OK. Let's talk UK GDP. First, the good. The UK economy began rebalancing in 2017, with manufacturing growing, and the current account deficit narrowing. The savings rate, which hit a low of 4% in the first quarter, also bounced back.
Now the bad.
The UK economy is still unbalanced, and we all exaggerate the effects of a rise in production. Let's start with Household Final Expenditure as a percentage of GDP.
(A quick segue. Consumer spending adds to the GDP number, even if we're borrowing to import Chinese goods. Imagine an iPhone enters the country, the iPhone? That too.)
UK 66.4 Italy 60.8 Spain 57.6 France 55.3 Germany 53.3 Netherlands 44.2
Because production, and specifically manufacturing, is a small proportion of our economy, changes to the consumer spending number will have a larger impact on GDP. By contrast, if Germany's manufacturing number disappoints, it's likely to have a very negative impact.
In Q4, consumer spending in the UK was bad.
Visa tracks spending (and does a pretty good job). They saw overall spending fall 1% year-over-year in December. Given consumer prices are up 2.7% y-o-y, that means December real spending fell 3.7%. Fortunately, the consumer spending mix is changing positively, with restaurants and bars (with the highest proportion of UK value add) rising, while those segments with the highest imported proportions (such as clothing and footwear) dropping the most.
Nevertheless, the biggest component of household consumption dropped in real terms. Manufacturing contributes just 10% to UK GDP. Even if manufacturing rose 20% in 2017 (which it didn't), that increase would still be swamped by the drop in consumer spending.
One final thought. Many people are using the strength of Q4 2016 (when the UK economy expanded 0.7%) to predict similar growth this year. This is what the Visa spending report had to say about that quarter: "Consumer spending continued to increase solidly in the run up to Christmas, rounding off the fastest growing quarter for two years. Average year-on-year growth for each month in Q4 2016 was 2.8%, twice as high as the average rates for Q2 and Q3."
My guess is that UK GDP for 2017 will end up at 1.6%, with upward revisions to the first quarter offset by weak consumer spending at the end of the year.
Re: insurance. The fact that women tend to have fewer accidents than men (which I do not dispute) should not be used as an excuse to charge individual men higher premiums. A very good male driver or very poor female driver get unfairly punished/rewarded by such a system.
It’s not reasonable to use characteristics of groups that you have no choice in whether you belong to them as a basis of price discrimination, especially for a product like car insurance that is essentially an compulsory purchase.
IMO that argument is entirely correct when it comes to interviewing people for jobs, determining guilt or innocence, handling university admissions etc. But, when it comes to pricing insurance, one is doing a rough and ready calculation, based on group characteristics, as well as individual records.
Re: insurance. The fact that women tend to have fewer accidents than men (which I do not dispute) should not be used as an excuse to charge individual men higher premiums. It’s not reasonable to use characteristics of groups that you have no choice in whether you belong to them as a basis of price discrimination, especially for a product like car insurance that is essentially an compulsory purchase.
Why is it not reasonable, particularly with an insurance model based on sharing the risk over an identifiable grouping? What are you suggesting, individual interviews?
Hands up if you are prepared more for your car insurance so that other people who are statistically more likely to claim can get a discount ?
For all those of you who do not recycle how would you cope with our local council (Conwy) agreeing to move to four weekly bin collections. And yes, we have had annual 5% council tax increases for years
I'm sure Dr Sally Baker wouldn't be at all surprised by that. Having said that, neither am I at all surprised by it. I could talk for hours about how poorly Conwy Council has operated over the years............
And having lived here for 53 years so could I but they have been good on recycling
Wow you've certainly seen it all there then, going back to the 1970's. I'm still asking my MP to see unredacted copies of the Jillings and Waterhouse reports - I live in hope............
Re: insurance. The fact that women tend to have fewer accidents than men (which I do not dispute) should not be used as an excuse to charge individual men higher premiums. It’s not reasonable to use characteristics of groups that you have no choice in whether you belong to them as a basis of price discrimination, especially for a product like car insurance that is essentially an compulsory purchase.
Why is it not reasonable, particularly with an insurance model based on sharing the risk over an identifiable grouping? What are you suggesting, individual interviews?
Hands up if you are prepared more for your car insurance so that other people who are statistically more likely to claim can get a discount ?
Out of interest did quote for the Muslim name first? In theory I could see the second application discriminated against.
Perhaps they priced one higher as there was potential fraud - Two apparently different people with the same date of birth, same occupation, same employer type, living at the same address, quoting for the same car they own, parking their car in the same place overnight, with the same claims history, same driving license period, same points and convictions, same accidents on the same dates, same marital status and just different names?
Re: insurance. The fact that women tend to have fewer accidents than men (which I do not dispute) should not be used as an excuse to charge individual men higher premiums. It’s not reasonable to use characteristics of groups that you have no choice in whether you belong to them as a basis of price discrimination, especially for a product like car insurance that is essentially an compulsory purchase.
Why is it not reasonable, particularly with an insurance model based on sharing the risk over an identifiable grouping? What are you suggesting, individual interviews?
Share the risk over groups that individuals have control over, such as criminal records. Individual pricing should be based on individual information.
Justifying makjng discriminatory decisions on individuals based on the the behaviour of the population at large is tantamount to tyranny.
