Jörg Ankel-Peters
@jrgptrs.bsky.social
development & environmental economist. energy access, climate policy, replication & meta-science. RWI & I4R. #FirstGen bit.ly/40e2aQj
Environmental science
29%
Energy
16%
Pinned
Jörg Ankel-Peters
@jrgptrs.bsky.social
· Oct 11
Thought about scientific consensus recently? We have a new DP @i4replication.bsky.social that probes into the famous replication debate between Acemoglu, Johnson & Robinson (AJR) and Albouy - and how experts assess this debate. We find that they disagree. 1/8 www.econstor.eu/bitstream/10...
But Michael, they don't suggest it is a "conspiracy theory". Neither do they claim it is a particularity of migration research. In fact, in my reading, they rather use it as an example to make a more general claim. And they even transparently discuss that it is hard to interpret the effect size.
...because they want "speaking" titles. Which I think is often too simplistic and unnecessarily provocative.
...normative biases. They use migration as an example, and do not say it is uniquely biased, as Clemens insinuates. I agree about the effect size statement, but the authors make this very transparent, I think. Perhaps it is wrong to publish such a controversial paper in a Science/Nature journal...
...named the normative prior of the respective author first, on both sides of the ideological Rubicon... I would be in favor. But here it is a one-sided questioning of credibility. IMO. The other thing is that Borjas/Breznau, in my reading, do not claim that migration is particularly affected by...
No I haven't. It is a deeper one. I generally admire Michael Clemens' work, so I want to be cautious. In any case, the thread is strangely ad hominem. Of course it matters if Borjas had an "anti-migration" agenda. But it also matters if someone has a pro-migration agenda, right? So, if we ALWAYS...
Important new paper on normative bias in research. In a many-analyst design estimates produced by pro- vs anti-immigration teams systematically differ, driven by specification choices. It’s exploratory analysis expanding on a PNAS paper, but findings seem plausible. www.science.org/doi/epdf/10....
Also for the reasons mentioned in that quoted thread you think, like, people being "randomy abused for posting regular research results"? Is that playing out in the econ community, too? (I'm in my curious-mode, not my contrarian mode :-))
As long as we ponder limits from both sides of the normative rubicon, all is good.
Older working-age individuals are more likely to experience chronic illness, to retire early, or to be discouraged if. In this context, the relatively high share of Helfertätigkeiten (low-skilled jobs) among migrants is particularly relevant. This raises a few legitimate follow-up questions.
Export agriculture and rural poverty: Evidence from Indonesian palm oil
This paper measures the impacts of Indonesia’s palm oil export expansion on district poverty and household expenditure from 2002 to 2015. Identificati…
www.sciencedirect.com
Two minor updates, friends:
- As of 1 January 2026, I'm an Associate Professor.
- As of today, the first paper I started writing on palm oil, about a decade ago, is in print in the Journal of International Economics.
Read, share, cite. Open access: www.sciencedirect.com/science/arti...
- As of 1 January 2026, I'm an Associate Professor.
- As of today, the first paper I started writing on palm oil, about a decade ago, is in print in the Journal of International Economics.
Read, share, cite. Open access: www.sciencedirect.com/science/arti...
..haha... well, I4R is agnostic about signs of effects and effect sizes. So far you only pre-specified your claim - as long as you don't change your hypothesis ex-post or p-hack your weight... there will be no I4R-fury.
I agree with that of course. He would probably even make his case stronger by engaging with the credibility revolution (and its shortcomings).
Interesting indeed. Of course there is and has always been good descriptive work in economics. Some economists are smart after all, and smart people know how to do description. But *the paradigm* is obsessed with causality, and hence 90% of papers are merely "causal", and it is how we train people.
As far as the econ part of the social sciences is concerned, we need more papers that ignore the [largely underpowered] "credibility revolution" and rather think deeply about good description. I am therefore looking forward to reading it.
Reposted by Jörg Peters
Description, Causal Explanation, and Policy Intervention in Sociology - KZfSS Kölner Zeitschrift für Soziologie und Sozialpsychologie
I seek to restate and to develop a position I have previously taken up on the relation between description and causal explanation in sociology, in part by reference to an ongoing controversy on this m...
link.springer.com
New paper by John Goldthorpe, in the Kölner Zeitschrift für Soziologie und Sozialpsychologie, restating and developing his position on causality. link.springer.com/article/10.1...
Reposted by Jörg Peters
23 years ago today, one of the greatest died
In memory of Joe Strummer 🖤
... gone but not forgotten...
📸 Anthony Saint James
#punk #punks #punkrock #punklegend #joestrummer #theclash #punkrockhistory #history #otd
In memory of Joe Strummer 🖤
... gone but not forgotten...
📸 Anthony Saint James
#punk #punks #punkrock #punklegend #joestrummer #theclash #punkrockhistory #history #otd
Right - but why are economists so obsessed with unobservables in OLS while they keep on ignoring them in applied IV work?
Thanks - and indeed! From your (vast) experience, why do you think people keep ignoring it?
Reposted by Jörg Peters
Bottom line: IVs need a different default. Start from a pessimistic prior – i.e. assume exclusion is violated - and try to vigoriously falsifiy that claim. We call this the revised-prior falsificationist approach, and apply it to a seminal topographic IV paper, Dinkelman (2011). 6/6
Weak IVs are the other structural issue. It’s well established—but widely ignored in practice—that inference is not credible when instruments are weak and exclusion is even subtly violated: under weakness, those violations can generate more bias than the original OLS confounding. 5/6
Collider bias is well established in econometrics. Yet in practice, IV aficionados rarely acknowledge it — even when conditioning on variables that are likely colliders. 4/6
So, topography is everywhere in socio-economic causal webs – and thus exclusion is the exception, not the default. The usual fix is adding controls. But other IV aficionados instrument (i.e. explicitly suspect endogeneity in) many of those controls - so controlling for them risks collider bias 3/6
Topography as IV is our case in point. We comb through top 120 econ journals (2011–2023) and find 161 studies using topographic variation. Here’s what the literature collectively implies: a dense causal web in which topography touches far more than the isolated channel any single IV paper wants. 2/6