CHAPTER 7
RANGE VOTING--MY DISSENT
The Following two items represent my response to range voting (RV) when I first learned about it in the Summer of 2007.
September 24, 2007
THE CASE FOR AV NOW
There are a number of attributes which a good voting system should have--there is no such thing as a perfect one as the economist Kennith Arrow showed in his seminal work which led to his Nobel prize. Among these I would include the following: 1) That the election outcome effectively reflects voters' preferences, 2) That it is transparent to voters in terms of understanding the results, and 3) That it is easy to administer. From this perspective I am not convinced that range voting (RV) is preferable to approval voting (AV).
Range voting looks a lot like utility theory in microeconomics where we use the concept of utility to explore the basic logic of consumer behavior in the market place. But there is one major problem with
this: We have no way of comparing interpersonal utility. This basically limits welfare comparisons to the idea of the Pareto optimum where we in essence use the opportunity cost of money to make comparisons between alternative situations and their social desirability.
We don't use utility in the benthemite sense of the greatest good for the greatest number because we have no way short of subjective evaluation of making such a comparison. But we can compare the price of things as opposed to the value of things to arrive at some economic conclusions or judgments. That is, economics has the concept of opportunity costs which can measure in terms of dollar values what people are willing and able to pay for something and what they would need to be payed to compensate them for an economic or, if you will, a welfare (utility) loss.
When we do use a notion of utility to analyze the logic of consumer behavior concretely we do so in the sense of indifference curves which reflect a given level of welfare or utility. That is, we do so in an ordinal sense. On this score, as an economist it does not surprise me that in all the literature I have seen on voting systems, the analysis is done in the ordinal sense of comparing preferences, not intensities of utility. Surely the transparency issue illustrates why this is so.
The Transparency Issue
Under range voting, as I understand it from the web site, one calculates an average score that voters assign to each candidate to arrive at the winner not unlike in some sporting events like ice skating. From what I know, often the judges in such events do not come up with identical scores for objectively the same performance.
This surely points to a problem here when one is calculating averages from the scorings of hundreds of thousands or even millions of voters.
How can one compare my 90 score for a candidate with John Doe's 90 score for the same candidate?
Perhaps range voting is nicely tractable, or easily handled, in a mathematical sense, but in a real world sense it is surely devoid of any precision. What for instance does it tell voters that a winning candidate has a 90 score as opposed to an 89 score or a 75 score as opposed to a 65 score for the runner up?
If in a plurality election with three candidates the winner gets 40% of the votes as opposed to 30% for each of the opponents, this surely tells us something concrete about the vote popularity or support level the candidates have that is understandable, or transparent, to most people. On this score, the discussion of the "nursery effect" in comparing RV to AV is surely questionable.
The Nursery Effect
It astonishes me that it can be asserted that RV would be any better for third parties than AV. The advantage of AV for third parties is that it totally, and I repeat totally, gets rid of the wasted vote syndrome. If a voter's first choice is not the lesser of evils, but is a candidate the voter truly approves of, then it always pays for a rational voter to give that candidate an approval vote. And in the election outcome one can meaningfully speak of a given candidate having an x% approval rate.
But in the case of RV, there is no way that I can see where the score, scale or index used tells us anything about support level as it is normally, and sensibly, understood. On this score, I cannot see how one can argue that RV would have any advantage over AV for third parties. In the case of RV there is no way that one can speak of an "over-20%-of-the-vote popularity" or an "up to 10-20% support" level from RV data. The scores used simply do not reveal such things and cannot.
The Advantage of the Ordinal Approach to Voting Analysis
Given the problems here, I don't see how one can reject what I would label as the ordinal approach to voting analysis. This surely was the basis on which both Condorcet and Borda raised legitimate questions about the voting system used by the French Academy to elect new members. And it surely is the basis on which one can concretely point to the considerable disadvantage of plurality voting.
It is my contention that the idea of the Condorcet candidate, where one would exist, is a reasonable equivalent in voting theory to the idea of a Pareto optimum in economics. That is, it is a reasonable standard by which to judge the degree to which a voting system accurately reflects voters' preferences in an ordinal sense.
On this score, I think that a strong case can be made for AV in two senses. First, in the sense that it does appear to have a tendency to choose the Condorcet candidate. And, second, when it does not, does come up with what I would regard as the most sensible second best criterion in reflecting voters preferences--the candidate which most voters would approve of.
This is clearly a value judgment, but at some point in judging voting systems one has to make a judgment as to reasonable starting criteria which cannot just emerge from some mathematical or computer simulations.
Sincere, Strategic and Manipulative Voting
In microeconomic theory economists analyze choice in terms of comparing preferences to constraints under the assumption that the interaction of the two is limited to a process of optimization alone.
While such an assumption may be useful for certain analytical purposes, it surely is not very reasonable in terms of real world behavior. In the real world it is hard to see how preferences can be determined separately from the nature of the constraints and tradeoffs that do exist.
