¹ Cooper, M. H. and Culyer, A. J., The Price of Blood, Institute of Economic Affairs, London, 1968.Google Scholar (At various points in this article, reference is made to results which are not reported in detail; all such data are available from the author.)

² Titmuss, R. H., The Gift Relationship, Allen and Unwin, London, 1970.Google Scholar

³ See for example comments on Titmuss's work made by the eminent American economists, Kenneth Arrow and Robert Solow. Titmuss was subsequently defended against Arrow by Singer. See also, in particular, the debate between Culyer, and Reddin, Michael in Johnson, D. (ed.), Blood Policy: Issues and Alternatives, American Enterprise Institute for Public Policy Research, Washington, D.C., 1976Google Scholar; Arrow, K., ‘Gifts and Exchanges’, Philosophy and Public Affairs, 1 (1972), 343–62Google Scholar; Solow, R., ‘Blood and Thunder’, Yale Law Journal, 80 (1971), 170et seq.CrossRefGoogle Scholar; and Singer, P., ‘Altruism and Commerce: A Defence of Titmuss against Arrow’, Philosophy and Public Affairs, 2 (1973), 314–20.Google Scholar There is a semantic issue which should be clarified: many American sources describe a ‘voluntary’ blood donor as anyone not selling his blood for cash payment; thus, participants in insurance and blood-banking programmes are deemed to be voluntary donors; in the study on which this article is based, a voluntary donor is contrasted to a market constrained donor, which would lump all the above categories under the second heading.

⁴ Havighurst advocates extension of the right to sue for receipt of ‘bad blood’ as a way of improving the American system – C. C. Havighurst, ‘Legal Responses to the Problem of Poor-Quality Blood’, in Johnson, op. cit. pp. 21–38.

⁵ Sapolsky, H. M. and Finkelstein, S. N., ‘Blood Policy Revisited: A New Look at The Gift Relationship’, Public Interest, 46 (1977), 15–27.Google Scholar

⁶ The problem is compounded by the fact that applied psychologists have given most attention to blood donation in the United States, while, in Britain, the major area of concern has been social policy and economics. One major American review of sixty-two separate studies of blood donation concluded that the overwhelming evidence indicated four major motivations for blood donation: humanitarianism or altruism; personal or family credit; peer or individual pressure; and replacement. However, of these four factors, two – credit and mandatory replacement – are particular to the United States, and would not be relevant in a different context such as Britain or Canada – Oswalt, R. M., ‘A Review of Blood Donor Motivation and Procurement’, Transfusion, 17 (1977), 124.CrossRefGoogle Scholar

⁷ See for example Alchian, A., The Economics of Charity, Institute of Economic Affairs, London, 1973Google Scholar; Boulding, K., The Economy of Love and Fear, Wadsworth, Belmont, California, 1973Google Scholar; Culyer, A. J., The Economics of Social Policy, Martin Robertson, London, 1973Google Scholar; and Phelps, E. S. (ed.), Altruism, Morality and Economic Theory, Sage Publications, New York, 1975.Google Scholar It was in 1951 that Arrow first identified the problems of aggregating individual preferences – Arrow, K., Social Choice and Individual Values, John Wiley, New York, 1951.Google Scholar

⁸ Macbeath, A., Can Social Policies be Rationally Tested?, Hobhouse, L. T. Memorial Trust Lecture no. 27, Oxford University Press, London, 1957, p. 1.Google Scholar

⁹ Cited in Titmuss, R. M., Social Policy: An Introduction, Allen and Unwin, London, 1974, p. 28.Google Scholar

¹⁰ Boulding, K., ‘The Boundaries of Social Policy’, Social Work, 12 (1967), 3–11.Google Scholar

¹¹ Titmuss, , The Gift Relationship, pp. 273–4.Google Scholar Titmuss assumed rather than ‘proved’ that people would act in the desired manner – that the requisite number of voluntary blood donors would, in fact, be forthcoming.

¹² Pinker, R., The Idea of Welfare, Heinemann, London, 1979, p. 10.Google Scholar Titmuss implicitly acknowledged the lack of utility in postulating a dichotomy; he described blood donation as ‘relatively altruistic’, because ‘ [no] donor type can…be said to be characterized by complete, disinterested, spontaneous altruism’ – Titmuss, , The Gift Relationship, p. 101.Google Scholar

¹³ Collard, D., Altruism and Economy: A Study in Non-Selfish Economics, Martin Robertson, London, 1978.Google Scholar

¹⁴ Ibid. p. ix.

¹⁵ Ibid. p. 5.

