{"items": [{"author": "Josh", "source_link": "https://www.facebook.com/jefftk/posts/671139179582?comment_id=671144997922", "anchor": "fb-671144997922", "service": "fb", "text": "> Instead of limiting prediction, however, if we could manage to allow unrestricted prediction but also implement inequality-reducing wealth transfers (tax the rich, give to the poor) then this should be better overall.
I like this conclusion. If more and better information leads to specific outcomes you don't like, look for other ways to tweak the outcomes. Keep the *system* clean and fair, and handle problems as problems, rather than making the system convoluted and unfair in a (completely futile) attempt to avoid problems.
(I'm not sure \"inequality-reducing wealth transfers\" is my personal favorite way to tweak the outcomes, but it seems like the right *approach*.)", "timestamp": "1407338359"}, {"author": "Hollis", "source_link": "https://www.facebook.com/jefftk/posts/671139179582?comment_id=671152048792", "anchor": "fb-671152048792", "service": "fb", "text": "Agreed. (It also feels worth noting that we're all coder-y sorts of guys and that that likely informs a desire for simplicity and transparency in systems. The \"make the system simple and then adjust the inputs so the system works fairly for everyone\" approach may be less intuitive for people who haven't ever had to invest time developing systems.)", "timestamp": "1407343092"}, {"author": "opted out", "source_link": "#", "anchor": "unknown", "service": "unknown", "text": "this user has requested that their comments not be shown here", "timestamp": "1407344419"}, {"author": "Arthur", "source_link": "https://www.facebook.com/jefftk/posts/671139179582?comment_id=671161375102", "anchor": "fb-671161375102", "service": "fb", "text": "Limiting the capacity of actuaries to assess risk isn't a kludge tacked on to the idea of insurance, it's an inherent part of what insurance is.
All insurance is fundamentally a redistributive act, taking money away from some people who don't need it and giving it to people who do need it.
The degree to which insurance underwriters can practice selection determines how redistributive insurance is.
Perfect selection (your insurance premia exactly match your predicted loss, which we predict with omniscient precision) would mean that insurance simply did not exist -- it balances out to zero.
By contrast, perfect anti-selection is socialist utopia -- literally \"from each according to his ability, to each according to his need\".
In practice insurance exists in an in-between state, partly because some things simply are not knowable and partly because we do in fact impose regulatory limitations on what we allow insurance companies to know -- the latter only partly due to direct socially-redistributive mandates but also partly due to other privacy concerns.
(I could give you a much more \"accurate\" life insurance rating system if I were allowed to have total and unrestricted access to stalk you all the time, but our government has decided it's a bad thing for any private entity to be able to do that to any other private entity in principle.)
I may not be explaining this very well, but the existence of a fuzzy area of uncertainty about how much you know about someone that allows you to justify taking money away from some people who turn out to not need it and give it to other people who do -- that is insurance.
The greater knowledge we all have of what our true risks are, the less anyone actually needs insurance -- and that's why even though greater predictive power theoretically could make insurance more profitable insurers don't always like it because it also makes insurance less necessary. (The way to square that circle is of course to give that information to the insurer but withhold it from their customers, and that's where it gets creepy. This is why the government also puts regulations in that require insurers to share actuarial data up to a certain point before it becomes a trade secret, and the Society of Actuaries makes openness about data one of their basic guild traditions.)", "timestamp": "1407348514"}, {"author": "David Chudzicki", "source_link": "https://www.facebook.com/jefftk/posts/671139179582?comment_id=671164863112", "anchor": "fb-671164863112", "service": "fb", "text": "Elliot, I'd love to name many more examples, but here's one that I feel is pretty clear-cut: When I worked in setting insurance rates, there were some states where we could only use information from previous accidents if the driver was determined to be \"at fault\". Non-at-fault accidents were also predictive, and I claim the upside to the system in letting us use it would outweigh the downside.", "timestamp": "1407350099"}, {"author": "Jeff Kaufman", "source_link": "https://www.facebook.com/jefftk/posts/671139179582?comment_id=671167777272", "anchor": "fb-671167777272", "service": "fb", "text": "@Elliot: \"I don't think the level of abstraction you're using to analyze this issue is productive.\"
I'll disagree by answering all your examples together. I think in each case we make the insurance market less efficient by grouping together people with different expected risk. (Though let me know if you think any of these are actually just preventing overt not-profitable discrimination.) On the other hand, each case also has a redistributive function, where compared to a world without these prohibitions it results in lower rates for people who are worse off at the cost of higher rates for people who are better off. I'm claiming that using the insurance market to make this transfer is awkward and inefficient, and we should let insurance do it's primary job (charging people in proportion to their risk) and do the redistribution directly.", "timestamp": "1407351635"}, {"author": "Arthur", "source_link": "https://www.facebook.com/jefftk/posts/671139179582?comment_id=671167842142", "anchor": "fb-671167842142", "service": "fb", "text": "It isn't awkward and inefficient. It's exactly how insurance is supposed to work. Insurance is a