Would you be happy with a tax system that charged pensioners a higher rate because they were statistically more likely as a whole to make use of state services?
For all those of you who do not recycle how would you cope with our local council (Conwy) agreeing to move to four weekly bin collections. And yes, we have had annual 5% council tax increases for years
I'm sure Dr Sally Baker wouldn't be at all surprised by that. Having said that, neither am I at all surprised by it. I could talk for hours about how poorly Conwy Council has operated over the years............
And having lived here for 53 years so could I but they have been good on recycling
Wow you've certainly seen it all there then, going back to the 1970's. I'm still asking my MP to see unredacted copies of the Jillings and Waterhouse reports - I live in hope............
Re: insurance. The fact that women tend to have fewer accidents than men (which I do not dispute) should not be used as an excuse to charge individual men higher premiums. It’s not reasonable to use characteristics of groups that you have no choice in whether you belong to them as a basis of price discrimination, especially for a product like car insurance that is essentially an compulsory purchase.
Why is it not reasonable, particularly with an insurance model based on sharing the risk over an identifiable grouping? What are you suggesting, individual interviews?
Hands up if you are prepared more for your car insurance so that other people who are statistically more likely to claim can get a discount ?
If Admiral are doing what they say they are doing which is taking into account appropriate risk factors, then sure. In the end all good drivers subsidise bad drivers regardless of premium.
If they are basing it on surname, they have picked an absurd metric. They deny this and we await a proper explanation.
Out of interest did quote for the Muslim name first? In theory I could see the second application discriminated against.
Perhaps they priced one higher as there was potential fraud - Two apparently different people with the same date of birth, same occupation, same employer type, living at the same address, quoting for the same car they own, parking their car in the same place overnight, with the same claims history, same driving license period, same points and convictions, same accidents on the same dates, same marital status and just different names?
The chance of that is about one in a million!
Some people did the "western" name first and some the "Islamic" name first.
Re: insurance. The fact that women tend to have fewer accidents than men (which I do not dispute) should not be used as an excuse to charge individual men higher premiums. It’s not reasonable to use characteristics of groups that you have no choice in whether you belong to them as a basis of price discrimination, especially for a product like car insurance that is essentially an compulsory purchase.
Why is it not reasonable, particularly with an insurance model based on sharing the risk over an identifiable grouping? What are you suggesting, individual interviews?
Share the risk over groups that individuals have control over, such as criminal records. Individual pricing should be based on individual information.
Justifying makjng discriminatory decisions on individuals based on the the behaviour of the population at large is tantamount to tyranny.
Would you be happy with a tax system that charged pensioners a higher rate because they were statistically more likely as a whole to make use of state services?
Well yes actually but that is a whole other argument!
It seems to me the average teenage boy has more testosterone than he can handle, a need to show off and a higher propensity for high speed, serious injury crashes. If I am being asked to insure a teenage boy I am entitled to take that propensity into account even if the individual in question has not been shown to have those tendencies.
Would you be happy with a tax system that charged pensioners a higher rate because they were statistically more likely as a whole to make use of state services?
Well, they have paid more because they've worked for longer.
Re: insurance. The fact that women tend to have fewer accidents than men (which I do not dispute) should not be used as an excuse to charge individual men higher premiums. It’s not reasonable to use characteristics of groups that you have no choice in whether you belong to them as a basis of price discrimination, especially for a product like car insurance that is essentially an compulsory purchase.
Why is it not reasonable, particularly with an insurance model based on sharing the risk over an identifiable grouping? What are you suggesting, individual interviews?
Hands up if you are prepared more for your car insurance so that other people who are statistically more likely to claim can get a discount ?
If Admiral are doing what they say they are doing which is taking into account appropriate risk factors, then sure. In the end all good drivers subsidise bad drivers regardless of premium.
If they are basing it on surname, they have picked an absurd metric. They deny this and we await a proper explanation.
I've worked in insurance companies for and with pricing units, I have EMBLEM on my CV, and I'm puzzled as hell with this "name" thing."Name" as an explanatory/independent variable field is *never* used in a model, and for good reason: It's a categorical variable with hundreds of thousands of possible values that is famously difficult to validate semantically, can't be easily grouped using rules, and would make any model using it grind to a halt.
Re: insurance. The fact that women tend to have fewer accidents than men (which I do not dispute) should not be used as an excuse to charge individual men higher premiums. It’s not reasonable to use characteristics of groups that you have no choice in whether you belong to them as a basis of price discrimination, especially for a product like car insurance that is essentially an compulsory purchase.
Why is it not reasonable, particularly with an insurance model based on sharing the risk over an identifiable grouping? What are you suggesting, individual interviews?
Hands up if you are prepared more for your car insurance so that other people who are statistically more likely to claim can get a discount ?
If Admiral are doing what they say they are doing which is taking into account appropriate risk factors, then sure. In the end all good drivers subsidise bad drivers regardless of premium.
If they are basing it on surname, they have picked an absurd metric. They deny this and we await a proper explanation.
Why is surname an absurd metric ?
Sorry it's actually both surname and first name.
They don't directly relate to the underlying cause of risk.
If for example Leeds has a lot of accidents, and I live in Leeds, then using that fact bears more relation to the actual risk.
