University of California, Berkeley
SRC
Survey Research Center
2538 Channing Way Berkeley,
CA 94720-5100

Working Paper
October 1998
[Comments Welcome]

PIE IN THE SKY WHILE YOU'RE ALIVE:

AMERICANS' BELIEF IN LIFE AFTER DEATH AND SUPPLY-SIDE RELIGION⁽¹⁾         Andrew M. Greeley University of Chicago University of Arizona Michael Hout University of California, Berkeley

	Priest: That was a fine funeral sermon, Rabbi. Your emphasis on life-after-death sounded just like what I would say.
	Rabbi: What can I do, Father? This is Chicago! My people hear about it all the time from your people. So now they expect to hear it from me. -True story.

_
Keep in touch...
Locally, and Globally! 

Check out
Andrew M. Greeley's Columns for the
Chicago SunTimes'
Daily Southtown.
_

.Abstract  

Belief in life after death has become more prevalent in the 1990s than it was in the 1970s according to data from the General Social Survey. Year-to-year changes are significant, but the increase is even more evident when we compare cohorts. Roughly 85 percent of Protestants believe in life after death in every cohort. The change occurred among persons from minority religions and persons with no religious affiliation. The proportion of Catholics believing in the afterlife rose from 67 percent of the cohort born 1900-09 to 85 percent of the cohort born 1960-69. Among Jews this belief increased from 17 percent of the 1900-09 cohort to 74 percent of the 1960-69 cohort. Immigration is a key factor in the increase. Many Catholics and Jews arriving from Europe were, with the exception of Irish and Polish Catholics, hostile or indifferent to organized religion. Their children and grandchildren, socialized in the United States, have come to adopt the prevalent view that each person has a soul that lives on after the death of the body.

The United States is more devout than most Western nations. Belief in God is widespread, church attendance is relatively high, and most people believe in heaven and hell (Herberg 1955; Greeley 1995). Although they are relatively high, these beliefs and practices are far from universal. For example, when the General Social Survey has offered its national sample of adults six alternative expressions of belief in god or a higher power, 62 percent of Americans have chosen the most certain response - "I know God really exists and I have no doubts about it" - while one-third of American adults have expressed some doubts and two percent have expressed outright disbelief. Nor are Americans' beliefs and practices immutable; although they have changed rather little (Greeley 1997), change remains a possibility.

Belief in life after death is a case in point. The Christian tradition teaches that the human person survives after the body dies. The New Testament contains many sayings attributed to Jesus on the subject of the life that comes after death. Other American religious traditions give less emphasis to the hereafter. For Jews, the Mishna mentions "resurrection and the world to come." This is not an equivalent doctrine; the Christian teaching is usually phrased in personal terms not directly echoed in the Mishna teaching. Nor is this teaching as central to the Jewish religion as "life everlasting" is to Christian traditions. So we expect variation among major religious traditions. Precisely because the fate of the person after death is a core religious question based on writings that are thousands of years old, it seems reasonable to hypothesize that beliefs about it will change little if at all.⁽²⁾

Survey data show that most Americans assent to the simple question "Do you believe there is a life after death?"⁽³⁾ Between 75 and 85 percent of American adults give an affirmative answer according to Gallup polls dating from the 1940s through the 1980s (American Institute of Public Opinion [AIPO] 1985) and the General Social Survey (Davis and Smith 1996).⁽⁴⁾ In 1996, 82 percent of American adults reported to the GSS that they believe in life after death. Belief in life after death was higher in 1996 than in previous years. In 1973, 77 percent of American adults reported a belief in life after death. The 1973 figure was no fluke. Gallup data consistently show similar percentages for earlier periods: 76 percent in 1944, 77 percent in 1952, and 75 percent in 1965 (AIPO 1985).

As expected, Christians are more likely to believe in life after death than non-Christians are. Our calculations from the GSS show that 86 percent of Protestants, 84 percent of Catholics, 54 percent of Jews, 68 percent of persons of other religions, and 64 percent of persons with no religious preference hold this belief. Although we expected some denominational variation, these numbers are surprising: 54 percent seems high for Jews and 64 percent is a very high rate of belief for persons who do not identify with any organized religion. The 1973 data show that belief in life after death increased for all groups except Protestants - 84 percent of Protestants, 76 percent of Catholics, 18 percent of Jews, and 46 percent of persons with no religious preference affirmed belief in life after death.⁽⁵⁾ The increases for Catholics, Jews, and persons with no religious preference are all statistically significant. Table 1 fills in the intervening years. We have grouped them because small numbers of Jews and people without affiliation make inference difficult when we use the annual observations. The monotonic increases among Jews and persons with no religious preference reinforce the impression of significant change garnered from comparing the endpoints. The trivial reversal among Catholics (78-77-78 percent in the 1978-82,1983-87, and 1988-92 periods, respectively) hardly detract from the impression that even Catholics' belief in life after death has increased be a modest amount.  #Table1

