The influence of remarriage on the racial difference in mother-only families in 1910

African American women are now, and historically have been, more likely than Euro-American women to be heads of families or heads of households with children (Gordon and McLanahan 1991; McDaniel 1994; Morgan et al. 1993; Pagnini and Morgan 1996). African American children were two to three times more likely than Euro-American children to live without one or both parents (usually without the father) from 1880 through 1960, and racial differences in family structure grew even larger after 1960 (Ruggles 1994; also see McDaniel 1994; Tucker and Mitchell-Kernan 1995). The reasons for these racial differences in household and family structure1 have been the subject of substantial scholarship and debate.

In most theoretical discussions of the causes of racial differences in family structure, the focus is on factors that are hypothesized to destabilize African American families disproportionately or otherwise to promote differentiation from normative Euro-American family patterns. For example, many early (and some contemporary) theorists point to the effects of slavery. They argue that instability and disorganization in African American families are the enduring legacy of slavery and the postwar political-economic institutions that emerged in the South, where most African Americans lived until well into the twentieth century (Du Bois 1989, 1908; Elkins 1963; Frazier 1939; Moynihan 1965; Myrdal 1944; Patterson 1998). Other scholars have focused on the recent increases in racial differences in family structure, drawing attention to factors such as the potential effects of welfare policies (Darrity and Myers 1995; Lichter, McLaughlin, and Ribar 1997; Murray 1984; Ruggles 1997a) or changing urban labor markets and the disadvantaged socioeconomic status of African Americans living in highly segregated urban areas (Massey and Denton 1993; Tolnay and Crowder 1999; Wilson 1987, 1996). Finally, others argue that sociocultural factors emanating from Euro-Americans' and African Americans' distinctive historical and personal experiences have differentially shaped family formation and dissolution processes as well as family structure (Cherlin 1998; Hill 1971; McDaniel 1994; Miller, Morgan, and McDaniel 1994; Pagnini and Morgan 1996; Patterson 1998; Ruggles 1994).

Although their relative contributions are still disputed, all of these distal social-institutional, socioeconomic, political-economic, and cultural factors ultimately influence racial differences in family structure through demographic processes. In particular, at the turn of the twentieth century, mortality and marriage patterns were key determinants of the prevalence of mother-only families. In that period, African American mortality was higher than that of Euro-Americans.2 According to Preston, Lim, and Morgan (1992), both partners survived to the end of the woman's reproductive years in only about 38% of African American marriages. In addition, around the turn of the twentieth century, African Americans were about twice as likely as Euro-Americans to be divorced and separated (Ruggles 1997a, 1997b). In general, African Americans had more fluid marriage patterns and a higher rate of marital turnover than did Euro-Americans (Pagnini and Morgan 1996; Preston et al. 1992).

In the literature on racial differences in family structure, these demographic, economic, cultural, and socialinstitutional forces are hypothesized to destabilize marriage differentially across racial groups, either by influencing marital dissolution or by preventing marriages from being contracted. In such ways, these factors increase racial differences in the prevalence of mother-only families.

It is vitally important to consider these potential explanations for racial differences in the prevalence of motheronly families; to date, however, there has been little theoretical discussion or empirical exploration of countervailing factors that might have reduced the size or changed the composition of the population of mother-only families. This is unfortunate, given that at least one mechanism, remarriage, plausibly could reduce the racial difference in the prevalence of such families. In a given period, racial differences in remarriage rates could reduce the racial difference in the prevalence of mother-only families in the population by differentially reintegrating women with children into dual-parent families (Elman and London forthcoming; Furstenberg 1980; Uhlenberg and Chew 1986).

One reason why remarriage has not been integrated into theoretical discussions in this manner is that contemporary rates of remarriage are lower among African Americans than among Euro-Americans; this difference is partly responsible for current racial differences in family structure (Patterson 1998; Sweet and Bumpass 1987; Tucker and Mitchell-- Kernan 1995). Although remarriage contributes today to the racial difference in mother-only families, the associations between race, remarriage, and the prevalence of mother-only families were not necessarily the same in earlier periods.

In this paper we use data from the Integrated Public Use Microdata Series (IPUMS) of the 1910 Census to fill an important gap in the literature. Specifically, we examine the contribution of racial differences in remarriage to the historical racial difference in mother-only families. The 1910 IPUMS provides a unique opportunity to do so because the 1910 Census was the first to partially enumerate marital histories: that is, whether persons were in first or higher-order marriages. Even though census data are cross-sectional, we can determine whether each currently married spouse was in a first or a higher-order marriage, and whether children from a prior union were present in the household at the time of the census. Using these data, we can estimate whether and how racial differences in choices of (re)marital partners contributed to the racial difference in the prevalence of mother-only families.