Using names must attempt to approximate cultural or ethnic backgrounds in which case you should be defending asking for and discriminating on them directly.
So there's your question: should white people get lower premiums?
Interesting analysis of the UK's EU contributions - after the rebate they were £235 million per week- even the gross figure before rebate is only £325 million per week:
Suffice there's a lot less dividend to play with than many in the thread below believe.
Oh no, only £235mn a week. What a shame.
The net figure of the last few years per this research is a whopping £137 million per week - FWIW the UK Gov budget is £15 billion a week - we're in rounding error territory here.
That's £7bn per year in unmatched spending. Some EU funding only comes if the government guarantees to match it. Add a billion for that and we're up to £8bn. I think if you said to people, would you rather give the EU £8bn per year or the NHS £8bn per year the answer would be fairly emphatic.
Gove's already promised farmers £3 billion per annum until 2024 with a new scheme to replace the existing method after this date. I'm sure there are a number of other vested interests who will be banging on the government's door for compensation for loss of EU funds.
If the NHS needs another £8 billion now there's nothing stopping us allocating these funds now - no need to wait till 2022.
That £3bn will replace EU CAP spending and not come from the saving out of the net contribution.
Where will the govt the farmers promised £3 Billion come from then?
From the gross payment to the EU - it's a like for like replacement.
There is roughly £18bn of gross payments, and £11bn of net payments. If the government was to replace all of the existing EU spending like for like then it would still have £11bn to spend on our priorities
Gove has already said that farm subsidies will be the same amopunt but allocated more on environmental grounds and less to large scale industrial farming.
It is not just how much but the criteria for receiving it. The EU CAP is all about supporting small to medium uneconomic (French) farmers.
Re: insurance. The fact that women tend to have fewer accidents than men (which I do not dispute) should not be used as an excuse to charge individual men higher premiums. It’s not reasonable to use characteristics of groups that you have no choice in whether you belong to them as a basis of price discrimination, especially for a product like car insurance that is essentially an compulsory purchase.
Why is it not reasonable, particularly with an insurance model based on sharing the risk over an identifiable grouping? What are you suggesting, individual interviews?
Share the risk over groups that individuals have control over, such as criminal records. Individual pricing should be based on individual information.
Justifying makjng discriminatory decisions on individuals based on the the behaviour of the population at large is tantamount to tyranny.
Would you be happy with a tax system that charged pensioners a higher rate because they were statistically more likely as a whole to make use of state services?
That would be fair if, but only if, the pensioners had been charged a lower rate when younger.
Re: insurance. The fact that women tend to have fewer accidents than men (which I do not dispute) should not be used as an excuse to charge individual men higher premiums. It’s not reasonable to use characteristics of groups that you have no choice in whether you belong to them as a basis of price discrimination, especially for a product like car insurance that is essentially an compulsory purchase.
Why is it not reasonable, particularly with an insurance model based on sharing the risk over an identifiable grouping? What are you suggesting, individual interviews?
Hands up if you are prepared more for your car insurance so that other people who are statistically more likely to claim can get a discount ?
If Admiral are doing what they say they are doing which is taking into account appropriate risk factors, then sure. In the end all good drivers subsidise bad drivers regardless of premium.
If they are basing it on surname, they have picked an absurd metric. They deny this and we await a proper explanation.
Why is surname an absurd metric ?
Sorry it's actually both surname and first name.
They don't directly relate to the underlying cause of risk.
If for example Leeds has a lot of accidents, and I live in Leeds, then using that fact bears more relation to the actual risk.
Using names must attempt to approximate cultural or ethnic backgrounds in which case you should be defending asking for and discriminating on them directly.
So there's your question: should white people get lower premiums?
I don't see why using names must attempt to anything. There are alternative explanations.
Hypothetically if the Jones family all keep crashing but the Smith family don't then would a new Jones be as likely to be in an accident as a new Smith? No culture or race involved.
Not saying that's what's happening, but hypothetically.
Re: insurance. The fact that women tend to have fewer accidents than men (which I do not dispute) should not be used as an excuse to charge individual men higher premiums. It’s not reasonable to use characteristics of groups that you have no choice in whether you belong to them as a basis of price discrimination, especially for a product like car insurance that is essentially an compulsory purchase.
Why is it not reasonable, particularly with an insurance model based on sharing the risk over an identifiable grouping? What are you suggesting, individual interviews?
Hands up if you are prepared more for your car insurance so that other people who are statistically more likely to claim can get a discount ?
If Admiral are doing what they say they are doing which is taking into account appropriate risk factors, then sure. In the end all good drivers subsidise bad drivers regardless of premium.
If they are basing it on surname, they have picked an absurd metric. They deny this and we await a proper explanation.
Why is surname an absurd metric ?
Because it has hundreds of thousands of possible values, is not stable through a person's lifetime, is difficult to validate and cannot easily be grouped. You could at a push use it for certain values - "Singh" being the obvious one - but what rules would you use to establish whether a name was a Muslim surname, a Jewish surname, a Christian one? An insurance dataset may easily contain hundreds of millions of records and I've dealt with some over three billion (which is a problem because that's bigger than 2^31 so you can't use it on 32-bit stuff). Pick out the Muslim surnames from that.
