Hacker News Board Hacker News, browsed like a classic message board
Live data from Hacker News

Want to be an actuary? Odds are, you’ll fail the test

wsj.com

Post reply on HN
Print view View original

301–310 of 349 posts

Re: Want to be an actuary? Odds are, you’ll fail the test

#301

Post by ogogmad »

HighChaparral wrote: post #288
ogogmad wrote: post #132

Earlier quoted context omitted.

In 3 steps: I. P(C) = 2 P(D) II. 0.15 = P(C and D) = P(C) P(D) {by independence} = 2 P(D)^2 {by I} implies P(D) = 0.27 III. P(~C and ~D) = P(~C) P(~D) {by independence} = (1 - P(C))(1 - P(D)) = 0.33 {by I and II}

Nicely laid out. I goofed on the last bit by assuming the final answer was: 1 - P(C) - P(D) which gives 0.178 ~ 0.18 or answer (A)

Worth remembering that:

  I.  P(A and B) = P(A) P(B)   {when independent}
  II. P(A or  B) = P(A) + P(B) {when mutually exclusive}
But in general:

  III.  P(A and B) = P(A) P(B|A)
  IIII. P(A or  B) = P(A) + P(B) - P(A and B)
The only way that I and II can both be true is that either P(A) = 0 or P(B) = 0. This does not imply though that A and B are mutual exclusive and independent, but it is a necessary condition. The conclusion is that if you're using both I and II, you're almost certainly doing something wrong.

Re: Want to be an actuary? Odds are, you’ll fail the test

#302

Post by throw0101a »

romwell wrote: post #59
asdfasgasdgasdg wrote: post #54

Earlier quoted context omitted.

While corruption may have been the source of this particular business, the fact that insurance is required by law is a good thing. The counterfactual involves a bunch of uninsured judgement proof drivers clogging the road and ruining everyone else's day/life. The problem is already bad enough now when it is illegal to do.

Hot take: if it's something that's universally a good thing that should be required of every driver, it should be provided by the state and paid for with taxes / annual registration fees. Otherwise, it's just an income-generating scheme for private parties. Oh, and fun fact: the liability insurance isn't required to pay up anything unless the covered driver loses in court . They don't have to follow that the police r…

> Hot take: if it's something that's universally a good thing that should be required of every driver, it should be provided by the state and paid for with taxes / annual registration fees.

Which it is in some jurisdictions, e.g., BC, Quebec. In the jurisdictions where it is done by the private sector, it is generally heavily regulated.

> Oh, and fun fact: the liability insurance isn't required to pay up anything unless the covered driver loses in court.

This depends on the jurisdiction: in the province of Ontario, with is a "no fault" area, there are few(er) court cases:

> Ontario has a "no-fault" car insurance system, but this does not mean that no one is at fault in an accident. The term "no-fault" insurance simply means if you are injured or your car is damaged in an accident, then you deal with your own insurance company, regardless of who is at fault. You don't have to go after the at-fault driver for compensation.

* https://mitchellandabbott.com/no-fault-insurance.php

> Essentially, no-fault insurance in Ontario is that in the event of an accident (without or without collision coverage), all drivers involved will process individual claims through their own insurance companies to get coverage for damages and injuries. It’s a system that has prioritized the claims process for drivers who need reimbursement so that the drivers aren’t kept waiting.

* https://www.ahainsurance.ca/car-insurance/no-fault-insurance...

* https://en.wikipedia.org/wiki/No-fault_insurance

There's even a specific regulation called "Fault Determination Rules" that covers the most common cases:

* https://www.ontario.ca/laws/regulation/900668

There are pros and cons for both public/government versus private insurance, and 'fault' versus no-fault insurance, regimes.

Re: Want to be an actuary? Odds are, you’ll fail the test

#303

Post by civilized »

solumos wrote: post #298
civilized wrote: post #294

Earlier quoted context omitted.

I've also worked with actuaries and my impression has been different. Some of them have some skills, but it's very hit-and-miss, and it's not clear to me that the average actuary would compete with the average Master's in Data Science / Stats / CS in terms of effectiveness as a data scientist. Their big weakness is computational/programming skills. Many of them are simply unable to function outside of Excel. The SOA…

this is true — as an analyst, I was severely reprimanded for writing data cleanup scripts since they couldn't be reviewed in vanilla Excel

Incredibly depressing. I hope you got out.

