iiinova — where your digital self keeps evolving

Dennis Alexis Valin Dittrich @davdittrich.fediscience.org.ap.brid.gy · 03/06/2026

Is Economics Finally Becoming Trustworthy? www.econlib.org/econlog/economics-f… "Economics, along with political science, looks relatively good by this measure, with about half of articles sharing data or code, compared to less than […] [Original post on fediscience.org]

A three-panel chart detailing replication rates across six academic fields (Political science, Economics, Sociology, Psychology, Business, and Education) and an "All fields" summary.

Panel a ("Attempts (%)"): Shows the percentage of studies where a reproduction was "Attempted" (dark blue) versus "Not attempted" (light grey). Political science has the highest attempt rate (around 20%), while all other fields remain under 10%.

Panel b ("Attempts (%)"): A stacked bar chart of attempt outcomes: "Precisely reproduced" (dark blue), "Approximately reproduced" (light blue), and "Not reproduced" (magenta). Economics and Political Science lead with precise reproduction rates over 50%, while Education shows 0% precise reproductions.

Panel c ("Counts"): A dot plot showing total study counts (0 to 150) color-coded by outcome, revealing that Psychology and Business have the largest total sample sizes in the dataset.

The chart highlights a dual-layered reality in social science replication. While fields like Economics and Political Science are much more proactive at attempting to reproduce past research (Panel a) and boast the highest rates of precise reproducibility when they do try (Panel b), the vast majority of published papers across all fields are still never attempted. When attempts are made, roughly half across all fields can be precisely reproduced, leaving a significant portion that can only be partially replicated or fail entirely.

012 2

Dennis Alexis Valin Dittrich @davdittrich.fediscience.org.ap.brid.gy · 03/06/2026

Remote Work Leaves Younger Workers Sidelined libertystreeteconomics.newyorkfed.o… "…youth #unemployment has risen dramatically since the pandemic—as has the prevalence of remote work … remote work […] [Original post on fediscience.org]

A line graph tracking changes in "Percentage points" on the y-axis (ranging from -2 to 4) over time on the x-axis (from 2015 to 2024). The graph compares two categories, each plotted with a line and a shaded confidence interval: "Remotable occupations" (black line with circular markers and grey shading) and "Non-remotable occupations" (blue line with triangular markers and light blue shading). Both lines converge at 0 in 2019. A vertical grey shaded block highlights the pandemic period around 2020 and 2021. In 2020, both lines spike sharply, but the blue line peaks much higher (around 2.5) than the black line (around 1.5). By 2021, the blue line drops sharply and remains lower, while the black line for remotable occupations steadily climbs upward from 2022 through 2024, ending near 1.5 percentage points.

The chart illustrates a distinct post-pandemic divergence in employment trends based on how adaptable a job is to remote work. While non-remotable occupations saw a massive, immediate spike in 2020 followed by a swift stabilization, **remotable occupations have experienced a sustained, upward trajectory in percentage points from 2022 through 2024**. This suggests a structural, long-term shift or growing challenge within the remote job market compared to traditional, in-person roles.

000 0

Dennis Alexis Valin Dittrich @davdittrich.fediscience.org.ap.brid.gy · 04/05/2026

The happiness crash of 2020 www.econstor.eu/bitstream/10419/339… Before 2020, Americans were a fairly cheerful group, but national #happiness has recently taken a historic crash. By analyzing decades of survey data, this study shows […] [Original post on fediscience.org]

A line graph titled "Happiness 1972-2018 and 2021-2024." The y-axis ranges from 0 to 30, and the x-axis shows years from 1970 to 2025. A black line tracks the percentage difference between people reporting they are "very happy" versus "not so happy." For decades (1972–2018), the data fluctuates mostly between 15 and 30, with a historical average of 22.01 indicated by a dashed horizontal line. There is a gap in the data for 2019–2020. In 2021, the line plummets to approximately -2, marking the first time the value is negative. While it recovers slightly by 2024, it remains historically low at around 6. Source: GSS variable happy.