Re: insurance. The fact that women tend to have fewer accidents than men (which I do not dispute) should not be used as an excuse to charge individual men higher premiums. It’s not reasonable to use characteristics of groups that you have no choice in whether you belong to them as a basis of price discrimination, especially for a product like car insurance that is essentially an compulsory purchase.
Why is it not reasonable, particularly with an insurance model based on sharing the risk over an identifiable grouping? What are you suggesting, individual interviews?
Hands up if you are prepared more for your car insurance so that other people who are statistically more likely to claim can get a discount ?
If Admiral are doing what they say they are doing which is taking into account appropriate risk factors, then sure. In the end all good drivers subsidise bad drivers regardless of premium.
If they are basing it on surname, they have picked an absurd metric. They deny this and we await a proper explanation.
Why is surname an absurd metric ?
Sorry it's actually both surname and first name.
They don't directly relate to the underlying cause of risk.
If for example Leeds has a lot of accidents, and I live in Leeds, then using that fact bears more relation to the actual risk.
Using names must attempt to approximate cultural or ethnic backgrounds in which case you should be defending asking for and discriminating on them directly.
So there's your question: should white people get lower premiums?
I don't see why using names must attempt to anything. There are alternative explanations.
Hypothetically if the Jones family all keep crashing but the Smith family don't then would a new Jones be as likely to be in an accident as a new Smith? No culture or race involved.
Not saying that's what's happening, but hypothetically.
By calling them "families" you import the idea they might pass traits from on to the other. In reality the 100,000 Khans - let alone everyone with the first name Mohammad, share nothing except (disproportionately) an ethnic background.
Anyway Admiral and M&S have denied. We await their explanation.
I've worked in insurance companies for and with pricing units, I have EMBLEM on my CV, and I'm puzzled as hell with this "name" thing."Name" as an explanatory/independent variable field is *never* used in a model, and for good reason: It's a categorical variable with hundreds of thousands of possible values that is famously difficult to validate semantically, can't be easily grouped using rules, and would make any model using it grind to a halt.
My daughter works for Admiral. Name is not a variable in their pricing model.
I've worked in insurance companies for and with pricing units, I have EMBLEM on my CV, and I'm puzzled as hell with this "name" thing."Name" as an explanatory/independent variable field is *never* used in a model, and for good reason: It's a categorical variable with hundreds of thousands of possible values that is famously difficult to validate semantically, can't be easily grouped using rules, and would make any model using it grind to a halt.
My daughter works for Admiral. Name is not a variable in their pricing model.
Re: insurance. The fact that women tend to have fewer accidents than men (which I do not dispute) should not be used as an excuse to charge individual men higher premiums. It’s not reasonable to use characteristics of groups that you have no choice in whether you belong to them as a basis of price discrimination, especially for a product like car insurance that is essentially an compulsory purchase.
Why is it not reasonable, particularly with an insurance model based on sharing the risk over an identifiable grouping? What are you suggesting, individual interviews?
Hands up if you are prepared more for your car insurance so that other people who are statistically more likely to claim can get a discount ?
If Admiral are doing what they say they are doing which is taking into account appropriate risk factors, then sure. In the end all good drivers subsidise bad drivers regardless of premium.
If they are basing it on surname, they have picked an absurd metric. They deny this and we await a proper explanation.
Why is surname an absurd metric ?
Because it has hundreds of thousands of possible values, is not stable through a person's lifetime, is difficult to validate and cannot easily be grouped. You could at a push use it for certain values - "Singh" being the obvious one - but what rules would you use to establish whether a name was a Muslim surname, a Jewish surname, a Christian one? An insurance dataset may easily contain hundreds of millions of records and I've dealt with some over three billion (which is a problem because that's bigger than 2^31 so you can't use it on 32-bit stuff). Pick out the Muslim surnames from that.
Well, OK, but if a job's worth doing it's worth doing badly. A rule which says "if this field contains mohamed or mohammed do something, otherwise do something else" would be easy to write and implement, and have roughly the intended result, even if it got things wrong in say 10% of cases, because getting things wrong is non-fatal, it just means the offer of insurance is a bit dearer or a bit cheaper than you intended.
I've worked in insurance companies for and with pricing units, I have EMBLEM on my CV, and I'm puzzled as hell with this "name" thing."Name" as an explanatory/independent variable field is *never* used in a model, and for good reason: It's a categorical variable with hundreds of thousands of possible values that is famously difficult to validate semantically, can't be easily grouped using rules, and would make any model using it grind to a halt.
My vague, uninformed suspicion (hell, if everyone else is guessing, why can't I have a go too!?) is this is a result of a marketing/sales dept tweak/cockup, rather than their main risk modelling.
Basically, giving an unconfirmed quote influenced by the users behaviour on their website.
I have little to back that up and i'm not familiar with insurance industry modelling etc, like you.
Whatever the case, admiral have some explainin' to do...
I've worked in insurance companies for and with pricing units, I have EMBLEM on my CV, and I'm puzzled as hell with this "name" thing."Name" as an explanatory/independent variable field is *never* used in a model, and for good reason: It's a categorical variable with hundreds of thousands of possible values that is famously difficult to validate semantically, can't be easily grouped using rules, and would make any model using it grind to a halt.
