to disease.

As a summary (although one that “hides” heterogeneity for the eleven states), Table 9 presents the actual 1860–1870 and the 1850–60 CCRs by age groups among all eleven Confederate states. Only in age group (50–54/40–44) does an actual 1860–70 CCR (.84791) exceed the expected 1850–60 CCR (.83761), and the difference is very small. It seems to us that one can draw three implications. First, across all eleven Confederate states, the war had a minimal impact on those aged 40–44 in 1860; second, there was no overall net in-migration of white males in either the 1850–60 period or the 1860–70 period; and, third, mortality was high among the age groups in which men were most likely to have served in the Confederate army. For example, the 1850–60 CCR for those aged 30–34/20–24 is .84575, which is well above the (actual) 1860–70 CCR of .58384. One expects to see CCRs far less than 1.00 for the older age groups (e.g., 60 and over). However, the presence of CCRs far less than 1.00 for those in the prime of life (and most likely to be of migratory age) indicates that they were affected by the American Civil War: 20–24 (.79237), 25–29 (.69639), 30–34 (.58384), and even 35–39 (in 1870). These results suggest that the impact from mortality alone was likely above 300,000, which is considerably more than the 258,000 found in the estimate provided by the National Park Service (n.d.). However, it is important to note that the deaths implied by our estimates include those of southerners who served with the Union Forces and exclude northerners who served with the Confederate forces. While these numbers are not huge overall, they do vary by state.

Table 9.  Comparison of (Actual) 1870/1860 CCRs to (Expected) 1860/1850 CCRs Across all 11 Confederate States Combined

CCR

1860–1870 (Actual)

1850–60 (Expected)

Difference: (Actual – Expected)

Percent Difference

20–24/10–14

0.79237

0.87511

–0.08274

–9.45%

25–29/15–19

0.69639

0.95921

–0.26281

–27.40%

30–34/20–24

0.58384

0.84575

–0.26191

–30.97%

35–39/25–29

0.68088

0.93417

–0.25329

–27.11%

40–44/30–34

0.62924

0.79801

–0.16878

–21.15%

45–49/35–39

0.74310

0.80577

–0.06267

–7.78%

50–54/40–44

0.84791

0.83761

0.01030

1.23%

Note: See text for details on calculations and data.

In general, it is useful to note that our results and interpretations pertaining to migration are consistent with findings regarding internal migration and immigration found, for example, in Ward (2001), who discusses immigration and regional internal migration prior to the civil war and after. Finally, we should also note that by using the 1870 census as the benchmark against which to compare the expected values generated by our method, we include postwar deaths in our estimated demographic effects of the Civil War on the Confederacy. Hacker’s estimates do this as well as he too used the 1870 census as the benchmark against which he calculated “excess deaths” due to the Civil War.

Concluding Remarks

Hacker states that “[t]he human cost of the Civil War was greater than historians have long believed” (2011b, 348). Our results suggest that the Civil War had a substantial demographic impact on the cohort of white males who were of military age (10–44 in 1860) in the eleven Confederate states during the War. The combined effects of mortality and suppressed net in-migration yielded numbers far short of the expected numbers for men aged 20–54 in 1870. The overall impact is that 345,802 fewer white males aged 20–54 were found in 1870 relative to what would have been expected had the 1850–60 trends not been so severely affected by the Civil War. We find that our numbers support the argument by Hacker that the conventional view, which has approximately 620,000 male deaths attributable to the Civil War, is far too low. Our estimate of the effects on white males in the eleven Confederate states alone would increase the total number of deaths by approximately 220,000 over the estimate of 126,476 provided by Fox (1889) and about 88,000 over the estimated guess of 258,000 provided by Faust (2006) and the National Park Service (n.d.). When added to Fox’s estimate, the additional Confederate deaths alone would bring the total male deaths to approximately 840,000, which approaches 850,000, the upper end of the estimate suggested by Hacker (2011b). If, in fact, the number is around 850,000, then the deaths of white males resulting from the Civil War exceed the deaths in every war and military action in which the United States has participated since the Civil War. Given our findings, we believe Hacker is correct in stating that “[t]he human cost of the Civil War was greater than historians have long believed” (2011b, 348).

While our analysis suggests that the war had a substantial demographic impact on the white male population of military age of the Confederacy during the war, an open question remains regarding its effect on white females as well as on the African American population, including those from the Confederate states who served in the Union Army.5

Our results lead us to suggest that the broader approach suggested by Hacker (2011b)—one that goes beyond direct and indirect mortality—should be used when estimating the demographic impact of the Civil War. The work of Holmes and Vinovskis (1992), as well as that of Neely (2007), provides a starting point. However, the research needs to be extended. Hacker (2014) suggests that we can do that by studying the effects of the War and its immediate aftermath on: (1) civilian deaths, especially among the African American population transitioning from slavery; (2) marriage and widowhood; (3) orphanhood; (4) family structure; (5) migration; and (6) the onset of the fertility transition.

