Paul Graham: Silicon Valley Favors Obama Over Romney 2 To 1
forbes.com
forbes.com
An interesting followup to that question may be to what degree do you think that either of the candidates can grow the economy? This would be to ascertain not just who may be better but also how good is the candidate. Do they believe that both candidates are bad and they are simply choosing the lesser of two evils? Lastly, I would ask, "what historical evidence do you rely on to make your decision?" I'd like to ask this to determine how these people are making their decisions. For example, do they have an explanation for why Obama hasn't grown the economy in the last four years or an explanation for why Romney's past performance qualifies him. Or is the explanation more humanistic: "Obama seems like a more trustworthy guy."
If Romney actually was what he tries to seem, he'd be more popular in the Valley. That's what makes this poll significant. The leaders of the startup community don't have the natural affinity they'd have for him if they believed his claims. Which in turn is fairly good evidence they're false.
These aren't small issues. They are reasons people vote the way they do. They have nothing to do with growth.
Also, just because you've run a successful investment firm or company, doesn't mean you understand well the government's role in creating or hindering growth. I consider those pretty different skill sets.
Also, the issues that are commonly identified as problems for growth in our industry aren't being addressed by either candidate. These include immigration reform, patent reform, copyright reform, updating regulations around privacy and health records, etc.
Also, those that are already rich have a smaller marginal utility for money, and don't personally care as much about taxes. As an illustrative example, for me personally, a lower cap gains rate directly translates to more angel investing.
The common mistaken opposition to low cap gains is that tax rates don't motivate people. The real reason to support low rates is how the increased return would be re-invested. This yields yet another idea no one is proposing: lowering income taxes to match cap gains rates (as opposed to the more common reverse to close the gap).
Finally, presidents don't really drive growth, so the entire premiss of your question is incorrect imho.
> If Romney actually was what he tries to seem, he'd be more popular in the Valley.
These two statements are contradictory.
Suppose a restaurant is unpopular, due to bad food, bad service, high cost, whatever. There's no contradiction in expecting it would become more popular if a movie star started to hang out there every night.
We're now arguing about basic logic. I've found threads are usually irrecoverable at that point, so if you don't mind, I'm done.
I'm sorry if you do not like these kinds of arguments as I try to involve as much base logic as I can. I think that base logic is to arguments what functions are to algorithms. Furthermore, sticking to logic results in an argument at least on the level of DH4 in your essay "How to Disagree."
Relevant > http://en.wikipedia.org/wiki/Multicollinearity*
___________
[1] A more detailed explantion is as follows:
Notwithstanding the substance (ie, the politics), we see an argument of the form:
Binary (0,1)=F(X1...Xn) | (challenger, incumbent)
ie
Vote = Function of (...)
So this equation is approximated:
(0,1) = f X1(party affiliation)+X2(vote in last election)...XN-1(Growth)...Xn(whatever else).
At first blush, its not clear ex-ante why this wouldn't be dominiated by the first two explanatory variables. At second blush, it seems obvious that any beliefs further out in the vector will be likely correlated with the general set of beliefs captured in the first two revealed preferences (ie, X1, X2).
Not only are thes Xn variables (eg, growth) a series of beliefs* (not actual observed data), those beliefs will be (predictably) co-lineated with other variable that are (exAnte) more likely to have greater predictive power.
So, you can't safely say Xn-1 is either independent or by itself stonger than X1 or X2. [#,##] And if that's the case, there really is not much to this. A poll of this type is the wrong form of argument to support your conclusion (as it is not strong enough).
There are other levels of logical and orthogonal attack on this general argument (will spare the ink), but in any event these too would be better avoided/adressed in a written essay on the subject. You may also have other levels of logical and orthogonal attack that support your general approach, which would be interesting to read.
_______________________________
[#] For the technical statistcal reference:
http://en.wikipedia.org/wiki/Multicollinearity
<Consequences of multicollinearity>
As mentioned above, one consequence of a high degree of multicollinearity is that, even if the matrix XTX is invertible, a computer algorithm may be unsuccessful in obtaining an approximate inverse, and if it does obtain one it may be numerically inaccurate. But even in the presence of an accurate XTX matrix, the following consequences arise: In the presence of multicollinearity, the estimate of one variable's impact on y while controlling for the others tends to be less precise than if predictors were uncorrelated with one another. The usual interpretation of a regression coefficient is that it provides an estimate of the effect of a one unit change in an independent variable, , holding the other variables constant. If is highly correlated with another independent variable, , in the given data set, then we only have observations for which and have a particular relationship (either positive or negative). We don't have observations for which changes independently of , so we have an imprecise estimate of the effect of independent changes in.
[##] With respect to the facts/data, the sub-population reported will predictavly over-represent in near proportion to reported data merely on pre-existing party affiliation (or: revealed preference). And historically, this sub-population will have already historically over-represented for the incumbent. So ex-ante, we expect nested logic and colinearity in Xbar.
So then comes the element of differentiation:
1) In general english, if you support a candidate you most likely support his position on Xn.
2) In general english, if you support a party you most likely support the Party Candidate.
3) In general english, if you support a party you most likely support the party position on Xn.
4) In general english, if a party supports a candidate, it most likely support the candidates position on Xn.
5) In general english, if you previously supported a candidate you most likely support his position on Xn.
Etc. While these may or may not be true, the issue is one needs a deeper analysis. The problem is one of differentiation and granularity from the data. If the sub-sample was neutral in ex-ante preference, this would help, but we know its not (admittedly). So its all muddled.
-- The Onion.