Belief in life after death among Protestants is the highest among American religious categories in each period - varying little around the figure of 85 percent. Classifying Protestant denominations as "liberal," "moderate," and "fundamentalist" based on their teachings about the Bible and other Christian fundamentals (Smith 1990) we find that Protestants from liberal denominations might be slightly less likely than other Protestants to believe in life after death, but not consistently so in all years and not significantly so in any year (details not shown).

Catholics' belief in life after death has caught up with Protestants' only in the most recent years despite the prominence of "life everlasting" in Catholic teaching. This is surprising considering the generally strong doctrinal orthodoxy of American Catholics on other religious matters such as the divinity of Jesus and the efficacy of the sacraments (cite?).

Among Jews belief in life after death has increased 142 percent from 19 percent in 1973-78 to 46 percent in 1994-96. The GSS samples are too small to yield many Jewish respondents in any one year, so the annual estimates for Jews are quite imprecise. Although none of the differences from one period to the next are significant, the trend is monotonic and the model of null association between year and belief is rejected. Thus the preponderance of evidence points to the conclusion that the increased belief in life after death among Jews in the GSS has a real-world counterpart (X² = 50.54; df = 4; p < .05).

Finally, the conviction that the human soul survives death increased among adults who have no religious affiliation - from 44 percent believing to 58 percent. The model of null association between the five time periods and belief is also rejected (X² = 18.62; df = 4; p < .05).

Thus, Catholics, Jews, and adults with no religious affiliation all significantly increased their belief in life after death between the early 1970s and the mid-1990s.
 

CONSIDERING WHETHER THE TREND IS AN ARTIFACT

Religious belief and behavior clearly relates to the life cycle - older persons are generally more religious than younger persons (e.g., Hout and Greeley 1987) - so trends during times of population aging might reflect that aging process rather than widespread changes in individuals' beliefs. Because of increased longevity and rapidly decreasing fertility from 1965 to 1974, the adult population of the United States did age during the period from 1973 to 1996. So some or all of the increased belief in life after death could be attributable to the aging of the population. First we take life cycle into account by comparing cohorts. The cohorts born early in the twentieth century are older but more likely to have been interviewed in the 1970s than later. Conversely most respondents born since 1960 are younger but most were interviewed in the 1990s. If life cycle and population aging are responsible for the cross-sectional increase in belief in the afterlife, then cohorts will be less different than time periods appear to be. The cohort data are in Figure 1. They show not only increases across the cohorts born this century but also more change among Catholics and Jews than the period perspective shows. The cohort data even show modest increases among Protestants. (Figure 1 about here)

Pursuing the issue of change among cohorts further, we ran a logistic regression with the log-odds of believing in life after death as the dependent variable and age, age-squared, cohort (coded as dummy variables for each decade), religion (also dummy variables), and the interaction between religion and cohort.⁽⁶⁾ We used the results of that regression to calculate the expected percentage believing in life after death for each combination of religion and cohort, setting age equal to 40 for all cells. The cohort main effect and its interaction with religion are significant. Thus the monotonic trends in belief in life after death statistically adjusted to age 40 are significant for all groups except Protestants.

Is it possible that the trend appears because the meaning of the question itself has changed across cohorts? Do young people understand "life after death" differently than older persons do? Fortunately, the 1983 and 1984 GSSs followed up the standard question about life after death with probes about what that afterlife might be like.

Of course nobody knows exactly what life after death would be like, but here are some interesting ideas people have had. [HAND CARD TO RESPONDENT.] How likely do you feel each possibility is? Would you say very likely, somewhat likely, somewhat unlikely, or very unlikely?

1. A life of peace and tranquility.
2. A life of intense action.
3. A life like the one here on earth only better.
4. A life without many things that make our present life enjoyable.
5. A pale, shadowy form of life, hardly life at all.
6. A spiritual life, involving our mind but not our body.
7. A paradise of pleasure and delights.
8. A place of loving intellectual communion.
9. Union with God.
10. Reunion with relatives.

