Changes in the past half century in the attitudes of the American public regarding euthanasia and suicide in the case of incurable disease have been dramatic, and they attest to the success of a social movement that has been in part a phenomenon “of the times” (which is what we say when we do not really understand why attitudes change as they do). But they are also in part a consequence of a highly visible social movement and vigorous deliberate actions intended to shape public opinion, moving it in a direction favourable to concrete changes in law, public policy, and precepts of the profession of medicine.

During much of this period, a bewildering plethora of polls has been conducted, related in one way or another to the question of how the public has viewed the issues about end of life decisions. End of life decisions have an important bioethical component, especially as they raise questions of who makes choices and how those choices can be made in an ethical manner. Does the individual have complete control of his or her own fate? Who can act as surrogate? What role, if any, should physicians play, and what are the interests of the state in such decisions? Anyone who wants to write a justification for, or promote advocacy of, any kind of policy regarding these issues ought to be seriously acquainted with trends in public attitudes. While public opinion cannot be the final arbiter of what is ethically correct, knowledge of it will help us to address the issues in a way that will best benefit the public discourse. Unfortunately, hardly any presentation of the polls related to the subject has attempted to reconcile the differing findings, to account for the differences, or to discern a pattern in the results.

Some commentators or advocates on either side of the debate have disregarded the poll data entirely in favour of unfounded speculation. Father Joseph Woodill, for instance, in his article, Euthanasia, physician assisted suicide, and the pursuit of death with dignity,¹ sketches a long view of the issues based on Philippe Ariès’s book, The Hour of Our Death,² but he seems to think nothing happened before the Karen Ann Quinlan case in 1976. To the contrary, polling data from before the Quinlan case indicate a good deal of support for euthanasia and suicide already (although it is entirely possible—perhaps even probable—that Quinlan, and later the 1990 Nancy Cruzan case, in which the Supreme Court ruled for the first time that competent adults have the right to refuse medical treatment, did have impacts on public opinion).

Others have settled for a haphazard citing of whatever polls they happened to have heard of, without any attempt to pull together the body of poll evidence to find out what has actually been going on. One case in point can be found on the website of Compassion and Choices (in part formerly the Hemlock Society) which uses poll results³ to support its position that “in the event of a terminal illness, we each should have the right to choose the time and manner of our death when the course is prolonged or agonized”.⁴

Another is an article by Jordan M Blanke, Physician assisted suicide and the right to die, which until recently was available on the internet. Blanke mentions some six or eight different polls on rather varied topics, but his only conclusion is that “during the last ten years [they] indicate general support for some sort of physician assisted suicide”. (A later version of this paper has appeared.⁵)

While such uses of polls may seem appropriate for advocacy and propaganda, they do little to help us understand how the American public really evaluates end of life choices. The problems are numerous. Sometimes what seem like inconsequential differences in the wording of questions produce dramatically different approval rates. A CBS News poll in 1998 registered 16% affirmative when respondents were asked: “If a terminally ill relative or close friend asked you to help him or her commit suicide to end his or her suffering, do you think you would or would not?” However, a nearly contemporaneous poll by Opinion Dynamics obtained 42% affirmative responses to the question: “If a member of your family were terminally ill and wanted to die, would you be willing to help him?”⁶ (All survey results cited in this study, with the exception of the World Values Survey mentioned in references 9 and 10, are available from iPOLL, an online database of US public opinion polls with national adult samples maintained by the Roper Center for Public Opinion Research, University of Connecticut.)

Muddying the waters further is a reliance by commentators on polls that may not lend themselves to systematic analysis no matter how carefully we compare results. Since most polls are carried out for commercial purposes and/or journalistic uses, their coverage of subject matter tends to be opportunistic. Perhaps a media pollster sees an opportunity for a good “story” in assessing, for instance, public reactions to a recent Jack Kevorkian incident. The timing of such a poll is adventitious, and there may be no baseline assessment for some prior time, which makes it less likely that the reader will read it as a reaction to a recent incident.

One notable exception to the otherwise sketchy treatment of polling data on euthanasia and assisted suicide is Poll trends: end of life issues, by John M Benson, which appeared in the summer 1999 issue of Public Opinion Quarterly.⁷ His comprehensive review of the survey results available on the subject at that time provides an excellent springboard for this discussion, which has the advantage of access to four more years of data and offers some analyses that extend beyond the scope of his coverage. Benson also presents some interesting hypotheses of his own which are not dealt with here. (A 1992 article by R J Blendon, et al, that appeared in JAMA looked at public opinion on physician assisted suicide but was not as complete in its coverage of polls as Benson, even for the earlier period.)⁸

