4 ms·

In a 12 year period. So what about a 15 year period? Or a 20 year period? Or in a 12 year period but 10 years ago? Results like that are generally bunk, or if

In a 12 year period. So what about a 15 year period? Or a 20 year period? Or in a 12 year period but 10 years ago? Results like that are generally bunk, or if they really cared, they would have made a broader statement (in these and those periods they performed like that). I have only looked at the Credit Suisse paper that I think started this meme, and they clearly use a cut off point that makes women led companies look good. If they start their comparison a couple of years earlier, suddenly it looks different. That's just standard methodology for companies like Credit Suisse whose main business is convincing people to buy stock. That same method is used to make stock market funds look good all the time (so that gullible people pay the 5% premium for the seemingly brilliant manager of that fund). Other explanations of course could be: categories of businesses were trending that were more attractive to women (like digital services vs mining or something like that). Or the market in the time favored a risk averse approach and women tend to be more risk averse (don't know that) - there was a bad economical recession in that period of time (as they claim women "manage risk better" - well there are times when more or less risk taking are called for). Don't get me wrong, I don't mind if women are CEOs and lead companies. I'm just wary of "narratives" that sound too good or too ideologically motivated.

therealdrag0 12y ago
I thought 12 was a red flag too. Not exactly the standard increment of 5 or 10.
- jsprogrammer 12y ago
  Because 5 or 10 is any less arbitrary?
  - robzyb 12y ago
    Yes. In this context 5 or 10 are less arbitrary. This is due to their regular usage.
  - Steuard 12y ago
    Certainly not. But they're common defaults. The worry is that if someone picks a non-standard interval, it suggests that they gave conscious thought to what interval to pick and came up with that one. The natural question in that case is why: Is there some economic reason that 12 years would be special, or some matching limitation to their data set? Or did they try every interval between 5 and 20 years and pick the one where the result was most favorable? Edit: But it sounds like it actually was due to the size of the dataset: https://news.ycombinator.com/item?id=9222416 https://news.ycombinator.com/item?id=9222416
tomlock 12y ago
Here's some more: http://www.asx.com.au/documents/asx-compliance/the_bottom_line_corporate_performance_and_womens_representation_on_boards_2004-2008.pdf http://www.asx.com.au/documents/asx-compliance/the_bottom_li... https://finsia.com/news/news-article/2015/03/13/boardroom-gender-diversity-improves-company-performance https://finsia.com/news/news-article/2015/03/13/boardroom-ge... http://www.newyorker.com/business/currency/do-more-women-on-the-board-mean-better-results http://www.newyorker.com/business/currency/do-more-women-on-...
- facepalm 12y ago
  Well I just looked at the first paper, and now look at the credit suisse paper I linked in another comment ( https://www.credit-suisse.com/newsletter/doc/gender_diversity.pdf https://www.credit-suisse.com/newsletter/doc/gender_diversit... ) which has charts for 2005 to 2011. There was a "crash" in mid 2007 that somehow reversed the trends for companies with or without women on the board. Before that, zero-woman companies were outperforming the woman-companies. Now it seems rather curious to me that they choose 2004-2008 for their comparison (in the first paper you linked to). So strangely, they catch exactly that crash point in their interval. It's always the same shenanigans...
sillysaurus3 12y ago
I have only looked at the Credit Suisse paper that I think started this meme, and they clearly use a cut off point that makes women led companies look good. If they start their comparison a couple of years earlier, suddenly it looks different. What does it look like?
- facepalm 12y ago
  I'm not sure which one I looked at (thought I would have saved it to hd but can't find it right now). Googling I find for example this (it looks different than the one I remember, but still): https://www.credit-suisse.com/newsletter/doc/gender_diversity.pdf https://www.credit-suisse.com/newsletter/doc/gender_diversit... Only skimmed it, but they have for example a chart for 2005 to 2011. Between 2005 and mid 2007the zero-women companies outperform the at-least-one-women companies, then there is some sort of crash and the roles reverse. Would be curious to see the chart for pre-2005...
choward 12y ago
Also, why are they comparing companies ran by women in the Fortune 1000 against the S&P 500? Is it because the Fortune 1000 has the best performance by female CEOs while in the same time period the S&P 500 had the worst performance they could find? Also, what percentage of companies in the S&P 500 where ran by women? I didn't see that anywhere.
- rnprince 12y ago
  The S&P 500 is the go-to index for a lot of people when they think about the "market return;" an important idea in finance. It's a portfolio of a large enough number of companies that the portfolio return converges pretty thoroughly to the market return. Not sure why the S&P 1000 isn't used as much, but using the S&P 500 is a convention. Why choose all companies with female CEOs from the S&P 1000? The S&P 500 as of last year had 26 female CEOs, and if I had to guess, I'd say that number is higher than it would have been in 2002. The portfolio of "S&P 500 companies with female CEOs as of 2002" wouldn't have been diversified enough to reliably converge to the "market return of companies with female CEOs" that they were trying to investigate.
refurb 12y ago
In a 12 year period. So what about a 15 year period? Or a 20 year period? Or in a 12 year period but 10 years ago? This is a great point. Anytime you see a study where the time period seems contrived, it's usually because it was used to tweak the outcome to get what the author wanted. Never invest in the stock market!!! (Source: S&P500 from Jan 02 to Jan 08)
- tomlock 12y ago
  One of the authors colleagues has posted in this thread stating that the selection of 12 years was because that was the amount of data available in the system they used.
qiqing 12y ago
1. A colleague of the author has commented in a lower thread that 12 years was the size of the data set available. 2. If n is very small in a several-year period, you wouldn't be able to do meaningful analyses on that period. How many women CEO's were there in the Fortune 1000 in the 80's? 90's? I'm going to guess epsilonically few. 3. Possible explanation that isn't hand-waving: selection bias, somewhat akin to the immigration effect. Because it's more challenging for women to get on a career path that eventually leads them to become a CEO, only the most determined/resourceful/hardworking will get there and be measured in the sample, whereas fewer men get weeded out. (e.g., IQ scores of H1B recipients from Asia is not representative of Asia as a continent, where IQ is a standin for whatever measurable trait you want here; and overall Asia and North America may well be quite comparable along most dimensions.)
- facepalm 12y ago
  So suppose they had done the study, and found no improved performance for women. Would they have published it? I guess the chance to get a result was roughly 50% (since stocks tend to go up or down, and the sample set of women-ceo-companies is probably small). As to your explanation 3: that's just the kind of thing people want to believe who cherish such studies. At the same time there were also women who failed as CEOs (as the article said). Sure, it's a valid hypothesis, but there would be many ways to test it. For starters, there must already be studies correlating IQ to CEO success. What do they say?

