On 30th June 2003, Marian Griffith wrote:
>> on Mon 30 Jun, Richard A. Bartle wrote:
>> 1) Perhaps fewer women are into the whole power-gaming thing
>> (that you describe as the reason people buy higher-level
>> accounts). This reduces the demand for female avatars, which in
>> turn reduces the price.
> If this is (part of) the explanation then the number of female
> avatars offered for sale should be decreasing steadily over time
I did say that I didn't necessarily agree with all these
possibilities, but I wasn't actually expecting anyone to go through
them with a fine-toothed comb. My point was that there are many ways
to interpret the data, of which "men value women lower than they
value men because of discrimination" is but one.
Still, I'll reply to your comments about my examples if you wish.
(Warning to bystanders: this is rather long).
> as those who create such characters for sale find that with the
> same amount of effort they can create a more valuable character.
It depends on what causes the supply. Example: it may be (I'm not
saying that it IS, just that it MAY BE) that many of these female
avatars come up for sale because they are the results of male
experiments in cross-gender playing. If that were the case, then the
supply would remain fairly constant - the characters weren't being
created to be sold, they were being created for other reasons but
later released when those reasons were no longer important. In this
situation, the supply would be constant, the demand would be lower
than for male characters (the premiss in this example) and therefore
the price would be lower. If the over-supply is not subject to
market forces (ie. people aren't creating characters purposefully to
sell them) then it wouldn't reduce to balance the lower prices paid
for the product.
>> 2) Women in the real world are paid less than men, on
>> average. Perhaps they can't afford to pay as much for their
>> characters, so the price falls until they can?
> This assumes that women are the primary buyers of the female
> avatars.
Yes, it does.
If you want to break the link between gender of player and gender of
avatar, that's fine by me: the general relationship between gender
of purchaser and gender of avatar is used in the paper to underpin
the suggestion that men value male avatars more than they do female
ones.
>> 3) Perhaps the people who manufacture characters so they can
>> sell them on eBay skew the market by their own preferences for
>> what character genders they play?
> That could easily be checked by looking at several games.
Certainly it could. However, it wasn't.
Remember that I'm merely proposing possible other interpretations
here to the suggestion that the evidence of the data points to men
undervaluing female avatars. It may be that men DO undervalue female
avatars, but where other reasonable possibilities exist they also
need to be explored.
>> perhaps it indicates an over-supply of female characters, in
>> which case the price could be expected to be lower than for male
>> characters.
> Without better demographics of both player and buyer population
> this argument actually could go either way.
It could. Again, I'm not saying this is what does happen, I'm saying
it could happen - and with a likelihood that makes leaping to the
conclusion of sexism in avatar sales rather suspect. If the argument
can "go either way", that means arguments based on the assumption
that it goes one particular way are for the moment merely
speculative.
>> 6) The price of a level for high-level characters is more than
>> for low-level characters, but your regression rules may not
>> fully capture the impact of this.
> From the little I have read of the article I had the impression
> that prices were compared for characters of similar level and
> equipment.
Well, kind of. The variation between characters is too great to
compare on an individual basis - races and classes come into play
too, for example. The paper applies three different metrics to try
to ascertain weights for the effect each character variable has on
the price of a character; they all come out with fairly similar
results, though, at least in terms of the order of importance
attributed to each variable.
The paper uses a process called "hedonic regression". I haven't come
across this before, and neither have any of the economic textbooks I
have (which isn't perhaps surprising, given I only have
two!). However, I have encountered something similar before in the
field of examination marking.
Warning: this is boring. The problem many university examiners have
is that for university courses it's hard to scale examination
results when there are small numbers of students - 10 or 20,
say. How do you know whether an exam was too easy or too hard? One
answer is to see what the pupils did across all the basket of exams
they took and arrive at some normalised weighting function. If a
student taking 4 exams got 84%, 89%, 80% and 50%, then the chances
are the last exam was too hard and all students' marks should be
scaled upwards. Students take different baskets of courses from one
another, but so long as there is some overlap then normalisation can
be a meaningful exercise. Performing the necessary analysis across
all exams for all students takes some hefty calculation, but we had
a numerical algorithms group that was into this kind of programming
at the time, so we got to apply it.
