Since I expect that my general overview points and criticisms of bad ways to analyze health data will run out long before the year does, I will intersperse some case studies of the style I would use in class. These will look at how the news offers lessons in understanding information about public health and public health policy. In UN49 I wrote about the concept of the dollar value used to assess whether a cost to save a statistical life is justified or too expensive. Currently there are hearings going on about the gas line leak that resulted in an explosion and fire in San Bruno, California last September which killed 8 people, injured many, and wiped out a neighborhood, with about 40 homes totally destroyed. A lot of attention is being devoted to the question of whether it is acceptable that Pacific Gas & Electric, the gas company that owned that pipe, and other companies elsewhere, do not install new valves that can be easily shut off in the event of a major leak or a fire, or even do so automatically.
The New York Times article about the hearings stated that PG&E had declared a few years earlier that the valves would not stop any destruction, presumably because they could only shut down after an explosion indicated there was a problem. But this does not seem entirely plausible, since the fires kept raging there and in a more recent similar event near here that burned all night before they could shut off the gas that was feeding it. Needless to say, some of the news reports were rather more incredulous about the claim that the values would do no good. Nevertheless, the NYT article ended by quoting someone who explained "automatic valves cost about 20 times as much as the value of the gas that would be saved if a ruptured pipeline could be sealed faster." I really find it difficult to imagine that reducing the loss of gas is really the biggest benefit from being able to stop runaway leaks one or more hours sooner. This seems like a good place to consider how to trade off lives saved and expense.
At that point, if I had a class, I would assign each study group to go research the costs and benefits and provide the best estimates they could (based on material I had taught, what they already knew from other classes, etc.). Lacking a class and the twenty person hours of time I would expect them to devote to the question, I decided to see if I could figure out a few interesting numbers more quickly. It turned out to be more of a challenge than I thought, with the press reports about the costs and benefits showing a clear tendency to not actually bother to try to report what the costs and benefits were.
I finally found this from the current hearings, an estimate that it would cost $600 million to replace half of the valves in the country. Since many of the pipelines are in sparsely populated areas where letting a leak and explosion burn for a few extra hours really would mostly just cause the loss of extra gas (a fact I picked up from various articles), it sounds like half would be roughly how many it would make sense to upgrade. Is the cost a big number? Well, reporting costs without knowing the benefits is common but pretty useless. A quick search through the discussion about this revealed one set of estimates in the range well over 100 "incidents" with 30 people killed or seriously injured per year and maybe $30 million in property damage. The latter is undoubtedly an underestimate: Tens of homes burned to the ground in a single incident adds up to $30 million pretty fast, to say nothing of the \figure probably only referring on to the insured price of marketable goods, ignoring the loss of a families' entire collection of personal possessions, things with basically zero market value that may be valued more highly than the structures.
The reports I read talked around the question of how many people die and are injured subsequent to the initial explosion (and thus could be saved by shutting off the gas), perhaps people who were trapped and could have been rescued or firefighters hurt dealing with the larger fire. If 10% of the deaths and injuries could be prevented, along with about half of the property damage, then the we are talking about roughly 1 death (about $9 million using the statistical life number I discussed earlier), 2 serious injuries (call it $1 million for medical care and suffering, probably an underestimate for serious burn victims), and $20 million worth of stuff (accounting for the underestimate). That sums to roughly $30 million per year. If the better valves could save 20% of the victims, it would be about $40 million.
You can see how the amount spent to save a statistical life thinking comes in. The claim that "no level of risk is acceptable" is obviously just as absurd as the claim that a gas leak is merely about lost gas and property damage. The loss of life has to enter the equation somehow, not zero (or we would never bother to protect people) and not infinity (or any public health intervention that saves any lives at all – be it as useful as chlorinating the drinking water or annoying as banning drop-side cribs or BPA – has the same benefit, infinity, making not just comparison to costs but also prioritization impossible).