Re: Want to be an actuary? Odds are, you’ll fail the test

#304

Post by devoutsalsa »

romwell wrote: post #57
d136o wrote: post #39

In this [1] interview by Barry Ritholtz, Markel CEO essentially told the origin story of an insurance giant. Basically a family member was a lawmaker, passed a law that drivers of the newly invented cars (maybe horse buggies?) must be insured. They simultaneously started a company that sold insurance. Over time they learned to make money from investing their customers capital. It blew my mind at how candid he was abo…

Side story about how satisfied people can be with the mandated insurance. I was on coast-to-coast road trip when I got into an accident in TX. Someone was trying to pass me on the left in a turning lane, veered into the incoming lane, and then hit my vehicle's front wheels from the back. The police ticketed the other driver, and told me that insurance will surely take care of that on my side. "Wow, it's worse than we…

It sucks that you couldn’t recoup any compensation, but I’m not surprised at all. You don’t have a relationship with the other person’s insurance. You have a relationship with your insurance. Your insurance covers you, and your insurance may try to collect from the other driver’s insurance company.

Also, this is why auto insurance offers uninsured & underinsured riders (add ons). There’s no guarantee that some jackass who hits your car will have enough insurance or any insurance at all.

Source: was an insurance agent

Re: Want to be an actuary? Odds are, you’ll fail the test

#305

Post by ActualActuary »

civilized wrote: post #300
ActualActuary wrote: post #299

Earlier quoted context omitted.

I think it’s an unfair comparison. Of course the average actuary wouldn’t even come close to the average days scientist when it comes to data science skills. Just like the average doctor wouldn’t, the average lawyer wouldn’t, and the average banker wouldn’t. The problems they solve require very different skills. The problems I solve on a day to day basis have no need for data science. There are certain areas in my co…

The question was about whether actuarial credentials should be positive signals in an application to a data science job. And my answer is: all else being equal, having more credentials is arguably good. However, all else is not equal. Having invested in actuarial credentials is a fairly strong signal of very limited computational/programming skills which are critical to a data science career. So the net signaling val…

I agree completely with your last paragraph. If I wanted to be a data scientist I’d be a data scientist. I don’t want my actuarial organisation wasting time and money trying to market us as “data scientists” when we are not data scientists. It’s a completely different job. It’s almost comparable to the profession deciding that we should now market ourselves as lawyers because we also contribute to writing contracts and treaties.

I would argue though that being an actuary doesn’t automatically make you bad at anything outside of Excel.

I’m not a data scientist, but I’m sure if I decided to go down that route my skill set would put me in a good position.

I have a very good grounding in statistics and probability.

I use R/Python quite heavily, regularly building models from scratch. I work with GLMs, Copulas, Monte Carlo simulation etc.

I deal with big volumes of data and have to write efficient algorithms to deal with it.

Most of my skills didn’t come from the exam path, but doing the exams also didn’t make me bad at all those other things.

Just being an actuary obviously doesn’t mean you have those skills though, which is what I was getting at with my “hit and miss” comment.

Re: Want to be an actuary? Odds are, you’ll fail the test

#306

Post by seibelj »

namdnay wrote: post #273
seibelj wrote: post #258

Earlier quoted context omitted.

This is the same model of all licenses, which is inherited from the guild system common throughout all of human history (the government (monarchs, oligarchies, democracies) protect favored industries from competition). I put them on a spectrum: - Licenses that anyone can obtain by showing the requisite skills. This is the least nefarious, and the clearest example would be driver's licenses. In a perfect world anyone…

> Licenses that anyone can get, but are purely graded on a curve in order to protect existing members. Finally we get to the actuarial licenses! Similar licenses would be Michelin star restaurants and wine sommeliers. Only a very small community appreciates these licenses, but the value to those that have them is high via the artificial scarcity. In order to keep them fair they offer the test(s) to anyone, but grade…

I would argue that there should be way more Michelin-star restaurants than exist, but they artificially cap the supply to keep them rare. It's a form of the same thing.