The chart highlights a "happiness crash" that broke a nearly 50-year trend:

Historical Stability: From 1972 to 2018, Americans were consistently more likely to say they were "very happy" than "not so happy," maintaining a steady average of about 22 points.

The Post-2020 Pivot: The data for 2021 shows a shocking reversal where, for the first time in the recorded history of this survey, the number of "not so happy" people actually outweighed the "very happy" group (resulting in a negative score).

The "New Normal": Although there is a slight upward tick between 2022 and 2024, the current happiness level (approx. 6) is still significantly lower than even the worst historical dips seen in the mid-80s or 2010. This suggests a lasting shift in the national mood following the events of 2020.

012 2

Dennis Alexis Valin Dittrich @davdittrich.fediscience.org.ap.brid.gy · 23/04/2026

Why did people stop responding to federal economic surveys? www.brookings.edu/articles/why-did-… Declining response rates challenge the precision and bias of economic indicators like […] [Original post on fediscience.org]

A line chart titled "Establishment survey response rates" shows data from 2015 to early 2025. The y-axis measures the percentage response rate from 0% to 80%. Five colored lines represent different Bureau of Labor Statistics (BLS) surveys: CPI-C&S (dark blue), CES (light orange), ECI (light blue), IPP-Exports (yellow), and IPP-Imports (dark orange). All five lines show a steady downward trend. In 2015, most rates were between 55% and 72%; by 2025, they have dropped significantly, with some surveys (IPP) crashing to around 30%. A sharp, synchronized drop is visible across most categories in early 2020. Source: Bureau of Labor Statistics / Brookings.

The chart reveals a growing "crisis of participation" in economic data collection:

The "COVID Cliff": Every survey experienced a sharp, permanent drop in early 2020. While we might have expected a recovery post-pandemic, the rates continued to stagnate or slide further, suggesting a fundamental change in how businesses prioritize government requests.

A Growing Blind Spot: The IPP (Import/Export Price Indexes) surveys are in the most trouble, plummeting from roughly 55% to 30% in just a decade. When only 1 in 3 businesses respond, the data becomes much "noisier" and less reliable for tracking inflation.

The "Gold Standard" is Fading: Even the CPI (Consumer Price Index) and ECI (Employment Cost Index)—critical tools for setting interest rates and wages—have fallen from the 70% range to the 45%–50% range.

000 0

Dennis Alexis Valin Dittrich @davdittrich.fediscience.org.ap.brid.gy · 23/04/2026

Our love of predictability may hinder adaptation to an unstable world jeroenvanbaar.substack.com/p/how-un… Many people feel a strong need to predict the future and get stressed when they cannot control events. While #anxiety […] [Original post on fediscience.org]

000 0

Dennis Alexis Valin Dittrich @davdittrich.fediscience.org.ap.brid.gy · 29/03/2026

robscale 0.5.3 is now on CRAN: A major update since the last public release (0.2.1). New: 11 robust estimators with confidence intervals, plus a variance-weighted ensemble combining seven scale statistics via bootstrap. Newton–Raphson replaces scoring […] [Original post on fediscience.org]

Three-panel speedup plot. Panel A: robScale, robLoc, adm vs revss (4–5× at small n, 2–4× at large n). Panel B: Qn, Sn vs robustbase (up to 9× at large n). Panel C: GMD vs GiniDistance, IQR vs stats (37× at small n), MAD vs stats (26× at small n). Sample sizes n=3 to n=10 million.

010 1

Dennis Alexis Valin Dittrich @davdittrich.fediscience.org.ap.brid.gy · 09/03/2026

Robust estimation demands highly efficient computation, especially in streaming anomaly detection where latency budgets are tight. While Rousseeuw & Croux's robust estimators ($Q_n$ and $S_n$), and Rousseeuw & Verboven's M-estimators of location and scale […] [Original post on fediscience.org]

A two-panel benchmark charting performance multipliers of the optimized C++ robscale package against legacy pure-R implementations across sample sizes from $n = 3$ up to $10^7$, with the vertical axis starting honestly at 0x. The left panel reveals a massive speedup for M-estimators (robLoc, robScale, adm vs. revss), pushing up to ~28x for robScale. The right panel tracks scale estimators ($Q_n$, $S_n$ vs. robustbase), showing the speedup curve upward from 1.6x, approaching 10x at large sample sizes. Shaded ribbons show 95% bootstrap confidence intervals, visually confirming dramatic computational efficiency.

