Memorandum by Professor Christine Godfrey
(AL 67)
RESPONSE TO INDUSTRY SPONSORED CRITICISMS
OF THE RAND AND SHEFFIELD STUDIES ON THE IMPACT OF PRICE AND CONSUMPTION
ON ALCOHOL RELATED HARMS
INTRODUCTION
Two of the papers are from the Centre for Economics
and Business Ltd. They claim to be independent consultants but
no names are attached to the report and it is difficult to assess
the credibility of the individual authors. Both studies were supported
by SABMiller. Oxford Economics review was sponsored by the Brewers
of Europe but the reviewers make no claim to independence, The
third body is the European Forum for Responsible Drinking.
CEBR CRITIQUE OF
THE RAND REPORT
Ineffectiveness of price on heavy drinkers
This critique seems to contain a fundamental
misunderstanding about price responsiveness. The argument put
forward is that raising prices may have little impact on alcohol
related harm or harmful consumption. This argument fails to take
account of the disproportionate share of consumption of hazardous
and harmful drinkers. In the UK in 2008, the 24% of hazardous
and harmful drinkers accounted for the consumption of over 75%
of the alcohol sold. This implies that even small impacts of price
on harmful and hazardous drinkers will have a large absolute impact
on consumption and alcohol related problems. Imply differences
in price elasticity—0.21 for hazardous/harmful from
the Sheffield study to—0.47 for moderate (see further
comments on CEBR critique of Sheffield modelling study) implies
that "alcohol consumption by heavy drinkers is much less
sensitive to price changes than for moderate drinkers"—seems
to exaggerate the evidence. As the Sheffield modelling exercise
on UK data also demonstrates taking cross price effects into account
also has an impact and the estimated benefits in reduce harm from
different pricing policies are substantial. For moderate drinkers
even if more price sensitive the impacts on consumption are modest
in absolute terms given their low level of drinking. It is interesting
that CEBR try to use the Sheffield study to reinforce their arguments
against the RAND study in March 2009 but then later bring
out a critique of the Sheffield modelling work.
Costs to moderate drinkers and other impacts
The argument made in this report but not detailed
is that policy changes that increase taxes impact on the welfare
of non harmful drinkers. The argument is that consumers will as
a result of the price changes loose the welfare they would have
gained from the higher consumption they would have had if prices
had not risen (the technical name is loss of consumer surplus).
This economic argument has some grounds and is true for any legislative
policy where changes in behaviour are not undertaken voluntarily.
The first counter argument that can be made
is that even moderate drinking is not completely risk free. Only
if consumers are fully aware of all the harms of alcohol and those
harms do not impact on third parties (externalities) can the loss
of consumer surplus be considered in full. Note is this type of
economic analysis no gains from the "trade" of alcohol—jobs
in the industry—as in claimed in this report. It should
be noted that any change in consumption would bring about changes
in employment and spending shifts—the overall impact in
any country on employment is hard to predict as it depends on
the labour intensity and import mix of the different consumer
goods. Studies of falls in tobacco consumption suggest that overall
the number of jobs in the economy rise in all countries other
than tobacco growing countries. While alcohol production is more
spread across the world it has become very capital rather than
labour intensive.
The ethical and economic arguments for public
health policies of these kind revolve around the public good and
the compensation moderate drinkers may enjoy from the drop in
third party alcohol related harm such a pricing policy may bring.
So if public drunkenness, alcohol related violence and accidents
reduce there are gains to moderate drinkers as there are if alcohol
related public expenditure on health care, criminal justice costs
etc etc reduce. If as well as such individualistic arguments there
is some public ethos (caring externalities) that the state does
have a stewardship role in individual behaviour there could be
gains even if the impact of the policy was only on improving the
quality and quantity of life of the hazardous and harmful drinker.
Given the very modest drop in consumer surplus (because of their
low absolute level of consumption) these arguments do not seem
very sound.
The CEBR report on the RAND study makes one
other claim for economic benefit from moderate drinking that moderate
drinkers have better health and higher work productivity. The
impact of alcohol on earnings (from which the claims of higher
productivity from moderate drinkers arises) is complex and this
is clearly a two-way relationship as the RAND study on affordability
demonstrates—higher income implies higher alcohol consumption.
The studies quoted are three from a much wider and more mixed
literature.
