‘It is a capital mistake to theorise before one has data’ Sherlock Holmes.
There as been various calls for a ‘cost benefit analysis’ of gambling in New Zealand. The expression ‘cost-benefit analysis’ (CBA) has a rigorous meaning in economics, and while there is no need for economics to insist that their meaning of the terms should be universally applied, it is helpful to recognise that the phrase is being used as a short hand for ‘an analysis of the costs and benefits’. Thus the CBA ends up with a single number – a total quantum of which summarises all the costs and benefits in an economy. But even were that quantum zero – so the costs and benefits netted out – there would still be considerable interest in the individual costs and benefits and their incidence. So until there is a consensus to the contrary, I propose to interpret the expression ‘cost benefit analysis’ to be synonymous with ‘the analysis of costs and benefits’ rather than economist’s technical term which in this text I shall refer to as ‘CBA’. I use the expression ‘cost benefit analysis’ to denote both.
Analysing the Costs and Benefits of Gambling
Even so, the best way of analysis the costs and benefits of gambling, or anything else, is to use the analytical framework that underpins a CBA. Of course like Humpty-Dumpty, we can make words mean whatever we want. The fallacy with his approach is that words are used to communicate with others, and if the speaker is their master, those who hear the words may not understand what it is intended to be conveyed. The fact is that expressions such as ‘costs’ and ‘benefits’ are used frequently in public conversation as if they refer to the economists’ notions. To Humpty-Dumpty the words would create confusion.
The core of the economists’ notion of cost is ‘opportunity cost’, that is the cost of forgoing the next relevant opportunity, as explained below. Benefits are treated in a parallel way, so we dont need to develop the economists’ notion of benefit.
The best way to think of an opportunity cost (or benefit) is to consider two situations which for the purpose of gambling might be called the ‘factual’ and the ‘counterfactual’. The factual is usually the current situation, but the counterfactual depends upon a precise formulation of the issue. In the following I am going to assume that we are interested in the question of how much harm – if any – does current levels gambling cause and the extent to which it is offset by benefits. The counterfactual might be a world in which there was no harmful gambling or perhaps a world in which there was no gambling. The choice of the counterfactual requires some judgement, but it depends upon the precise issue of concern.
(Another issue might be a concern as to what will happen if opportunities for gambling in a region were extended. The factual is the current situation and the counterfactual is the situation where the additional outlets or whatever exist.)
Now, in principle, we can compare the various features with the factual situation with the counter-factual scenario. Some will be exactly the same – the sun will get up in the morning in each. Some will be different but not enough to worry about. And there may be some very substantial differences. In principle we make a list of all these differences.
Notice the construction of this list does not really involve economists, but a host of other social scientists (and epidemiologists and accountants). In my experience the economist’s work is very dependent upon the quality of this list, both in terms of the precision of the definition of items on the list and their quantification. Very often people ask whether X or Y or Z can be taken into account in a CBA. The most common response is that if the questioner can specify and quantify the phenomenon adequately on the list, then the economist can do something (economists’ caveats below).
The list is likely to be a very long one. In practice it is shortened by aggregation but even so the list will still remain long and heterogeneous. At this stage the economist applies a valuation to each item of the list. The valuation is done by a set of rules which are based on that the cost/price equals the social marginal cost which is (broadly) the amount of resources that society would exchange for having the opportunity to avoid or incur that particular item. The costing can be tricky and sometimes involves guessing because the data base does not exist.
(There are also some contentious issues of valuation – see the bibliography for papers which discuss some of the disputes. I do not want to minimise these contentions, but I would say that as important as they are, I have been more often let down by enormous gaps in the data which characterises the list.)
If all the items on the list are valued in the same monetary units, they may be added together and that gives (just about) the outcome of a CBA. Over the years I have become uneasy about this simple aggregation. One reason is that it is not always obvious that the aggregation is like with like. For instance, how does one add together goods and services with human mortality and morbidity. (I have argued why it has to be done elsewhere, and could do so again. My concern here is that the public does not always understand sufficiently what has been done, and misunderstanding a statistic undermines its usefulness.)
Perhaps for gambling a greater concern is that the aggregate may obscure the distribution differences between a factual and counterfactual. For instance (harmful) gambling might make children worse off (because the parents gamble away money that should be used for the children’s welfare) and casino shareholders (say) better off. That details of the redistribution may be far more important than the net impacts of the two. CBAs are underpinned by what is sometimes called ‘Hume’s Law’: ‘ a dollar is a dollar, is a dollar’ and it does not matter who gets the dollar, or who loses it. (There are some more sophisticated measures such as the Atkinson-Stiglitz index, but I have never seen them used in a CBA.)
If may not be a full aggregation. If instead of aggregating to a single monetary amount but finishing with a number of separate items, we have what might be called ‘an analysis of costs and benefits’. Note that the components need not be valued in the same units – a great relief to economists for we are often stretched by the valuation problems.
The Implication for Non-economist Researchers
While the above has not understated the difficulties of applying the economic analysis, it has also drawn attention to that the real need is for a careful, quantified differentiation between a properly specified counterfactual scenario and the factual one.
The ‘industry’ sector part of the difference can, usually, be dealt with relatively easily. Data bases are typically well placed to quantify the differentiation, and there are straight-forward well-attested rules for valuation. In some parts of the industry side there may be gaps, although overseas studies may be helpful. In my experience with health applications, the health sector data bases are more problematic, although sometimes epidemiologists have slaved away to provide credible data about hospital outlays. The same is almost certainly true for the criminal justice system.
The government sector part of the difference is usually straight-forward too (other than public sector industries – see above). It is a mainly matter of the application of known tax rates checked against their revenue in the government accounts.
It is the household sector is the complicated one, because the data is usually grossly inadequate. There are two broad issues.
