Greg Faletto

Greg Faletto@GregoryFaletto

0

1

316

Greg Faletto@GregoryFaletto·24 Şub

I'm excited to announce that {fetwfe} is now available on CRAN! cran.r-project.org/web/packages/f… You can now install {fetwfe} by simply using ```install.packages("fetwfe")``` While preparing the package for CRAN, I also fixed a couple little bugs and edge cases from the original version

NEW R PACKAGE: I created the {fetwfe} package to implement fused extended two-way fixed effects (FETWFE), my estimator for difference-in-difference under staggered adoptions. Link to GitHub and short thread below:

English

0

3

221

Greg Faletto@GregoryFaletto·21 Oca

You don't have to be part of some billionaire's political agenda. Come on over to the other site, it's great. You can find me with the same profile photo, and my handle there is my website domain

English

83

Greg Faletto@GregoryFaletto·18 Oca

@njhochman @mattyglesias Many people on the left use the word "privilege" to refer to the same thing that you're labeling "good fortune"

English

1

129

Greg Faletto@GregoryFaletto·9 Oca

@KhoaVuUmn

QME

129

Khoa Vu@KhoaVuUmn·8 Oca

Out of all ridiculous responses to the meme, this is the wildest one.

English

16

6

342

21.8K

Greg Faletto@GregoryFaletto·6 Oca

@haroldpollack Thanks Harold!

English

Greg Faletto@GregoryFaletto

0

21

Harold Pollack@haroldpollack·6 Oca

Looks like a valuable new resource.

NEW R PACKAGE: I created the {fetwfe} package to implement fused extended two-way fixed effects (FETWFE), my estimator for difference-in-difference under staggered adoptions. Link to GitHub and short thread below:

English

0

3

1K

Greg Faletto@GregoryFaletto·6 Oca

(By the way--I'm mostly on the other site now. My username there is my website URL. Same profile pic.)

English

65

Greg Faletto@GregoryFaletto·6 Oca

There are links to the paper and many more resources for FETWFE on the GitHub page. Please let me know if you have any questions or feedback! github.com/gregfaletto/fe…

English

0

72

Greg Faletto@GregoryFaletto·6 Oca

So, in my opinion, FETWFE has the best of both worlds: what we like about classical estimators (simple to use, asymptotically valid inference) and ML methods (flexible, adapts to your data to get the best possible estimates).

English

0

78

Greg Faletto@GregoryFaletto·6 Oca

It gets better--unlike many ML methods, the FETWFE estimates are asymptotically Gaussian, so you can get standard errors and valid confidence intervals. It's also relatively simple to implement--it only has a single tuning parameter.

English

0

60

Greg Faletto@GregoryFaletto·6 Oca

FETWFE uses machine learning to figure out which of these restrictions are okay to keep around and which aren't. By learning from the data which restrictions are justified and which aren't, FETWFE preserves unbiasedness while improving efficiency as much as possible.

English

0

2

67

Greg Faletto@GregoryFaletto·6 Oca

This ensures unbiasedness regardless of whether the restrictions hold, but at a steep price to efficiency. Probably the TE for cohort 3 at time 4 has a lot to do with the TE for cohort 3 at time 5. Most of the new methods discard that information.

English

0

1

70

Greg Faletto@GregoryFaletto·6 Oca

This can also be seen as a large number of small homogeneity assumptions, or restrictions (the TE for cohort 3 at time 3 equals the TE at time 4 AND at time 5 AND...) The newer estimators drop all these restrictions.

English

0

74

Greg Faletto@GregoryFaletto·6 Oca

Here's an explainer of what FETWFE brings to the table and why I think you should consider {fetwfe} instead of other estimators: TWFE is perfectly fine under staggered adoptions IF (and only if) a strong assumption on homogeneity of treatment effects holds.

English

0

71

Greg Faletto@GregoryFaletto·6 Oca

Here's the complete code for the data application from my paper. FETWFE is one of the many new estimators that is unbiased for difference-in-differences with staggered adoptions (unlike the classic two-way fixed effects estimator).

English

0

112

Greg Faletto@GregoryFaletto·6 Oca

github.com/gregfaletto/fe… {fetwfe} is really easy to get started with. Just provide a panel data set (formatted as described in the documentation), and you'll get an estimate of the overall ATT, the ATT for each cohort, and standard errors for all of these.

English

0

1

316

Greg Faletto@GregoryFaletto·6 Oca

NEW R PACKAGE: I created the {fetwfe} package to implement fused extended two-way fixed effects (FETWFE), my estimator for difference-in-difference under staggered adoptions. Link to GitHub and short thread below:

English

5

21

3.2K

Greg Faletto@GregoryFaletto·22 Ara

@otis_reid Yes it's been largely forgotten but there was a theory in the late 2000s that the only way to save the music industry from piracy was to make legal streaming more convenient (and affordable) and Spotify basically proved it right

English

Greg Faletto@GregoryFaletto

3

54

Otis Reid@otis_reid·22 Ara

This chart of the sources of music industry revenue also is super interesting — for all its faults, streaming has saved the industry

English

3

2

35

1.8K

Otis Reid@otis_reid·22 Ara

The cultural weight of the music industry relative to its financial size is pretty wild — it’s just a very small business

English

12

43

424

40.7K

Greg Faletto@GregoryFaletto·10 Kas

@asmartbear @tautologer And yes, that wealth can be partially tapped into in various ways without moving out. All I'm saying is that, one way or another, homeowners capture at best a fraction of the increase in the value of their home. First-time buyers feel every last penny. x.com/GregoryFaletto…

@asmartbear @tautologer Point being, I think part of why the affordability side has had snowballing success is that in a very literal, dollars and cents way, new home buyers who want to keep housing costs down have more to gain than incumbent homeowners who want their investment to appreciate in value.

English

0

2

22

Greg Faletto@GregoryFaletto·10 Kas

@asmartbear @tautologer Yes, the value doesn't evaporate at death. But if homeowners are thinking about what matters for their kids, the current state of affairs isn't great. Their kids can't buy a home now in their 20s/30s, it doesn't do a lot of good that one of them will inherit one in their 50s.

English

0

2

45

tautologer@tautologer·7 Kas

the fundamental contradiction at the heart of American politics is that housing costs can't go down while home values go up because they are the exact same thing

English

16

15

275

11.9K

Greg Faletto@GregoryFaletto·9 Kas

@asmartbear @tautologer Point being, I think part of why the affordability side has had snowballing success is that in a very literal, dollars and cents way, new home buyers who want to keep housing costs down have more to gain than incumbent homeowners who want their investment to appreciate in value.

English

50

Greg Faletto@GregoryFaletto·9 Kas

@asmartbear @tautologer Well in (a) the appreciation never meant anything to you, it was just paper wealth that stayed on paper until you died. So what did it matter? People do hope for (b). But of course everyone can't "beat the market," there can only be as many winners as there are losers.

English