We would add greater specificity to these suggestions by noting that: (1) the combination of war-based mortality and suppressed levels of in-migration among soldiers may have reduced fertility through a “marriage squeeze” (in this case due to fewer males in the reproductive age groups than females) and related social mechanisms;6 (2) the war may have reduced immigration to the US as a whole as well as immigration and domestic in-migration to the states particularly affected by it; and (3) the war may also have increased out-migration rates from states particularly affected by it. While it may be difficult to isolate the war’s impact on fertility, mortality, and migration separately, we believe that the approach illustrated here provides reasonable estimates of the war’s impact on these variables of population change taken as a whole. As such, our approach provides a new perspective on the war’s demographic impact.

Finally, along with Hacker (2014), we believe that Marshall (2014) is wrong in his assessment, which argues that the impact of death in the Civil War is greatly exaggerated. Clearly, General Robert E. Lee was very sensitive to its impact as evidenced by his systematic and intentional undercounting of Confederate deaths after 1863 (Faust 2006, 253). Even today, the US military seeks to keep casualty counts low among both its members and civilians because, among other consequences, high numbers of casualties can turn public opinion against military action (Hays and Myers 2009; Larson and Savych 2006). Finally, our estimates of deaths and suppressed migration provide even more support for those who argue that there were no economic benefits to the US from the Civil War, much less for the former states of the Confederacy (Goldin and Lewis 1975; Ransom 2001).

In conclusion, recall the 48,834 deaths by all causes to active duty service members from 1980 to 2010 reported by the Defense Manpower Data Center (2011). These numbers represent an annual average of 1,628 deaths (1,628 = 48,834/30) from 1980 to 2010. If this same rate of casualties continued from 2010 to 2020, an additional 16,280 more deaths would have been recorded. When added to all of the US military deaths resulting from wars and military actions subsequent to the Civil War through 2010, the total number of deaths would still not likely exceed the 840,000 Civil War deaths that this research suggests. It is even less likely that the number of deaths would reach 850,000, the upper limit suggested by Hacker (2011b).

References

Baker, Jack, David A. Swanson, Jeff Tayman, and Lucky M. Tedrow. 2017. Cohort Change Ratios and Their Applications. Dordrecht, Heidelberg, London, and New York: Springer B.V. Press.

Defense Manpower Data Center. 2011. “Active Duty Military Deaths by Year and Manner, 1980–2010.” Accessed May 20, 2017. https://dcas.dmdc.osd.mil/dcas/pages/report_by_year_manner.xhtml.

Faust, Drew G. 2006. “Numbers on Top of Numbers: Counting the Civil War Dead.” Journal of Military History 70 (4): 995–1009.

———. 2008. The Republic of Suffering. New York: Vintage Books.

Fox, William F. 1889. Regimental Losses in the American Civil War, 1861–1865: A Treatise on the Extent and Nature of the Mortuary Losses in the Union Regiments, with Full and Exhaustive Statistics Compiled from the Official Records on File in the State Military Bureaus and at Washington. Albany, NY: Albany Publishing Company.

George, M. V., Stanley K. Smith, David A. Swanson, and Jeff Tayman. 2004. “Population Projections.” In The Methods and Materials of Demography, Condensed Edition, Revised, edited by Jacob S. Siegel and David A. Swanson, 561–601. Los Angeles: Academic/Elsevier Press.

Goldin, Claudia D., and Frank D. Lewis. 1975. “The Economic Cost of the American Civil War: Estimates and Implications.” The Journal of Economic History 35 (2): 299–326.

Hacker, J. David. 1999. The Human Cost of War: White Population in the United States, 1850–1880. Unpublished PhD diss. Minneapolis: University of Minnesota.

———. 2011a. “Recounting the Dead.” New York Times, September 20, 2011. Accessed April 5, 2018 http://opinionator.blogs.nytimes.com/2011/09/20/recounting-the-dead/?_r=0.

———. 2011b. “A Census-Based Count of the Civil War Dead.” Civil War History 57 (4): 307–348.

———. 2014. “Has the Demographic Impact of Civil War Deaths been Exaggerated?” Civil War History 60 (4): 453–458.

Hamilton, C. Horace, and Josef Perry. 1962. “A Short Method for Projecting Population by Age from One Decennial Census to Another.” Social Forces 41 (2): 163–170.

Hayes, Andrew F., and Teresa A. Myers. 2009. “Testing the ‘Proximate Casualties Hypothesis’: Local Troop Loss, Attention to News, and Support for Military Intervention.” Mass Communication and Society 23 (4): 379–402.

Holmes, Amy E., and Maris A. Vinovskis. 1992. “The Impact of the Civil War on American Widowhood.” In The Changing American Family, edited by Scott J. South and Stewart E. Tolnay, 62–85. Boulder: Westview Press.

Land, Kenneth C. 1986. “Methods for National Population Forecasts: A Review.” Journal of the American Statistical Association 81 (396): 888–901.

Larson, Eric V., and Bogdan Savych. 2006. Misfortunes of War: Press and Public Reactions to Civilian Deaths in Wartime. Santa Monica: RAND Corporation.

Marshall, Nicholas. 2014. “The Great Exaggeration: Death and the Civil War. Journal of the Civil War Era 4 (1): 3–27.