BACKGROUND AND ANALYTIC STRATEGY

Background

Social scientists from various disciplines continue to debate about the extent of racial differences in family patterns, when they took shape in American history or whether they have always been present, and how to account for them. The contemporary manifestation of this debate emerged primarily in response to the Moynihan report (1965), which attributed disorganization and deterioration in African American communities to "pathology" and instability in African American family structure resulting from slavery. Critical responses to Moynihan have taken many forms; some of the primary challenges to his conclusions are based on empirical historical research on family structure and marriage.

Recent reviews of the literatures on African American family structure (Ruggles 1994) and the decline of marriage among African Americans (Tucker and Mitchell-Kernan 1995) have provided two divergent but overlapping frameworks for organizing scholarship in these related areas. Ruggles (1994:136-37) provides a chronological framework; he notes that from the turn of the century to the time of the Moynihan report (1965), many social theorists argued that slavery had destabilized the African American family and was the root cause of many of the problems faced by African Americans. He suggests that critical responses to the Moynihan report fall into two broad categories delineated primarily by the period on which they focus.

The first group of scholars, which concentrates on the contemporary period, argues either that Moynihan reversed the relationship between socioeconomic disadvantage and family instability3 or that Moynihan's emphasis on the "pathology" of the African American family obscures its strength and resilience (Allen 1979; Hofferth 1984; McAdoo 1998; Stack 1974). The second group, "the revisionists," focus on the past, arguing that Moynihan overemphasized differences in family structure. Using cross-sectional data, these authors demonstrate that African American families were nuclear, male-headed, and essentially similar to Euro-American families in the late nineteenth and early twentieth centuries (Furstenberg, Hershberg, and Modell 1975; Gutman 1976).

Ruggles concludes this review by identifying a third group, "the neo-revisionists," who have sought to revise the original revisionists' conclusions. These scholars argue that historical and contemporary African American family patterns are marked by greater continuity than is generally recognized, and that high levels of single-parent families are evident in the African American population in both periods (Gordon and McLanahan 1991; Morgan et al. 1993).

In a second conceptual framework, Tucker and Mitchell-- Kernan (1995:6-7) draw on the work of Allen (1978) and Walker (1988) to characterize the literature as dominated by three "ideological perspectives": cultural equivalence, cultural deviance, and cultural variation. The cultural equivalence perspective draws in part from the work of the original "revisionist" scholars, such as Gutman (1976), who found few differences between Euro-American family structure and African American family structure, and in part from work that attributes racial differences in family structure to class factors. This perspective deemphasizes distinctions between Euro- and African American families. In contrast, the cultural deviance perspective, epitomized by the Moynihan report (1965), emphasizes deviations from the nuclear pattern and their presumed dysfunctions (also see Patterson 1998). A third perspective, the cultural variation perspective, is broadly consistent with the neorevisionist scholarship identified by Ruggles (1994). This perspective recognizes the distinctiveness of African American family structure, but constructs it positively as a reflection of resilience and adaptation to differing sociocultural and socioeconomic contexts.

It is problematic that the historical studies that inform these two conceptual frameworks have used cross-sectional data and have not examined family structure differences in light of race differences in the prevalence of remarriage. Information about the events of marriage, marital dissolution, and remarriage over the course of marital trajectories is conflated in cross-sectional data; therefore we do not fully understand the relationship between marital processes and family structure in either historical or contemporary periods. For example, there is considerable historical evidence that the marriage pattern was more fluid (Frazier 1939; Pagnini and Morgan 1996; Preston et al. 1992; Stevenson 1995) and remarriage rates were higher among African Americans than among Euro-Americans (Elman and London forthcoming; Gutman 1976; Manfra and Dykstra 1985; Preston et al. 1992; Stevenson 1995).' This evidence is potentially consistent with various, conflicting interpretations: it is consistent with finding racial differences in family structure (i.e., a neorevisionist/cultural variation perspective that emphasizes adaptation and resilience, or a cultural deviance perspective that emphasizes disorganization), but it is also consistent, in cross-sectional data, with finding a predominantly nuclear family structure (i.e., an original revisionist or cultural equivalence perspective) resulting from remarriage and family reconstitution.

The key issue we address here is the effect of remarriage on historical racial differences in family structure. Although the 1910 IPUMS is cross-sectional, the data are rich enough to allow an examination of (re)marital configurations at the point where the marriage was contracted, in order to learn where each spouse was located in the cycle of marriage, marital dissolution, and remarriage, and whether the woman brought at least one child from a prior marriage into her current union. By focusing on (re)marital configurations, we can determine whether racial differences in remarriage choices influenced the racial difference in the prevalence of motheronly families in this period, and, if so, in what direction.