221 2

Dennis Alexis Valin Dittrich @davdittrich.fediscience.org.ap.brid.gy · 05/01/2026

Malthus www.smbc-comics.com/comic/malthus #smbc #economics …was reminding me of: "Cependant, qui sait? La terre a des limites, mais la bêtise humaine est infinie!"

WHY DO PEOPLE KEEP BELIEVING IN MALTHUSGIAN THEORIES WHEN THEY KEEP NOT PANNING QUT?
THE NUMBER OF PROBLEMS THAT SEEM MALTHUSIAN GROWS GEOMETRICALLY, WHILE THE NUMBER OF MALTHLS DEBOUNKINGS
ONLY GROWS LINEARLY.

011 1

Dennis Alexis Valin Dittrich @davdittrich.fediscience.org.ap.brid.gy · 29/10/2025

Things versus People: Gender Differences in Vocational Interests and in Occupational Preferences www.sciencedirect.com/science/artic… HT @SteveStuWill "On average, men are more interested in working with things, whereas women […] [Original post on fediscience.org]

003 3

Dennis Alexis Valin Dittrich @davdittrich.fediscience.org.ap.brid.gy · 23/09/2025

The narrative around transformative productivity gains from #AIassisted coding seems not supported by empirical evidence, and is causing real harm. Despite widespread claims of 10x #productivity gains from #AI tools, there seems to be no evidence of a surge […] [Original post on fediscience.org]

010 1

Dennis Alexis Valin Dittrich @davdittrich.fediscience.org.ap.brid.gy · 23/06/2025

Large Language Models, Small Labor Market Effects bfi.uchicago.edu/wp-content/uploads… "#AI chatbots have had no significant impact on earnings or recorded hours in any occupation, with confidence intervals ruling out […] [Original post on fediscience.org]

This figure displays the average perceived earnings impact of AI chatbots among adopters, categorized by workers’ occupations and whether their employers encourage AI. They are all close to zero.

001 1

Dennis Alexis Valin Dittrich @davdittrich.fediscience.org.ap.brid.gy · 27/05/2025

A not so great observation for #experimentalEcon: "Results indicate that misunderstanding is common: the proportion of participants who misunderstood ranged from 22 % (Dictator Game) to 70 % (#Trust Game) in the online samples and from 22 % (Dictator Game) to […] [Original post on fediscience.org]

Comprehension in economic games

Many disciplines rely on economic games to measure prosocial behavior. However, there is a concern that participants may misunderstand these games, complicating interpretation of results. This study combines online and laboratory data (total n = 1568) to assess subject comprehension of five standard economic games: the Dictator Game, Ultimatum Game, Trust Game, Public Goods Game, and Prisoner's Dilemma. The online and lab data collections are carried out separately and for the online data collection we collect data for two separate platforms (Prolific and CloudResearch’s MTurk Toolkit). Within each data collection participants carry out all five games, and are randomized to comprehension questions with or without incentives for correct answers. Results indicate that misunderstanding is common: the proportion of participants who misunderstood ranged from 22 % (Dictator Game) to 70 % (Trust Game) in the online samples and from 22 % (Dictator Game) to 53 % (Public Goods Game) in the lab sample. Incentivizing the comprehension questions had no significant impact on misunderstanding, but numeracy was associated with lower misunderstanding. Misunderstanding also predicted increased prosocial behavior in several of the games. Our findings suggest that misunderstanding may be important in explaining prosocial behavior, making it more complicated to draw clear inferences about social preferences from experimental data.