redistribution of money -- insurance that doesn't do that isn't insurance.", "timestamp": "1407351703"}, {"author": "Arthur", "source_link": "https://www.facebook.com/jefftk/posts/671139179582?comment_id=671167931962", "anchor": "fb-671167931962", "service": "fb", "text": "The destabilizing effect of everyone being able to throw in any risk factors they want into a pricing model would be on net harmful to the insurance industry as it exists, which is why they don't fight that hard against rating restrictions. To have a recognizable \"insurance industry\" you need, to paraphrase Douglas Adams, \"rigidly defined areas of doubt and uncertainty\".", "timestamp": "1407351781"}, {"author": "Jeff Kaufman", "source_link": "https://www.facebook.com/jefftk/posts/671139179582?comment_id=671168186452", "anchor": "fb-671168186452", "service": "fb", "text": "@Arthur: Insurance has historically worked by charging people in proportion to their risk. If you don't have regulations restricting insurers from charging more in riskier cases that doesn't cause the collapse of insurance because we can't predict all risk.", "timestamp": "1407351933"}, {"author": "David Chudzicki", "source_link": "https://www.facebook.com/jefftk/posts/671139179582?comment_id=671168241342", "anchor": "fb-671168241342", "service": "fb", "text": "Doesn't have to be redistribution in expected value, before any insured events happened.", "timestamp": "1407351960"}, {"author": "Arthur", "source_link": "https://www.facebook.com/jefftk/posts/671139179582?comment_id=671168361102", "anchor": "fb-671168361102", "service": "fb", "text": "Maybe I wasn't clear enough -- the most \"efficient\" possible insurance model by far is simply nobody having any insurance. Your total premium input becomes exactly equal to your total predicted lifetime loss, which we predict with a probability of 1 through the simple expedient of waiting until after it happens to predict it.
\"If you don't have regulations restricting insurers from charging more in riskier cases that doesn't cause the collapse of insurance because we can't predict all risk.\"
Yes, but an insurance market exists because an area of uncertainty exists -- and the actuary's job is actually easier the more \"rigidly defined\" that area of uncertainty is. Screwing around trying to explore new and exciting boundaries in risk factors is quite likely to throw a lot of money away in terms of costly failed experiments and in terms of the side-effect social costs of invading privacy.", "timestamp": "1407352040"}, {"author": "Jeff Kaufman", "source_link": "https://www.facebook.com/jefftk/posts/671139179582?comment_id=671168391042", "anchor": "fb-671168391042", "service": "fb", "text": "@Arthur: Consider workers comp insurance. This is a market where we really don't want less risky places to subsidize riskier places, because that cuts down on the economic incentive for places not to hurt their workers.", "timestamp": "1407352056"}, {"author": "David Chudzicki", "source_link": "https://www.facebook.com/jefftk/posts/671139179582?comment_id=671168440942", "anchor": "fb-671168440942", "service": "fb", "text": "Arthur, in most cases I'd be all for your extreme example of perfect prediction and not needing insurance. Eg you'd only drive if you weren't going to hit and kill someone.", "timestamp": "1407352064"}, {"author": "Arthur", "source_link": "https://www.facebook.com/jefftk/posts/671139179582?comment_id=671168615592", "anchor": "fb-671168615592", "service": "fb", "text": "Note: I'm ideologically in favor of the maximal socialist position, i.e. 100% anti-selection, so that's my bias.
I think that single-payer health care is the only moral way to do health care, for instance, and would think so even if it weren't also true that it's simply less costly because the act of an insurance company trying to determine risk classes is itself a costly activity.", "timestamp": "1407352244"}, {"author": "opted out", "source_link": "#", "anchor": "unknown", "service": "unknown", "text": "this user has requested that their comments not be shown here", "timestamp": "1407352289"}, {"author": "Jeff Kaufman", "source_link": "https://www.facebook.com/jefftk/posts/671139179582?comment_id=671168870082", "anchor": "fb-671168870082", "service": "fb", "text": "@Arthur: \"The destabilizing effect of everyone being able to throw in any risk factors they want into a pricing model would be on net harmful to the insurance industry as it exists\"
The existing insurance industry is used to working on a tightly limited set of information about applicants, and they're pretty good at this which sets up a barrier to entry, so they're happy.
But if we got rid of these restrictions and then new \"disruptive\" companies would be able to more accurately predict risk and decrease charges for the less risky applicants. In some cases like workers comp this is obviously good and in others like fire and flood it's mostly good. People would switch around how they did things so that they were less exposed to risk in the situations where the risk was not necessary, for example by being less likely to build in places that would flood.", "timestamp": "1407352496"}, {"author": "Arthur", "source_link": "https://www.facebook.com/jefftk/posts/671139179582?comment_id=671168964892", "anchor": "fb-671168964892", "service": "fb", "text": "In a complicate way I feel like all of this \"Let's maximize free-market libertarianism AND have a guaranteed minimum income paid for by redistributive taxes\" stuff is very appealing to nerds for exactly the same reasons it never actually happens.
It's a lot like Isaac Asimov's deeply flawed argument that of course we'll have humanoid robots, because making one robot with a human-like brain and body is going to be so much easier than putting a different, separate computer in every single machine.
The robot nerds greatly overestimated the ease of making one human-like smart robot, while underestimating how easy it would eventually be to computerize most tools because individual computers are so cheap.