Re: Want to be an actuary? Odds are, you’ll fail the test

#307

Post by civilized »

ActualActuary wrote: post #305
civilized wrote: post #300

Earlier quoted context omitted.

The question was about whether actuarial credentials should be positive signals in an application to a data science job. And my answer is: all else being equal, having more credentials is arguably good. However, all else is not equal. Having invested in actuarial credentials is a fairly strong signal of very limited computational/programming skills which are critical to a data science career. So the net signaling val…

I agree completely with your last paragraph. If I wanted to be a data scientist I’d be a data scientist. I don’t want my actuarial organisation wasting time and money trying to market us as “data scientists” when we are not data scientists. It’s a completely different job. It’s almost comparable to the profession deciding that we should now market ourselves as lawyers because we also contribute to writing contracts a…

> I would argue though that being an actuary doesn’t automatically make you bad at anything outside of Excel.

I agree, I just think it's statistically a net negative signal if conditioned only on years of experience. I would expect a generic technical BS or MS in a DS-relevant field to be more qualified than someone who passed actuarial exams, if the two are at a similar point in their careers.

That said, competent employers shouldn't be relying on unconditional signals, they should be interviewing and testing and getting more information on the candidate. For such employers, the signaling value of the actuarial credential ought to be neutral.

Re: Want to be an actuary? Odds are, you’ll fail the test

#308

Post by seibelj »

mch82 wrote: post #277
seibelj wrote: post #258

Earlier quoted context omitted.

This is the same model of all licenses, which is inherited from the guild system common throughout all of human history (the government (monarchs, oligarchies, democracies) protect favored industries from competition). I put them on a spectrum: - Licenses that anyone can obtain by showing the requisite skills. This is the least nefarious, and the clearest example would be driver's licenses. In a perfect world anyone…

Licenses actually mean a lot and are integral to many industries. Software isn’t traditionally regulated because society still doesn’t understand the danger of defective software. Buildings, electronics, phones/radios, vehicles, food, medicine, and serious industries are all licensed.

Some software is dangerous, like airplanes and medical software that aids in live surgeries. But the majority of software is not dangerous. You shouldn't need to be licensed to make a video game or browser extension.

Re: Want to be an actuary? Odds are, you’ll fail the test

#309

Post by gruez »

nwiswell wrote: post #250
gruez wrote: post #222

Earlier quoted context omitted.

>The solution is to change laws about redress. For example, statutory punitive damages and attorneys' fees in lawsuits against insurers where the final judgement is significantly different from the insurer's initial offer. 1. I you're supposed to have negotiated with the counterparty in good faith prior to a lawsuit. If not, that will be looked negatively by the judge. https://www.sbwllp.com/rule-68-offers-of-judgmen…

Part 1 is just a matter of public policy. This is all armchair theory anyway. Regarding Part 2, I actually don't see the problem. If the result is causing insurance companies to be more cautious about lowball offers when there's a range of reasonable outcomes, is that bad? The plaintiff still must succeed at proving damages, so it's not like that outcome is unforseeable. It does increase "settlement risk" for insurer…

>but if premiums increase because insurers are actually paying more in fair settlements and/or punitive judgements because they failed to pay a fair settlement, again I cannot see the problem.

But the problem is that in the example I given, the lowball option is fair given the probabilities/expected value.

Re: Want to be an actuary? Odds are, you’ll fail the test

#310

Post by lazide »

SilasX wrote: post #254
lazide wrote: post #244

Earlier quoted context omitted.

It seems more like the poster was complaining that actually following through was hard and a pain in the ass, therefore the game is rigged. Really it’s just they’re used to someone else doing all the hard work and don’t want to shell out or expend effort for when something doesn’t go smoothly. Having dealt with contractors and others who decided doing what they were supposed too was clearly too hard (and then making…

I don't know what you think you're adding to the conversation here. The point of the auto insurance system as mandated by governments, including the "required liability" part, is to rise a bit above (or a "smidge above" as you might say) this folksy "it is what it is" nihilism.

If no one pushes the button that is provided to do anything about it, what do they expect to happen?

The original comment seemed to be complaining that no one was pushing the button for them, and the other side wasn’t just doing something without them pushing the button.

Well duh.

Post reply on HN