This is a follow-up to an old post on reduced fire risk under the Australian Home Insulation Program (HIP). Blathering critics of that program have suggested that, in addition to “causing more fires” it also “killed Aussie workers” because under the HIP it is known that 4 installation workers died, two of heat exhaustion and two of electrocution. This has led to (mainly political) claims that the HIP was badly implemented and dangerous. I say “political” claims because no one making these claims has bothered to find out whether people died installing home insulation before the program was introduced, or how many people died. People often die in workplace accidents; the question is whether suddenly increasing the rate of home insulation led to an increase in bad workplace practices that might have led to reduced safety and higher death rates. This question is best answered using statistical methods. We can’t do this properly because no one making the claims has presented any data about deaths before the scheme[1]. But we can use the estimates of risk exposure in my previous post, along with the known death rate under HIP, to make some plausible estimates of what data would be required in order to claim that the HIP led to an increase in deaths. As we’ll see, with “only” 4 deaths under HIP, it’s highly unlikely that one can make a strong claim that the HIP led to an obvious increase in risk.

Methods

Using the most basic assumptions about risk from my last post[2], I calculated the death rate per 1000 installations under the HIP program. Under the assumption that deaths are a rare event and poisson distributed, and assuming the best possible situation in the pre-HIP program of 0 deaths, I calculated a confidence interval for the 0 death case using a simple online calculator. I then estimated how many years of risk exposure would be required for the upper bound of this confidence interval to fall below the point estimate of the death rate under HIP. This indicates 95% confidence that the death rates in the pre-HIP era were lower than under HIP.

Results

There were 1,100,000 home insulation installations in 200 days under HIP, giving a death rate of 4/1,100,000=0.36 per 100,000 installations.

There were 70,000 home insulation installations per year pre-HIP, giving a death rate of 0 per 100,000 installations. The upper bound of the 95% confidence interval for 0 deaths under the poisson distribution is 3.7, which calculates to 3.7/70,000=5.3 per 100,000 installations, much higher than the HIP rate. For this confidence interval to fall below the point estimate of the death rate under HIP, we need to observe 0 deaths for 5.3/0.36=14.7 years.

This result arises because the rate of death under HIP is very small. Under a poisson distribution with 0.36 deaths per 100,000 installations, the probability of 0 deaths in 70,000 observations can be calculated as 78% [3]. So under the null hypothesis that the HIP death rate is the same as the pre-HIP rate, you need to go back well past 1 year before you can find a significantly elevated risk of death with any degree of confidence.

Conclusion

In order for critics of the HIP to claim that it led to an increase in death rates, they need to claim that there were no deaths in the home insulation installation business for more than 14 years preceding the implementation of the HIP. This is a pretty tall claim to make in an industry which is known to have had dodgy installations prior to HIP. The CSIRO’s report on the HIP makes this latter fact patently obvious in its risk profiling section, and although it’s possible that there was never a home insulation death in Australian history, I find the possibility quite remote.

Of course this whole issue is a complete furphy, since HIP is a federal program and workplace safety is a state issue. If critics of the HIP want to argue that federal laws can affect workplace safety, then they need to recognize that for the 10 years preceding the HIP, federal workplace safety laws were set by the current critics of HIP, and any risk accruing to workers under HIP has to at least partly be blamed on the previous 10 years of Federal laws. But the likelihood is that there was no increase in workplace deaths, and just as in the case of fires, if the data were available we would probably find a reduction. Criticisms of HIP on this basis essentially use publicity over 4 tragic deaths for purely political aims, with no interest in either workplace safety or greenhouse gas abatement on the part of the critics.

fn1: I wonder why? Could it be because they don’t really care to?

fn2: I did a lot of work to establish different risk exposure profiles in that last post, but in the preceding time period no one who is critical of the HIP program’s “dangers” has bothered to advance the state of knowledge one whit. So I’m feeling a lot less charitable a second time around, and can’t be bothered with further shenanigans in the interests of conservative analysis. If some conservative hack has an interest, they’re welcome to find proof of the data they need to make their case. They won’t.

fn3: I did this calculation very quickly in excel and could be wrong, so don’t quote me