If some images of the afterlife are more prevalent among persons from one tradition than among others, then we would have to worry that the standard item is not comparable across denominational bounds. If some images are more prevalent in recent cohorts than in earlier ones, while others are more prevalent among older cohorts than among recent ones, then we can say that the meaning of the standard question has probably changed. Figures 2 and 3 allay those concerns. Figure 2 shows that - among persons who believe in life after death - images of what that life might be like differ very little from one religious tradition to another. Nearly all Christians think that union with God, peace and tranquility, and reunion with relatives are very likely or likely to await them in the afterlife. Between 85 and 95 percent of Christians also see a place of loving intellectual communion and a spiritual life involving mind not body. Smaller majorities of Christians endorse the "garden of Eden" imagery - paradise of pleasures and delights and like her on Earth only better. A minority of Christians thinks that action is likely in the afterlife (40 percent) and that the afterlife is likely to be pale or shadowy (20 percent). Jews rank most of the images the same way Christians do (although they see eight of the ten as slightly less likely than Christians do), except a higher proportion of Jews than Christians think that an intellectual afterlife (87 percent) and a pale and shadowy existence (49 percent) are likely. People with no religious affiliation have views that are very similar to liberal Protestants' views, but they are less certain (not shown in the figure is the fact that for each item, a larger percentage of liberal Protestants chose "very likely" and a larger percentage of those with no religion chose "somewhat likely"). (Figures 2 and 3 about here)

Images of the afterlife differ very little across cohorts. To the extent to which any differences emerge, they suggest that younger people have simpler images of the afterlife than their elders. The average number of "somewhat likely" and "very likely" responses among persons born since 1960 is one less than the number of those responses among persons born prior to 1920. None of the images become more popular among the recent cohorts.

Thus the trends toward increasing belief in life after death cannot be dismissed as mere artifacts of changing meanings. The major traditions agree on most elements. Christians are slightly more likely than Jews to imagine a personal or recognizable existence, while half of Jews but only a fifth of Christians see a vague existence as likely. Older and younger people have very similar images of the afterlife.

INSTITUTIONAL COMPETITION AS A SOURCE OF CHANGE

In recent years scholars have come to attribute America's extraordinary religiousness not to characteristics of the American people but to the competition among religious denominations in the United States (Finke and Stark 1992; Warner 1993). The "supply side" theory of religious behavior proposes that American religiosity is a special case of a general tendency for competition among churches to stimulate religious practice. We do not propose to test this proposition. We are not even sure whether the theory can generate a testable hypothesis about beliefs such as that in life after death; the theory may only be relevant for behavior and practice. We are interested in using the theory to stimulate our thinking about the subject.

The institutional structure of American religion has changed little since the 1920s. The fundamentalist and Pentacostal movements swept through Protestantism early in the century. Since then there has been a steady movement of individuals out of the so-called mainline Protestant denominations and into the fundamentalist and Pentacostal churches (Wuthnow 1989; Finke and Stark 1992). But that was all within Protestantism. Catholics have been 25 percent of the country's population at least since the 1940s and probably since immigration was curtailed in 1924. Jews have constituted between 3 and 5 percent of the population throughout this century.

Immigration has dramatically altered the composition of religious communities. Among Catholics born before 1920, 70 percent are immigrants or the children of immigrants. Among Catholics born since 1940, 75 percent are natives born to native parents. Among Jews 91 percent of those born before 1920 are immigrants or children of immigrants; 75 percent of those born since 1940 are natives born to native parents.⁽⁷⁾ Thus only a minority of the Catholics and Jews from the earliest cohorts received their religious socialization in the United States and half of them were living with parents who had been educated abroad. In contrast, the vast majority of Catholics and Jews from more recent cohorts were not only socialized in the United States, their parents were also. The immigrants came from countries that had far less religious competition.

Our hunch, loosely derived from the supply side theory, is that religious competition for the hearts and souls of the immigrants led to a more vigorous religious socialization in this country than youth in most of the countries that sent emigrants to this country. The upshot is greater orthodoxy among the children and grandchildren of immigrants than existed among the immigrants themselves. The statistical test of our hunch will come when we contrast the beliefs of immigrants with those of native-born persons. But it is not enough to show merely that belief in life after death is more prevalent among native-born than among immigrant Catholics and Jews. The crucial test will be whether significant differences among cohorts are spurious, i.e., whether they disappear when we control for immigrant status. Immigrants are a larger share of earlier than later cohorts, as we documented, so the possibility that we might obtain such a result is real. If complications arise, as they always do, our strategy will be to add explanatory factors that pertain to the actions of organized religion, e.g., the establishment of church-affiliated organizations and schools, to the regression equation to see how far the supply-side argument will take us. We will also consider contact with Protestants - through intermarriage - as another possible mechanism.
 

