Friday, 10 July 2009

Rejoinder to BERL

On Monday, BERL released a statement rejecting criticism of their study. We have completed our reply, available here. Let's hit the highlights.

BERL levies three main critiques of our work. Let's take them in turn.

Reply Number One:

They say that we misinterpret their brief and consequently fault them for things that were never within their remit: specifically, that they do not include the benefits of alcohol.

It is certainly true that benefits were not within the RFP. We noted as much throughout our report, and in our executive summary. However, counting benefits as being precisely equal to zero is what allows BERL to count private costs as social costs. As BERL correctly notes at page 173 of their report:
When measuring the social cost of harmful AOD use, known private costs should generally be excluded…because private costs are offset by private benefits, so there is no net social cost
They there cite Collins and Lapsley, their primary source, as warning against the counting of private costs. They then go on:
In the case of harmful drug use, however, individual decisions are not necessarily made on a rational basis, that is, a decision where the consumer equates their costs and benefits. We argue that the consequences of irrational consumption decisions lead to private costs that are borne by the rest of society, and hence should be included as social costs… We assume that it is irrational to drink alcohol to a harmful level and that harmful alcohol use has zero private benefit.
Without that assumption, BERL could not count private costs as socially relevant. Their entire method hinges critically on that they have decided to assume zero gross benefits to drinkers of their drinking. So while benefits were outside of the RFP, they have taken a very strong position on the absence of benefits: a position without support in the economic literature. It consequently is fair game to critique BERL for counting the benefits as being equal to zero. And we do not understand how BERL can say with a straight face "we cannot accept criticism for not covering issues that were outside the project's brief" when their entire method is built on their having brought it into the project's brief.

If benefits are outside the remit, the proper approach is to consider only external costs. At minimum, BERL should have apportioned its cost tally between private and external costs. Instead, they termed all costs as social costs.

Reply Number Two

BERL notes two errors in our original document. Adrian Slack pointed these out to me at the NZAE meetings, I checked into it, and emailed him the day before he posted his rebuttal informing him that we were adjusting our costs upwards slightly to correct for this. We planned on incorporating the fix to a revised version of our paper, but once BERL's rebuttal went up, I instead quickly issued our errata here. Correcting the errors Adrian pointed out added $36 million to our measure of external costs; at the same time, we added in $197 million in excise-equivalent duties collected by the Customs Service and left out of our initial analysis. We now find net external benefits rather than costs, but would say rather that external costs roughly match external benefits.

Interestingly, in re-reading the section from BERL's source, Rayner, on the employment costs of harmful alcohol use, we also find that Rayner provides a strong argument against BERL's multiplier. Again, BERL multiplies all forgone wages by 1.87: the ratio of GDP to wages. Rayner says that total wages are an upper bound estimate of labour costs; the lower bound estimate is transitional costs in replacing a worker. BERL goes 1.87 times above their cited source's upper bound estimate.

We're reasonably pleased that a rather complicated reverse-engineering job on BERL's report resulted in objections amounting to only $36 million.

Reply Number Three

BERL argues we use assumptions with a cost-deflating bias motivated by world view.

The first striking bit of evidence against us? That we're willing to consider that maybe, just maybe, prisoners and hard core alcoholics would have labour market characteristics somewhat worse than the average Kiwi even if alcohol had never been invented. We don't assume anything here: we go to the relevant literature, find some reasonable relevant estimates of comorbidity of alcoholism and other mental health disorders and of the labour market characteristics of prisoners prior to incarceration, and adjust the figures accordingly. The adjustment has little effect on our overall figures in the grand scheme of things: it knocks about $69 million off of the total cost figures and much less than that off of the external cost figure (since most wage costs are internal). But we thought it was something that should be checked.

Second, we count only as external costs of forgone labour ten percent of the total forgone wages. In this category, BERL begins by obfuscating about computers not running themselves if you're home hung-over. If you go through BERL's numbers, like we have, you'll find that the vast bulk of their forgone labour costs come from unemployment, followed by reduction in workforce through premature mortality. In the first case, BERL's argument about complementarities is a total nonsense. In the second, firms will bear transitional costs of hiring and retraining. BERL's numbers require that these costs equal the forgone worker's total wages times 1.87. Hiring costs are not 187% of annual salaries. Absenteeism, reduced productivity, and injury-related wage costs amount to about 13% of BERL's tabulated total paid labour costs. It's only in those cases where the complementarities story can start making sense as being a substantial fraction of the overall wage bill. Again, recall that BERL's cited source, Rayner, argues for a multiplier of no more than 1.0. We apply a multiplier of 1.1, and apportion the 0.1 as an external cost. Note also that even counting that ten percent as external is more than a little controversial; others question whether there can be any externality at all.

Finally, BERL asserts that alcohol reduces human capital formation. We cited some evidence to the contrary in our report; BERL cites nothing. Frankly, BERL's turning to this issue, and complementarities, smells like ex post rationalization of a dodgy multiplier. In the initial report, this is all that BERL provides to justify its multiplier:
The value to society of lost output is considerably larger than lost earnings alone, for example, in addition to lost wages there is also lost profit. As such, the earnings profiles were scaled up to reflect the difference between wages and residual value added. The resulting output profiles were based on the assumption that the average GDP per FTE (BERL Forecast Database) is 1.87 times the average wage income (StatsNZ). These output profiles by age, gender and workforce status are used in the calculations below.
BERL specifically viewed the 1.87 multiplier as reflecting residual value added. We went to the empirical literature, found estimates of 1.05-1.1 for value added, and applied the upper bound of that range. It seems that, in BERL's view, actually thinking to consult the empirical evidence counts as a cost-deflating downwards bias.

BERL concludes by rehashing their favorite trope: world-view and rationality. Never mind that we have, ad nauseum, shown that none of our results hinge on assuming perfect information, perfect foresight, or perfect rationality. The prior link is the most exhaustive coverage, but the same issue has been covered here and here as well. And I re-iterated it at the NZEA conference with both Nana and Slack in attendance. Instead of providing any reason why the analyses above are wrong, BERL instead reiterates the perfect rationality tropes and throws in personal attacks by suggesting that we think drink driving and murder are perfectly ok. Is this really how senior economists at an established consultancy are going to defend their report?

In the same section, BERL laments our lack of citation on a claim that alcohol saves more lives than it costs.
The ‘result’ that “alcohol saves many more lives than it takes” is an assertion that requires evidence. And the idea that addiction has rational foundations clearly indicates a model view of a consumer that would be at variance with others' views, including those of some economists. Such a value judgement would not have been appropriate for an independent study such as ours.
On that point, as noted in our report (at page 37), we're drawing again on one of BERL's preferred sources, Collins and Lapsley, who at page 38 of their report (Table 11) present figures from Chikritzhs, Stockwell et al (2002) finding that across all categories of drinking, alcohol saved about 2363 lives in Australia in 1998; even among high-risk drinkers, where total lives lost outweigh total lives saved, 557 individuals were estimated to have had their lives saved. If there's any citation issue here, it's that we should have noted that Collins and Lapsley here were citing the work of Chikritzhs et al.

Our full reply, linked to at the top of this post, has a few other bits of interest, including a summary of issues we have raised that BERL thus far has utterly failed to address. Do read the whole thing.

No comments:

Post a Comment