OTHER CBER ARGUMENTS
ON THE
RAND REPORT
The CBER report makes two other arguments about
the RAND study:
1. Affordability is a misleading measure
It is not clear what point is being made about
the composite measure of affordability. On one hand the report
claims that only price is the important factor but this belies
the very large evidence base that both price and income are important
determinants of alcohol consumption and thereby alcohol related
consequences. The policy issue is that taxes set according just
to price changes do not take into account such income changes
and the consequent impacts on public health. This point in itself
does not invalidate any of the RAND results. The CBER Figure 1 just
illustrates the changes in affordability over the period of the
study which was a period of income growth.
The unreliable statistics arguments are a collection
of points. The first point concerns that use of a combined index
of affordability rather than disaggregating the impacts into price
and income effects. However, the critique which suggest the RAND
report is univariate is misleading in that the analysis was a
multivariate one with other factors entering the regression analysis.
The different factors influencing trends in Europe over this time
period mean it is not surprising that simple graphical analysis
of the relationship between affordability and consumption vary—indeed
this is the whole reason for more sophisticated statistically
modelling. They do suggest other faults of the regression specification
concerning error terms which may have some validity. While statistical
tests do exist to examine whether these are or are not a problem,
these are inconsistently applied in peer reviewed published academic
studies. Without further testing of the RAND regression model
this claim cannot be repudiated or substantiated.
The final argument made seems to be a repudiation
of the relationship between overall alcohol consumption and overall
patterns of harm. This seems to be based on two arguments, that
moderate drinkers have no harms and indeed may have health benefits.
These are essentially non economic arguments which have been fully
explored elsewhere in the literature.
OXFORD ECONOMICS
CRITIQUE OF
THE RAND
STUDY
The main thrust of this report is to criticise
the link between alcohol consumption and harm and to suggest the
report is biased and selective in reporting evidence. There is
no real evidence to suggest the report has been selective in evidence.
The Oxford Economics critique of the harm data selected also is
very questionable but outside my specific sphere of expertise
and I am sure there are other experts to hand to explore these
claims.
This report also claims the RAND study fails
to quantify the benefits of the alcohol industry. As stated above
in most normal economic analysis of policy while any impacts of
changes in consumption could have local impacts as such these
are not usually impacts factored into economic evaluations—in
part because changing consumption of one consumer good is generally
replaced with other consumption (less any increase in savings)
with other consumption with its own economic impacts. Previously
industry sponsored studies eg the Ernst and Young study have grossly
exaggerate the jobs dependent on alcohol and other so called economic
benefits of the alcohol industry. This report also mentions the
consumer surplus argument (see above).
They also quote selectively on the literature
about price effects and problem drinkers, arguing that moderate
drinkers are "hit ..far more" (page seven) than harmful
drinkers. "hugely impact on the rights of moderate consumers"—This
again fails to equate proportionate effects with absolute effects
and seems to equate lower price responsiveness with no price responsiveness.
They do mention the potential cross price effects which in the
Sheffield modelling study implies that harmful and hazardous drinkers
have higher overall impacts (see discussion on CEBR critique of
Sheffield study below) see quote on page six without realising
it is actually a counter argument to their own. This is their
main conclusion (page nine) but does not stand up to the international
evidence base unlike their claim on page 10 ("consensus
in the academic evidence (which is not quoted in the paper) shows
those who drink excessively have a very low price elasticity".
Some specific points:
Price shifting and taxes—This is
an area of very little research and the only peer reviewed published
studies would suggest that in the US, tax changes are over shifted.
That is the retail price is higher than that would have resulted
from the increase in the tax rate. Detailed analysis of UK tax
changes which occur annually may give a European perspective but
this is a very under-researched area. It seems a minor part of
the major argument about tax changes in Europe and their real
value.
Total income versus income per drinker population—The
argument made here is for a refinement in the income variable
used. The argument is that income should be first per capita and
then refined to per potential drinker. (Check RAND study for data
used)
The arguments made about young people
being less price sensitive are similar to those for heavier drinkers.
The policy issue is less about relative price sensitivity across
different drinking or age groups but more about the impact on
harm. Those drinking moderately will not suffer major absolute
changes in consumption because of their low level of consumption.
The skewed nature of alcohol consumption with hazardous and harmful
drinkers accounting for the vast majority of alcohol consumed
implies seems to have been completely ignored by these commentators,
and emotive language used to make claims about small absolute
impacts.
Cross border effects. There is an interesting
quote in the study that "to help eliminate cross-border consumption
requires equalization of tax rates" (page 11). However, smuggling
of tobacco products was for many years more prevalent in low tax
than high tax countries across Europe. There is also a need to
consider illegal smuggling and tax unpaid trade within and between
countries as well as cross border legitimate shopping. However,
I also was not totally convinced by the arguments in the RAND
report. A specific problem in the UK is not one necessarily around
the South East but in low income areas where the illegal trade
in alcohol may be impacting on the poor and the young drinker,
although the evidence is mainly anecdotal.