The numbers in this case seem like a fairly close call (which is actually kind of rare – most of these news-generating cases are really quite lopsided, even though they are portrayed as having legitimate arguments on both sides). One useful comparison is that if you invested $600 million at a 5% rate of return, that would be the $30 million per year. This is not to confuse real resources with bank accounts and finance, but it is a convenient way to think. The actual translation is the $600 million are real resource costs and we would hope to be able to get a reasonable rate of return in terms of real resources – stuff, health, lives saved. The dollars are just a measurement instrument, and failing to understand that is why lots of people cannot think rationally about health tradeoffs.
My numbers are way too rough to figure out whether it comes in a bit above or a bit below the rate of return we would want, and I have no idea if the sources I found had good estimates. I would expect my students to have more than a single newspaper article as the source of quantification, even for a back-of-the-envelope calculation like this. What is clear is that it is not completely obvious which way it cuts, so anyone who reports otherwise did not bother to spend an hour running through the numbers. But it is also the case that if they report that we just have absolutely no idea, they are even more wrong, as I just demonstrated.
As a final thought, if we were asking the government to pay for the upgrades we would have to balance it against other valuable infrastructure improvements or public health measures. (Do not bother to say "we could just take it out of pointless military expenditures" – a rational discussion must include the realization that the many horribly wasteful expenditures in our society are not the ones that are traded off against public health measures.) Government expenditure directed at fixing pipelines people needs to be balanced against numbers like the $9 million allowed per statistical life for other expenditures. On the other hand, if we are asking private companies to stop putting innocent bystanders at risk, as we are in this case, it seems eminently reasonable to require a bit more from them. If we simply had the right level liability (see UN28) for the damage that their pipelines did, it would theoretically give them the right incentives to fix the infrastructure where that was efficient and leave the risk in place where it was not. On the other hand, since corporate managers typically fixate on their one-year balance sheets and investing in safer pipelines will have high current costs and long-term benefits, it seems unlikely that we can trust them to make the right decision based on the financial incentives alone.
[Update:
]
The New York Times article about the hearings stated that PG&E had declared a few years earlier that the valves would not stop any destruction, presumably because they could only shut down after an explosion indicated there was a problem. But this does not seem entirely plausible, since the fires kept raging there and in a more recent similar event near here that burned all night before they could shut off the gas that was feeding it. Needless to say, some of the news reports were rather more incredulous about the claim that the values would do no good. Nevertheless, the NYT article ended by quoting someone who explained "automatic valves cost about 20 times as much as the value of the gas that would be saved if a ruptured pipeline could be sealed faster." I really find it difficult to imagine that reducing the loss of gas is really the biggest benefit from being able to stop runaway leaks one or more hours sooner. This seems like a good place to consider how to trade off lives saved and expense.
At that point, if I had a class, I would assign each study group to go research the costs and benefits and provide the best estimates they could (based on material I had taught, what they already knew from other classes, etc.). Lacking a class and the twenty person hours of time I would expect them to devote to the question, I decided to see if I could figure out a few interesting numbers more quickly. It turned out to be more of a challenge than I thought, with the press reports about the costs and benefits showing a clear tendency to not actually bother to try to report what the costs and benefits were.
I finally found this from the current hearings, an estimate that it would cost $600 million to replace half of the valves in the country. Since many of the pipelines are in sparsely populated areas where letting a leak and explosion burn for a few extra hours really would mostly just cause the loss of extra gas (a fact I picked up from various articles), it sounds like half would be roughly how many it would make sense to upgrade. Is the cost a big number? Well, reporting costs without knowing the benefits is common but pretty useless. A quick search through the discussion about this revealed one set of estimates in the range well over 100 "incidents" with 30 people killed or seriously injured per year and maybe $30 million in property damage. The latter is undoubtedly an underestimate: Tens of homes burned to the ground in a single incident adds up to $30 million pretty fast, to say nothing of the \figure probably only referring on to the insured price of marketable goods, ignoring the loss of a families' entire collection of personal possessions, things with basically zero market value that may be valued more highly than the structures.