¹⁶ Most empirical studies of altruism cannot go beyond the cash transfer. Yet, even traditional charitable giving is described as exacting a measure of social control, the quid for the quo; failure by the recipient to observe certain societal norms will be followed by withdrawal of the charitable ‘gift’.

¹⁷ Titmuss identified eleven attributes of voluntary blood donorship which mark it as an act for the community rather than the self – Titmuss, , The Gift Relationship, pp. 84–5.Google Scholar These were then summarized into three key criteria, the absence of rewards, the absence of sanctions for the non-performance of the act, and anonymity between giver and recipient – ibid. p. 101. The first two differentiate the gift from the bilateral exchange, but it is really the anonymity – ‘to be free to choose to give to unnamed others’ – which Titmuss believed to be the essential condition – ibid. p. 272.

¹⁸ Collard, op. cit. p. 5.

¹⁹ Budgetary considerations dictated a total sample of approximately 4,000.

²⁰ Alexander, G. and twelve others, Voluntary Blood Donorsliip: A Study of Active and lapsed Donors, Faculty of Social Work, University of Toronto, 1978Google Scholar, unpublished.

²¹ The addresses of blood donors are not updated on the Red Cross master list, and many individuals had become unreachable since the date of their last donation. The student project (Alexander et al., op. cit.) had found that, for each questionnaire returned by the post office as ‘undelivcrable’, there were 2.6 individuals who could not be contacted (as distinct from refusals to participate). Applying the same ratio to the 466 forms returned by the post office in the present study, it was estimated that 2,748 questionnaires were actually received by the intended donor. This produced a net response rate of 64.9 per cent.

²² Certain limited summary data are maintained by the Red Cross concerning the total population of all donors for the relevant time period (n=159,891). If this information is compared to the corresponding results for the sample of respondents, there appear to be no particular biases in terms of the available demographic variables of age, sex or blood type. However, the donor histories do indicate the sample to be over-represented with recent or active donors. The total number of donations among the respondents is 11.3 (s = 12.8) compared to 6.9 (s = 5.2) for all donors, and the mean number of months since the last donation is considerably lower (9.8 compared to 16.6). Thus, the findings reported in the study refer to a large number of blood donors, a sample with a relatively high response rate, but a group disproportionately weighted in favour of recent and frequent blood donors.

²³ Collard notes that ‘Altruism is concerned with tastes or preferences and not yet with action…Altruistic feelings may exist but be too weak to lead to action’ – Collard, op. cit. p. 8.

²⁴ The anonymity of the questionnaire used in the study may limit conscious misreporting by respondents, though there is, of course, an opposite literature which suggests that validity is enhanced through the use of personal interviews. See Selltiz, C., Jahoda, M., Deutsch, M. and Cook, S., Research Methods in Social Relations, revised edition, Holt, Rinehart and Winston, New York, 1959, ch. 7.Google Scholar

²⁵ A table showing the detailed demographic information for the respondents, along with the corresponding data, where available, on the population of all blood donors as drawn from the Red Cross master tape, is available on request from the author. A comparison of the demographic data for the present sample to the findings of previous research is not useful because the sampling designs of the various studies differ widely. Many samples, for example, are drawn from ‘captive’ groups such as college students, in which the educational levels would necessarily be relatively high. Oswalt does present, in tabular form, summary demographic data on ten American and Australian studies–Oswalt, op. cit.

²⁶ Statistics Canada, 1976 Census of Canada, Ottawa, 1978Google Scholar – Report 92–826, ‘Population: Demographic Characteristics: School Attendance and Level of Schooling’, p. 27.Google Scholar

²⁷ Statistics Canada, Income Distribution by Size in Canada, 1977, Ottawa, 1979Google Scholar, Report 13–207, Table 5. Higher socio-economic groups tend to reply to postal questionnaires with greater frequency than do the less educated and less affluent. As the population parameters for all blood donors are not available for the socio-economic variables (specifically education and income), the extent of this non-response bias – if any – cannot be measured. What can be observed, however, is that respondents in the present study were far better educated than the community at large, though the distribution of family incomes was virtually identical to that for metropolitan Toronto in 1977.