My vague, uninformed suspicion (hell, if everyone else is guessing, why can't I have a go too!?) is this is a result of a marketing/sales dept tweak/cockup, rather than their main risk modelling.
Basically, giving an unconfirmed quote influenced by the users behaviour on their website.
I have little to back that up and i'm not familiar with insurance industry modelling etc, like you.
Whatever the case, admiral have some explainin' to do...
I mean I got a final quote within £5 of the number of the website so they can't be playing around with it much! But yes, maybe it is marketing that's the problem...
I've worked in insurance companies for and with pricing units, I have EMBLEM on my CV, and I'm puzzled as hell with this "name" thing."Name" as an explanatory/independent variable field is *never* used in a model, and for good reason: It's a categorical variable with hundreds of thousands of possible values that is famously difficult to validate semantically, can't be easily grouped using rules, and would make any model using it grind to a halt.
My daughter works for Admiral. Name is not a variable in their pricing model.
Pleased to see The Shape of Water getting so many noms at the Oscars. Also really pleasantly surprised at Phantom Thread getting so many. Probably the film that stayed with me most after watching. It is wonderfully strange, and Daniel Day-Lewis must be really spitting tacks that Gary Oldman's Churchill portrayal came out this year.
Biggest snub (for me) is Denis Villeneuve not getting a director nod for Bladerunner 2049. He really deserved one.
Agreed. He has made 3 films that I have seen so far (Sicario, Arrival and BR 2049) and all of them deserve Oscar nominations for Direction.
Well, OK, but if a job's worth doing it's worth doing badly. A rule which says "if this field contains mohamed or mohammed do something, otherwise do something else" would be easy to write and implement, and have roughly the intended result, even if it got things wrong in say 10% of cases, because getting things wrong is non-fatal, it just means the offer of insurance is a bit dearer or a bit cheaper than you intended.
So it would go like this
1 You have a field called "name" 2 You apply a rule to that field to create another field ("if name= "mohamed" or "mohammed" or... then set name_ethnicity to "M", else...) 3 You then use name_ethnicity as a field in your model, and your writeup justifies its inclusion based on its behavior in the model 4 Your supervisor doesn't notice the field, either in the model or the writeup (it gets checked) 5 The people who signoff the risk factors don't notice it either.
Pleased to see The Shape of Water getting so many noms at the Oscars. Also really pleasantly surprised at Phantom Thread getting so many. Probably the film that stayed with me most after watching. It is wonderfully strange, and Daniel Day-Lewis must be really spitting tacks that Gary Oldman's Churchill portrayal came out this year.
Biggest snub (for me) is Denis Villeneuve not getting a director nod for Bladerunner 2049. He really deserved one.
Agreed. He has made 3 films that I have seen so far (Sicario, Arrival and BR 2049) and all of them deserve Oscar nominations for Direction.
Interesting analysis of the UK's EU contributions - after the rebate they were £235 million per week- even the gross figure before rebate is only £325 million per week:
Suffice there's a lot less dividend to play with than many in the thread below believe.
Oh no, only £235mn a week. What a shame.
The net figure of the last few years per this research is a whopping £137 million per week - FWIW the UK Gov budget is £15 billion a week - we're in rounding error territory here.
That's £7bn per year in unmatched spending. Some EU funding only comes if the government guarantees to match it. Add a billion for that and we're up to £8bn. I think if you said to people, would you rather give the EU £8bn per year or the NHS £8bn per year the answer would be fairly emphatic.
Gove's already promised farmers £3 billion per annum until 2024 with a new scheme to replace the existing method after this date. I'm sure there are a number of other vested interests who will be banging on the government's door for compensation for loss of EU funds.
If the NHS needs another £8 billion now there's nothing stopping us allocating these funds now - no need to wait till 2022.
That £3bn will replace EU CAP spending and not come from the saving out of the net contribution.
Where will the govt the farmers promised £3 Billion come from then?
From the gross payment to the EU - it's a like for like replacement.
There is roughly £18bn of gross payments, and £11bn of net payments. If the government was to replace all of the existing EU spending like for like then it would still have £11bn to spend on our priorities
Gove has already said that farm subsidies will be the same amopunt but allocated more on environmental grounds and less to large scale industrial farming.
It is not just how much but the criteria for receiving it. The EU CAP is all about supporting small to medium uneconomic (French) farmers.
France are a net contributor to CAP.
...but that doesn't fit with the Brexit narrative, so it should be ignored.
Well, OK, but if a job's worth doing it's worth doing badly. A rule which says "if this field contains mohamed or mohammed do something, otherwise do something else" would be easy to write and implement, and have roughly the intended result, even if it got things wrong in say 10% of cases, because getting things wrong is non-fatal, it just means the offer of insurance is a bit dearer or a bit cheaper than you intended.
So it would go like this
1 You have a field called "name" 2 You apply a rule to that field to create another field ("if name= "mohamed" or "mohammed" or... then set name_ethnicity to "M", else...) 3 You then use name_ethnicity as a field in your model, and your writeup justifies its inclusion based on its behavior in the model 4 Your supervisor doesn't notice the field, either in the model or the writeup (it gets checked) 5 The people who signoff the risk factors don't notice it either.
OK, somebody's going to get fired.