Just like a lot of policy nerds hold out this dream for a Fundamentally Just Economic System that simply can't and won't happen, while also being opposed to tweaking individual pieces of the economy to make each individual one of them more just because they overestimate how hard that is or just have some vague aesthetic objection to things being \"kludgey\".
(Cf. the annoying argument that \"You shouldn't pressure this one company to raise its wages in the absence of fixing all of capitalism!\" when pressuring one company after another to raise its wages is the only thing we've ever done to \"fix capitalism\" that actually got anywhere.)", "timestamp": "1407352565"}, {"author": "Arthur", "source_link": "https://www.facebook.com/jefftk/posts/671139179582?comment_id=671169029762", "anchor": "fb-671169029762", "service": "fb", "text": "Anyway that's my philosophical defense of why I'm in favor of the thing that all the nerds on the Internet regardless of Left or Right affiliation seem to be against, i.e. a giant bureaucracy with lots of little individual regulations that don't try to be particularly elegant or consistent", "timestamp": "1407352616"}, {"author": "Jeff Kaufman", "source_link": "https://www.facebook.com/jefftk/posts/671139179582?comment_id=671169064692", "anchor": "fb-671169064692", "service": "fb", "text": "@Arthur: \"I think that single-payer health care is the only moral way to do health care, for instance\"
I don't think health care is something we should set up through insurance because people have very little control over their health costs.", "timestamp": "1407352623"}, {"author": "Jeff Kaufman", "source_link": "https://www.facebook.com/jefftk/posts/671139179582?comment_id=671169369082", "anchor": "fb-671169369082", "service": "fb", "text": "@Elliot: \"Problem is the redistribution never actually happens.\"
So we should fix this by lobbying for redistribution then. Luckily I'm in favor of redistribution even if we don't change how insurance works, so it's not a case where we can only do one if we do both of them.
(And in cases where insurance is currently redistributing expected value we should not relax that without increasing redistribution more overtly, unless it's currently just horribly inefficient or something.)", "timestamp": "1407352779"}, {"author": "Arthur", "source_link": "https://www.facebook.com/jefftk/posts/671139179582?comment_id=671169668482", "anchor": "fb-671169668482", "service": "fb", "text": "Usually when we get \"cleaner\" or more \"comprehensive\" social-safety-net programs it's because those programs step in to take over the job of a kludgey mass of regulations, because people got used enough to the guarantees those regulations represented they were unwilling to give them up even when the kludginess hit critical mass.
Avoiding kludges from the outset and designing systems to be elegant from the start simply doesn't work. I'd argue it doesn't even actually accurately describe how things happen in software engineering, which is the metaphor we're using, much less in politics.
For instance, the Canadian single-payer healthcare system is arguably not insurance anymore, but it's still thought of as insurance because that's what it started out as.
Most Americans who criticize Obamacare are not aware that Canada only has single-payer because it used to have private health insurers who found themselves bound by an increasing morass of Obamacare-like regulations, guarantees, mandates, etc. and ultimately they decided it would be cheaper to abandon their assets to the government than to keep on trying to make a profit while following the law.
The only path to there is through here, in other words -- I'm annoyed by Obamacare as much as anyone else but anyone who thinks there's any possibility of getting to something better by backing away from Obamacare instead of going forward through it is deluded.", "timestamp": "1407353021"}, {"author": "Jeff Kaufman", "source_link": "https://www.facebook.com/jefftk/posts/671139179582?comment_id=671171849112", "anchor": "fb-671171849112", "service": "fb", "text": "@Arthur: You're saying that to get to a desired simple generous policy, you should push for tweaks to existing policies that make them inefficiently provide benefits similar to what the desired policy would provide, and then at some point we can get fed up, throw the whole mess away, and implement the desired policy. And with health \"insurance\" I can totally see this happening: you make the system more complicated and restrictive, getting rid of insurers' ability to charge based on risk until \"throw the whole thing away and go single payer\" is clearly better.
And I can also see this happening for guaranteed minimum income: expand welfare, make it complex, make it confusing, give people $X to spend on fruit and $Y to spend on vegetables and $Z to spend on bedsheets and then when people get fed up we can throw it away and just pay people $W/month and be done with it.
In cases where the desired policy is a clean \"government provides benefit\" then it seems like it could work. But it seems like it can only push us in that direction. Which is fine if you're always in favor of the \"maximal socialist position\", but there are a lot of cases where I don't think that's a very good outcome, like this insurance example that started us off. I *want* more dangerous drivers to be charged more for car insurance so they drive less.", "timestamp": "1407353780"}, {"author": "Arthur", "source_link": "https://www.facebook.com/jefftk/posts/671139179582?comment_id=671172023762", "anchor": "fb-671172023762", "service": "fb", "text": "I'd actually argue that the welfare state and the insurance industry have a really closely tied shared history, and that philosophically speaking the for-profit insurance model is a kind of stepping stone to an actual social safety net.
Insurers aren't acting out of any kind of altruism, but both market forces and policy are kludgily steering them in that direction, and in our particular culture our social welfare programs are going to come out of the idiom of insurance and cloaked as insurance. (Look at how we still pretend Social Security is an annuity rather than a transfer.)
The overall trend in the insurance industry is really toward *less* grouping by risk class, not more, even though it's not a straight linear progression.