Similarly it seems to me that the distinction between sincere and strategic voting while useful for certain analytical purposes is not a meaningful distinction when looking at voting systems and the role of preferences in election processes. Clearly an accurate assessment of voting possibilities in terms of realistic choices must inform voters'
choices in terms of the preferences they reveal in the voting process.
The great advantage that AV has is that it encourages and allows voters to vote both sincerely and strategically at the same time to effectively reveal their preferences without manipulating the expression of the preferences of others to affect an election's outcome. The same is certainly not true of IRV where a candidate's supporters can certainly manipulate the elimination process by their ranking of other candidates to impact that process in their candidate's favor.
Surely an important distinction in judging voting systems is that between sincere and strategic voting together on one hand to reveal a voter's preference and manipulative voting to obscure the expression of other voters' preferences on the other hand. On this score, in addition to the arbitrariness that Lord Alexander pointed to in his dissent to the Jenkins report concerning IRV and the chaotic and bad aspect of the single transferable vote (on which IRV is based) that Sir Michael Dummett referred to in a recent Social Choice and Welfare article, AV is certainly much superior to IRV.
Can One Really Say That RV is Superior to AV?
There is a widely used model in trade theory in economics which purports to prove the factor price equalization theorem. The problem with the model is that it in facts assumes what it claims to prove.
My impression is that in utility simulations analysis you have a similar situation. Given the problem of interpersonal utility comparisons, I fail to see how one can use ranking scores to say that something like RV is superior to AV. Here it would seem to me one really is mixing apples with oranges without any price to weight their comparability. One may be able to analyze the scores that arise in RV in a cardinal sense of assigning specific numerical numbers, but I do not see how such numbers can meaningfully be assigned a specific meaning whereas the ordinal results, if you will, of AV do give specific concrete percentages in terms of votes that are ordinal in nature and can clearly be consolidated into a number that expresses a
percentage of approval for specific candidates.
Given this and given the obvious difficulties of instituting RV with our existing voting infrastructure in the broadest sense, I do not see how AV can not be considered the better alternative. In fact it would be my asserting in terms of the criteria set out at the beginning of this note that it probably is the best system that is available to us. In adopting a new voting system one has to realize that it will take time for our people and institutions to adjust to such a system. On this score, the compelling case for election reform surely ought to be made in terms of approval voting.
UTILITARIAN VOTING--A DISSENT
John Howard Wilhelm
As part of an exchange I had with proponents of range voting, I was referred to an article titled "Utilitarian voting" by Professor Claude Hillinger of the University of Munich. In the article, Professor Hillinger argues that "With the advent of MARGINALISM (i.e., neoclassical microeconomics), utilitarianism became a victim of ideological battles. He goes on to identify two additional attacks on the idea of cardinal measure-ability of utility: the first "as a result of the conflict between orthodox economics and socialism." and the second as a result of "the ideological battles... in the early days of the Cold War." As part of the latter he asserts that the Rand Corporation along with other American institutions, especially the Ford Foundation, were involved in an effort "to fight communism...also on the ideological front" and ties that effort to Kenneth Arrow's 1951 book SOCIAL CHOICE AND INDIVIDUAL VALUES. He also makes the claim that "Among economists it became a mantra to say 'Of course, interpersonal comparisons of utility are impossible!' in order to signify one belonged to the orthodoxy."
The difficulty I have with such a discourse is that it is a polemical, not an analytical, argument in that it simply reduces "orthodox"
economics to a question of ideology. Such a reductionist approach implies that those supporting the "orthodox" position here have neither the intellectual honesty nor a logical basis for their position, a position that I would reject. It is a position which frequently characterized the Marxist polemics against Non-Marxists with their disasterious consequences in the 20th Century.
As an example of this, I would cite the response to the publication in
1921-22 in the Soviet Union of the article on the Socialist economic system by the great Russian economist Boris Brutzkus which I helped to get republished there in 1990 during the period of glasnost'. For publishing his critique, Brutzkus was arrested and expelled along with other intellectuals from the Soviet Union in late 1922.
In his essay, which clearly anticipated 60 years of scholarship on the Soviet economic system, Brutzkus was very sceptical of the socialist position which Hillinger describes in his article. In the piece Hillinger states that, "The socialists and their allies on the left believed in the possibility and desirability of a strong state that would use scientific principles to achieve desirable social goals."
He goes on to point out that "They realized that such policies had to be based on measurement, with the result that there was an explosion of work on social and economic statistics."