Systematic examinations of cross-national data are also hard to come by, with variations in question wordings and survey methodologies in different countries making direct comparisons impossible. One important exception is the World Values Survey, conducted since 1981 in some 80 societies by a network of social scientists at universities throughout the world. In the twenty one countries for which data on attitudes toward euthanasia and suicide are available for 1981, 1990, and 2000, trends show the same growing acceptance of euthanasia as seen in the US in our study. Attitudes toward suicide are more mixed, but for the most part also reflect a sharp decrease in disapprobation of the act. (Some notable exceptions to this trend show marked movement in the opposite direction since 1990. These include Ireland, Hungary, Finland, and, especially, Mexico, and warrant further investigation.)^9,10

DATA SOURCES

Typically ignored by commentators on end of life issues is the General Social Survey (GSS), a long running survey of social, cultural, and political indicators conducted in the US by the National Opinion Research Center (NORC). Founded in 1972, the GSS is widely used by social scientists to study ongoing trends. It was conducted more or less annually until it became a biennial survey in 1994. Surveyed are national probability samples of non-institutionalised adults, with sample sizes of 1,500 for the first nineteen surveys, and 3,000 after the survey became biennial.

When interpreting GSS data and the data from public opinion polls that will be discussed here, it should be recognised that all surveys are subject to sampling error. Results may differ from what would be obtained if the whole population of adults or voters had been interviewed. A sample size of 1052 will, with a 95 per cent degree of confidence, have a statistical precision of approximately plus or minus three percentage points.

In the case of the GSS, data are collected by means of in person interviews lasting approximately ninety minutes. Response rates are high, ranging from 82 per cent in 1993 to 70 per cent in 2002, as calculated by the RR5 method described in the American Association for Public Opinion Research’s Standard Definitions: Final Dispositions of Case Codes and Outcome Rates for Surveys. Margins of error are at the 95 per cent confidence level, plus or minus one percentage point.^11,12

The GSS offers several advantages over even the most diligent compilation of findings from other, more disparate surveys.¹³

First, in as much as one of its primary purposes is to provide measures of social change, care is taken in the GSS to ask exactly the same questions year after year, with no changes in wording or mode of presentation, so that differences between surveys taken at different times cannot be attributed to such factors.

Second, the GSS is an open source. One can get immediate access to the data now accumulated for over 40,000 respondents, each of whom is asked some subset of a thousand or so different questions. The availability of this cumulative file makes it easy not only to plot trends in attitudes, but also to identify sectors of the public in which change is more or less rapid than the overall trend.

FINDINGS OF THE GENERAL SOCIAL SURVEY

Two questions directly relevant to the issue of “death with dignity” appear in the GSS. The first asks: “When a person has a disease that cannot be cured, do you think doctors should be allowed by law to end the patient’s life by some painless means if the patient and his family request it?”

The second reads: “Do you think a person has a right to end his or her own life if this person has an incurable disease?”

During the quarter century spanned by the data, 1977–2002, the GSS asked these questions in 16 years, compiling a cumulative data file consisting of records of some 22,700 interviews.

What do the survey data tell us?

Figure 1 shows the year by year percentages of respondents approving euthanasia and suicide, respectively. Here and throughout, the per cent approving is computed from the total of persons answering yes, no, or don’t know.

• Download figure
• Open in new tab
• Download powerpoint

Figure 1

 Per cent approving euthanasia and suicide, General Social Survey: 1977–2002.

As the figure indicates, Americans’ approval of euthanasia rose after 1978 to a peak in 1990–91 and then dropped somewhat, with readings for 1994–2002 resembling those obtained back in 1985–89. As of now, one might say tentatively, approval of euthanasia has levelled off at roughly two thirds of the population. But we cannot assume that situation will continue. Approval could undergo a downward trend or resume an upward trend, depending on history yet to happen.

Figure 1 also indicates a spectacular upward movement in suicide approval from 38% in 1977–78 which, apart from year to year fluctuations, seems to have peaked at 61% in 1994–98. A partial recovery in 2002 from the drastic drop registered in 2000 leaves the question of future trends wide open.

Until 1994, approval of suicide was rising much more rapidly than approval of euthanasia, thereby closing the gap between them. But the gap has remained at six to eight percentage points since then. Bear in mind that the responses to the two questions were being made by the same respondents. The convergence, therefore, suggests that more people were changing their minds about suicide than about euthanasia.

COHORT ANALYSIS

A useful follow up to this observation is to carry out cohort analysis, a strategy not widely known among lay consumers of survey findings and not well understood even by most analysts of survey data.

The main results of detailed cohort analyses are adequately conveyed by a simple classification of the population by birth years: before 1924, 1924–59, and after 1959. We refer to the three categories as the old, the middle, and the new cohorts. The new cohorts entered the adult population only after GSS began to ask the end of life questions in 1977. The old cohorts had very nearly vanished from the population by 2002. In 1977, 33% of the sample were members of the old cohorts, and 67% were included in the middle cohorts. By 2002, only 4% were in the old cohorts and nearly half (47%) were in the new cohorts, and the other half (49%) were in the middle cohorts.