therealdrag0 12y ago

I thought 12 was a red flag too. Not exactly the standard increment of 5 or 10.

jsprogrammer 12y ago
Because 5 or 10 is any less arbitrary?
- robzyb 12y ago
  Yes. In this context 5 or 10 are less arbitrary. This is due to their regular usage.
- Steuard 12y ago
  Certainly not. But they're common defaults. The worry is that if someone picks a non-standard interval, it suggests that they gave conscious thought to what interval to pick and came up with that one. The natural question in that case is why: Is there some economic reason that 12 years would be special, or some matching limitation to their data set? Or did they try every interval between 5 and 20 years and pick the one where the result was most favorable? Edit: But it sounds like it actually was due to the size of the dataset: https://news.ycombinator.com/item?id=9222416 https://news.ycombinator.com/item?id=9222416

tomlock 12y ago

Here's some more: http://www.asx.com.au/documents/asx-compliance/the_bottom_line_corporate_performance_and_womens_representation_on_boards_2004-2008.pdf http://www.asx.com.au/documents/asx-compliance/the_bottom_li... https://finsia.com/news/news-article/2015/03/13/boardroom-gender-diversity-improves-company-performance https://finsia.com/news/news-article/2015/03/13/boardroom-ge... http://www.newyorker.com/business/currency/do-more-women-on-the-board-mean-better-results http://www.newyorker.com/business/currency/do-more-women-on-...