CATHOLIC IMMIGRANTS AND THEIR CHILDREN AND GRANDCHILDREN

We begin with Catholics. They are far more numerous in this country (and in the GSS data file) so we can explore a richer set of tests without running out of cases. Since 1977 the GSS has been asking respondents where they were born, whether their mother and father were born in the United States or some other country, and how many of their grandparents were born in the United States. From the answers to these questions we have constructed a "generation" variable that distinguishes among immigrants born overseas to foreign parents, second-generation Americans who were born in the United States but whose father or mother was foreign-born, third-generation Americans whose parents were born in the United States but who have at least one foreign-born grandparent, and fourth-generation Americans who have no foreign-born grandparents.
 

STATISTICAL TESTS OF THE IMMIGRATION ARGUMENT

We need go no farther with our exploration if immigrants are not, in fact, less likely than natives to believe in life after death. Table 2 presents this first test: the proportion of Catholic adults from each cohort who believe in life after death broken out by the number of generations the family has lived in the United States. In recent cohorts there is very little difference between the beliefs of immigrants and natives, but for Catholics in general, and for Catholics from the oldest cohorts in particular, belief in life after death increases across generations. In the oldest cohorts barely half of immigrants, two-thirds of the second generation, three-fourths of the third generation, and five-sixths of the fourth generation believe in life after death.  #Table2

Having established that both cohort and generation are associated with belief in life after death, we turn to logistic regression analysis to assess the extent to which generation accounts for the cohort pattern. The proportions in Table 2 suggest that the relationship between generation and belief varies among cohorts. Our strategy is to establish the additive baseline first, then move on to interactions. We begin with cohort and age, add controls for potentially confounding factors (education, family size, and marital status), and then consider the additive and interacting effects of generation. The cohort patterns are not linear, so we need to use dummy variables to model them properly. That much is standard. But we would like a single measure to assess the strength of the nonlinear association. Goodman (1986) provides such a tool, the standard deviation of the logistic regression coefficients that represent the cohort effect (also see Hout, Brooks, and Manza 1995). We call this index "kappa" and write it in the rest of this article. Selected statistics from our logistic regression analysis are reproduced in Table 3.  #Table3

The baseline model returns significant effects for cohort, age, and age-squared. The baseline is .700. Educational attainment and divorce have risen over this time span; family size has fallen. We control for the potential for these variables to confound our attempts at inference in the second model. Education proves to be the key confounding factor. As with many other indicators of religiosity and doctrinal orthodoxy (Greeley, McCready, and McCourt 1976; Greeley 1977), education increases Catholics' adherence to the doctrine of life everlasting. Belief in life after death among Catholics who graduated from college runs about 11 percentage points ahead of its level among Catholics who stopped their educations at the end of high school, and 16 percentage points ahead of Catholics who dropped out of high school.⁽⁸⁾ Furthermore this strong effect of education accounts for the net association between age and belief (net of cohort), so age is not significant when education is in the equation. Education also accounts for one-third of the association between cohort and belief in life after death. The younger cohorts are more likely to believe in part because they have more education and education increases orthodoxy.

The third model adds generation. Although the additive effects of generation are substantial and significant, they do not change the net cohort difference very much at all. The value of decreases by a trivial .006 and the Wald test (which is distributed as chi-square with seven degrees of freedom) is still significant after generation is in the equation. But as the pattern in Table 2 suggested, there is a significant interaction between cohort and generation. We experimented quite a lot with the functional form of the interaction and concluded that the simplest form that communicated the full impact of the interactions is to distinguish between the fourth generation and the other three (taken as a group). The pattern in Table 2 actually makes the third and fourth generations look similar and the first two different, but once education is in the equation, it is the fourth generation alone that is distinct. For the fourth generation, there is a slight, non-significant relationship between cohort and belief in life after death. The value of for the fourth generation is .260, i.e., one-third its original size and 56 percent as large as in the additive model. For generations one through three, on the other hand, the cohort trend is actually stronger than it appears in the additive model ( = .580). Thus some of the increase in belief in life after death among Catholics has come about because Americans whose families have been in the United States for at least three generations have always been orthodox in this way; now they are a larger proportion of the Catholic population than they used to be. But there is more to the story, because the belief in life after death shows a residual increase in generations one through three not accounted for by contact with America.