Aggregation versus disaggregation. Clearly
the RAND study had a specific purpose to conduct a cross national
empirical study. At this level of aggregation of data the sub
analysis suggested in the Oxford Economics critique cannot be
undertaken. Different studies using different types of data are
for different purposes. Ideally for proving specific causal relationships
the finer and more disaggregated data where all confounders can
be modelled statistically is preferred. However, for policy simulations
aggregation of data is required especially at a European level.
The RAND results are clearly in line with international reviews
of the literature and these arguments have less credibility in
these circumstances. However, as a single study it would not come
high in an evidence hierarchy.
EFRD COMMENTS ON
RAND'S FINAL
REPORT
This critique is long and focuses on very specific
factors and data within the RAND study. I have not gone through
each point but attempted to pick out any additional economic arguments
that were not covered in other critiques. In this report the repeated
claim is about young people and price. Again lower than older
price responsiveness is equated to no effect a number of times
or again proportionate rather than absolute effect arguments are
made.
Overall there seems to be some confusion in
this report between the relationship between alcohol consumption
and affordability and trends in data in specific countries in
the individual trends. Also they exaggerate price changes require
for consumption changes—suggesting elasticity of—0.22 requires
a 900% price rise to reduce consumption by 19.8%. It is not clear
how this was calculated but must include some adjustment for income
rises over the last decade. In reality the policy recommendations
is that tax changes should take account of income changes over
the immediate period not for adjustments to be made at 6-10 year
intervals.
Some points are made about differences in on
and off trade consumption and prices. Different tax structures
with different proportions of value added and specific rates can
favour different selling methods. Also as has been seen in some
European countries, alcohol can be sold at a loss in multi-product
outlets such as supermarkets unlike on-license premises where
the scope for such sustained price cuts are less. However, the
aggregate analysis in the RAND report was never capable of tackling
this issue nor are data available other than general trade sources.
The report also suggests that there may not
always be a perfect relationship between tax rates and price.
They are also concerned that RAND did not look at other factors
influencing relative prices between countries. However, of more
interest to policy is whether changes in tax rates would
or would not be passed onto the consumer. In situations where
there is less competition and drinks producers and retailers have
excess profits there would be scope for taxes to be under-shifted
or over-shifted. The report only reports evidence on levels from
a WHO report (page 10). The potential of the industry to distort
prices is of concern as alcohol producers and retailers become
more concentrated across Europe especially as under shifting may
subvert public health objectives. Also in general there can be
health benefits in setting taxes as a specific rate as unlike
value added taxes these do not decrease if prices are cut. The
issue is the low minimum rate of these specific duties across
Europe.
Cross border shopping is also mentioned in this
report and binge drinking in the North East and North West mentioned.
This is an historic difference in drinking patterns but also these
areas have larger concentration of poor and may be the target
for smuggling rather than legitimate cross border shopping. Large
urban poor populations mean that it is easier and cheaper for
criminal gangs to organise distribution and sale of cheap illegal
alcohol.
They also refer to selective evidence on taxation
but as with other critiques their claims are not borne out by
the international review evidence.
There are some interesting quotes about two
for one offers which were also offered as evidence at the House
of Commons Select Committee evidence. The claim is that all these
offers are on higher price products. They do not provide evidence
for this claim. The House of Commons Select Committee did ask
the trade bodies for such evidence.
Finally, the report ends to suggest education
measures should be added to the list of effective tools. Their
claim is that other factors influence consumption but do not distinguish
between those that can and those that can not be influenced by
policy. Other than education they do not point towards other policy
measures to reduce alcohol harm.
CEBR CRITIQUE OF
THE SHEFFIELD
MODELLING STUDY
This paper is longer than the CBER critique
of the RAND report and interestingly put together there are some
inconsistencies in their comments especially concerning aggregation
of data.
The most important claim they make is that the
elasticity estimates of the Sheffield study are double the estimates
they think are reasonable. Their analysis is however based on
a fundamental misunderstanding of the Sheffield research. The
Sheffield model is complex and so these arguments superficially
can look plausible.