Any change in the percentages approving for the old and middle cohorts must have been primarily due to individuals changing their minds. Effects of selective immigration and differential mortality were investigated, and we concluded that they cannot have been major factors shifting the approval rates.

Figures 2 and 3 support these observations:

• Download figure
• Open in new tab
• Download powerpoint

Figure 2

 Per cent approving suicide, for cohorts, General Social Survey: 1977–2002.

• Download figure
• Open in new tab
• Download powerpoint

Figure 3

 Per cent approving suicide, for cohorts, General Social Survey: 1977–2002.

• The approval rate for all respondents moves pretty much in parallel with that for the middle cohorts. The trends described for all respondents above hold for the middle cohorts with only slight changes.

• For both questions, the highest approval rates are observed for the new cohorts, the lowest for the old cohorts. Hence, as the new replace the old cohorts, the aggregate approval rate will tend to go upward for this reason alone, if for no other.

• The initial pronounced upward trend for each of the three groupings which is evident for suicide is greatly muted in the data for euthanasia.

Keep in mind that changes within each grouping can be alternatively described as showing what happens when people get older or as what happened in the course of historical time, 1977–2002, perhaps owing to changes in the climate of opinion. Is it plausible, however, that ageing, as such, accounts for the rapid increase in suicide approval for the old cohorts between the periods 1985–86 and 1996–98 and is primarily due to more—and more serious—meditation on the predicament of seniors sensing the proximity of life’s end? More plausible, perhaps, is the conjecture that ageing sensitises people to the increasingly frequent news from outside that suicide is a thinkable alternative to a natural death. The highly deviant data point for old cohorts in 2000–2002 (which is the average of their responses in the two successive surveys and is shown in figure 2 as 2001) defies any but an ad hoc explanation. But the survivors to these years are a tiny minority of those followed in the preceding period.

In any event, what is not a matter of conjecture is that change within cohort groupings is much more prominent for suicide than for euthanasia, and it accounts for the convergence of the two series shown in figure 1. Correlatively, we can say that cohort replacement is relatively more important for euthanasia than for suicide.

An illustrative calculation will make this observation concrete. Suppose that for each cohort grouping there was no change between 1977 and 2002; each has throughout the period the average over the whole period that is shown by the data. For suicide the averages for old, middle, and new cohorts respectively are 34, 53, and 62%; for euthanasia, 54, 65, and 72%. Holding these fixed and letting the relative numbers in the three groupings change, we can calculate the expected percentages approving, shown as the light lines in figure 1. If cohort replacement were the sole source of change, the light and heavy lines would, of course, coincide. This is close to being the case for euthanasia. For suicide, however, the actual trend is much more pronounced than the trend that would be produced by cohort succession alone.

A clear contrast of this sort between two rather closely related opinion questions is something of a rarity. A partial explanation is the observation (which itself is in need of an explanation that we cannot supply) that suicide approval is much more strongly related to educational attainment than is approval of euthanasia, and historically each new cohort has completed more years of schooling than its predecessors. For respondents with postgraduate schooling, suicide approval lags behind euthanasia approval by only a few percentage points, whereas euthanasia is much more frequently approved than is suicide among those who did not finish high school.

THE GSS FINDINGS IN CONTEXT

Having delineated some clear patterns in the GSS, we now can regard them as benchmarks for assessing the relevance of additional survey results. Other polls using the same or different questions as the GSS provide modest but convincing evidence of a longer term trend in approval of euthanasia. Like the GSS, these are all national probability sample polls of US adults. With the exception of Gallup and Harris polls during 1973 that used personal interviews, all were conducted by telephone. Three of the CBS News polls had sample sizes between 573 and 916 respondents, but samples for the rest ranged from about 1000 to about 1500. The same cautions pertaining to sampling error that were mentioned above with reference to the GSS should be applied to these surveys. It should also be noted that, unlike the GSS, the sample data from these polls are weighted, using parameters from the most recently available census data to compensate for known biases in survey derived estimates resulting from differences in non-response.

We see in figure 4 that the Gallup question on euthanasia is the same as the GSS question; indeed, GSS adopted Gallup’s wording on this as on many other topics in order to permit meaningful long run comparisons. Gallup used this wording in nine polls dating from 1947. Unfortunately, there is a hiatus between the 1950 and 1973 polls, so it is not possible to ascertain exactly when the rise from 1947–50 to 1973 actually got underway. The Gallup data record another substantial increase from 1973 to 1990. GSS data covering most of this period are reasonably consistent with the linear change implied by the Gallup data.