One of the key changes in the composition within the first three generations is the change over time in the source of immigrants (e.g., Lieberson and Waters 1987). Nineteenth-century Catholic immigrants were predominantly from Ireland and Germany. Late in the nineteenth century that began to shift as immigration from Italy, Poland, and elsewhere in Southern and Eastern Europe swamped the continuing influx of Irish. Some of these sources of immigration - notably Ireland and Poland - are devout countries; others - notably Italy and Mexico - are not. Table 4 summarizes this state of affairs. The Irish of all generations are highly likely to believe in life after death (even more than Irish in Ireland in 1991). Catholics from Italy, the Germanic countries, and Latin America show sharp increases from the immigrant generation to the second generation; for the European groups the trend continues on to the third and fourth generations as well.  #Table4

We add ethnicity to the mix of factors in our logistic regression model to understand the first three generations. Indeed the changing ethnic composition of immigrants and their offspring across cohorts is responsible for some of the trend in belief in life after death. Adding ethnicity to the equation reduces from .580 to .360. The other important aspect of the changing ethnic mix is the mix that emerges within families through intermarriage (e.g., Pagnini and Morgan 1990; Hout and Goldstein 1994). We add a term for having married a Protestant (i.e., religious intermarriage, not ethnic intermarriage) to our model. To our surprise it plays almost no role in increasing belief in life after death. This is a very important null result for understanding the source of change. After all, the Protestants always exhibit very high rates of belief in life after death. If it was contact with Protestants that has led to the increased faith (or is it hope?) among Catholics, then this form of intimate contact should intensify the effect. That we observe neither an effect of intermarriage on the cohort differences nor even a main effect of marrying a Protestant on that person's belief in life after death indicates to us that we need to eschew the current fashion of finding Protestantism in every American institution (George 1997; Bellah 1998) and look within the Catholic community for mechanisms of orthodoxy.

Hints of Catholic community abound in ethnographic evidence. They are scarce in the GSS, unfortunately. Two powerful ones we have at our disposal are membership in church-affiliated organizations and attendance at Catholic schools. As American Catholics built their institutional base - and the main actors in the building were Irish (more on that in the next section) - they first established the parish as village within the city. Then they build the schools. Adding participation in a church-affiliated organization to the logistic regression model shows that this form of parish involvement brought immigrants and their offspring closer to the orthodox view about life after death. Members of church-affiliated organizations are much more likely than others to hold the orthodox belief (b = 1.047). More importantly for our purposes, the growing participation of the non-Irish in these organizations accounts for their increasing orthodoxy in recent cohorts. The value of decreases from .360 to .288 (half its original value of .560 for this group); the Wald statistic of 7.81 with seven degrees of freedom is not significant at conventional levels.⁽⁹⁾

When we add attendance at Catholic schools (measured in years of Catholic schooling) to the logistic regression model we discover that the Catholic education wipes out all other variables in the model. We do not present the details of this part of the analysis for three reasons. The test is weak because the church schooling item was available in only three years, leaving us with just 411 cases. The years are close together (1988, 1989, and 1991) so we cannot separate age and cohort. It is also difficult to separate the effects of church schooling and education per se with so few cases. Thus we simply note that the simple correlation between our dichotomous dependent variable and years of Catholic education (among Catholics in generations one through three, including Irish) is .28.
 

IF NOT THE IRISH, THEN WHO?⁽¹⁰⁾

The non-Irish Catholic immigrant groups have clearly acculturated with generation into a level of belief in the survival of the human person after death comparable to both that of their Irish fellow religionists and of their fellow Americans who are Protestant. The question then arises as to whether the cultural norms leading to this change are the norms of Irish Catholic Americans or Protestant Americans. The issue is one which cannot be completely resolved.

We incline to an Irish Catholic explanation. For an explanation to be acceptable, it must be more than a vague appeal to cultural osmosis; rather it must specify (as best it can) the mechanisms by which acculturation takes place. That kind of Protestant influence which might most directly influence a modification of belief - marriage to a Protestant - seems to have no effect. However, a Catholic influence - membership in a church-affiliated organization - does seem to play an important part in the increase in belief in life after death across generation lines. Moreover, for most immigrants there were no such things as religious organizations to join in their countries of origin. The strategy of creating a web of organizations to protect immigrants from Protestant proselytizers was developed by Irish church leaders who, well into the second half of this century, were mostly Irish - half the bishops and the third of the priests were Irish as recently as 1970. The later immigrant clergy were quick to adopt the same strategy (Dolan 1975, 1985, 1987; Gleason 1987).