Analytic Strategy

At the outset we considered that racial differences in remarriage, at the turn of the century, could have either increased or decreased the racial difference in the prevalence of mother-only families in the population, or could have had no effect on it. Because very little research has been conducted on this issue, we devised an analytic strategy that allowed us to evaluate each of these possibilities. Underlying our approach is the assumption that men were the primary initiators of (re)marriage around the turn of the century. In the absence of gender equity in decision making, men were the key (albeit not the sole) decision makers. Thus we focus our analyses on men's (re)marital choices and on whether and how men's partner choices influenced the racial difference in the prevalence of mother-only families.

First, we considered the possibility that African American males' remarital choices were different from those of Euro-American males in a manner that contributed to racial differences in the prevalence of mother-only families in the population. In one manifestation of this influence, previously married African American men would be more likely than previously married Euro-American men to remarry nevermarried women. Because many previously married women were raising children on their own (McDaniel 1994; Miller et al. 1994; Ruggles 1997a), this racial difference in men's remarriage choices would tend to contribute to the racial difference in the prevalence of mother-only families. Like the contemporary retreat from marriage among African Americans, evidence showing that remarriage processes contributed to the racial difference in mother-only families would be consistent with a cultural deviance perspective (see, for example, Patterson 1998).

A second possible outcome we considered was that African American men (either previously married or previously single) were more likely than Euro-American men to (re)marry previously married women, specifically women with children from a previous marriage. Other things being equal, this would tend to decrease the proportion of African Americans in the population of women eligible for remarriage. If it could be shown specifically that previously single and/or previously married African American men were more likely to (re)marry previously married women with children, then we could conclude that racial differences in men's (re)marriage choices reduced racial differences in the prevalence of mother-only families in the population. Evidence supporting this outcome could be interpreted as an indicator of adaptation and response to the many factors contributing to higher rates of marital dissolution and more fluid marriage patterns among African Americans in this period; such evidence would be consistent with a cultural variation perspective.

In addition to the two possible outcomes described above, we considered the possibility of no significant racial differences in remarriage choices. On the basis of available evidence regarding racial differences in remarriage (Elman and London forthcoming; Manfra and Dykstra 1985), we believed that this outcome was unlikely. Evidence of no difference, however, if found, would be consistent with a cultural equivalence interpretation.

Individual Characteristics and Social Contexts in Remarriage Markets

Our examination of family structure focuses on men's (re)marital partner choices; thus we must take into account both the characteristics of the men who were (re)marrying and the social contexts in which they lived as they relate to partner selection. Remarriage, as a strategy, reflects the propensity and the need to remarry, which are affected by individual circumstances, as well as the opportunity to remarry, which is affected by contextual conditions (Coombs 1993; Grigg 1977; Thornton 1977; Uhlenberg and Chew 1986; van Poppel 1995). We extend this conceptualization by suggesting that men's (re)marital partner choices reflect (1) their personal characteristics, such as age, prior family structure, and socioeconomic standing; and (2) local surroundings, including local marriage markets (Cready, Fosset, and Kiecolt 1997; Lichter et al. 1997; South and Lloyd 1995) and the availability of women with particular characteristics.

Personal characteristics. Age at first marriage is one of the most robust predictors of remarriage (Grigg 1977; van Poppel 1995). Men's ages at (re)marriage shaped remarriage risks differentially, and likely influenced the size and the composition of the pool of potential mates they could attract. African American men entered first marriages at significantly younger ages' than Euro-American men (Tolnay 1999) and were at greater risk of remarriage (Elman and London forthcoming; Gutman 1976; Preston and McDonald 1979; Ruggles 1997b). Although data on racial differences in the incidence of remarriage are rare for this period, the risk of experiencing marital dissolution, via divorce or widowhood, was higher among African American men than among EuroAmerican men at every age over the life course.6 Yet because of early marriage and higher maternal mortality (Preston and Haines 1991), their risk would be especially high at younger ages. Fluidity in marriage patterns reflects processes of marital dissolution and reconstitution. If couples' timing of separation and divorce in 1910 was similar to that of couples today, young African American men would be more likely to be divorced than their Euro-American age peers. With regard to mortality, Euro-American sex differences in life expectancy widened just after the turn of the twentieth century (i.e., Euro-American women began to live longer), whereas this effect of the epidemiologic transition was delayed for African Americans (Ewbank 1987). As a result, young African American men in 1910 also would be more likely to be widowed than their Euro-American age peers.

Men's ages at (re)marriage shaped spousal matching because persons at different stages of the life course could differentially attract mates with particular qualities from the pool of eligibles. Around the turn of the twentieth century, men experienced declines in health and functional status from age 20 on (Elman and Myers 1997), as well as in other hard-tomeasure personal qualities. Thus we expect to find that, net of other factors, men with younger ages at (re)marriage would be more able to select spouses with the qualities they preferred, including (but not limited to) the types of spouses that were most socially acceptable. Assuming that never-married women were preferable as spouses, we therefore hypothesize that men who were younger at (re)marriage would be less likely to (re)marry previously married women.