020 2

Dennis Alexis Valin Dittrich @davdittrich.fediscience.org.ap.brid.gy · 26/04/2025

The salary bump that people who switch jobs used to command has vanished (in the US) archive.ph/McbH0 #wages #LaborMarkets

000 0

Dennis Alexis Valin Dittrich @davdittrich.fediscience.org.ap.brid.gy · 06/04/2025

via ‪@economeager.bsky.social‬ today we will all read imbens 2021 on statistical significance and p values, which is a strong contender for having the best opening paragraph of any stats paper pubs.aeaweb.org/doi/pdf/10.1257/jep… #statistics #economics #significance

Statistical Significance, p-Values, and the Reporting of Uncertainty

Theodor Geisel, better known by the nom de plume Dr. Seuss, published The Sneetches in 1961. In this children’s story, the Star-Belly Sneetches viewed themselves as superior to the Plain-Belly Sneetches. When the character of Sylvester McMonkey McBean arrives with a machine that can add or remove belly stars (for a modest fee), social upheaval results. In empirical work in economics, stars have long been attached to numbers in tables and figures to indicate the level of statistical significance: one star typically refers to an estimate that is statistically significant at a 10 percent level; two stars, the 5 percent level; and the coveted three stars, the 1 percent level. In the word of Dr. Seuss: “Those stars weren’t so big. They were really so small./You might think such a thing wouldn’t matter at all./But, because they had stars, all the Star-Belly Sneetches/Would brag, ‘We’re the best kind of Sneetch on the beaches.” In empirical studies, estimates with one, two, or three stars are often viewed as superior to those without such adornments.

140 4

Dennis Alexis Valin Dittrich @davdittrich.fediscience.org.ap.brid.gy · 17/03/2025

Half-Past Four is the New Five O’Clock in More Efficient Workday archive.ph/PNdyP The average American workday now ends 42 minutes earlier than two years ago. Despite the shorter workday, overall productivity has increased by about 2%. #laborEconomics #work

000 0

Dennis Alexis Valin Dittrich @davdittrich.fediscience.org.ap.brid.gy · 13/02/2025

via @flowingdata Dishonest charts are coming from all directions, and they're only going to hit harder. Unless we push back. So, an interactive guide: Defense Against Dishonest Charts flowingdata.com/projects/dishonest-… #dataviz

001 1

Dennis Alexis Valin Dittrich @davdittrich.fediscience.org.ap.brid.gy · 27/01/2025

Work Cohorts Americans with the same education, income, work hours, and commute as you do. flowingdata.com/2025/01/27/work-coh… #LaborMarkets #dataviz #rstats

000 0

Dennis Alexis Valin Dittrich @davdittrich.fediscience.org.ap.brid.gy · 09/01/2025

What We’ve Learned About Money and Happiness econlife.com/2025/01/money-and-happ… "Predictably, our short-term #happiness provides information about our preferences (and I assume our demand). By contrast, looking at the long run takes us to […] [Original post on fediscience.org]

000 0

Dennis Alexis Valin Dittrich @davdittrich.fediscience.org.ap.brid.gy · 23/12/2024

via @MiguelHernan: Upgrade your #causalinference arsenal. A revision of our book "Causal Inference: What If" is available at miguelhernan.org/whatifbook Thanks to everyone who suggested improvements, reported typos, and proposed new citations and […] [Original post on fediscience.org]

003 3

Dennis Alexis Valin Dittrich @davdittrich.fediscience.org.ap.brid.gy · 16/12/2024

via Florian Ederer: Return-to-office (RTO) mandates lead to abnormally high employee #turnover, especially for female/more senior/more skilled employees. It also takes significantly longer to fill these job vacancies after #RTO mandates […] [Original post on fediscience.org]