This is partly because of policy, like Obamacare, and also partly because of the forces in the industry that push companies to seek competitive advantage simply by having an ever-bigger pool of insureds and \"make it up on volume\" with minimum overhead -- and doing more underwriting both reduces your volume and adds to your overhead.
I mean, this frequently goes too far to be sustainably profitable -- there was a mini-collapse in the insurance industry in the '90s that Warren Buffett blamed on \"irrational exuberance\" and \"underwriters forgetting their job is to actually underwrite\" -- but it's still where the overarching trend is going.
Look at how many insurers used to have a brand identity built around narrowly targeting their product to one very specific risk class who aren't doing that anymore. GEICO tehcnically stands for Government Employees Insurance Company, etc.
For that matter if you want to get all historical the fact that there's a commercial \"insurance industry\" is itself a massive ratcheting-down of expectations about risk management.
Before there were insurance companies there were fraternal organizations, who pooled risk on the understanding that they were doing so among people they actually personally knew, among whom they had strong social bonds, who were carefully evaluated on a personal level before joining and had to make a strong personal/social commitment to the community to join the risk pool, and who were subject to ongoing social pressure to act in the good of the group.
Compared to that level of \"risk management\" a big corporate underwriting department is practically doing nothing at all. And yet commercialized insurance has rapidly displaced the local fraternal organization as our preferred risk-pooling method -- even people who are still in the Freemasons get their insurance from Progressive.
(That's not just a comparison I pulled out of my ass, by the way, the first chartered life insurance company in the US was a spinoff of the Freemasons. The Acacia Life Assurance Company used the acacia tree, a Masonic symbol, as their trademark and for a while in America the acacia tree was therefore synecdoche for insurance. That's a cool tidbit for any conspiracy theorists out there)", "timestamp": "1407353830"}, {"author": "Jeff Kaufman", "source_link": "https://www.facebook.com/jefftk/posts/671139179582?comment_id=671172857092", "anchor": "fb-671172857092", "service": "fb", "text": "@Arthur: I wonder if some of this progression over time is that historically we haven't been able to collect enough data and do statistics well enough to actually turn that data into good risk predictions? Or that the profits from this have been much lower than the profits from simply signing up more people? But \"big data\" changes this?", "timestamp": "1407354182"}, {"author": "Arthur", "source_link": "https://www.facebook.com/jefftk/posts/671139179582?comment_id=671173181442", "anchor": "fb-671173181442", "service": "fb", "text": "It could well be, though I think as long as buying insurance is a voluntary choice the boost in profitability you get from being high-volume by having very limited underwriting is still significant. (No one's going to buy a product that promises \"significantly lower rates!\" if they, personally, are pretty uncertain if they'll even be approved when they sign up for it.)
But yeah, Big Data might be a countervailing current in this trend, though Big Data so far hasn't lived up to most of its promises and also you're going to face basic social resistance to full deployment of Big Data.
The joke used to be that underwriters for auto insurance would want to sit in the backseat of your car for a week before issuing you a policy, but now they actually can, by making you download a motion-sensing app on your iPhone. The problem -- or blessing, depending on your POV -- is that the law makes it very hard for them to actually require you to do this.", "timestamp": "1407354376"}, {"author": "Jeff Kaufman", "source_link": "https://www.facebook.com/jefftk/posts/671139179582?comment_id=671173640522", "anchor": "fb-671173640522", "service": "fb", "text": "@Arthur: \"No one's going to buy a product that promises 'significantly lower rates!' if they, personally, are pretty uncertain if they'll even be approved when they sign up for it.\"
In future-everything-public-world you could see an ad targeted only at you with exactly the rate you would be charged if you signed up.
\"though Big Data so far hasn't lived up to most of its promises and also you're going to face basic social resistance to full deployment of Big Data.\"
True. Though some of this is you mostly just hear about the failures, when a big company jumps on silly buzzwords and fails miserably.", "timestamp": "1407354615"}, {"author": "Arthur", "source_link": "https://www.facebook.com/jefftk/posts/671139179582?comment_id=671173710382", "anchor": "fb-671173710382", "service": "fb", "text": "AFAICT the \"motion-sensing app\" idea is only even being implemented as a completely voluntary way to apply for a discount in Europe and the UK. Over here they let you download it just for shits and giggles but no state government has allowed them to actually use it (and the ability to still download it and be told \"This is how much you COULD save!\" is a pretty blatant, aggressive bit of political advertising).", "timestamp": "1407354661"}, {"author": "Arthur", "source_link": "https://www.facebook.com/jefftk/posts/671139179582?comment_id=671173989822", "anchor": "fb-671173989822", "service": "fb", "text": "\"In future-everything-public-world you could see an ad targeted only at you with exactly the rate you would be charged if you signed up.\"
That would require them to get all the info they need about me BEFORE I say I'm interested in their product and explicitly give them permission to get that info. Which is the whole problem. (Or, again, blessing.)", "timestamp": "1407354710"}, {"author": "Jeff Kaufman", "source_link": "https://www.facebook.com/jefftk/posts/671139179582?comment_id=671174054692", "anchor": "fb-671174054692", "service": "fb", "text": "@Arthur: If a \"completely voluntary way to apply for a discount\" becomes widespread enough then it's de-facto mandatory. You have the expensive insurance no one buys unless they have to, and the cheap insurance that everyone has to opt into if they want reasonable rates.", "timestamp": "1407354749"}, {"author": "Arthur", "source_link": "https://www.facebook.com/jefftk/posts/671139179582?comment_id=671174234332", "anchor": "fb-671174234332", "service": "fb", "text": "One reason insurance sales is such a fraught field is that sales and underwriting are fundamentally at odds with each other, and the law forces them to be married in a pretty uncomfortable way.