The first empirical sociological observation of a Catholic immigrant neighborhood is in Thomas and Zaniecki (1927). Reporting on the situation during the second decade of this century in a Polish immigrant neighborhood in Chicago, they describe the parish a "cocoon" which surrounds the whole life of its parishioners to protect them from the outside world. Such cocoons did not exist in Poland because they were not necessary to protect against Protestant encroachments. Once the cocoons came into existence, however, they also became structures for socializing immigrants' children into orthodox beliefs and behavior, which, in practice meant beliefs and behavior acceptable to Irish Church leaders. As Lopreato (1970) observes: "The Italians, for two thousand years the chief trustees of Roman Catholics, have adopted in America a form of Catholicism that socially and ritually bears little resemblance to what they knew in Italy.... The Italians join in prayer their Polish, German and especially Irish co-religionists and send their children to the same schools." Greeley (1974) and others speak of the "Hibernization" of Catholic immigrants based on Irish dominance of the Catholic Church and Irish models (the network of organizations and parochial schools) as the preferred models for protecting the faith of the immigrants. Moreover these cocoons, urban villages of a sort (McGreevy 1996), have persisted long after the initial immigration. Even to the present they provide a means of social location for third and fourth generation Catholics (McMahon 1995). We judge that this strategy, which not only insulated to a considerable extent Catholics from Protestant influence but which also reinforced Catholic orthodoxy, is compatible with the influence of organizational membership in table 5. On this historical record and its trace in our statistical evidence we incline to attribute the phenomenon reported in this paper largely to the Irish influence in the American Catholic Church.
 

BELIEF IN LIFE AFTER DEATH AMONG JEWS

Belief in life after death is not a question of orthodoxy for Jews as it is for Christians. As belief in the afterlife increases for Jews, they are innovating at least as much as they are conforming. So we will be surprised if we discover that the same mechanisms we have found for Catholics promote hope in the Jewish community. We have already seen that Jews have slightly different images of the afterlife (union with God and reunion with family are the predominant images but slightly less prevalent). Jews also have more doubts about the afterlife; Jews are twice as likely as Christians to say that they "don't know" if there is a life after death (17 percent compared to 8 percent). Finally, persons who grew up Jewish are more likely than Christians to prefer no religion in adulthood. Both uncertainty and distance from organized religion have to be incorporated into the statistical analysis.

Our strategy in addressing the rise in Jews' belief in life after death is the same exploratory one, guided by insights from supply-side theory, that we used to study Catholics. Our tactics differ slightly from the approach we took to Catholics, though. We include all persons who were raised Jewish in the analysis and then use "currently Jewish" as a factor in the analysis. We use an ordered logit model that allows three outcome categories, taking into account the rank ordering of them with respect to belief; the order is noncontroversial: no belief, can't say, believes. The ordered logit model resembles logistic regression, but it assumes an unobserved latent trait that is "cut" at two points. These cut-points are parameters of the model that are estimated along with the coefficients for the substantive variables. The results are in Table 5.  #Table5

The first model is the simple demographic one containing just cohort, age, and age-squared.⁽¹¹⁾ Recall from Figure 1 how the upswing in Jewish belief in life after death begins with cohorts born after the 1930s. That is evident here as well. The first three cohorts are virtually identical. The 1940-49 cohort is one point higher on the logit scale, the 1950-59 a half-point higher than that, and the last cohort about two-thirds of a point higher yet (for an accumulated gain of between 2.25 and 2.30 on the logit scale). This is the trend to be explained. It results in a of .89. Age and age-squared are not significant.

For Jews increased education reduces belief in life after death. Belief in life after death among college graduates is between 7 and 11 percentage points lower than that of high school graduates, depending on cohort. Part of the trend across cohorts is actually suppressed until we control for the sharp upward trend in education. Thus the for model 2 is 1.03.

Immigrant generation does not account for the increase in Jews' belief in life after death. The bivariate association between generation and belief is not significant (X² = 6.96; df =6). Nor is the net effect of immigrant generation when it is added to model 2 (see model 3).

Contact with Christian and other groups is important for identifying which Jews are likely to believe in life after death. Persons who were raised Jewish but converted to Protestantism are far more likely to believe in life after death. Those who prefer no religion are significantly less likely than persons who are currently Jewish to believe in life after death. The main source of this Christian conversions among Jews is marriage. But the spouse's religion has no effect on the respondent's beliefs unless the respondent converts (model 5).

None of these factors - and no other factors we could identify - explain the cohort differences. The kappa values do not diminish when we add current religion or spouse's religion to the equation. We tried a number of measures that might be thought of as proxies for contact with Christians - residence in major metropolitan areas, central city versus suburban residence, and regional mobility. We also considered the role of strength of religious identification and membership in religious organizations. None of these variables proved to be significant.

Thus we have an enigma. Belief in life after death increased dramatically for Jewish Americans - those raised Jewish and those who are currently Jewish. Intermarriage and conversion to Christianity are important for beliefs in the cross-section, but they do not explain the trend.
 

DISCUSSION

We have observed a modest increase in belief in life after death among Americans. Surprising to us is the fact that Protestant have not changed their (already high) beliefs. The modest change in the cross-section is due to sharp increases across cohorts among Catholics, Jews, and persons with no religious affiliation.