I highly doubt it would be anything like that if name was used.
If it was used (and they say its not) it would be along the lines of
1 You have a field called "name" 2 Get actuarial data on the risk applied to that name. 3 Run your maths accordingly.
Nobody would sit there pre-determining "good" or "bad" names, an algorithm would do it.
Interesting analysis of the UK's EU contributions - after the rebate they were £235 million per week- even the gross figure before rebate is only £325 million per week:
Suffice there's a lot less dividend to play with than many in the thread below believe.
Oh no, only £235mn a week. What a shame.
The net figure of the last few years per this research is a whopping £137 million per week - FWIW the UK Gov budget is £15 billion a week - we're in rounding error territory here.
That's £7bn per year in unmatched spending. Some EU funding only comes if the government guarantees to match it. Add a billion for that and we're up to £8bn. I think if you said to people, would you rather give the EU £8bn per year or the NHS £8bn per year the answer would be fairly emphatic.
Gove's already promised farmers £3 billion per annum until 2024 with a new scheme to replace the existing method after this date. I'm sure there are a number of other vested interests who will be banging on the government's door for compensation for loss of EU funds.
If the NHS needs another £8 billion now there's nothing stopping us allocating these funds now - no need to wait till 2022.
That £3bn will replace EU CAP spending and not come from the saving out of the net contribution.
Where will the govt the farmers promised £3 Billion come from then?
From the gross payment to the EU - it's a like for like replacement.
There is roughly £18bn of gross payments, and £11bn of net payments. If the government was to replace all of the existing EU spending like for like then it would still have £11bn to spend on our priorities
Gove has already said that farm subsidies will be the same amopunt but allocated more on environmental grounds and less to large scale industrial farming.
It is not just how much but the criteria for receiving it. The EU CAP is all about supporting small to medium uneconomic (French) farmers.
France are a net contributor to CAP.
...but that doesn't fit with the Brexit narrative, so it should be ignored.
How much does France net contribute to CAP compared to Britain?
Well, OK, but if a job's worth doing it's worth doing badly. A rule which says "if this field contains mohamed or mohammed do something, otherwise do something else" would be easy to write and implement, and have roughly the intended result, even if it got things wrong in say 10% of cases, because getting things wrong is non-fatal, it just means the offer of insurance is a bit dearer or a bit cheaper than you intended.
So it would go like this
1 You have a field called "name" 2 You apply a rule to that field to create another field ("if name= "mohamed" or "mohammed" or... then set name_ethnicity to "M", else...) 3 You then use name_ethnicity as a field in your model, and your writeup justifies its inclusion based on its behavior in the model 4 Your supervisor doesn't notice the field, either in the model or the writeup (it gets checked) 5 The people who signoff the risk factors don't notice it either.
OK, somebody's going to get fired.
If this is happening at all, it is being imposed from the top, so your supervisor has ex hypothesi been told to keep quiet. I have no reason to think it is happening, but equally it is no argument to say that big responsible companies don't decide at a high level to do that sort of thing. VW executives have, after all, sat down and decided in principle to cheat on the emission tests, and must then have passed the instruction all the way down to the engineers responsible for implementing the fraud.
Interesting analysis of the UK's EU contributions - after the rebate they were £235 million per week- even the gross figure before rebate is only £325 million per week:
Suffice there's a lot less dividend to play with than many in the thread below believe.
Oh no, only £235mn a week. What a shame.
The net figure of the last few years per this research is a whopping £137 million per week - FWIW the UK Gov budget is £15 billion a week - we're in rounding error territory here.
That's £7bn per year in unmatched spending. Some EU funding only comes if the government guarantees to match it. Add a billion for that and we're up to £8bn. I think if you said to people, would you rather give the EU £8bn per year or the NHS £8bn per year the answer would be fairly emphatic.
Gove's already promised farmers £3 billion per annum until 2024 with a new scheme to replace the existing method after this date. I'm sure there are a number of other vested interests who will be banging on the government's door for compensation for loss of EU funds.
If the NHS needs another £8 billion now there's nothing stopping us allocating these funds now - no need to wait till 2022.
That £3bn will replace EU CAP spending and not come from the saving out of the net contribution.
Where will the govt the farmers promised £3 Billion come from then?
From the gross payment to the EU - it's a like for like replacement.
There is roughly £18bn of gross payments, and £11bn of net payments. If the government was to replace all of the existing EU spending like for like then it would still have £11bn to spend on our priorities
Gove has already said that farm subsidies will be the same amopunt but allocated more on environmental grounds and less to large scale industrial farming.
It is not just how much but the criteria for receiving it. The EU CAP is all about supporting small to medium uneconomic (French) farmers.
France are a net contributor to CAP.
...but that doesn't fit with the Brexit narrative, so it should be ignored.
How much does France net contribute to CAP compared to Britain?
Interesting analysis of the UK's EU contributions - after the rebate they were £235 million per week- even the gross figure before rebate is only £325 million per week:
Suffice there's a lot less dividend to play with than many in the thread below believe.
Oh no, only £235mn a week. What a shame.
The net figure of the last few years per this research is a whopping £137 million per week - FWIW the UK Gov budget is £15 billion a week - we're in rounding error territory here.
