Tim Watson

I have a new paper out today. Gen AI shows a pattern: it excels at some tasks but fails at seemingly similar ones. This "jagged" performance creates a fundamental problem for users trying to decide when to rely on AI. I explain why this happens and what it means. 👇🧵

@BenPhillips_ANU I got it funded and reestablished in the 2018 Women’s Economic Security Statement- But I don’t think it has ongoing funding at present. The next release is listed as ‘unknown’

@TimxWatson ‘The 2024 Time Use Survey (TUS) was conducted by the Australian Bureau of Statistics (ABS) between August and October 2024. Previous surveys were conducted in 1992, 1997, 2006 and 2020-21. ‘

ABS time use survey results out. Women doing more unpaid labour, men doing more paid labour. Women more rushed for time regardless of age - even more so as kids. WFH more likely to do unpaid labour at home (is it really unpaid!?) #data-downloads" target="_blank" rel="nofollow noopener">abs.gov.au/statistics/peo…

Forthcoming in the JEL: "How Do Central Banks Control Inflation? A Guide for the Perplexed" by Laura Castillo-Martinez and Ricardo Reis. aeaweb.org/articles?id=10…

🚨 New Paper and Public Good🚨 "The Global Macro Database: A New International Macroeconomic Dataset", joint with @chenzix, Mohammed Lehbib, and Ziliang Chen. We built the most comprehensive macro database ever—covering 243 countries from 1086-today, integrating 110 sources. 🧵

DeepSeek and what happened yesterday: Probably the largest positive tfp shock in the history of the world.

Don't miss out on this guide on how to run surveys: on creating your own identifying variation: more than 80 pages and an Appendix with examples including social desirability, vignettes, information treatments and more.

I live in a world not built for me, here is what I have learned buff.ly/4itPlKm

TREASURY HAS RELEASED ITS LATEST ESTIMATES OF THE COST OF TAX CONCESSIONS Exciting, huh? (I have a low excitement threshold. I am, after all, an economist.) To me the standout is that Treasury has hugely revised up its estimates of the cost of superannuation tax concessions. The cost of those concessions next year is now estimated at $64bn - some $9.3bn more than estimated less than a year ago. The bulk of the higher costs are due to a huge lift in the estimated cost of tax concessions on earnings within super. To be clear, it's hard to properly estimate the tax concessions given to super. To be even clearer, the concessions given to earnings within super are cray cray. That’s because parts of the tax system that make sense in isolation collide in silly ways inside super: (1) The tax system should encourage saving – and ours does through super. (2) The tax system should recognise that inflation can artificially inflate capital gains – and ours (indirectly) does through a capital gains tax discount. (3) And the tax system should avoid taxing the same income twice – as ours does through franking credits. But super funds get the benefit of all three of these. And super has a very heavy reliance on Australian shares, many of them with fully franked earnings. The upshot is that we’ve massively overdone the concessions. Super funds make earnings on which they have to pay 15% in tax, but then those same earnings often come with a 30% tax credit attached. Nice little earner, huh? No wonder that Treasury has had to massively revise up its estimates of the concession ... (More ranting here ... x.com/ChrisEconomist… and here x.com/ChrisEconomist…)

📢Very important new research from @unimelb's Dr @_yi_yang, Peter Summers, @ZoeAitken11 @Kavanagh_AM @gdisney_melb looking at mortality inequalities between people with and without disability. Targeted interventions and policy reforms are needed now. thelancet.com/action/showPdf…

#TTPI #Seminar Parenting Payment Single is a welfare payment for low-income single mothers in Australia. Does changing the payment eligibility have employment and earnings effects? @SobeckKristen taxpolicy.crawford.anu.edu.au/news-events/ev…

My truly wonderful friend and mentor – Geoff Carmody – has passed away. This lovely story by @Johnkehoe23 – afr.com/policy/economy…– gets to the heart of it: Geoff had an absolute passion for a better Australia. That meant no phone call from Geoff was ever short: he brimmed with ideas on everything from tax to super to aviation and tourism policy and the energy transition. (And no phone call ever finished without a joke – the man had an endless supply!) A better Australia is built on the back of people like Geoff and the passion they have. I miss him already.

@BlairWilliams26 You’ve just got to change companies- I get these notices all the time for insurance. All the companies do it, and you just have to switch

Some news coverage today suggesting the COVID-19 SME Cashflow Boost measure has never been formally evaluated- it has been! onlinelibrary.wiley.com/doi/10.1111/14…

@SHamiltonian The consistent approach to claiming Gov/ APS did worse than almost all independent research suggests is somewhat peculiar. My Cashflow Boost research is the least afflicted by external validity/ methodological concerns and shows even better results than JK! What to make of that?

Nice to see the COVID-19 Response Inquiry Report released today citing my research on jobs saved by JobKeeper and the SME Cashflow Boost Measure as an authoritative reference. A 🧵… pmc.gov.au/resources/covi….

@lachlanvass All the papers have irreducible external validity concerns of varying degree due largely to data availability at time of writing. With ours- some firms did better, some worse. We also present a linear structural model that simulates our wage results well. Yes context matters!

@TimxWatson I'm not sure saying "our numbers are similar to others" answers the external validity question raised by Borland and Hunt (2023). Our paper also has external validity concerns, which we flag transparently and apply a range to estimates to reflect

@lachlanvass External validity would be a concern if the results from my analysis were significantly different from other papers with different identification strategies. In short- maybe we should be having a discussion about why your results are so low, not why everyone else’s are higher! 😉

@TimxWatson Isnt the issue on the JK analysis the counterfactual construction - based on industry likely endogenous. And then also a question of external validity, as your paper notes you contstrain the sample to "employers that were significantly negatively affected"

Always happy to chat about research findings with fellow public servants if they are interested in an accurate reporting and contextualisation of the results. I think this is actually important for future crisis responses.

3. When you take out the control for SME Cashflow Boost my results are effectively the same as Bishop and Day (2020). 🤷‍♂️🤷‍♂️🤷‍♂️ In short, just because my JobKeeper results are towards the upper end of what others have found plausible does not imply they are ‘overstated’.