wat0n wrote:I don't see why not. Abadie et. al. (2010 and 2003) project beyond 10 years. Abadie's the guy who came up with the method.
The intervention period in Abadie's paper is 1989-2000. They directly quote Abadie for justifying the 10-year period:
"Moreover, a decade-long period after the passage of Proposition 99 seems like a reasonable limit on the span of plausible prediction of the effect of this intervention."
I personally find it quite generous given the pre-treatment period is only 22 years.
wat0n wrote:All the ones that are selected by the optimizer.
Abadie (2004) gives a reason for restricting the donor pool:
"Even if there is a synthetic control that provides a good fit
for the treated units, interpolation biases may be large if the
simple linear model presented in this section does not hold over
the entire set of regions in any particular sample. Researchers
trying to minimize biases caused by interpolating across regions
with very different characteristics may restrict the donor pool to
regions with similar characteristics to the region exposed to the
event or intervention of interest."
It's clearly motivated by Abadie (2014) using OECD countries with the same variables though. There's a more detailed discussion in that paper:
https://onlinelibrary.wiley.com/doi/epd ... ajps.12116wat0n wrote:Synth controls are really cool but it seems like a waste to limit the donor pool this much. One may also wonder the same about the donor variables, save for those that are related to the outcome by construction (e.g. including nominal GDP per adult as a predictor for real GDP per adult). The only real reason to do it, if I were him, is to get done with my undergrad thesis ASAP - because a high-dimensional matrix would take forever to optimize and he's also supposed to do all the placebo tests, which would take even longer. But I'd then be more skeptical about it.
The problem doesn't seem large enough for computing power to be an issue. As mentioned, they use the growth predictors recommended by Abadie (2014).
wat0n wrote:Yeah, Saez and Zucman have been pushing for this since forever. I'll check it out later, but quite frankly even the literature on wealth inequality is rather... Unclear. Depending on how you define wealth - and there are many valid ways to do it - you can conclude anything about it's evolution in the US. The key issue is to decide whether Social Security should be counted as being a form of wealth or not - the money is taxed from you (so from that point of view it's not yours, since you don't have an individual account), yet the law does mandate that people who contribute are to be paid (usually the money others owe as counts as an asset - why wouldn't the same hold for Social Security? Isn't that a Government liability? If so, then it's someone else's asset). And this is important because a big motivation is to address the "disproportionate increase in wealth inequality".
Um...I don't see how social security is relevant for wealth taxes.
wat0n wrote:This also raises issues on how would "wealth" be defined for taxation purposes.
There might be practical issues with measuring all types of wealth. But isn't a capital income or gains tax incomplete in that regard as well?
wat0n wrote:I wouldn't count consumption taxes as a tax on labor. But I agree, there are also payroll taxes but those are generally following the "broad base and low rate" principle (usually the opposite is done with capital, too).
I think in Europe labor is effectively taxed at a higher rate than capital, can't find the source at the moment.