Abstract
Is a larger population of people living good lives a better population, all else equal? This question is central to population issues in social welfare, ethics, and policy. Many answers in the philosophical literature argue that if a policy choice results in the birth of additional people living good lives, these extra lives are irrelevant to any evaluation of the policy. This paper applies the questionnaire-experimental method of empirical studies of social choice to investigate participants’ policy choices and social orderings with respect to population size and average well-being. In general, heterogeneous responses depended on the quantitative and qualitative properties of the question. In particular, an experimentally manipulated increase in population size caused an option to be more likely to be selected, on average. Overall, responses suggest that population size is not neutral to social welfare. Many participants, although not all, reported that a larger population of people living good lives could be strictly preferable, at small or no costs to average well-being.
This is a preview of subscription content, log in via an institution to check access.
Similar content being viewed by others
Notes
This is because alternative future climate paths are likely to cause large, long-lasting changes in the size and composition of future populations.
See Pressman (2015), for a recent example. An implicit Average Utilitarianism may also be implied by economists’ typical focus on averages such as GDP per capita or rates such as literacy rates or infant mortality rates.
Results in Sect. 3 verify that almost all participants choose according to these two assumptions, in a case when the questionnaire was structured to permit them not to. Very many major theories of social welfare in economics accept transitivity and same-number Pareto for risk-free cases.
“At a high quality of life” is important here. The Negative Expansion Principle, which Blackorby et al. describe as essential, holds that “a desirable property of a population principle is that the addition of a person to a utility-unaffected population should be ranked as bad if the utility level of the added person is negative” (p. 161). Other theories rank addition of positive-utility lives as bad if they extra lives are not good enough, meaning below a critical level.
Studies 1–4 received human subjects clearance from the RICE Institute, Inc. Institutional Review Board, which is registered with the U.S. National Institutes of Health.
Amazon Mechanical Turk provides this information anonymously by providing each task submitter with a unique “Worker ID.”
This is useful because we are in part responding to claims about intuition in the population ethics literature and because it is a standard approach in empirical social choice: “most studies are careful in pointing out that the respondents involved were not acquainted with the theory before the experiment started” (Gaertner and Schokkaert 2011, p. 23).
Additionally, the final survey question asked which country the respondent was in; all respondents selected “United States” from a list of countries.
These guarantees were intended to promote respondents making a ceteris paribus decision that takes seriously the stipulated levels of wellbeing (see supplementary appendix section S1.1). However, results throughout the paper show that responses vary according to the quantitative features of the survey question, its qualitative properties and framing, and properties of the respondents, so future studies could find that the text of this guarantee also impacts the result.
This is true for two reasons. One concerns internal validity: because participants were randomized to levels of \(n^h\), and because participants assigned to higher levels of \(n^h\) were more likely to choose Policy B, many of the participants assigned to low \(n^h\) would, assuming randomization succeeded, choose Policy B at higher \(n^h\). The other reason requires an assumption of external validity: because choosing Policy B is shown to be monotonically increasing in \(n^h\), the results suggest an out-of-sample prediction where eve more respondents may choose Policy B at some higher \(n^h\).
Study 1 posed the question written in Sect. 2.2.2, but in slightly less detail. Instead of the three introductory paragraphs, the prompt before the presentation of the two policies read only:
You are an economic advisor for a middle-income country which has been benefiting from fast economic growth and development. It is your job to choose one of two policies to recommend, either Policy A or Policy B. Whichever policy you choose, your country is stable and secure enough that there will be no extreme poverty or violence; people will have basic freedom to choose how to live their lives; there is enough land for everyone to live comfortably; and everyone will have access to food, healthcare, and other basic needs. The only difference between the two policies is in the size and economic well-being of the population in 50 years.
This text was modified in subsequent studies (to the text in Sect. 2.2.2) for additional clarity.
ANOVA similarly rejects that population size does not influence policy choice with \(p = 0.036\). A logit model containing only indicators for random treatment assignment correctly predicts 62% of participants’ policy choices.
For example, none of Studies 1–3 consider bad lives.
Study 5 was completed while the author was a visitor at Princeton University’s Woodrow Wilson School in the Spring of 2016. For human subjects approval, the RICE Institute, Inc. IRB approved Study 5 and this approval was adopted for Princeton University in a letter from the Princeton IRB.
In class discussions 3 weeks later, these two participants claimed that they had intended to submit a social ordering and simply made an error.
Hurley et al. (2011) specifically emphasize misunderstanding as a mechanism: “The pattern of judgments is consistent with the hypothesis that subjects do not fully understand the relationship between the conceptual meaning of the principles (as described verbally) and their implied quantitative divisions.” This seems less relevant in the case of Study 5, as the text does not describe complex distributions, but rather merely levels of quality of life. Moreover, descriptions are accompanied by a graph; Hurley et al report that “judgments based on both quantitative and verbal information match more closely those made with only quantitative information.”
Study 5b may have shown more perverse responses than Study 5a because of a discipline encouraged by binary choice in 5a, or because 5a was administered on paper in a classroom setting, that may have resembled a quiz to some students. However we do not make any statistical comparisons between Study 5a and 5b because the 5a sample is so small that its standard errors are large (over 15 percentage points) and therefore statistical inference is not possible in the ordinary sense.