²⁸ In the analysis of fifteen individual motivators, there is a potential problem of construct validity. Each item may have limited validity, in that the particular terminology used may evoke only part of the total concept intended by the researcher, or, alternatively, a theoretical concept may consist of many parts requiring many questions. A common solution is to develop more meaningful measures by combining specific items into aggregate indices, with a resulting increase in the overall validity of responses. At the same time, there is considerable subjectivity inherent in such aggregation, and, as a result, it is normal also to examine the correlation matrix to observe the extent to which the responses to the various questionnaire items are, in fact, statistically associated with one another. Responses to the Likert-type items for the fifteen motivations in Table 1 were recoded dichotomously (as either important or not important), and Pearson product moment correlation coefficients were produced. No bivariate correlation yielded a coefficient as high as 0.5, and all but four of the values were below 0.3. Hence, one may conclude that, at least for the present sample, responses to the questionnaire items are statistically independent of one another, using a dichotomous distinction. The grouping of items into categories may be useful for purposes of exposition, but will not increase the overall validity of the responses. The phi coefficient is often used to indicate the degree of relationship between two dichotomous variables. However, it ‘actually amounts to a simplification of the Pearson correlation when…both variables are dichotomous’. For computational ease, Pearson product moment correlations were used throughout the present study. See Harshberger, T. R., Introductory Statistics: A Decision Map, Macmillan, New York, 1971, pp. 396Google Scholaret seq.

²⁹ Reddin has indicated the importance of this factor in Britain as well: ‘Simple convenience- how easily people can get to places to give blood – is a major determinant of much blood donation in the United Kingdom’ – M. Reddin, ‘Commentary’, in Johnson, op. cit. p. 63.

³⁰ There may be an associated need for peer support at the point of first undertaking a process which is essentially unknown and anxiety-inducing in the mind of the potential donor. It may also be noted that many donors begin their blood donation histories while they are at school or university, and continue after graduation.

³¹ As was indicated above, such categorization must be based on logical grounds, and cannot be supported on a statistically rigorous basis, as all bivariate correlation coefficients yielded values below 0.5.

³² Arrow, op. cit. p. 348.

³³ As with Table 1, there is a potential problem of construct validity in dealing with seventeen different motivators. All responses to the Likert-type questions were again recoded dichotomously as important or not important, and a matrix of Pearson product moment correlation coefficients was produced. Though the correlation levels here were somewhat higher than those in Table 1, only four coefficients were above 0.5, and, of these, three were between 0.5 and 0.55. The remaining coefficient – between ‘rude staff’ and ‘incompetent staff’ at the clinic – was above 0.8, suggesting probable multicollinearity in this case. The two questionnaire items could be combined to produce a more general measure describing ‘problems with staff at the clinic’, or one category could be dropped entirely, and the other used to represent both. Creation of an index would hinder the comparison across factors which is the focus of Table 2, and so the relatively high inter-correlation should simply be noted in interpreting the results. Because of problems of statistical bias and multicollinearity, an index is later developed for use in the regression analysis.

³⁴ ‘Apathy’ is, of course, a concept difficult to explicate. It is used in this study primarily for its outcome measure, that is, withdrawal. It may include notions of anger, hostility and rejection.

³⁵ Other definitions such as six months or eighteen months could also be used, and were tested. Although the results differ in detail, the general thrust of the findings is the same as it is for the results reported here. In a study of blood donors in Wales, Oborne's report drew what may be an important theoretical distinction between what it described as ‘real ex-donors’ – those who have made a positive decision to stop donating blood – and ‘pseudo ex-donors’ – those ‘who may have drifted out of donation but fully intended to give blood again at some time in the future’. The distinction was tested by directly asking past donors ‘if they considered that they would donate again’ – Oborne, D. J.Aspects of Blood Donor Motivation, report financed by the Social Science Research Council, University College of Swansea, p. 5.Google Scholar The difficulty with this approach, for the present purposes, is that many who are labelled ‘pseudo ex-donors’ on the basis of their self-reported expectations will in fact become ‘real ex-donors’ the problems of predicting future behaviour on the basis of present attitude remains unanswered.

³⁶ A 10 level of significance is often reported as well, though acceptance of this standard is left to the reader. It is arguable whether 10 significance should be accepted with a sample of the present size; however, given the scarcity of clear theory (and, hence, strong expectations) it was felt appropriate to at least mention results at the 10 level.

³⁷ The total number of donations for the lapsed and active groups was, on average, 7.9 (s = 9.3) and 14.2 (s = 14.0) respectively (p ≺.001). However, there was no significant difference between the average ages at which these donors first donated blood, or in the distribution of blood types. Eight categories were used for blood type in the first instance. An attempt was also made to distinguish dichotomously between ‘rare’ and ‘common’ blood types, in the expectation that the former group would be less likely to lapse. The Red Cross has indicated that there is no absolute definition of rare and common blood, but, rather, specification is in market terms, based on demand and available supply. The two most common blood types, O POS and A POS, were combined into one category containing 66 per cent of the sample, with all others also combined; the result was a 2 X 2 table, which, again, showed no significant difference between lapsed and active donors.