A family's number of generations in America accounts for the change among Catholics. Immigrants, it turns out, are the Catholic group least likely to believe in life after death. Persons who are the grandchildren of immigrants and those whose grandparents were born in the United States are the most likely to believe. The fall-off in immigration between 1924 and 1980 means that recent cohorts have a larger proportion native born and therefore a larger proportion believing in life after death.

This explanation was suggested to us by the supply-side theory of religious competition. We make no claim that our exploration has confirmed the supply-side theory. We merely acknowledge that the theory suggested a clue that turned out to unlock the explanation for Catholic belief and part of the explanation for Jewish belief. Our findings contradict the widely-held but incorrect assumption that immigrants are more religious than native-born Americans. Quite the contrary, immigrants from most parts of the world (Ireland and Poland are exceptions) where faith is weaker than it is in the United States. Most do not give up the scepticism they bring. But their children and grandchildren -- socialized in the more devout American environment -- do come to conform to the American levels of belief.

Thus, the change in Catholics' belief in life after death is rooted in demographic trends. It would be a misreading of the evidence to cite it as the stirrings of a religious revival in the United States. Rather this increased is appropriately viewed as the socialization of less-than-devout immigrants (Catholic ethnics) into a more devout society than the one from which they came. The competitive religious environment in the United States tended to view the immigrants as potential converts. Some did convert to another religion. Most, however, did not. They had offspring who responded to the new religious culture by exploring the faith of their fathers more deeply. For Catholics they came away with more orthodox beliefs - including hope in an afterlife.

If we had not begun the investigation of the apparent cross-sectional increase in belief in life afer death with the supply-side theory in mind, it is most unlikely that we would have discovered this phenomenon of cross-generational socialization that brings the descendants of immigrants in line with the native-born co-religionists of an earlier time. The supply-side theory has generated testable hypotheses that have been investigated by others. Here it gets credit for forcing analysts to explore areas that might otherwise have been missed. The theory turns out to be not only predictive but also useful.

REFERENCES

American Institute of Public Opinion [AIPO]. 1985. Fifty Years of Religious Change. Princeton: AIPO.

Bellah, Robert N. 1998. "Is There a Common American Culture?" Journal of the American Academy of Religion 66: 613-625.

Davis, James A., and Tom W. Smith. 1996. General Social Survey Cumulative Codebook, 1972-1996. Storrs CT: Roper Center.

Dolan, Jay P. 1975. New York's Irish and German Catholics 1815-1865. Baltimore: Johns Hopkins University Press.

---------- 1985. The American Catholic Experience. Garden City, N.Y. Doubleday.

---------- 1987. The American Catholic Parish. New York: Paulist Press.

Finke, Roger, and Rodney Stark. 1992. The Churching of America, 1776-1990: Winners and Losers in Our Religious Economy. New Brunswick: Rutgers University Press.

George, Francis. 1997.

Gleason, Philip. Keeping the Faith: American Catholicism, Past and Present. Notre Dame Ind.: University of Notre Dame Press.

Goodman, Leo A. 1986. "The Analysis of Cross-Classified Data Having Ordered and/or Unordered Categories: Association Models, Correlation Models, and Asymmetry Models for Contingency Tables With or Without Missing Entries." The Annals of Statistics 13: 10-69.

Greeley, Andrew M. 1974. Ethnicity in the United States. New York: Wiley.

Greeley, Andrew M. 1977. The American Catholic: A Social Portrait. New York: Basic Books.

Greeley, Andrew M. 1995. Religion as Poetry. New Brunswick NJ: Transaction Books.

Greeley, Andrew M. 1997. Religious Change in America. Paperback edition. Cambridge: Harvard University Press.

Greeley, Andrew M., William C. McCready, and Kathleen McCourt. 1976. Catholic Schools in a Declining Chruch. Kansas City: Sheed and Ward.

Herberg, Will. 1955. Protestant-Catholic-Jew. Garden City, NJ: Doubleday.

Hout, Michael, Clem Brooks, and Jeff Manza. 1995. "The Democratic Class Struggle in U.S. Presidential Elections, 1952-1992." American Sociological Review 60: 805-828.

Hout, Michael, and Joshua R. Goldstein. 1994. "How 4.5 Million Irish Immigrants Became 40 Million Irish Americans: Demographic and Subjective Aspects of the Ethnic Composition of White Americans." American Sociological Review 59: 64-82.

Hout, Michael, and Andrew M. Greeley. 1987. "The Center Doesn't Hold: Church Attendance in the United States, 1940- 1984." American Sociological Review 52: 325-345.