The reports I read talked around the question of how many people die and are injured subsequent to the initial explosion (and thus could be saved by shutting off the gas), perhaps people who were trapped and could have been rescued or firefighters hurt dealing with the larger fire. If 10% of the deaths and injuries could be prevented, along with about half of the property damage, then the we are talking about roughly 1 death (about $9 million using the statistical life number I discussed earlier), 2 serious injuries (call it $1 million for medical care and suffering, probably an underestimate for serious burn victims), and $20 million worth of stuff (accounting for the underestimate). That sums to roughly $30 million per year. If the better valves could save 20% of the victims, it would be about $40 million.
You can see how the amount spent to save a statistical life thinking comes in. The claim that "no level of risk is acceptable" is obviously just as absurd as the claim that a gas leak is merely about lost gas and property damage. The loss of life has to enter the equation somehow, not zero (or we would never bother to protect people) and not infinity (or any public health intervention that saves any lives at all – be it as useful as chlorinating the drinking water or annoying as banning drop-side cribs or BPA – has the same benefit, infinity, making not just comparison to costs but also prioritization impossible).
The numbers in this case seem like a fairly close call (which is actually kind of rare – most of these news-generating cases are really quite lopsided, even though they are portrayed as having legitimate arguments on both sides). One useful comparison is that if you invested $600 million at a 5% rate of return, that would be the $30 million per year. This is not to confuse real resources with bank accounts and finance, but it is a convenient way to think. The actual translation is the $600 million are real resource costs and we would hope to be able to get a reasonable rate of return in terms of real resources – stuff, health, lives saved. The dollars are just a measurement instrument, and failing to understand that is why lots of people cannot think rationally about health tradeoffs.
My numbers are way too rough to figure out whether it comes in a bit above or a bit below the rate of return we would want, and I have no idea if the sources I found had good estimates. I would expect my students to have more than a single newspaper article as the source of quantification, even for a back-of-the-envelope calculation like this. What is clear is that it is not completely obvious which way it cuts, so anyone who reports otherwise did not bother to spend an hour running through the numbers. But it is also the case that if they report that we just have absolutely no idea, they are even more wrong, as I just demonstrated.
As a final thought, if we were asking the government to pay for the upgrades we would have to balance it against other valuable infrastructure improvements or public health measures. (Do not bother to say "we could just take it out of pointless military expenditures" – a rational discussion must include the realization that the many horribly wasteful expenditures in our society are not the ones that are traded off against public health measures.) Government expenditure directed at fixing pipelines people needs to be balanced against numbers like the $9 million allowed per statistical life for other expenditures. On the other hand, if we are asking private companies to stop putting innocent bystanders at risk, as we are in this case, it seems eminently reasonable to require a bit more from them. If we simply had the right level liability (see UN28) for the damage that their pipelines did, it would theoretically give them the right incentives to fix the infrastructure where that was efficient and leave the risk in place where it was not. On the other hand, since corporate managers typically fixate on their one-year balance sheets and investing in safer pipelines will have high current costs and long-term benefits, it seems unlikely that we can trust them to make the right decision based on the financial incentives alone.
[Update:
Pacific Gas & Electric estimated that it would cost $225 million to replace 300 manual valves throughout its gas distribution system in Northern California with automated technology. It agreed this week to promptly replace 12 of them.Also, I was reminded of this fact that governs many owners of pipelines:
Under California’s utility rules, customers pay for safety improvements, a tab that includes an 8.79 percent profit margin for PG&E. The safety measures already identified by the company could cost customers nearly $100 apiece in rate increases, which would be added to bills over several years if approved by the California Public Utilities Commission.This suggests that perhaps my last observation was misguided, the one about the difference between what we would expect a private company to pay for exposing people to involuntary risk versus what society as a whole should pay to fix to pay. Since utility customers pay the costs, it is basically the same as a government taxing and spending paying for it (that is, in a society like the U.S. where the taxes are regressive since utility bills are regressive -- i.e., the richer you are the less you pay as a percentage of your income/wealth). Thus this should probably be judged under the same standards as we would judge any government expenditure on public health.
]
No comments:
Post a Comment
Note: Only a member of this blog may post a comment.