OK, well the point is that there are some assumptions inherent in
this procedure that may or may not hold true. For example, it works
best if the variables are independent: if (all else being equal) a
female elf costs $400, a male elf costs $500 and a male human costs
$400, you would expect a female human to cost $300. Although this
independence is indeed the case for avatar gender (at least at the
programming level), it isn't for other combinations. For example, 3
of the 4 permutations of dwarf/halfling fighter/spellcaster might
have low values but one (dwarf fighter) a high one. This can throw a
regression overall if it happens too much (even if it doesn't happen
for the particular variables in which you're interested).
Another assumption is that the contribution of any particular
variable to price follows a well-behaved function. It might not.
Suppose, for example, that large numbers of players felt they wanted
to break the level-50 barrier personally, so they could say they had
"made level 50". I'm not saying they DO that, but suppose they
did. The effect of it would be that the price for level 49 avatars
could well be higher than that of level 50 avatars, which again
would throw an overall regression. The functions aren't likely to be
as awkward as that, of course, but there could be sudden step-change
rises in prices for a particular variable, or other badly-behaved
functionality. The regression strategies applied can dampen these
effects, but they dampen them all the same way; it may be that some
should be damped one way but not others.
What I was suggesting in my original statement of the problem was
that it might be that for higher-level characters, the effect of
level on price may be way, way, WAY stronger than for characters
just a few levels lower. In this case, since most female avatars are
for sale at lower levels than most male avatars, the contribution of
the top-level sales to avatar value would be factored into them less
than for male avatars.
Again, though, this is just a possibility. I'm not saying it's how
it does happen; just that it's an alternative to saying it's all
down to men not valuing women.
>> 7) Perhaps female players rate a character's appearance higher
>> than male players? They'll pay premium rates for an exotic
>> female dark elf, but they rate other races much lower than do
>> male players. If they won't pay for female characters in
>> general, the price will drop, although it's "balanced" by the
>> extra they'll pay for the few combinations they do like. Again,
>> the mathematics you use only approximates the effect, though,
>> and may do so badly.
> While this may true, it does not quite explain, why male players
> do not pick up such bargain characters instead.
It doesn't have to explain that. It's saying that perhaps gender
itself isn't necessarily independent as a variable (in this example,
on the part of women; men could do it too, though).
> The point of the article was that men were willing to pay more so
> they would not have to play a female character.
That was A point of the article, although now it seems to have
become THE point of it.
> That is in a sense discriminatory and not discredited by this
> explanation.
It's not discredited by this argument, no, but it's exposed as more
speculative by it. What I'm basically saying is that the math may
have missed another possible explanation for why male characters
appear to cost more than female characters. That being the case,
perhaps it's not true that male characters cost more than female
ones, in which case saying that men are willing to pay more so they
don't have to play a female character would be invalid as men
wouldn't actually be paying more.
I suspect that in this particular instance the influence on price is
not so great as to skew the results much, but that doesn't mean
there aren't other examples where it might.
>> Maybe the equipment that typically comes with female avatars
>> is, for some reason, not as attractive to potential buyers as
>> that which typically comes with male avatars?
> Given that there should be no meaningful difference between male
> and female equipment (or the abilities of the characters). This
> is an unlikely explanation of the observed price difference.
The whole point is that "there should be no meaningful difference",
yet there (apparently) is. Can we be sure that the "meaningful
difference" isn't in the character, but rather is in the equipment
that the character carries? It's not like it would be TOO hard to
check.
Example: let's suppose that in general men and women have different
overall aims when playing virtual worlds. Let's say that the
equipment that men purchase for their characters reflects their
aims, and the equipment women purchase for their characters reflects
their aims. Let's further suppose that most female characters are
played by female players and that most male characters are played by
male players. Given this, a male player might be perfectly happy
with a female character but not with her equipment. Is he going to
pay the same for that character as for an otherwise identical male
character that has the equipment he wants? No.
>> 10) Perhaps more women are TOS-abiding than men, or believe
>> that buying characters is wrong anyway, or regard auctions as
>> some kind of competitive male thing. This would mean that fewer
>> of them would be available to buy avatars, therefore anyone who
>> wants a female avatar can expect to pay less to get one.