I can't legally grab the info from you I need to underwrite you until I've already convinced you to sign a contract giving me permission, by convincing you to buy the insurance. But I can't actually know whether the insurance will be a good bargain for you -- or if you can buy it at all -- until I underwrite you. It's a tough catch-22.", "timestamp": "1407354823"}, {"author": "Jeff Kaufman", "source_link": "https://www.facebook.com/jefftk/posts/671139179582?comment_id=671174259282", "anchor": "fb-671174259282", "service": "fb", "text": "@Arthur: \"That would require them to get all the info they need about me BEFORE I say I'm interested in their product and explicitly give them permission to get that info.\"
Yes. I think there's a large chance of this happening. It's more likely that we could prevent the companies from using this information from charging you different amounts than we could legally prevent people from being able to gather it.", "timestamp": "1407354862"}, {"author": "Jeff Kaufman", "source_link": "https://www.facebook.com/jefftk/posts/671139179582?comment_id=671174359082", "anchor": "fb-671174359082", "service": "fb", "text": "@Arthur: \"I can't legally grab the info from you I need to underwrite you until I've already convinced you to sign a contract giving me permission, by convincing you to buy the insurance\"
Which is an example of how current regulations on what information insurers can have (no pulling credit reports to make quotes to serve ads) make the insurance market less efficient.", "timestamp": "1407354943"}, {"author": "Arthur", "source_link": "https://www.facebook.com/jefftk/posts/671139179582?comment_id=671174488822", "anchor": "fb-671174488822", "service": "fb", "text": "Well, right now whether or not there's someone looking at your credit score is part of the algorithm they use to generate the credit score (too many credit pulls means your credit is being questioned a lot for some reason which makes you riskier). \"Everything public all the time\" would disrupt the system on a very fundamental level, not just in the way you're naming.", "timestamp": "1407355051"}, {"author": "Jeff Kaufman", "source_link": "https://www.facebook.com/jefftk/posts/671139179582?comment_id=671174833132", "anchor": "fb-671174833132", "service": "fb", "text": "@Arthur: \"whether or not there's someone looking at your credit score is part of the algorithm they use to generate the credit score\"
Which is nuts. Someone looking at your credit score because they're trying to decide whether to offer you a job makes you more creditworthy. Even marking you as less creditworthy because lenders are looking at your credit score only makes sense because our loans take a long time to be issued and run through the system, and this is a hack to keep people from applying for 100 loans at once and then getting approved for all 100 as if they were the only loan you were currently applying for.", "timestamp": "1407355285"}, {"author": "Arthur", "source_link": "https://www.facebook.com/jefftk/posts/671139179582?comment_id=671176175442", "anchor": "fb-671176175442", "service": "fb", "text": "Anyway a strong argument against credit checks as a rating mechanism -- it's banned in some states but not others, part of the fun regulatory patchwork insurance lives in -- is that credit bureaus are already extremely secretive, shadowy, unelected entities with immense power to destroy individual people's lives, if not through malice then through simple incompetence. (Look at how there's a cottage industry basically shaking people down for protection money -- paying hundreds of dollars a month to make sure there's no random errors on your credit report.)
The more stuff you use credit scores for, the more pressure you put on the credit bureaus to be right and the more damage they do when they're wrong -- and the more incentive they have to profit by being evil.
There's a real danger in trusting any individual actor in a competitive free-market system to set the terms by which everyone else competes in that system.
Part of why we place a limitation on what insurers can pull data from for rating is so that we place a limitation on how many points of failure the system has, how many institutions we have to police for fraud. (The medical profession has way, way more social and legal controls placed on who doctors can talk to about what than the vast majority of other professions. Health insurance fraud on both the underwriting and claims side is still rife.)
The problem with any form of benchmarking is the Heisenberg effect, that once it becomes known that it is a benchmark this fact affects the measurement process and leaves it vulnerable to manipulation.
This is true even when the benchmarking is the point of the benchmark and its accuracy is all that the company has to sell. It's an open secret if you ask anyone at all in the advertising and media industry that the Nielsen company is incredibly, ridiculously corrupt -- in bed in every way with the Big Media companies that are much richer than it that it purports to measure -- and yet everyone remains obsessed with the Nielsen benchmark because it's there and the system was built around it.
Look at how badly Moody's and S&P's securities ratings systems' credibility was tarnished by the fact that those ratings systems pretty much outright lied in a way that outright caused the 2009 financial collapse -- and did so blatantly in order to profit from near-term kickbacks from the entities they were rating -- but we're all still using them for lack of a better alternative.
If this is a risk that comes about with benchmarking institutions that were explicitly created for that purpose then it's a million times worse when you take a benchmark from somewhere else and repurpose it to put more weight on it. Credit rating was scarily fraught enough when it just determined whether you could take out a home loan -- now people are tracking down credit scores to evaluate each other's suitability as a romantic partner.