1. At issue is not the proportion of the population which is subject to a particular phenomenon, but the difference between the proportions in the two scenarios. For instance, suppose the counterfactual differs from the factual by there being no tobacco consumption. What we are interested in then is, say, the proportion who have lung cancer under current circumstances (when there is smoking) to the proportion in the counterfactual (when there is not). Now there will still be some lung cancer even if there were no tobacco consumption, and epidemiologists put a lot of effort into estimating what they call the ‘attributable fraction’ which estimates the additional lung cancer cases as a result of tobacco consumption.
The prevalence and/or incidence data used to calculate the fractions (and also the community levels) is not always as detailed as even the collectors would wish, and often subject to high margins of error. My impression is that there is not sufficiently high quality data for New Zealand, in part because some of the most serious conditions (say criminal behaviour engender by gambling) is, fortunately, a relatively low occurrence and so difficult to quantify with precision. (There is a more on incidence and prevalence in an appendix. I have noticed that non-epidemiologists, myself included, get confused by the distinction unless we close our eyes and try very hard.)
2. Additional to these population figures we need how the phenomenon impacts on the person and those associated with her or him. It is not enough to say that a certain proportion of smokers die from lung cancer caused by tobacco consumption. They go through a harrowing time before that, and their families and friends suffer too, both emotionally and perhaps financially.
The gambling impact is probably more complicated. What we want to know is the various differences in material possessions, health and welfare and so on between family groups with some members involved in harmful gambling and those with none (supposing that is the counterfactual). The exercise of measuring the differences is challenging, but it may not be impossible. While any results will not be entirely satisfactory they should represent progress. (We need not be too despondent if the progress seems limited. Ten years ago I was at a seminar which thought doing the cost of crime attributable to alcohol and illegal drugs was hopeless. Today they are being included in estimates of the social costs of drugs.)
Note that while the data is needed for an analysis of cost and benefits, it is also extremely valuable for all sorts of policy purposes. Think of the implications if we knew something systematic about how harmful gambling by a parent impacted on the children’s education and health.
This leads to a conclusion, but before doing so a word about regional issues. Cost benefit analyses usually cover a national jurisdiction, not least because that is the way the data is usually collected. There is a small literature which looks at the economic and social impacts of gambling, say, on a region or sub-national area. (The procedure could be generalised to look at some other sub-national group such as an ethnic minority or age cohort.) In principle such cost benefit analyses do not involve any new principles compare to national ones. In practice the balance of applications, and therefore potentially the conclusions, is different. Even more practically, the data is even more inadequate. (The two scenarios may be different. For instance the factual may be the region without a casino, the counter-factual may be the region if the casino is introduced.)
Hardly any of this is about the technical side of doing a cost benefit analysis, issues which are covered in papers listed in the bibliography. Rather the paper has tried to provide for economists a robust account of what economists are trying to do, and how ultimately economists are dependent upon what others define and provide. In particular the definition of the factual and counterfactual needs to be given carefully thought out, and the differences of between them carefully measured.
The other point made is that a cost benefit analysis, be it a CBA or an analysis of costs and benefits, is a systematic way of thinking about the entirety of the issue. The value of building one up is more than the numbers that come out at the end. For it forces us to think about the factual and the counterfactual comprehensively, and so move towards an understanding and perhaps policy directions, from a holistic perspective. The economist needs always to keep this in mind too.
Appendix Prevalence vs Incidence Based Approaches
The following is an excerpt from International Guidelines for Estimating the Costs of Substance Abuse: Second Edition, 2001.
Estimates of the economic costs of substance abuse may be either prevalence-based or incidence-based. Prevalence-based studies estimate the number of cases of death and hospitalisations attributable to substance abuse in a given year and then estimate the costs that flow from those deaths or hospitalisations (as well as other costs, such as prevention, research and law enforcement costs). Incidence-based studies estimate the number of new cases of death or hospitalisation in a given year and apply a lifetime cost estimate to these new cases. Thus, prevalence-based estimates generally measure the costs of substance abuse in the present and the past in a given year, while incidence-based studies generally estimate the present and future costs of substance abuse in a given year. For ongoing health and social problems such as illicit drug use, the results of prevalence-based and incidence-based estimates are often similar. For health problems that are declining in magnitude (such as smoking in some countries), prevalence-based estimates will generally be lower than incidence-based estimates. For emerging health issues such as epidemics of HIV or Hepatitis infection, incidence-based estimates generally provide higher estimates than prevalence-based estimates, because many infected persons may still be in the latency phase of the diseases.
Bibliography: Some References on Social and Economic Impacts of Gambling
Australian Institute for Gambling Research, 2001 :Social and Economic Impacts of Gambling in New Zealand, Final Report.
Banks, G. (2002) The Productivity Commission’s Gambling Inquiry 3 years on.
Collins, D. & H. Lapsley (2003) ‘The Social Costs and Benefits of Gambling: an Introduction to the Economic Issues’ Journal of Gambling Studies, 19(2):123-147.
Rankine, J. & D. Haigh (2003) Social Impacts of Gambling in Manukau City. A report for Manuaku City Council July 2003.
Single, E., D. Collins, B. Easton, H. Harward, & H. Lapsley International Guidelines for Estimating the Costs of Substance Abuse: Second Edition, 2001 (The report will shortly be published by World Health Organisation.)
Walker, D. M. (2003) ‘Methodological Issues in the Social Cost of Gambling Studies’, Journal of Gambling Studies, 19(2):149-183.
Wynne, H. J. & M. Anielski, (2000) The Whistler Symposium Report. The first international symposium on the economic and social impact of gambling..On this Website
Index of articles on evaluation in healthcare.
Gambling in New Zealand: And Economic Overview.