³⁸ At the 10 level, ‘support for the work of the Red Cross’ and ‘crisis or emergency’ were found to be statistically significant. The former could be interpreted as ‘internal’, similarly to ‘sense of duty’, while the latter could be ‘external’, associated with ‘need for rare blood’. The data support these expectations, as relatively more active donors found ‘support for the Red Cross’ to be initially important or very important, while more lapsed donors so identified ‘crisis or emergency’. It is possible that the important distinction was dichotomous, in other words, that a factor was either important or unimportant. The motivations were recoded dichotomously and cross-tabulated by donor status. All five positive motivations in Table 3 continued as statistically significant, and, in addition, ‘company of a friend’ and ‘crisis or emergency’ now became significant at the 05 level, with relatively more lapsed donors reporting each of these factors as important in the initial decision. Both of them would be described as external to the respondent; the former is undoubtedly associated with ‘encouragement’ and ‘support of others’, as described in the text, while, as indicated, the interpretation of the latter would be similar to that of ‘need for rare blood’.

³⁹ As in footnote 38, the negative motivations were recoded dichotomously as important or not important. The identical factors were found to be statistically significant, except that ‘discomfort during donation’ and ‘unpleasant after-affects’ dropped to a 10 level of statistical significance.

⁴⁰ It is also possible that an unpleasant clinic setting was associated with a variety of the other more specific deterrents, none of which by itself was strong enough significantly to influence subsequent behaviour; only the most general statement around which a variety of dissatisfactions could converge became strong enough to yield a significant coefficient. At the same time, it should be recalled that none of the bivariate correlation coefficients between an unpleasant clinic setting and any other variable was as large as 05.

⁴¹ See Collard, op. cit. p. 8.

⁴² Boulding, , ‘The Boundaries of Social Policy’, p. 8.Google Scholar

⁴³ Titmuss, , The Gift Relationship, p. 101.Google Scholar

⁴⁴ For a discussion of the statistical limitations of using a dichotomous dependent variable, see Ashenfelter, O., ‘Some Statistical Difficulties in Using Dummy Dependent Variables’, in Bowen, W. G. and Finegan, T. A., The Economics of Labor Force Participation, Princeton University Press, Princeton, 1969, pp. 644Google Scholaret seq. An identical methodology to that used in this study may be found in Bowen and Finegan, op. cit. and also in Lightman, E., ‘Attrition of Applicants to the Canadian Armed Forces’, Armed Forces and Society, 3:3 (1977).CrossRefGoogle Scholar

⁴⁵ Although the independent variables are assumed to be distributed independently of one another, there remains a potential problem of multicollinearity with a large number of independent variables. Examination of the total correlation matrix produces only one bivariate coefficient – that between ‘rude staff’ and ‘incompetent staff’ at the clinic -with a value above 0.55. High multicollinearity raises the standard errors in a regression, thereby reducing the t-values and lowering the probability of a coefficient being significantly different from 0. One ‘solution’ is to drop one variable and rerun the regression. This, however, produces a bias, in that part of the effect properly attributable to the omitted variable will, in fact, be picked up by the included variable. In the present case, the positively signed correlation between the two variables in question implies upward bias if one of them is dropped. An alternate approach to multicollinearity is to combine the variables to form an index, and this was done: each respondent was assigned a score equal to 0 if neither of the two variables was reported as important or very important, equal to 1 if one of the two variables was important or very important, and equal to 2 if both variables were identified as important or very important. The result was the creation for each respondent of a continuously distributed integer score, with a sample mean of 0.46 and a standard deviation equal to 0.8. The regression was rerun with this continuous variable replacing the two dichotomous variables of ‘rude staff’ and ‘incompetent staff’ at the clinic. However, this change had no effect on the regression or on the statistical significance of the other variables, and, as a result, it is not reported here.

⁴⁶ The low multiple correlation suggests that relatively little of the variance in the dependent variable is explained by the independent variables. Regression analysis using observations across individuals is not expected to produce a high r (or R²); what is relevant is that certain of the independent variables are systematically associated with changes in the dependent variables. See Bowen and Finegan, op. cit.

⁴⁷ As it has been shown that under no circumstances are any of the demographic variables statistically significant, no restriction need be placed on them in terms of prediction.

⁴⁸ In principle, the probability of being an active donor could produce a value greater than 1.0 by this method, an outcome resulting from the linearity assumption. In practice, values greater than 1.0 would be read as equal to 1.0. The method described here refers to a point estimate; interval estimation could also be used.