It may be a frivolous example, but this is exactly why sports journalists demanded that the BCS stop using journalist polling in order to determine seasonal standings. Turning a journalist's opinion into an *official* opinion that has real, concrete impact on what happens to a team makes it impossible for that opinion to be unbiased anymore.
It politicizes the giving of the opinion, it politicizes the act of getting a job as a journalist in order to offer that opinion, it creates a massive incentive for corruption/bribery/intimidation where it didn't exist before, and it pretty much ruins the perceived objectivity of the journalistic rankings that was the reason to make the poll into a benchmark in the first place.
I feel like this is a principle that has wide application elsewhere. In a competitive society it's very hard for a system of measurement to remain both competent and non-evil when you make it Important.", "timestamp": "1407356118"}, {"author": "Jeff Kaufman", "source_link": "https://www.facebook.com/jefftk/posts/671139179582?comment_id=671176434922", "anchor": "fb-671176434922", "service": "fb", "text": "@Arthur: Isn't this an argument for switching from \"independent\" ratings agencies to doing this calculation in-house? My understanding is insurance companies use credit ratings because they have to, but they'd much rather collect data from as many sources as they could and use that to make predictions.", "timestamp": "1407356327"}, {"author": "Arthur", "source_link": "https://www.facebook.com/jefftk/posts/671139179582?comment_id=671176634522", "anchor": "fb-671176634522", "service": "fb", "text": "This thing I'm calling the Heisenberg principle is also part of the clustering effect you're talking about. Risk analysis can only be objective if the people doing that analysis have no power to, themselves, affect the risk. Incentivizing being right only sometimes incentivizes accuracy -- other times it incentivizes interfering in order to make yourself right.
Take the stock market, and the argument that stock prices are a valuable signal of the inherent soundness of a company, and that short selling is an indispensable tool for communicating judgments about a company's unsoundness.
That's all well and good except that a short-seller's massive economic incentive to be right about their pessimism about a company gives them a massive incentive to *cause* themselves to be right if they aren't already. And, conveniently, sending negative signals about a company can and do affect that company's actual financial soundness. It's an ugly and vicious cycle -- look up the history of Wall Street \"bear chases\" to learn more.
This is why Wall Street finance reporters or analysts are required to disclose the position they have in companies they report on, but this is very difficult to enforce -- especially when you have big octopuses with their tentacles in a lot of pies (an analyst works for a company that has an unwritten deal to get some future financial benefit from an investor who has a short position in the company she's attacking).
Even when they do disclose it's not always easy to separate chicken from egg -- I mean, shorting a stock because you genuinely do think the company is overvalued and then telling people why you think it's overvalued is exactly what an honest, consistent actor would do, and it's also exactly what an evil person out to get that company would do.", "timestamp": "1407356443"}, {"author": "opted out", "source_link": "#", "anchor": "unknown", "service": "unknown", "text": "this user has requested that their comments not be shown here", "timestamp": "1407356935"}, {"author": "Arthur", "source_link": "https://www.facebook.com/jefftk/posts/671139179582?comment_id=671177482822", "anchor": "fb-671177482822", "service": "fb", "text": "It's not my personal field, but my sense is that mortgages are a much less automated field than, say, the volume dealt with by auto insurers.", "timestamp": "1407357005"}, {"author": "Arthur", "source_link": "https://www.facebook.com/jefftk/posts/671139179582?comment_id=671177587612", "anchor": "fb-671177587612", "service": "fb", "text": "Maybe it's the left-wing anti-credit-bureau memes I marinate in all day but I'm intensely paranoid about my credit score, check it often, go out of my way to do things that will be positive for it (like reporting my rent payments to Experian), etc.", "timestamp": "1407357064"}, {"author": "opted out", "source_link": "#", "anchor": "unknown", "service": "unknown", "text": "this user has requested that their comments not be shown here", "timestamp": "1407357223"}, {"author": "Arthur", "source_link": "https://www.facebook.com/jefftk/posts/671139179582?comment_id=671179369042", "anchor": "fb-671179369042", "service": "fb", "text": "I'm perfectly willing to TAKE risks with my credit, which is what travel hacking is all about, but on principle the idea of having my credit dinged for something that I didn't actually do/didn't actually happen really pisses me off", "timestamp": "1407358079"}, {"author": "opted out", "source_link": "#", "anchor": "unknown", "service": "unknown", "text": "this user has requested that their comments not be shown here", "timestamp": "1407358738"}, {"author": "Arthur", "source_link": "https://www.facebook.com/jefftk/posts/671139179582?comment_id=671181025722", "anchor": "fb-671181025722", "service": "fb", "text": "That's still not as crazy/weird as the story of Fairfax Financial, where the hedgies who shorted it ended up pulling a Nixon and hiring a