Total utilitarian or CLGU respondents may strictly prefer a larger population living bad lives if they believe lives at this level are nevertheless worth living or above a critical level, although participants’ written responses indicated that they generally interpreted the “bad” lives to not be worth living; if any respondents thought this way, then excluding their responses with the apparently perverse responses are excluded would cause support for total-type utilitarianism to be understated.
As a consistency check, note that 100% of non-perverse respondents replied that \(B \succ C\): a larger excellent population is strictly preferred to a smaller good population.
Indeed, responses in these categories would also be related to further generalized social welfare functions which recognize larger populations with excellent lives as better, but which also include number-sensitive or welfare-sensitive critical levels (Ng 1989), such as the highly flexible Expected Prioritarian Equally Distributed Equivalent social welfare function of Fleurbaey and Zuber (2015). This study was not designed to separate social preferences with a fixed critical level from social preferences with a context-specific critical level, or to otherwise make precise identifications of functional form; as an initial empirical study of social population preferences, this research concentrated on logically prior questions, such as whether population size enters social ordering decisions at all, and whether population size and well-being interact at all (as would be implied by a positive critical level). Future research should investigate this question of functional form.
An important caveat is that no finite set of questions can rule out that a participant has a vague set of critical levels that is between levels of quality of life that are tested. For example, a participant could have a vague set of critical levels spanning from just above bad to just below good, and still report \(D \succ C\) and \(E \succ F\). In this sense, Study 5 can only identify “Strict or Vague CLGU,” which here is called “Strict.”
This is analogous to an interaction in a statistical regression model, where the coefficient on one variable depends on the level of another variable.
Relative to the economic frame, the fertility frame moves about 13 percentage points of participants from reporting that more good lives is better to reporting that more good lives is neutral (a difference that is not statistically significant); there is no difference in the fraction reporting that more good lives is worse.
References
Amiel Y, Cowell FA (1992) Measurement of income inequality: experimental test by questionnaire. J Public Econ 47(1):3–26
Amiel Y, Michele B, Frank C, Valentino D (2013) Do we value mobility? Soc Choice Welf 44:1–25
Appiah KA (2008) Experiments in ethics. Harvard, Cambridge
Blackorby C, Bossert W, Donaldson D (1995) Intertemporal population ethics: critical-level utilitarian principles. Econometrica 63:1303–1320
Blackorby C, Bossert W, Donaldson D (2005) Population issues in social choice theory, welfare economics, and ethics. Cambridge University Press, Cambridge
Broome J (2004) Weighing lives. Oxford University Press, Oxford
Broome J (2012) Climate matters: ethics in a warming world. Norton, New York
Buhrmester M, Kwang T, Gosling SD (2011) Amazon’s mechanical turk: a new source of inexpensive, yet high-quality data? Perspect Psychol Sci 6(1):3–5
Connelly MJ (2009) Fatal misconception: the struggle to control world population. Harvard University Press, Massachusetts
Fleurbaey M, Zuber S (2015) Discounting, beyond utilitarianism. Econ Open Access Open Assess E J 9:1–52
Gaertner W, Schokkaert E (2011) Empirical social choice: questionnaire-experimental studies on distributive justice. Cambridge University Press, Cambridge
Greene J (2014) Moral tribes: emotion, reason and the gap between us and them. Atlantic Books, London
Hurley J, Buckley NJ, Cuff K, Giacomini M, Cameron D (2011) Judgments regarding the fair division of goods: the impact of verbal versus quantitative descriptions of alternative divisions. Soc Choice Welf 37(2):341–372
McMahan J (1981) Problems of population choice. Ethics 92(1):96–127
Narveson J (1973) Moral problems of population. Monist 57:62–86
Ng Y-K (1989) What should we do about future generations? Econ Philos 5(02):235–253
Paolacci G, Chandler J, Ipeirotis PG (2010) Running experiments on Amazon mechanical turk. Judgm Decis Mak 5(5):411–419
Parfit D (1984) Reasons and persons. Oxford University Press, Oxford
Pressman M (2015) A defence of Average Utilitarianism. Utilitas 27(04):389–424
Tarroux B (2015) Comparing two-dimensional distributions: a questionnaire-experimental approach. Soc Choice Welf 44(1):87–108
Yaari ME, Bar-Hillel M (1984) On dividing justly. Soc Choice Welf 1(1):1–24
Author information
Authors and Affiliations
Corresponding author
Ethics declarations
Conflict of interest
The author declares that there are no relevant potential conflicts of interest with this research. The data collection in Studies 1–4 was funded by r.i.c.e. (RICE Institute, Inc.), a non-profit 501(c)(3) research organization, online at http://www.riceinstitute.org. Study 5 was logistically supported by Princeton University’s Survey Research Center.
Electronic supplementary material
Below is the link to the electronic supplementary material.
Rights and permissions
About this article
Cite this article
Spears, D. Making people happy or making happy people? Questionnaire-experimental studies of population ethics and policy. Soc Choice Welf 49, 145–169 (2017). https://doi.org/10.1007/s00355-017-1055-7
Received:
Accepted:
Published:
Issue Date:
DOI: https://doi.org/10.1007/s00355-017-1055-7