In addition to controlling for men's ages at (re)marriage, we control for men's chronological ages. Accounting for chronological age, net of the effect of age at marriage, locates men in time. This step is important because in 1910, men who had remarried at age 25 might have been anywhere from age 25 to age 64 (the oldest persons in our sample). Chronological age, as opposed to age at marriage, provides information about cohort membership: that is, when men were born (Ryder 1965). Because of the increasing importance of companionship in marriage (Degler 1980), we hypothesize that men born into younger cohorts would be less likely to marry previously married women, particularly those with children.

African American men not only married at younger ages than did Euro-American men, but also started having children earlier in the life course (Tolnay 1999). A pattern of early childbearing is important because quantitative and qualitative reports suggest that men who experienced marital dissolution early in the life course and had children generally needed to remarry, and to remarry quickly. New spouses provided necessary assistance in child care, household management, and economic supports because biological children who were at home were often too young to do the work done by the former spouse (Kleinberg 1989; Laslett 1971; Schlissel 1992; van Poppel 1995). Remarriage also augmented the number of economically active coresident kin when household dependency ratios were high (Elman and London forthcoming). In addition, the presence of children from a previous marriage likely shaped spousal matching and the pool of potential mates. Men with children may not have been able to attract women who had not yet started their own families; conversely, they may have been particularly able to attract women with a corresponding need to take care of their own children. There may have been strong practical and normative pressures for homogamy along this dimension of family life. Thus we hypothesize that men who had children from a previous marriage would be more likely to remarry previously married women, particularly those with children.

Socioeconomic status shaped remarriage probabilities as well as the pool of potential mates (van Poppel 1995). In the recent past, remarriage was a strategy that reflected economic need (Coombs 1993; Schlissel 1992) because the economic functioning of families-especially families that had experienced marital dissolution and lived on the margins of economic sufficiency-was based on the allocation of multiple relatives, including children and more distant kin, into paid and nonpaid household tasks for the purpose of gathering resources (Hareven 1982; Moen and Wethington 1992).7 Remarriage improved the economic functioning of families of both Euro- and African American remarried men, although this was not the case for remarried Euro- or African American women (Elman and London forthcoming). The economic advantages associated with remarriage for men and the relatively high prevalence of remarriage among men (especially African American men) in 1910 suggest that, for all men, being remarried was preferable to remaining widowed or divorced. Yet it is likely that economic need also shaped the matching process. Persons with higher socioeconomic status and greater human capital probably were able to attract larger pools of potential mates and to select spouses with the qualities they preferred. Also, higher socioeconomic status allowed men the time to wait for an optimum match by affording them the means to hire household service (van Poppel 1995). Thus we hypothesize that men of higher socioeconomic status were less likely to remarry previously married women, with or without children.

Contextual factors. Contextual conditions influenced the social organization of families, the strategies they used to cope with their environment, including remarriage, and men's likelihood of finding partners for remarriage. This latter point is not trivial; just as the availability of potential spouses is a matter of concern in African American communities today (Patterson 1998; Tucker and Mitchell-Kernan 1995), it was a concern in the past (Tolnay 1997).

One mechanism by which local communities shaped remarriage was by providing incentives for men to marry early (Tolnay 1999). An early age of first marriage was correlated highly with rural residence and the strategy of generating a large pool of economically active kin, especially in the rural South and Midwest (Coombs 1993; Patterson 1998; Riley 1991; Tolnay 1999). Larger families meant greater work productivity and higher living standards. Young persons in local communities with relatively high rural land costs and/or blocked opportunity for landownership tended to make transitions into early marriage (Easterlin, Alter, and Condran 1978; Landale 1989; Landale and Tolnay 1991; Tolnay 1999); they supported their families by resorting to cash or noncash tenancy (especially in the Midwest) and/or by sharecropping (primarily in the South) (Moneyhon 1997; Tolnay 1999; Wright 1978). Both kinds of tenancy, and the economic benefits of large families, apparently were associated with high rates of remarriage in the Midwest and in the South (Coombs 1993; Elman and London forthcoming; Manfra and Dysktra 1985; Schlissel 1992).

In urban areas as well, however, family reconstitution was necessary to maintain household earnings. Multiple earners rather than single working individuals determined urban families' socioeconomic status at the turn of the twentieth century; urban men rarely earned a "living wage," even in their prime working years (McClymer 1986). This was especially the case for urban African American males: their wages were kept artificially low, they were last in line in labor queues, and so their families needed more than one earner, including wives (Goldin 1980; Patterson 1998).