Return to Office Mandates and Brain Drain

By tracking over 3 million tech and finance workers’ employment histories reported on LinkedIn, we analyze the effect of S&P 500 firms’ return-to-office (RTO) mandates on employee turnover and hiring. We find that these firms experience abnormally high employee turnover following RTO mandates. The increase in turnover rates is more pronounced for female employees, more senior employees, and more skilled employees. Further, it takes significantly longer time for these firms to fill their job vacancies after the mandates. Their hire rates also significantly decrease. These results are consistent with firms losing their best talent and female employees and facing greater difficulties with talent attraction after RTO mandates. Our study highlights brain drain as a significant cost of RTO mandates even for the largest firms in the world. Keywords: Return-to-office Mandate, Employee Turnover, Employee Hiring, Human Capital

010 1

Dennis Alexis Valin Dittrich @davdittrich.fediscience.org.ap.brid.gy · 05/11/2024

“Before the #experiment, managers estimated #hybrid would reduce productivity by 2.6%. After the six-month experiment they estimated it increased #productivity by 1%. Those working under the hybrid model had a higher satisfaction rate, and 35% lower […] [Original post on fediscience.org]

000 0

Dennis Alexis Valin Dittrich @davdittrich.fediscience.org.ap.brid.gy · 24/10/2024

via @ckrafftc Research in Labor Economics (RLE) is planning a volume highlighting research on New Developments in Labor Economics. The editors Solomon Polachek and @ben_elsner are soliciting up to ten new papers showcasing new developments in #laborEconomics […] [Original post on fediscience.org]

CALL FOR SUBMISSIONS:
Research in Labor Economics (RLE)

Submission Deadline
November 29, 2024
Notification Deadline
December 15, 2024
Research in Labor Economics (RLE) is planning to publish a volume highlighting research on New Developments in Labor Economics.

The editors Solomon Polachek (SUNY Binghamton and IZA) and Benjamin Elsner (University College Dublin and IZA) are soliciting up to ten new papers showcasing new developments in labor economics. They welcome submissions tackling new questions as well as submissions that look at traditional labor economics questions with new methods and/or interesting new data.

The editors seek to publish papers from a broad range of topics, for example:

The impact of new technologies on the labor market
New findings on the influence of labor market institutions
New developments in research on migration
Measuring labor market dynamics with new data
Taxation, labor supply, and inequality
Education, skill formation, and the labor market
Gender and the labor market
New developments in personnel economics
In order to meet the publication deadline, please submit a detailed abstract (full papers are even more desirable) by November 29, 2024, to the RLE website:

000 0

Dennis Alexis Valin Dittrich @davdittrich.fediscience.org.ap.brid.gy · 10/10/2024

A New Indicator of Labor Market Tightness for Predicting Wage Inflation libertystreeteconomics.newyorkfed.o… The "Tightness Index of quits & vacancies per searcher […] [Original post on fediscience.org]

000 0

Dennis Alexis Valin Dittrich @davdittrich.fediscience.org.ap.brid.gy · 03/10/2024

#CfP: Cyborg workers 2.0 The past, present & future of automated labour - the social & environmental costs of big tech, Feb 13-14, Brussels www.etui.org/events/cyborg-workers-… The main goal & ambition of the conference is to challenge the framing […] [Original post on fediscience.org]

International Conference
Cyborg workers 2.0
The past, present and future of automated labour - the social and environmental costs of big tech
13-14 February 2025

000 0

Dennis Alexis Valin Dittrich @davdittrich.fediscience.org.ap.brid.gy · 30/09/2024

via Nature: Scaling up & shaping up large language models increased their tendency to provide sensible yet incorrect answers at difficulty levels humans cannot supervise, highlighting the need for a shift in #AI design towards #reliability according to […] [Original post on fediscience.org]

This is figure 1, which shows key indicators for several models in GPT (OpenAI), LLaMA (Meta) and BLOOM (BigScience) families.

000 0

Dennis Alexis Valin Dittrich @davdittrich.fediscience.org.ap.brid.gy · 25/09/2024

The Arrival of Field Experiments in Economics conversableeconomist.com/2024/09/23… "… experiments shouldn’t be designed with the “petri dish” mindset, but instead can be designed to think in advance […] [Original post on fediscience.org]

000 0