The Stimulus Is A Failure

[KrazyHorse]

That's probably because you're working with models which are fairly well behaved.

Model uncertainty in HE is NOT a systematic error. Different plausible models give such wildly different answers that it's unreasonable to include model uncertainty in standard error analysis.

[KrazyHorse]

There is such a thing as theory uncertainty, but that derives from an understood limitation to current calculations...

[KrazyHorse]

The whole point of generating a list of multiplier values for different policy options (such as issuing food stamps or spending on roads) is to reduce the effect of those policies on economic activity to a single, easy to compare number.

THIS IS PRECISELY THE POINT.

CURRENT ESTIMATES OF FISCAL MULTIPLIERS ARE SO BAD THAT TO USE THEM FOR THIS PURPOSE IS RETARDED.

If you need a confidence interval on top of that number, then you have generated the multiplier number incorrectly.

This makes absolutely no sense. A fiscal multiplier has a precise definition, but is not directly observable. Even directly observable quantities require confidence intervals. ****.

The purpose of the multiplier number is comparison.Higher numbers yield more economic activity. Period. No other prediction is meaningful.

You have no idea what you're talking about. The purpose of a fiscal multiplier is PREDICTION. It answers the question "if the government spends an additional dollar in such a way, then what is overall economic activity compared to a counterfactual world where the government does not spend that additional dollar".

If your goal is to create the largest amount of additional economic activity for a set number of dollars (technically, this number must be small, as the multiplier is a measure of marginal elasticities) then using the fiscal multiplier to compare various policy options is, of course, ideal.

However, I have no idea why you think that the above is the ONLY meaningful application of a fuscal multiplier.

Are there possible outcomes where a lower multiplier generates more activity than a higher multiplier? Of course.

No. In fact, by definition this is impossible. You ****ing ******.

If we were going to choose only one single form of stimulus, then confidence might be important.

Confidence intervals are ALWAYS important. Understanding of how good your knowledge is is JUST AS IMPORTANT AS THE KNOWLEDGE ITSELF.

But we are not going to do that. The purpose of this multiplier is to establish relative priority of stimulus policies.

Even if this claim was true (which it is not; there are multiple possible uses for knowledge of fiscal multipliers, among which dollar for dollar stimulus direction is only ONE) then understanding of how good your knowledge is IS KEY. The policy implications for the statement that "food stamps have a fiscal multiplier of 1.74 +/- 0.1 while these tax cuts have a fiscal muliplier of 1.24 +/- 0.2" is FAR DIFFERENT than if you multiply both of those uncertainty estimates by ten. Not to mention the fact that the confidence interval may be asymmetric, AND the fact that different methods of estimating fiscal multipliers may even generate different multiplier orderings. So by publishing something as monumentally information-free as the statement "the fiscal multiplier for food stamps is 1.74" Moody's is failing epically in its duty as an informative source.

Zandi and Barrons both used the number in this manner, as an index for comparison. There is nothing wrong with their precision, even though it may not be accurate.

Let me ask you a question: what is your highest level of experience working in a quantitative, empirical field? Because this sentence tells me that you have no appreciation whatsoever for actual error analysis and are instead operating off of some sort of legalistic view of what constitutes proper scientific discipline.

On the other hand, there is no reason to think it isn't accurate, especially in the short run. Which is what we are concerned with.

So the fact that different estimates arising from different methodologies/models of fiscal multipliers are off by an order of magnitude isn't a good enough reason to think that your favourite estimate isn't accurate?

You ****ing dimwitted troglodyte.

[Lorizael]

****ing cool it, KH.

[Jon Miller]

I am well aware that you can get very different results depending on the model (This is true in Nuclear too). That is why I said that if you can't include the model uncertainty in your calculation, that you present the model that was used and estimate the systematic uncertainty that would be associated with it.

Whether you include it in the uncertainty that you quote or not, it is a systematic uncertainty.

JM

[KrazyHorse]

**** no.

[KrazyHorse]

I am well aware that you can get very different results depending on the model (This is true in Nuclear too). That is why I said that if you can't include the model uncertainty in your calculation, that you present the model that was used and estimate the systematic uncertainty that would be associated with it.

Whether you include it in the uncertainty that you quote or not, it is a systematic uncertainty.

JM

Again, this doesn't make sense in the context of HE. The whole point there is to compare and contrast the different models. To do this you use data to compare to predictions made by each model. Saying that the model "uncertainty" impacts your systematic errors is ridiculous.

[KrazyHorse]

Jon, there are numbers which you report for data under a certain choice of viable models which don't even have a meaning in other models.

[KrazyHorse]

For example, electroweak predictions of the Higgs mass under SM assumptions. How would you include "systematic error" on this "measurement" due to the fact that we may live in a universe where some version of technicolor is true?

[Jon Miller]

Obviously you are looking at comparing the predictions of each model.

The difference between the models relates to the systematic uncertainty in the results from that model. For example.

I have a feeling that you aren't understanding my posts.

Nowhere did I say that you should include the systematic uncertainty from the model as part of the quoted systematic uncertainty.

JM

[Jon Miller]

For example, electroweak predictions of the Higgs mass under SM assumptions. How would you include "systematic error" on this "measurement" due to the fact that we may live in a universe where some version of technicolor is true?

As I said, you give some estimation if you can. Obviously in some cases you can't give an estimation.

A nuclear physics example being QWeak (Which I am not looking to take part of, but have considered). If the SM is incorrect for the measurement, it and some of the other data wouldn't even be measuring the same thing.

JM

[KrazyHorse]

The difference between the models relates to the systematic uncertainty in the results from that model.

I think that we're talking at cross purposes. The idea is that a systematic error leads to an estimate of a quantity which diverges from the long run mean of measurements of that quantity.

If a quantity has no meaning in model A but does in model B then it seems ridiculous to me to claim that an estimate of that quantity from model A is subject to systematic error due to the possibility that model B is more correct.

Nowhere did I say that you should quote the systematic uncertainty as part of the quoted systematic uncertainty.

I have no idea what this is supposed to mean. I shouldn't quote systematic uncertainty as part of systematic uncertainty?

[KrazyHorse]

As I said, you give some estimation if you can.

Estimation? The higgs boson doesn't EXIST in technicolor. How the **** is that a "systematic uncertainty" on estimates of the SM higgs mass derived from precision EW data?

[KrazyHorse]

examples of systematic errors as usually understood in this community are theory uncertainty (from calculation limits) and experimental systematics (for example, a miscalibrated calorimeter or tracking system)

[Kuciwalker]

KH, is it possible to actually explain why we need the Higgs boson at roughly my level of understanding? All the news ever prints is "it explains why particles have mass", if they even say that much. I have no idea why we would need to explain that, if the explanation is even partially true.