To capture these contextual effects we use variables denoting region of residence, urban, town, or rural residence, and farm tenancy, all measured at the household level. We hypothesize that persons living in the West, Midwest, and South, as well as farm tenants, would be less concerned about a woman's prior marital status and would be more likely to remarry previously married women. This outcome is due to the indirect effects of early age at marriage in these contexts and, more directly, to the economic benefits of having more potential workers or wage earners, including stepchildren. We also expect to find that rural men had higher likelihoods of (re)marrying women who had been married previously because of their greater need for assistance with farming and household duties.

Local marriage markets also shaped marital opportunity structures. Although we know of no historical studies that examine the direct effects of local marriage markets on remarriage, there is strong evidence that when many potential partners are available for (re)marriage, the opportunity costs of remaining unmarried increase, while the costs of searching for a new spouse decrease (Lichter et al. 1997; South and Lloyd 1995). Therefore we expect that the more potential marriage partners exist in a local market, the more likely it is that remarriage takes place. Beyond this, we expect that when a large number of potential partners for men exists, the men would be less likely to (re)marry a previously married woman, especially one with children. Probabilities of (re)marriage, however, may have been shaped not only by the relative availability of potential partners, but also by the perceived eligibility of partners with prior marital and childbearing histories. We expect that in markets containing a greater proportion of previously married women to eligible women (i.e., both never-married and previously married), men will be more likely to marry previously married women.

DATA AND METHODS

Analytic Sample

Data for this study were obtained from the 1910 Integrated Public Use Microdata Series (IPUMS) (Ruggles and Sobek 1997). The 1910 IPUMS is a self-weighting 1-in-250 sample of households in the United States drawn from microfilm copies of original census documents. Our analytic sample includes currently (re)married 20- to 64-year-old Euro-American and African American (including mulatto) men (N = 57,827).8 We truncated the age distribution because preliminary analyses indicated that remarriage before age 20 was rare, and underreporting of remarriage for those over age 65 is known to be substantial (Kramarow 1995). Kramarow (1995) estimated that underreporting of remarriage for males at all ages between 20 and 64 was less than 7%.9

We also used data from the thirteenth Decennial Census of the United States (1910) to construct marriage market and eligibility indicators at the county level. (Below we provide details about the methods used to construct these indicators and our assessment of their quality.) We linked macro- and micro-level data sets by sorting and merging by state and county codes.

Dependent Variable

To examine whether and how race influenced (re)marital partner choice, we used multinomial logistic regression. For this analysis, we constructed a trichotomous dependent variable to reflect women's statuses at the time of contracting their current marriages to the men in our sample: (1) entering a first marriage; (2) previously married with no child from a previous marriage; and (3) previously married with children from a prior marriage.

To create this dependent variable, we had to link the data records of coresident biological children to those of their mothers, and then link information about women and their children to the data records of their coresident husbands. This procedure allowed us to determine whether a woman's coresident biological children were the product of her current marriage or a previous marriage." For each mother, we created a separate count variable to indicate the number of coresiding biological children, and the exact age of each child. We then compared the coresident children's ages with the duration, in years, of their mother's present marriage. Mothers with at least one biological child older than the duration of their current marriage were coded as having had a child from a previous marriage at the time of contracting their current marriage."

We then used the IPUMS variables denoting marital status and marriage number to identify men in first and higherorder marriages, and used the IPUMS variable SPLOC to locate spousal records for linkage. Variables denoting spouse's marriage number and the presence of children from a previous marriage were attached to the records of married and remarried men.

Primary Independent Variables

The primary independent variable of interest in our analyses is race, which we measure with a dummy variable (African American = 1, Euro-American = 0). Mulattos were included with African Americans. In addition, we distinguished men in first marriages from those in second or higher-order marriages (higher-order = 1, in first marriage = 0), and determined whether men had a child from a prior marriage at the time of contracting the current union (previous child = 1; no previous child = 0).12 Because these two latter variables are empirically interrelated, and because the effects of both on (re)marital partner choice potentially differ by race, we examined the two- and three-way interactions among these variables in our analyses.

Other Social and Contextual Control Variables

In the model, we included a set of variables to control for potentially confounding social and contextual effects. The model includes age at (re)marriage, a continuous variable calculated as current age minus current marital duration, in years,13 and age in 1910, measured as a continuous variable ranging from 20 to 64 years. We also included a dummy variable for literacy (literate = 1, no = 0), a robust indicator of socioeconomic status that tends to occur early in the life course before marriage or remarriage. The 1910 IPUMS does not provide information about the level of educational attainment or income. It furnishes information about occupation; in that period, however, men's occupations varied relatively little, especially in the West, Midwest, and South, where agriculture and farming predominated.