Brutzkus, who, according to what I was told by an Italian economist, was one of the foremost statisticians in Russia prior to the Revolution, took the position that such "scientific" principles were useless in the absence of a market test for summing up and determining consumer demand in contemporary economies. One cannot read Aron Katsenelinboigan's account in his book SOVIET UNION of attempts to introduce quantitative techniques into Soviet planning without realizing the Brutzkus had great insight on this issue. It is clear from Katsenelinboigen's account, as well as those from others such as Igor Birman, that while such efforts taught Soviet economists a great deal, they led to no practical consequences for the system. When Katsenelinboigen told me that it was Kantorovich who taught him that Marx was wrong and then added that Kantorovich was quite conservative, I had understood from the comment that he was referring to linear programming, not to Kantorovich himself. And indeed, Katsenelinboigen makes that very clear in his book.
As an "orthodox" trained economist one cannot read Brutzkus's 1921-22 essay without seeing that he owed a lot in his analysis to the insights of neoclassical economics, or if you will Marshallian economics. And it is clear from work done by the Japanese economist Masashi Morioka that Brutzkus knew of Alfred Marshall's works and valued him a great deal as an economist. It would be my assertion that in his analysis Brutzkus was operating very much within the framework of those who had reservations about interpersonal utility comparisons. On this score, I do not see how one can dismiss such reservations as cavalierly as Hillinger appears to do in his Utilitarian voting essay. As the late Alec Nove, hardly a right-wing orthodox economist, wrote to me in a personal note, Brutzkus's essay was "A very good analysis already in 1921-22!" There surely are some valid and useful insights that so called "orthodox" economics does give us and they surely cannot be as simply dismissed as Hillinger does in the case of its reservations about interpersonal utility comparisons.
Hillinger's statement that "Macroeconomics and welfare economics, the two principal areas that require cardinal measurement, were outside their (the marginalists' or orthodox economists') focus," surprised me a great deal. In the case of welfare economics, it certainly is true that most economists eschew the idea of interpersonal utility comparisons. But this does not exclude cardinal measurement in that area of welfare comparison. We use the idea of a marginal rate of substitution of money, as the late Kennith Boulding labeled it, for a good or service to measure consumers' surplus in making welfare comparisons and arriving at economic conclusions. And surely such considerations underlies most cost benefit analyses. In doing so we assume that given the utility of a quantity of money a person has they can compare that to the utility they find at the margin of a given good or service. But as we surely know, the utility of monetary income at the margin is likely to be different for a person if they are rich or poor. This surely raises serious problems for the idea of interpersonal utility comparisons in economic analysis.
In the case of the idea of utilitarian voting, which as Hillinger maintains encompasses range voting, the issue is what evidence do we have of the validity of interpersonal utility comparisons in judging voting outcomes under alternative voting systems. In an answer to this, Hillinger cites the area of happiness research. In his piece on Utilitarian voting he states that "A contemporary revival of utilitarianism is associated with the burgeoning interdisciplinary field of happiness research. Happiness researchers measure the determinants of individual happiness, or life satisfaction, on cardinal scales."
In reading Daniel Haybron's recent piece "Do We Know How Happy We Are?
On Some Limits of Affective Introspection and Recall" (Nous, Vol 41, No. 3, 2007, pp 394-428), one cannot but be struck with the fact that there are a number of conceptual problems here which raise serious issues about such inter and even intra-personal comparisons. In particular Haybron cites the problem of scale norming. He points out (p. 403) that in asking the question of "How pleasant would you say your experience is, on a scale of one to ten?"...people can use widely varying scales when rating their experience, a phenomenon known as scale norming." Therefore, he points out that, "one person's ten might be another's six" and that such scale norming undermines interpersonal--and even INTRApersonal--comparisons.
How such a situation can be different in voting systems as in other cases is hard for me to see. Indeed, I have frequently found it difficult to be consistent in responding to such scales on identical questions when they are asked in different places in consumer surveys I have filled out. Grading is clearly another area where scale norming appears which is why we frequently resort to standardized tests to try to measure things in this area more accurately. In the case of range voting, Hillinger makes a very perceptive case when he points out that range voting "with the 1-100 'temperature' scale, or even with a ten point scale, raises difficult problems," since such temperature scales provide "more incentives for strategic voting, and hence falsification of preferences." His reasoning in favor of a simpler scale because "in political elections people are ill-informed so that they are hardly in a position to offer a more differentiated judgment" seems to me rather perceptive.
While Hillinger proposes the scale {+1,0,-1}, I have reservations about that as well. As opposed to a system like approval voting, such a scale can obstruct transparency in terms of how people understand voting results which most see in terms of percentages of voters in favor of individual candidates. When one takes all these considerations into account one cannot have very much confidence in mathematical proofs of the advantages of range voting in terms of their robustness. It would be my contention that the traditional approach to welfare economics which separates economic judgments from value judgments, surely important in dealing with policy decisions, is much more sound than an utilitarian approach. Similarly, it seems to me that analysis of voting systems based on ordinal considerations would also have to be given more credibility than any based on cardinal considerations in terms of the robustness of the conclusions reached.