McPherson, James M. 1982. Ordeal by Fire: The Civil War and Reconstruction. New York: Knopf.

Minnesota Population Center. 2008. North Atlantic Population Project: Complete Count Microdata. Version 2.0 [Machine-readable database] Minneapolis: Minnesota Population Center.

National Archives. n.d. Black Soldiers in the Civil War. Accessed May 5, 2017. https://www.archives.gov/education/lessons/blacks-civil-war.

National Park Service. n.d. The Civil War. Accessed on April 4, 2018. https://www.nps.gov/civilwar/facts.htm.

Neely, Mark E. 2007. The Civil War and the Limits of Destruction. Cambridge, MA: Harvard University Press.

Ransom, Roger L. 2001. “Economics of the Civil War.” EH.Net Encyclopedia, edited by Robert Whaples. Economic History Association. Accessed April 4, 2018. http://eh.net/encyclopedia/the-economics-of-the-civil-war/.

Ruggles, Steven, Katie Genadek, Ronald Goeken, Josiah Grover, and Matthew Sobek. 2010. “Integrated Public Use Microdata Series: Version 5.0 [Machine-readable database].” Minneapolis: University of Minnesota.

Smith, Stanley K., Jeff Tayman, and David A. Swanson. 2013. A Practitioner’s Guide to State and Local Population Projections. Dordrecht, Heidelberg, London, and New York: Springer B.V. Press.

Swanson, David A. 2008. “The Demographic Effects of Hurricane Katrina on the Mississippi Gulf Coast: An Analysis by Zip Code.” Journal of the Mississippi Academy of Sciences 53 (4): 213–231.

Swanson, David A., Jerome N. McKibben, Lynn Wombold, Richard G. Forgette, and Mark V. Van Boening. 2009. “The Socio-Demographic and Environmental Effects of Katrina: An Impact Analysis Perspective.” The Open Demography Journal 2: 1–11.

Swanson, David A., Alan Schlottmann, and Bob Schmidt. 2010. “Forecasting the Population of Census Tracts by Age and Sex: An Example of the Hamilton–Perry Method in Action.” Population Research and Policy Review 29 (1): 47–63.

Swanson, David A., and Jeff Tayman. 2012. Subnational Population Estimates. Dordrecht, Heidelberg, London, and New York: Springer B.V. Press.

Swanson, David A., and Lucky Tedrow. 2012. “Using Cohort Change Ratios to Estimate Life Expectancy in Populations with Negligible Migration: A New Approach.” Canadian Studies in Population 39 (1–2): 83–90.

Vaupel, James W., and Anatoli I. Yashin. 1985. “Heterogeneity’s Ruses: Some Surprising Effects of Selection on Population Dynamics.” The American Statistician 39 (3): 176–185.

Vinovskis, Maris A. 1989. “Have Social Historians Lost the Civil War? Some Preliminary Demographic Speculations.” Journal of American History 76 (1): 34–58.

Ward, David. 2001. “Population Growth, Migration and Urbanization, 1860–1920.” In North America: The Historical Geography of a Changing Continent. 2nd ed., edited by Thomas F. McIlwraith and Edward K. Muller, 285–306. New York: Rowman and Littlefield.


1. The authors are grateful for comments from the editors and two anonymous reviewers and to Jerry McKibben and Jeff Tayman for suggestions. We also thank the Minnesota Population Center for assembling the data and online tools that made this study possible.

2. The eleven Confederate states are Alabama, Arkansas, Florida, Georgia, Louisiana, Mississippi, North Carolina, South Carolina, Tennessee, Texas, and Virginia. In order to separate the population of interest (white, non-Hispanic) from the Hispanic population, we implemented an algorithm developed by the Minnesota Population Center, which is available from either the authors or the Center (https://usa.ipums.org/usa/). The 1850 and 1860 data for Virginia do not include the 55 counties that separated to become West Virginia in 1863 or the two that were assigned to West Virginia immediately following the Civil War.

3. This method can be run in reverse to move a population back in time. A discussion of this form of “backcasting” along with examples (e.g., an estimate of the population of Hawaii in 1778, the year of first European contact) is found in Baker et al. (2017: 151–164).

4. An online version of Fox’s work can be found at a site maintained by Tufts University (Gregory Crane is the editor in chief of a series of historical documents, under which this report is found): http://www.perseus.tufts.edu/hopper/text?doc=Perseus%3Atext%3A2001.05.0068%3Achapter%3D16 . The estimate of Confederate casualties from all causes by state is taken from a compilation assembled by US General James B. Fry, which can be found at https://books.google.com/books?id=UQdPUmWfdM4C&pg=PA301&dq=%22james+b+fry%22+Confederate+states+of+america%22&hl=en&sa=X&ved=0ahUKEwj867XC8ozKAhVB5GMKHdy6As0Q6AEIHTAA#v=onepage&q=%22james%20b%20fry%22%20Confederate%20states%20of%20america%22&f=false.

5. An estimated 40,000 African Americans died while serving in the Union Forces (National Archives n. d.).

6. Hacker (1999: 103–104) provides an estimate of the birth deficit related to the Civil War, a phenomenon that suggests a marriage squeeze resulted from it as well.