• Download figure
• Open in new tab
• Download powerpoint

Figure 4

 Per cent approving suicide, various polls, 1947–2003 (GSS as in figure 1).

An indication of the more or less random fluctuations to which all poll data are subject is given by the comparison of the two Gallup surveys taken less than three months apart in 1996 but differing by six percentage points. Strict comparisons between two polls are compromised when (as is usually true) they are not carried out at the same time. Each is subject to sampling variation, and both may pick up real but transitory changes. The graph helps to keep such phenomena in perspective.

There are other stray results that fit into this picture pretty well:

• An early (1936) Gallup question about “mercy deaths” showed 46% approval, a bit higher than the 1947 and 1950 figures with the question now in use. Given the difference in wording, it seems best to regard this simply as confirmation that approval of euthanasia was rather low in the distant past by comparison with present day readings. An NBC News figure of 63% approval in 1988 is similar to the nearest GSS and Gallup readings, as is the 63% likewise reported from a poll by the Kaiser Family Foundation, the Harvard School of Public Health, and the Boston Globe. The latter used the Gallup wording, whereas the NBC poll had a somewhat differently worded question.

• Three Harris poll questions provide a reading reasonably in agreement with Gallup and the GSS. A close look shows that Harris B (shown in figure 4) records decreases in approval between 1993 and 2001; GSS does not confirm them. Gallup reports somewhat comparable numbers between 1996 and 2001, but the approval ratings spike back upward in 2002–2003. So there is reason to keep open the question of whether the long run upward trend is coming close to a ceiling or even entering a period of declining approval.

• It was anticipated that questions specifically mentioning physician assisted suicide would get higher approval than the GSS question that posited only the right to suicide. If, however, the five CBS News readings and the eight from Gallup (shown in figure 5) were regarded as a single series, it would register approximately the same general trend as the GSS, ignoring erratic variations.

• The three Harris poll questions asked between 1994 and 2001 about reactions to a 1994 Oregon law allowing physician assisted suicide in the case of terminally ill patients (Harris D in figure 5) could be read as supporting either of the GSS questions for the same period. The spread between the two in GSS is unmistakable, however, and the Harris questions introduce another issue. So all we can say is that there is no major inconsistency between the responses to questions that specify a legislative issue and those that only ask about personal opinions.

• Download figure
• Open in new tab
• Download powerpoint

Figure 5

 Per cent approving suicide, various polls, 1947–2003 (GSS as in figure 1).

The long term Gallup series fits into a continuous trend with the GSS, covering 55 years and showing euthanasia approval in 1947 just where suicide approval was in 1977, and with suicide now catching up. With little strain the Harris data as well as the GSS data confirm this trend. The historical evidence would lead to the expectation that the two questions will get about the same level of approval in the not so distant future. Like all trend extrapolations, however, this one is to be entertained only as a possibility, not used to exclude other possibilities we cannot now foresee. Looking backward, it seems likely that with more data, the Gallup trend, shown as a straight line from 1950 to 1973, would have begun some time in the 1960s, prefiguring the rise in suicide approval shown in figure 5.

CONCLUSIONS

What conclusions can physicians draw about trends in American public attitudes toward physician assisted suicide (PAS) based upon the data presented here? Neither of the two relevant GSS questions by itself addresses the issue directly. One asks about suicide but does not mention physician involvement in it. The other does mention doctors but pertains to euthanasia rather than suicide.

One might expect the latter question to be more germane to the issue of PAS because of the inclusion of doctors, but it is GSS suicide that seems to work best as a proxy to a question on PAS, in as much as the CBS News and Gallup data in figure 5 that do stipulate physician assistance seem to follow the same trend.

This evidence is not overwhelming, but it is nevertheless worth some emphasis. Physicians might dismiss the GSS findings as not relevant to their primary interest if PAS is more acceptable to them than outright euthanasia. By establishing the GSS suicide question as a rough proxy for a measurement of attitudes toward PAS, we show them how to get a decent estimate (in the absence of other evidence) of how public sentiment in favour of PAS has grown over a quarter century. Moreover, since the approval levels indicated by the GSS euthanasia and suicide questions are converging, the difference between them may before long turn out to be minor. Of course, this conjecture is not guaranteed. Differences in wording of the relevant questions that were not salient in the past could become so in the future.

In as much as we can expect the GSS to continue reporting in the future, the parties interested in PAS can be reassured that it is reflecting the trend and fluctuations of public opinion on this topic reasonably well. Poll watchers should start looking to the GSS and not merely to the Gallup releases and such. Perhaps the strongest suggestion we can offer is that the debates about end of life decisions should accept as a first premise that public opinion on these matters is still fluid, as is suggested indeed by the dramatic changes registered by the GSS, Gallup Poll, and other survey questions on euthanasia and suicide over the last half century reported in our graphs.