That's £7bn per year in unmatched spending. Some EU funding only comes if the government guarantees to match it. Add a billion for that and we're up to £8bn. I think if you said to people, would you rather give the EU £8bn per year or the NHS £8bn per year the answer would be fairly emphatic.
Gove's already promised farmers £3 billion per annum until 2024 with a new scheme to replace the existing method after this date. I'm sure there are a number of other vested interests who will be banging on the government's door for compensation for loss of EU funds.
If the NHS needs another £8 billion now there's nothing stopping us allocating these funds now - no need to wait till 2022.
That £3bn will replace EU CAP spending and not come from the saving out of the net contribution.
Where will the govt the farmers promised £3 Billion come from then?
From the gross payment to the EU - it's a like for like replacement.
There is roughly £18bn of gross payments, and £11bn of net payments. If the government was to replace all of the existing EU spending like for like then it would still have £11bn to spend on our priorities
Gove has already said that farm subsidies will be the same amopunt but allocated more on environmental grounds and less to large scale industrial farming.
It is not just how much but the criteria for receiving it. The EU CAP is all about supporting small to medium uneconomic (French) farmers.
France are a net contributor to CAP.
...but that doesn't fit with the Brexit narrative, so it should be ignored.
How much does France net contribute to CAP compared to Britain?
Interesting analysis of the UK's EU contributions - after the rebate they were £235 million per week- even the gross figure before rebate is only £325 million per week:
Suffice there's a lot less dividend to play with than many in the thread below believe.
Oh no, only £235mn a week. What a shame.
The net figure of the last few years per this research is a whopping £137 million per week - FWIW the UK Gov budget is £15 billion a week - we're in rounding error territory here.
That's £7bn per year in unmatched spending. Some EU funding only comes if the government guarantees to match it. Add a billion for that and we're up to £8bn. I think if you said to people, would you rather give the EU £8bn per year or the NHS £8bn per year the answer would be fairly emphatic.
Gove's already promised farmers £3 billion per annum until 2024 with a new scheme to replace the existing method after this date. I'm sure there are a number of other vested interests who will be banging on the government's door for compensation for loss of EU funds.
If the NHS needs another £8 billion now there's nothing stopping us allocating these funds now - no need to wait till 2022.
That £3bn will replace EU CAP spending and not come from the saving out of the net contribution.
Where will the govt the farmers promised £3 Billion come from then?
From the gross payment to the EU - it's a like for like replacement.
There is roughly £18bn of gross payments, and £11bn of net payments. If the government was to replace all of the existing EU spending like for like then it would still have £11bn to spend on our priorities
Gove has already said that farm subsidies will be the same amopunt but allocated more on environmental grounds and less to large scale industrial farming.
It is not just how much but the criteria for receiving it. The EU CAP is all about supporting small to medium uneconomic (French) farmers.
France are a net contributor to CAP.
...but that doesn't fit with the Brexit narrative, so it should be ignored.
There is no inconsistency between "The EU CAP is all about supporting small to medium uneconomic (French) farmers" and "France are a net contributor to CAP." The interests of France, and of French farmers, are by no means 100% aligned.
I've worked in insurance companies for and with pricing units, I have EMBLEM on my CV, and I'm puzzled as hell with this "name" thing."Name" as an explanatory/independent variable field is *never* used in a model, and for good reason: It's a categorical variable with hundreds of thousands of possible values that is famously difficult to validate semantically, can't be easily grouped using rules, and would make any model using it grind to a halt.
My vague, uninformed suspicion (hell, if everyone else is guessing, why can't I have a go too!?) is this is a result of a marketing/sales dept tweak/cockup, rather than their main risk modelling.
Basically, giving an unconfirmed quote influenced by the users behaviour on their website.
I have little to back that up and i'm not familiar with insurance industry modelling etc, like you.
Whatever the case, admiral have some explainin' to do...
Indeed (and for the avoidance of doubt I work with pricing, not for them: I have built naive and saturated models, made suggestions based on forward selection, cleaned the data and loaded it up and I understand the underlying theory, but it's the pricing people who take the models forward and formally select the variables and decide the "weights" (the word pricing people use to describe the numbers you multiply the variable by to build the model), based on how the model changes as the variables change
It's more art than science: you can automatically generate a model and given enough data you can generate a model that fits perfectly[1], but that's useless. The pricing people decide the weights based on how the model changes as the variables change, so there's lots of visual inspection of graphs. It's a very qualitative process.
[1] if you have enough variables with a sufficiently large range of values in them, they exceed the number of datapoints in your training data and you end up creating a model that effectively gives one unique result for each unique example, so just thru brute force it fits perfectly.
Well, OK, but if a job's worth doing it's worth doing badly. A rule which says "if this field contains mohamed or mohammed do something, otherwise do something else" would be easy to write and implement, and have roughly the intended result, even if it got things wrong in say 10% of cases, because getting things wrong is non-fatal, it just means the offer of insurance is a bit dearer or a bit cheaper than you intended.
So it would go like this
1 You have a field called "name" 2 You apply a rule to that field to create another field ("if name= "mohamed" or "mohammed" or... then set name_ethnicity to "M", else...) 3 You then use name_ethnicity as a field in your model, and your writeup justifies its inclusion based on its behavior in the model 4 Your supervisor doesn't notice the field, either in the model or the writeup (it gets checked) 5 The people who signoff the risk factors don't notice it either.