mentally unstable lowlife \"fixer\" to stalk/harass the CEO and associates to try to get the company to \"crack\"", "timestamp": "1407358913"}, {"author": "Arthur", "source_link": "https://www.facebook.com/jefftk/posts/671139179582?comment_id=671181849072", "anchor": "fb-671181849072", "service": "fb", "text": "One of the most important reasons to avoid ambition and social climbing is to never end up in a position of authority where powerful entities will have some vested financial interest in ensuring that you, as an individual and on a personal level, \"crack\"", "timestamp": "1407359448"}, {"author": "David Chudzicki", "source_link": "https://www.facebook.com/jefftk/posts/671139179582?comment_id=671182173422", "anchor": "fb-671182173422", "service": "fb", "text": "It sounds like Arthur (and Elliot?) are questioning the usefulness (but not necessarily the correctness) of Jeff's advocating more redistribution paired with allowing insurance/credit discrimination that arguably makes sense with and only with more redistribution. It's true that you probably can't literally pair those things together in today's legislative environment, but I don't think that's the point. More to the point would be debating whether there's value in building a market-oriented coalition that also likes wealth redistribution.", "timestamp": "1407359567"}, {"author": "David Chudzicki", "source_link": "https://www.facebook.com/jefftk/posts/671139179582?comment_id=671182837092", "anchor": "fb-671182837092", "service": "fb", "text": "Arthur \"All insurance is fundamentally a redistributive act, taking money away from some people who don't need it and giving it to people who do need it\":
I think you're making a mistake in not distinguishing more the *kinds* of redistribution involved. You (and I) like redistribution from rich to poor, from systematically privileged to systemically disadvantaged, etc. Not letting insurers use some factor in pricing fundamentally just redistributes from risky to less risky. That may happen to coincide with the kind of redistribution you like, that's not guaranteed and it will differ on a case-by-case basis.", "timestamp": "1407359938"}, {"author": "David Chudzicki", "source_link": "https://www.facebook.com/jefftk/posts/671139179582?comment_id=671182886992", "anchor": "fb-671182886992", "service": "fb", "text": "Arthur \"Note: I'm ideologically in favor of the maximal socialist position, i.e. 100% anti-selection, so that's my bias.\"
Would you really want the state to just cover everyone's car-related liability (perhaps with a larger bureaucracy deciding who's allowed to drive at all? If so, we're way further apart on this stuff than I'd ever imagined. At the moment I won't say any of the reasons I think that'd be terribly, since even we disagree vehemently you're more than smart enough to predict the first few things I'd say.", "timestamp": "1407359955"}, {"author": "opted out", "source_link": "#", "anchor": "unknown", "service": "unknown", "text": "this user has requested that their comments not be shown here", "timestamp": "1407360159"}, {"author": "David Chudzicki", "source_link": "https://www.facebook.com/jefftk/posts/671139179582?comment_id=671183246272", "anchor": "fb-671183246272", "service": "fb", "text": "Arthur \"Look at how many insurers used to have a brand identity built around narrowly targeting their product to one very specific risk class who aren't doing that anymore. GEICO tehcnically stands for Government Employees Insurance Company, etc.\" <== My sense from a couple years in the industry is that this is largely just a combination of (a) what Jeff said (we can distinguish prices better now, so might as well open it to everyone), and (b) everything starts small, and if you're starting small it may be a good idea to start out marketing to particular groups.", "timestamp": "1407360224"}, {"author": "David Chudzicki", "source_link": "https://www.facebook.com/jefftk/posts/671139179582?comment_id=671183460842", "anchor": "fb-671183460842", "service": "fb", "text": "Elliot: Yes, I agree with you on that one. I wouldn't want insurance companies explicitly distinguishing based on race regardless of what redistribution I could pair that with. (Implicitly distinguishing on race is a much more complicated matter.) We may be disagreeing with Jeff's OP on this.", "timestamp": "1407360360"}, {"author": "David Chudzicki", "source_link": "https://www.facebook.com/jefftk/posts/671139179582?comment_id=671183740282", "anchor": "fb-671183740282", "service": "fb", "text": "Arthur said \"the boost in profitability you get from being high-volume by having very limited underwriting is still significant\": I assume \"underwriting\" here means \"decisions about who you won't insure at all\". In that case, the ideal is no underwriting: You'll insure anyone, if you can charge them the right price. My understanding is companies have been moving in that direction.", "timestamp": "1407360514"}, {"author": "David Chudzicki", "source_link": "https://www.facebook.com/jefftk/posts/671139179582?comment_id=671184094572", "anchor": "fb-671184094572", "service": "fb", "text": "Arthur said 'AFAICT the \"motion-sensing app\" idea is only even being implemented as a completely voluntary way to apply for a discount in Europe and the UK'
I don't know anything about the motion-sensing app, but FWIW several companies are moving forward with the devices you plug into your car to report telematic data. I was involved in that project briefly before I left the industry a couple years ago.