Because we are interested partly in determining the effects of contextual factors on family reconstitution, it is appropriate to control for whether men were tenant farmers or sharecroppers because these occupations strongly influence family formation processes (Tolnay 1999). We estimate this effect by including a dummy variable that measures whether the household was a non-owned farm (tenant farmer, cash or noncash = 1, no = 0).14 We also include a set of dummy variables for region (North, Midwest, and West respectively versus South = 0), and location type (city and town respectively versus rural = 0).

(Re)marriage market conditions are indicated in our model as county-level sex ratios. As noted by South and Lloyd (1995), two issues arise in examining marriage markets: (1) how to delimit the relevant market area, and (2) how to estimate the number of partners in each area that might be "suitable" for remarriage.

With regard to delimiting markets, we used counties as macro-level units of analysis in this study because their boundaries are clear and because they are the smallest geopolitical units available in the census. Yet in determining the number of females available for remarriage in a particular county, we faced a constraint: county-level information about subgroups of men or women demarcated by marital status or age (except for African American and Euro-American men of voting age) are not available in the 1910 census data. Thus we cannot know the number of currently "available women" (i.e., unmarried women) or "age-appropriate women" available for (re)marriage in each county by using county-level data. An alternative option is to aggregate data in each county, using the IPUMS, by sex, age, and race. We did not opt for this approach because the sampling fraction in the 1910 IPUMS is I in 250 and may provide less robust sex ratios than the county-level measures. Another alternative strategy would have been to focus on selected cities or large counties and to aggregate upward. Yet because most African Americans lived in the rural South, this strategy would not have been helpful. For these reasons we decided to employ race-specific sex ratios at the county level in direct form as our measure of (re)marriage markets." IMAGE FORMULA 34IMAGE FORMULA 35

RESULTS

Sample Description

Table 1 describes the characteristics of the 20- to 64-yearold currently married men in our analytic sample. Overall, 8.0% were African American, 9.5% were remarried, and almost all were literate." These men, on average, were 40.2 years old in 1910, and had entered their current marriages when they were approximately 26 years old. When they entered their current marriages, 4.7% had coresiding children from a previous marriage. More than half lived in rural areas, and 19.3% were tenant farmers. Approximately onethird of the sample each lived in the North, the Midwest, and the South; only 7.2% lived in the West. Overall, 93.2% of currently married men had married women who were entering their first marriages, 3.9% had married women who were previously married without children from a previous marriage, and 2.9% had married previously married women with children from a previous marriage.

As seen in Table 2, panel A, a substantially higher percentage of Euro-American men than of African American men married women who were entering first marriages, regardless of whether the men were in their first marriages, were remarrying with no children from a prior marriage, or were remarrying with children from a prior marriage. Correspondingly, as shown in panels B and C, African American men were substantially more likely than Euro-American men to have married previously married women, many of whom brought children from a previous marriage into their current union. The only exception is seen in panel C: among remarried men with no children from a previous marriage, African American men were somewhat less likely than Euro-American men to have remarried previously married women with children (13.15% versus 15.46%). Despite this one exception to the overall pattern, African American men were more than twice as likely as Euro-American men to have (re)married women with children from a previous marriage (7.36% versus 2.52%). These estimates provide evidence that racial differences in (re)marital partner choices, particularly the reintegration of previously married women with children into nuclear families, apparently made the racial difference in the prevalence of mother-only families less than it would have been in the absence of such racial differences in (re)marriage.

To further investigate the influence of race on (re)marital partner choices, and to determine whether racial differences in (re)marriage choices could be explained by men's other characteristics or by their social contexts, we estimated a multinomial logistic regression model that included both individual and contextual variables as controls. Among currently married 20- to 64-year-old men, the model provides estimates of the odds of being married to a previously married woman without children from a previous marriage or to a previously married woman with children from a previous marriage, relative to the odds of being married to a woman who was currently in her first marriage.21 As seen in Table 3, in both equations, men's race, previous marital status, and whether they had a child from a previous marriage at the time of contracting the current marriage were all statistically significant, as were the two- and three-way interactions among these variables. In addition, all of the control variables were significantly associated with (re)marital partner choices; these effects, however, were slightly less consistent in magnitude, direction, and statistical significance across the two equations. In short, the effects of most of the variables in the model were similarly large and in the same direction across the two equations; this fact suggests that these factors distinguished men who married previously married women (regardless of whether they did or did not have children from a previous marriage) from those who married women who were entering their first marriages.21

Using the estimated coefficients from the model presented in Table 3, we computed the relative odds that men (re)married previously married women with and without children from a previous marriage, by race, men's (re)marital status, and whether men had children from previous marriages. As seen in Table 4, after adjusting for other factors, the odds of (re)marrying previously married women were consistently higher for African Americans than for EuroAmericans across all contrasts presented in panels A and B.