> But it would also mean *far* fewer would be available for sale.
Why "*far*" fewer? It would only take a small percentage not to
regard auctioning as a good thing for it affect the price. Also, it
may be that the ones for sale come mainly from men, rather than from
women.
> One point that supports the original conclusion is the observation
> of many female players that their female avatar is not taken
> seriously as male ones.
That may support it, yes, but that's not what the paper is
describing. The paper says that female characters cost less than
their male equivalents. It speculates that this may be because men
value female avatars less than they do male avatars. It hints that
this may be because men are carrying over real-life discrimination
into the virtual world.
I am suspicious of the proof that female characters actually do cost
less than their male equivalents, but I suspect its findings are
more or less correct; I am wary of the suggestion that this is
because men value female characters lower than male characters, but
again I suspect that they probably do; I am very skeptical of the
implication that this stems from RL discrimination.
It's interesting to speculate on what would have been said if female
avatars had turned out to cost more than male avatars.
> All this suggests that indeed a strong gender bias affects how
> players perceive avatars, and that the lower price for female ones
> is because they are seen as less capable.
There are many reasons why male players might prefer to pay more for
a male character than for an otherwise identical female one. Sexism
on the part of some male players probably is one such reason, but
there are plenty more.
The Roberts & Parks survey of cross-gender play in MOOs [Lynne
D. Roberts and Malcolm R. Parks, "The Social Geography of
Gender-Switching in Virtual Environ-ments on the Internet". In
Eileen Green and Alison Adam, "Virtual Gender: Terminology,
Consumption and Identity". New York, Routledge, 2001] found that
people cited the following reasons for not playing cross-gender:
1) They had no desire to be the opposite sex (23%)
2) They felt that gender-switching is dishonest and deceitful
(14.6%)
3) They desired to present their "real" self accurately (12.9%)
4) They had a strong identification with own gender (12.4%)
5) They doubted they could successfully play their opposite gender
(11.8%)
6) They saw no benefit to gender-switching (11.8%)
These are across all players (those who have played cross-gender but
have since stopped; those currently playing cross-gender; and those
who have never played cross-gender). There were attitudinal
differences between them: of the 40% of role-play MOOers and 60% of
social MOOers had never played cross-gender, many saw it as
dishonest, deceitful and manipulative, and got upset and
uncomfortable when a character did not match the biological sex of
its player.
Of course, these figures might not apply to EQ, as it's not a
MOO. There are demographic differences, too: in the (random) MOOs
surveyed there was an almost 50/50 split between male and female
players, as opposed to 92.4/7.6 in favour of men in EQ. However,
biological gender was not found to be a factor in determining
whether a player would switch genders in a MOO, and doesn't appear
to be in EQ either: 12.3% of men and 10.4% of women do it with their
EQ main character.
Of the figures in the main table above, it looks like maybe the
11.8% of 6) might consider gender-switching if they perceived a
benefit; lower prices for female avatars would presumably count as a
benefit for many of them. It's possible that some of 1) might
switch, too (as "no desire" to switch isn't the same as "a desire
not" to switch). The rest, though, aren't going to play a character
of the opposite gender unless their attitudes change.
In other words, they don't devalue female avatars because such
avatars are taken less seriously (even if they are); rather, they
devalue them because for social and psychological reasons they don't
want to play a character that is a gender other than their own.
Note that none of the respondents in the MOOs claimed they didn't
want to switch genders because of perceived inadequacies in the new
gender. That isn't to say such a view doesn't apply in EQ, though.
> When games are set up so that the gender of the avatar makes no
> difference for its abilities this is discrimination against
> females and it does indeed put a price tag (with a considerable
> margin of error) on it.
So because men have to pay more if they want to buy an avatar of
their biological sex, that's discrimination against women?
These sexism arguments never cease to amaze me.
> Better quantifying and better blind testing of the data should
> allow more confidence in the results
I agree that the results could be made more robust, although in the
end I expect that female avatars really do cost less to buy at
auction than equivalent male characters, at least for a
male-dominated player base.
> The various alternative explanations of course should be examined
> to see if, and by how much, they contribute to the price
> difference
Again, I agree. Unfortunately, it doesn't really matter now that a
"men devalue women" headline has got loose.
Richard