Lieberson, Stanley, and Mary C. Waters. 1987. From Many Strands: Ethnic and Racial Groups in Contemporary America. New York: Russell Sage Foundation.

Lopreato, Joseph. 1970. Italian Americans. New York; Random House.

McGreevy, John T. 1996. Parish Boundaries. Chicago: University of Chicago Press.

McMahon, Eileen. 1995. What Parish are You From?. Lexington, Ky. University of Kentucky Press.

Pagnini, Deanna, and S. Philip Morgan. 1990. "Ethnic Intermarriage at the Turn of the Century." American Journal of Sociology 95: 405-432.

Smith, Tom W. 1990. "Classifying Protestant Denominations." Review of Religious Research 31: 225-245.

Thomas, W. I. and Florian Znaniecki. 1922. The Polish Peasant in Europe and America. Chicago: University of Chicago Press.

Warner, R. Stephen. 1993. "Work in Progress Toward a New Paradigm for the Sociology of Religion." American Journal of Sociology 98: 1044-1093.

Wuthnow, Robert. 1989. The Restructuring of American Religion. Princeton: Princton University Press.

ENDNOTES

1. Thanks to Margaret L. Anderson, Paula England, Michael McGarry, and Melissa Wilde for helpful comments on a previsous version of this paper. We received financial support from the UC-Berkeley Survey Research Center.
 

2. Of course advocates of the secularization hypothesis might predict sharp declines in belief in life after death. After all, there is no credible scientific evidence that a soul exists. As more people base important decisions on scientific evidence, they may come to abandon belief in life after death.

3. This question does not explicitly mention "the soul" or other Christian concepts. It might be ambiguous whether this question has any religious content or refers to something else were it not part of a battery that includes other religious questions. Given the context of the question, we are confident that we are not reading anything into the positive and negative answers to it that respondents did not intend.

4. All of the original calculations in this paper come from the General Social Survey (GSS). Begun in 1972, the GSS is a nationally representative sample of American adults (18 and over) who speak English well enough to understand the interview. The 1972 GSS used block quota sampling. Since 1975, full probability sampling has been used (1973 and 1974 were transition years that split the sample between quota and probability sampling). Until 1993 the average completed sample size was just under 1,500 interviews. Design changes in 1994 (repeated in 1996) resulted in a "double sample" of nearly 3,000 interviews in the even-numbered years but no survey in the odd-numbered years. Since the introduction of full probability sampling, the GSS has maintained a 77 percent response rate. See Davis and Smith (1996) for details or consult the GSS Data and Information Retrieval System (GSSDIRS) website: www.icpsr.umich.edu/gss. The question about belief in life after death has been asked 16 times beginning in 1973. Since 1989 it has been on two of the three "ballots," i.e., asked of two-thirds of respondents.

5. For this calculation and for the remainder of the paper we delete the amorphous "other religion" category because we cannot discern enough about this group to understand changes in it or differences between it and the more specific categories.

6. Specifically:    y_i = ln(p_i / 1 - p_i) = ₀ + ₁Age_i + ₂Age_i² + _kkX_k + _kkX_k+_{kk k,k}X_kX_k  

where p_i is the probability that person i believes in life after death, y_i is the result of the logit transformation of p_i, Age_i is person i's age (in years), k = 3,...,10 indexes the dummy variables for cohorts categorized as before 1900, 1900-09, 1910-19, ..., 1970-78 (with the first cohort as the reference category), k = 11, 12, 13 indexes the dummy variables for religions (Protestant, Catholic, Jewish, and None with Catholic as the reference category), and the s are coefficients to be estimated from the data.

7. The statistics in this paragraph are our calculations from the GSS. Neither the U.S. Bureau of the Census nor the Immigration and Naturalization Service collects data on religion.

8. This calculation is based on the formula dp/dx = bp(1-p), where p is the percentage believing in life after death, x is education, b is the logistic regression coefficient for x, and dp/dx is the partial derivative of p with respect to x, i.e., the effect of education on the percentage believing. Taking p = .75 and b = .143, we get .143*.75*.25 = .0268, which is the expected change in p for a one-year increase in x, evaluated at p = .75. College graduates have four years more education than high-school graduates, so we multiply .0268 by four to get the expected difference between them; college graduates have six years more education than the typical high-school dropout, so we multiply .0268 by six to get that difference.

9. We note in passing that about one-fourth of the effect of education is also indirect through membership in church organizations (compare .099 and .077).

10. The source of this question is Gleason (1987).

11. We had to combine the first three cohorts into a single category and the last two cohorts into another single category because we only have 496 cases (and lose between 20 and 25 percent of them when we add other variables).

Table 1

Table 2

Table 3

Table 4

Table 5