Especially relevant here, however, are the results presented in panel B. These findings indicate that African American men, whether they were in their first or a higherorder marriage, and whether or not they had a child from a previous marriage, were more likely than Euro-American men with the same statuses to have married previously married women with children. The effect of race persisted in the presence of controls for a broad array of individual and contextual conditions; this fact provides compelling evidence that the racial difference in the prevalence of mother-only families in 1910 would have been larger than is currently recognized, in the absence of these patterns of partner choice. Moreover, the persistence of a strong race effect indicates that theoretical models of (re)marriage processes at the turn of the century must account for race differences in terms other than those measured in this model.

All of the control variables included in the model presented in Table 3 were associated with (re)marital partner choices, although not always as hypothesized. Except for age and region, the effects of these variables were in the same direction and of approximately the same magnitude across the two equations. The odds of (re)marrying previously married women with and without children from a previous marriage respectively were significantly greater for men who were older at marriage, men living in cities (relative to men in rural areas), and Euro-American men living in counties with higher eligibility ratios (i.e., a greater supply of previously married women). The African American eligibility ratio was not significant in either equation. The odds were decreased significantly for men who were literate, were occupied in tenant farming, and lived in areas with higher sex ratios (i.e., more women than men). Age was significant in both models, but had a positive effect on the odds of (re)marrying previously married women without children and a negative effect on the odds of (re)marrying previously married women with children. Living in the Midwest and the West increased the odds of (re)marrying previously married women without children from a previous marriage (relative to living in the South), but had no effect on the odds of (re)marrying previously married women with children from a previous marriage.

In a supplemental analysis, we stratified the sample by race to examine how strongly these individual and contextual effects on partner choice exerted similar influences among Euro- and African Americans. We reestimated the same model presented above minus the dummy variable for race and three of the four interaction terms (all of which included race as one of the components). As seen in Table 5, except for region and tenancy, the effects of the individual and contextual variables on (re)marital partner choices were remarkably similar among Euro- and African Americans. For both groups, we found a consistent pattern of (re)marital partner choices by (re)marital status and the presence of children from a prior marriage. Being previously married and having children from a prior marriage increased the odds of remarrying a previously married woman with children from a previous marriage, but decreased the odds of remarrying a previously married woman without children from a previous marriage.

In addition, for both Euro- and African American men, the odds of (re)marrying previously married women without children from a previous marriage (versus women entering their first marriages) were significantly higher among older men, men with older ages at marriage, men living in cities (as opposed to rural areas), and men living in counties with higher race-specific eligibility ratios; the odds were significantly lower among those living in counties with higher sex ratios. Similarly, the odds of (re)marrying previously married women with children from a prior marriage were significantly higher among those who were older at marriage and those living in cities; the odds were significantly lower among older men and men living in counties with higher sex ratios. Yet despite the remarkable consistency of these effects across race, we must repeat that these factors do not account for race differences in (re)marital choices.

In a final supplemental analysis, we estimated the prevalence of mother-only families observed in 1910, by race, and compared those estimates with adjusted prevalence estimates computed by including in the numerators remarried women with children from a prior marriage. Among the 60,955 EuroAmerican women with children observed in our sample, 4,589 (7.53%) were heads of mother-only families. In contrast, 12.07% (595/4,928) of the African American women with children headed mother-only families. When we added the number of remarried women with children from a prior marriage to the numerators, by race, the adjusted prevalence of mother-only families increased to 9.98% of Euro-Americans and 19.36% of African Americans. Thus, in the absence of remarriage for women with children from previous marriages, the racial difference in mother-only families would have been double the observed difference.

DISCUSSION

In a paper on the origins of African American family structure, Ruggles (1994) states:

Race differences in family structure have expanded throughout the twentieth century, especially over the past three decades. But the fundamental differences in the percentage of children residing without parents began well over a century ago. The critical question remains: What is the source of this distinctive African American pattern of single parenthood? Recent economic changes can be invoked to explain the growing differential between black family structure and white family structure, but they cannot explain why blacks started from a higher base. (p. 147)

Our results indicate that the racial difference in African American family structure at the turn of the century would have been even larger without the countervailing influence of remarriage. We find no evidence that racial differences in remarriage choices contributed positively to the racial difference in mother-only families. Instead we find that African American males in both first and higher-order marriages were more likely than their Euro-American counterparts to have (re)married previously married women generally, and those with children specifically. The persistence of this large racial difference, net of a wide array of social and contextual factors, is compelling evidence that (re)marriage in that period was influenced strongly by other, unmeasured factors associated with racial background.