OK, somebody's going to get fired.
I highly doubt it would be anything like that if name was used.
If it was used (and they say its not) it would be along the lines of
1 You have a field called "name" 2 Get actuarial data on the risk applied to that name. 3 Run your maths accordingly.
Nobody would sit there pre-determining "good" or "bad" names, an algorithm would do it.
Yes, but there would still be a field in the model representing that risk. You could get round it by calling it a fiddle factor or something, but it would still be there. It would be easier for the actuaries to manually manipulate the results after the modeling was done.
Well, OK, but if a job's worth doing it's worth doing badly. A rule which says "if this field contains mohamed or mohammed do something, otherwise do something else" would be easy to write and implement, and have roughly the intended result, even if it got things wrong in say 10% of cases, because getting things wrong is non-fatal, it just means the offer of insurance is a bit dearer or a bit cheaper than you intended.
So it would go like this
1 You have a field called "name" 2 You apply a rule to that field to create another field ("if name= "mohamed" or "mohammed" or... then set name_ethnicity to "M", else...) 3 You then use name_ethnicity as a field in your model, and your writeup justifies its inclusion based on its behavior in the model 4 Your supervisor doesn't notice the field, either in the model or the writeup (it gets checked) 5 The people who signoff the risk factors don't notice it either.
OK, somebody's going to get fired.
If this is happening at all, it is being imposed from the top, so your supervisor has ex hypothesi been told to keep quiet. I have no reason to think it is happening, but equally it is no argument to say that big responsible companies don't decide at a high level to do that sort of thing. VW executives have, after all, sat down and decided in principle to cheat on the emission tests, and must then have passed the instruction all the way down to the engineers responsible for implementing the fraud.
I've worked in insurance companies for and with pricing units, I have EMBLEM on my CV, and I'm puzzled as hell with this "name" thing."Name" as an explanatory/independent variable field is *never* used in a model, and for good reason: It's a categorical variable with hundreds of thousands of possible values that is famously difficult to validate semantically, can't be easily grouped using rules, and would make any model using it grind to a halt.
My daughter works for Admiral. Name is not a variable in their pricing model.
Comments
nor an.
Choose another insurance if you are offended that Admiral are giving discounts to those less likely to claim.
That's how much I* had to pay for the first nine months of my car insurance on a Ford Fiesta.
* By "I", I mean my dad.
The chance of that is about one in a million!
Share the risk over groups that individuals have control over, such as criminal records. Individual pricing should be based on individual information.Justifying makjng discriminatory decisions on individuals based on the the behaviour of the population at large is tantamount to tyranny.
Would you be happy with a tax system that charged pensioners a higher rate because they were statistically more likely as a whole to make use of state services?
If they are basing it on surname, they have picked an absurd metric. They deny this and we await a proper explanation.
It seems to me the average teenage boy has more testosterone than he can handle, a need to show off and a higher propensity for high speed, serious injury crashes. If I am being asked to insure a teenage boy I am entitled to take that propensity into account even if the individual in question has not been shown to have those tendencies.
They don't directly relate to the underlying cause of risk.
If for example Leeds has a lot of accidents, and I live in Leeds, then using that fact bears more relation to the actual risk.
Using names must attempt to approximate cultural or ethnic backgrounds in which case you should be defending asking for and discriminating on them directly.
So there's your question: should white people get lower premiums?
Hypothetically if the Jones family all keep crashing but the Smith family don't then would a new Jones be as likely to be in an accident as a new Smith? No culture or race involved.
Not saying that's what's happening, but hypothetically.
Anyway Admiral and M&S have denied. We await their explanation.
Basically, giving an unconfirmed quote influenced by the users behaviour on their website.
I have little to back that up and i'm not familiar with insurance industry modelling etc, like you.
Whatever the case, admiral have some explainin' to do...
https://www.politico.com/story/2018/01/23/mitch-mcconnell-russia-obama-joe-biden-359531?lo=ap_a1
TSE even compared it for us...
1 You have a field called "name"
2 You apply a rule to that field to create another field ("if name= "mohamed" or "mohammed" or... then set name_ethnicity to "M", else...)
3 You then use name_ethnicity as a field in your model, and your writeup justifies its inclusion based on its behavior in the model
4 Your supervisor doesn't notice the field, either in the model or the writeup (it gets checked)
5 The people who signoff the risk factors don't notice it either.
OK, somebody's going to get fired.
If it was used (and they say its not) it would be along the lines of
1 You have a field called "name"
2 Get actuarial data on the risk applied to that name.
3 Run your maths accordingly.
Nobody would sit there pre-determining "good" or "bad" names, an algorithm would do it.
It's more art than science: you can automatically generate a model and given enough data you can generate a model that fits perfectly[1], but that's useless. The pricing people decide the weights based on how the model changes as the variables change, so there's lots of visual inspection of graphs. It's a very qualitative process.
[1] if you have enough variables with a sufficiently large range of values in them, they exceed the number of datapoints in your training data and you end up creating a model that effectively gives one unique result for each unique example, so just thru brute force it fits perfectly.
https://twitter.com/peston/status/955903219762331649