facepalm 12y ago
Well I just looked at the first paper, and now look at the credit suisse paper I linked in another comment ( https://www.credit-suisse.com/newsletter/doc/gender_diversity.pdf https://www.credit-suisse.com/newsletter/doc/gender_diversit... ) which has charts for 2005 to 2011. There was a "crash" in mid 2007 that somehow reversed the trends for companies with or without women on the board. Before that, zero-woman companies were outperforming the woman-companies. Now it seems rather curious to me that they choose 2004-2008 for their comparison (in the first paper you linked to). So strangely, they catch exactly that crash point in their interval. It's always the same shenanigans...

sillysaurus3 12y ago

I have only looked at the Credit Suisse paper that I think started this meme, and they clearly use a cut off point that makes women led companies look good. If they start their comparison a couple of years earlier, suddenly it looks different. What does it look like?

facepalm 12y ago
I'm not sure which one I looked at (thought I would have saved it to hd but can't find it right now). Googling I find for example this (it looks different than the one I remember, but still): https://www.credit-suisse.com/newsletter/doc/gender_diversity.pdf https://www.credit-suisse.com/newsletter/doc/gender_diversit... Only skimmed it, but they have for example a chart for 2005 to 2011. Between 2005 and mid 2007the zero-women companies outperform the at-least-one-women companies, then there is some sort of crash and the roles reverse. Would be curious to see the chart for pre-2005...

choward 12y ago

Also, why are they comparing companies ran by women in the Fortune 1000 against the S&P 500? Is it because the Fortune 1000 has the best performance by female CEOs while in the same time period the S&P 500 had the worst performance they could find? Also, what percentage of companies in the S&P 500 where ran by women? I didn't see that anywhere.

rnprince 12y ago
The S&P 500 is the go-to index for a lot of people when they think about the "market return;" an important idea in finance. It's a portfolio of a large enough number of companies that the portfolio return converges pretty thoroughly to the market return. Not sure why the S&P 1000 isn't used as much, but using the S&P 500 is a convention. Why choose all companies with female CEOs from the S&P 1000? The S&P 500 as of last year had 26 female CEOs, and if I had to guess, I'd say that number is higher than it would have been in 2002. The portfolio of "S&P 500 companies with female CEOs as of 2002" wouldn't have been diversified enough to reliably converge to the "market return of companies with female CEOs" that they were trying to investigate.

refurb 12y ago

In a 12 year period. So what about a 15 year period? Or a 20 year period? Or in a 12 year period but 10 years ago? This is a great point. Anytime you see a study where the time period seems contrived, it's usually because it was used to tweak the outcome to get what the author wanted. Never invest in the stock market!!! (Source: S&P500 from Jan 02 to Jan 08)

tomlock 12y ago
One of the authors colleagues has posted in this thread stating that the selection of 12 years was because that was the amount of data available in the system they used.

qiqing 12y ago

1. A colleague of the author has commented in a lower thread that 12 years was the size of the data set available. 2. If n is very small in a several-year period, you wouldn't be able to do meaningful analyses on that period. How many women CEO's were there in the Fortune 1000 in the 80's? 90's? I'm going to guess epsilonically few. 3. Possible explanation that isn't hand-waving: selection bias, somewhat akin to the immigration effect. Because it's more challenging for women to get on a career path that eventually leads them to become a CEO, only the most determined/resourceful/hardworking will get there and be measured in the sample, whereas fewer men get weeded out. (e.g., IQ scores of H1B recipients from Asia is not representative of Asia as a continent, where IQ is a standin for whatever measurable trait you want here; and overall Asia and North America may well be quite comparable along most dimensions.)

facepalm 12y ago
So suppose they had done the study, and found no improved performance for women. Would they have published it? I guess the chance to get a result was roughly 50% (since stocks tend to go up or down, and the sample set of women-ceo-companies is probably small). As to your explanation 3: that's just the kind of thing people want to believe who cherish such studies. At the same time there were also women who failed as CEOs (as the article said). Sure, it's a valid hypothesis, but there would be many ways to test it. For starters, there must already be studies correlating IQ to CEO success. What do they say?