This conclusion is consistent with the qualitative evidence provided by Pagnini and Morgan (1996), who argue that the institution of marriage was different for EuroAmericans than for African Americans. For example, although this comment is not related explicitly to remarriage, they quote one African American respondent as follows:

You see, Mama didn't want me to marry her because she had a little boy when I married her. Mama wouldn't have nothing to do with her because she thought she was common, and now Mama has come to live with us .... I left and married her, in spite of all they could say about it. I loved her and reckon I did a good day's work when I married her. Her folks was tickled, and I don't see why mine won't because having the baby won't against her so far as other Negroes was concerned. My folks had to try to be like white folks, though. (Pagnini and Morgan 1996:1,707)

Whatever the participants' intent, the institution of remarriage was viable; it reintegrated vulnerable women and children into dual-parent families, and provided men with a means to sustain economic activity, their households, and their connection to family life. Remarriage countered the disparate demographic, economic, cultural, and social-- institutional forces that disproportionately destabilized African American marriages, and contributed to the more fluid marriage pattern among African Americans that has been noted by others (Manfra and Dyskstra 1985; Pagnini and Morgan 1996; Preston et al. 1992). In this period, remarriage apparently operated as an adaptive response to the high levels of marital disruption among African Americans.

These findings highlight two gaps in the literature on racial differences in family structure. First, remarriage rates were higher for African Americans than for Euro-Americans at the turn of the century, but now are lower. Therefore the remarriage transition was greater for African Americans than for Euro-Americans, and may have been more consequential for other aspects of family and community life. Potentially, by understanding more fully what happened to remarriage over the course of the twentieth century and how changes varied by race, we could gain important insights into several other domains of investigation, such as childbearing outside marriage, utilization of welfare, and the feminization of poverty. Although remarriage recently has declined for both Euro- and African Americans (Patterson 1998),22 our evidence suggests that, over the long course, it did so more for African Americans than for Euro-Americans. We believe that researchers need to focus more explicitly on the long course of remarriage transitions in the United States and on explaining how these transitions affected change in other domains.

Our research also exposes a second gap in the literature. There is little research on the influences, if any, of remarital partner choices on family structure today. Even though African Americans are less likely than Euro-Americans to marry and remarry, it may be that (re)marital partner choices either increase or decrease the prevalence of mother-only families relative to what would be found in the absence of such choices. It is important to explore both of these gaps in the literature because historical and contemporary differences in remarriage choices may exert important but still largely undocumented influences on racial diversity in family structure in the twentieth century.

Similar factors led African American and EuroAmerican men to remarry; this point is at least partly consistent with an interpretation of cultural equivalence in family structure. We believe, however, that the results presented here provide stronger evidence in support of a cultural variation perspective (see Tucker and Mitchell-Kernan 1995) and with findings emerging from the "neo-revisionist" response to the Moynihan (1965) report (e.g., Gordon and McLanahan 1991; McDaniel 1994; Morgan et al. 1993). Our results are consistent with evidence that African Americans had a more fluid marriage pattern (Manfra and Dykstra 1985; Pagnini and Morgan 1996; Preston et al. 1992; Stevenson 1995); at the same time they are consistent with finding relatively high levels of dual-parent families in cross-sectional data. The key point is that Euro-American and African Americans achieved dual-parent families by decidedly different mechanisms in the early twentieth century. Although our research cannot address the causes of this greater fluidity in marital arrangements (which probably were multiple), it appears that remarriage among African Americans in that period was not a sign of social dysfunction. Rather, it was an important adaptive strategy, which countered the effects of the forces that destabilized African American communities and families at the turn of the century.

Our results are not only consistent with a perspective that acknowledges racial differences in family structure without attaching a negative valence to those differences, or using a "deficit model" to interpret them. They also challenge the arguments of Moynihan (1965), Patterson (1998), and others who argue that slavery "destroyed" the African American family and undermined men's commitment to family life. Our evidence suggests that African Americans sought to respond to the factors that contributed to marital instability and to construct their own institutions (McDaniel 1994). Our findings suggest that both previously single and previously married African American men viewed remarriage as an acceptable and viable option to help meet their own needs, those of their own children (whom they often had to care for after the mother's death), and those of previously married women (with and without children) in their communities. This evidence, we believe, lends itself to the conclusion that, despite slavery, Jim Crow, and economic deprivation, African American men at the turn of the century had strong communal ties and commitments to families and marriage (Litwack 1979).

Although our findings indicate that remarriage strategies were adaptive in 1910, a single cross-sectional data set cannot address the long-term change in the African American remarriage pattern. Still to be explained is how and why this adaptive strategy for family reconstitution declined so drastically among African Americans. We do not accept the notion that there was racial dysfunction or equivalence in marriage patterns in 1910; instead we argue that racial differences in the prevalence of mother-only families increased over time, at least in part because remarriage among African Americans was undermined. Therefore we contend that a fuller understanding of remarriage, specifically the factors that contributed to the decline of remarriage among African Americans, would provide valuable insights into many domains of investigation. The results presented here suggest the utility and importance of additional work to disaggregate remarriage from marriage in general in historical and contemporary research on racial differences in family structure and other aspects of family functioning.