The CEBR critique fails to recognise that the
Sheffield model is a disaggregated one and all the simulations
are based on the disaggregated equations not any artificially
aggregated estimates. The models take account for each group not
only of all the cross price effects of other alcoholic drinks
but also the impact of a change in alcohol prices on the consumption
of other non alcohol goods. The authors of the Sheffield clearly
lay out the reasons why individual aggregate price elasticities
for alcohol by different drinking groups may not be meaningful
on page 51 of their report. They also explain why their model
results which show own price elasticity estimates which increase
with mean quantity of alcohol consumed in their disaggregated
model are consistent with the literature which at that highest
level of aggregation show hazardous and harmful drinkers are less
price elastic than moderate drinkers. That is the Sheffield results
are logically consistent with previous literature. Interestingly
the paragraph explaining these results and that in the Sheffield
study finds that impacts rise with level of consumption is quoted
in full in the Oxford Economics critique.
The CEBR by attempting to draw relationships
between estimated effects at the aggregate level leads to misleading
results because they are not taking the disaggregated cross price
and income effects into account. Their estimates—using half
the estimated aggregate effect are not only distorted in overall
size but also in distribution. The Sheffield model takes into
account that cross price impacts vary in a very complex way between
moderate and hazardous/harmful drinkers and across the different
beverage and price groups of goods.
These arguments invalidate all of their empirical
estimates in the report.
More specific points:
Figure 1.2—The argument made here is that
younger people face higher price per unit on average than moderate
drinkers. However, these figures are not broken down by source
of alcohol unlike Figure 1.1. The proof of the Sheffield modelling
report is in the estimated effects for different policy simulations
and this point has no relevance other than some younger people's
drinking may not be impacted by minimum prices.
Table 2.1—again the authors only quote
the aggregate elasticity estimates which the Sheffield study indicates
are misleading.
Dependency—there is also an "intuitive"
appeal made about lower price elasticities for dependent drinkers
on page 20. There is no evidence for this and the risk of dependency
as a reason for more stringent restriction on young people's drinking
is not mentioned in arguments about the benefits of moderate drinking
and consumer surplus later in the report.
The focus in this report as others is in interpreting
low price elasticities as no impact and they suggest the Chief
Medical Officer as misinterpreting the Sheffield study (page 22).
In fact they have not understood the Sheffield results as explained
above.
The evidence from Canada is all on graphical
analysis and there is no statistical analysis. They also claim
that a switch from beer to spirits would not be desirable as it
implies people are switching to higher strength products. But
the health argument must be whether individuals are consuming
more pure alcohol as a result of the policy and no evidence is
given for this. Where graphic analysis would seem to support the
minimum pricing policy as in rates of violent crime it is dismissed
as not taking account of other factors.
The main points of Chapter 4 are covered above
and the first three bullet points on page 30 arise from the
CBER's lack of understanding. The fourth point is interesting
as it is suggesting that the aggregate (average) results for moderate
drinkers is not representative as although the average is five
to six units per week some moderate drinkers have higher alcohol
consumption. This is a surprising analysis as clear all averages
are just that averages as well as the group who will have higher
impacts there are an equal number with lower effects.
Some comments are made about crime harms and
how these were introduced into the Sheffield model. However, the
Sheffield report used a study from the Home Office and the AAFs
were taken from this study not by the Sheffield authors. This
would be considered best available evidence for the model but
obviously there is need for more research on the link between
alcohol and crime and how risk functions vary with level of consumption.
The rest of Chapter 4 of the CEBR report is
not valid as explained above.
In Chapter 5 the authors use a model of measuring
benefits to individual drinkers using the consumer surplus arguments.
They do suggest in this report that hazardous and harmful drinkers
may not value the harm to themselves in making their consumption
decisions. However, they fail to put any value on this other than
estimated alcohol related harms. They fail to account for potential
dependency and its costs nor to they seem to value life years
lost in anyway although the graphs are somewhat obscure. One Quality
Adjusted Life Year (QALY) is generally considered to be worth
between £20,000 and £30,000 in the UK. The
CEBR model also assumes there are no risk to moderate drinking.
The rest of the analysis in the report is again
meaningless as they do not understand the Sheffield model and
their own estimates are distorted.
The focus on Chapter 7 is the profitability
and competitiveness of firms. This is interesting given the evidence
provided to the House of Commons Inquiry which suggests that competitiveness
arguments may be a barrier to effective public health measures.
However, there is little evidence of a highly competitive alcohol
production market with increased globalisation. More likely would
be a change in alcohol products with producers switching to lower
alcohol content higher quality products which may have additional
public health benefits. There has already been some movement towards
this in the UK with changes in the cider market.
Professor Christine Godfrey
The University of York
8 June 2009
| 