It'll always be \"optional\" of course, but in my view it would eventually get such widespread adoption that the non-telematics version would essentially be a market for lemons, and prices would become too high for most people.", "timestamp": "1407360701"}, {"author": "David Chudzicki", "source_link": "https://www.facebook.com/jefftk/posts/671139179582?comment_id=671184483792", "anchor": "fb-671184483792", "service": "fb", "text": "Jeff \"My understanding is insurance companies use credit ratings because they have to, but they'd much rather collect data from as many sources as they could and use that to make predictions.\"
For the most part insurance companies don't use the actual credit ratings. They develop their own version based on the same sorts of data that's more tailored to the particular purpose. You could imagine directly using that data in the model instead, but I think the biggest reason to still make it a \"rating\" relates to credit being such a touchy regulatory subject.", "timestamp": "1407360890"}, {"author": "David Chudzicki", "source_link": "https://www.facebook.com/jefftk/posts/671139179582?comment_id=671184578602", "anchor": "fb-671184578602", "service": "fb", "text": "Sorry for the barrage, everyone!", "timestamp": "1407360963"}, {"author": "opted out", "source_link": "#", "anchor": "unknown", "service": "unknown", "text": "this user has requested that their comments not be shown here", "timestamp": "1407363676"}, {"author": "David Chudzicki", "source_link": "https://www.facebook.com/jefftk/posts/671139179582?comment_id=671193954812", "anchor": "fb-671193954812", "service": "fb", "text": "Yes, of course... in some cases (health) the ideal is more like a single-payer universal health care. It's sad you couldn't get health insurance, and I'm glad we have this goofy alternative situation that accomplishes some of the same goals single payer would. (Hopefully it helps you out.)", "timestamp": "1407363866"}, {"author": "David Chudzicki", "source_link": "https://www.facebook.com/jefftk/posts/671139179582?comment_id=671194044632", "anchor": "fb-671194044632", "service": "fb", "text": "(I'm also glad Arthur pointed out that such a goofy system was one step toward Canada's single payer. I didn't know that.)", "timestamp": "1407363903"}, {"author": "David Chudzicki", "source_link": "https://www.facebook.com/jefftk/posts/671139179582?comment_id=671194174372", "anchor": "fb-671194174372", "service": "fb", "text": "... but that's just a weird example b/c no one in this thread thinks it should really be an insurance market at all. But as far as I know, everyone in this thread agrees with having a private market for auto insurance. In response to your question above, I already named a specific example (not-at-fault accidents) of a rule where I think the price should be allowed to more closely match the risk.", "timestamp": "1407364040"}, {"author": "opted out", "source_link": "#", "anchor": "unknown", "service": "unknown", "text": "this user has requested that their comments not be shown here", "timestamp": "1407387335"}, {"author": "opted out", "source_link": "#", "anchor": "unknown", "service": "unknown", "text": "this user has requested that their comments not be shown here", "timestamp": "1407389852"}, {"author": "David Chudzicki", "source_link": "https://www.facebook.com/jefftk/posts/671139179582?comment_id=671243460602", "anchor": "fb-671243460602", "service": "fb", "text": "Yeah. It's a good general rule that you'll make more money the more accurate your pricing model. Because of what Elliot said and for similar reasons with less extreme examples.", "timestamp": "1407391537"}, {"author": "David Chudzicki", "source_link": "https://www.facebook.com/jefftk/posts/671139179582?comment_id=671243470582", "anchor": "fb-671243470582", "service": "fb", "text": "That was my whole job for a couple years.", "timestamp": "1407391551"}, {"author": "David Chudzicki", "source_link": "https://www.facebook.com/jefftk/posts/671139179582?comment_id=671243525472", "anchor": "fb-671243525472", "service": "fb", "text": "A good simplification: For any given customer, if you correctly price them lower than your competitors, you win the business and make more money. (yes, your margins are lower than your competitor's would have been, but the counterfactual isn't that you get those margins. It's that your competitor does).", "timestamp": "1407391701"}, {"author": "David Chudzicki", "source_link": "https://www.facebook.com/jefftk/posts/671139179582?comment_id=671243545432", "anchor": "fb-671243545432", "service": "fb", "text": "If you incorrectly price them lower than your competitors, you lose money.", "timestamp": "1407391755"}, {"author": "David Chudzicki", "source_link": "https://www.facebook.com/jefftk/posts/671139179582?comment_id=671243615292", "anchor": "fb-671243615292", "service": "fb", "text": "Either way, in a competitive market (and it is reasonably competitive), this is a drive to price everyone as precisely as possible.", "timestamp": "1407391835"}, {"author": "David Chudzicki", "source_link": "https://www.facebook.com/jefftk/posts/671139179582?comment_id=671243650222", "anchor": "fb-671243650222", "service": "fb", "text": "There are other factors at play, but that story paints a fairly accurate picture.", "timestamp": "1407391871"}, {"author": "David Chudzicki", "source_link": "https://www.facebook.com/jefftk/posts/671139179582?comment_id=671243720082", "anchor": "fb-671243720082", "service": "fb", "text": "Wayne - - the point about losing statistical power is wrong. The way to get a lot of distinctions in price isn't as simple as making a lot of groups and using the mean for each group, however noisy that is. We had better statistical models.", "timestamp": "1407392078"}, {"author": "Arthur", "source_link": "https://www.facebook.com/jefftk/posts/671139179582?comment_id=671254398682", "anchor": "fb-671254398682", "service": "fb", "text": "Wang, I work for an insurer, though not in the actuarial field.
Insurers do save a lot of overhead when they don't have to underwrite individuals, which is the whole point of group insurance. But they have to prevent adverse selection into the group, which is the core problem of group underwriting (and why it has to be a \"true group\", a group that you can't join just to get insurance).", "timestamp": "1407413305"}]}