The Unassuming Economist

From a paper by by Adela Karapici and Arjeta Vokshi:

“This paper estimates the short-run relationship between real output growth and unemployment in Albania over 1992–2023 and asks how sensitive that relationship is to the way crisis episodes are dated. Albania offers a demanding test: a transition economy with high informality, extensive self-employment and sustained emigration, whose sample contains three major shocks in three decades. We estimate first-difference specifications in which the change in the unemployment rate is regressed on output growth, with a lagged dependent variable capturing persistence and separate indicators for the three crisis episodes. Because conventional standard errors are unreliable at this sample size, we report alternative estimators alongside bootstrap inference and a full set of residual and stability diagnostics. Our main practical result concerns crisis dating. Crisis dummies are conventionally dated to the year a crisis began internationally, but for a small open economy whose transmission is indirect, the year in which a shock reaches domestic activity need not be that year. Albania did not contract in 2008 or 2009. We date the financial-crisis indicator from an external chronology of transmission—credit growth, non-performing loans, exports, investment, remittances and the fiscal stance—assembled from contemporaneous institutional sources and fixed before estimation, which places the arrival of the crisis in the Albanian economy in 2009 and 2010. The estimated crisis coefficient is highly sensitive to that choice: a 2008-dated indicator produces the best-fitting specification in this paper on every information criterion together with a large negative coefficient; the conventional 2008–2009 window straddles the turning point and returns an estimate indistinguishable from zero; the 2009–2010 window returns a positive one. These coefficients are individually imprecise, and their intervals overlap, so we do not claim that the data identify any one dating as correct, and we do not treat the sign of an estimate as evidence for or against a window. Our claim is narrower: the substantive conclusion a reader would draw about the crisis is determined by a specification choice that the data cannot settle, so the window has to be fixed ex ante on transmission evidence and its consequences reported. We demonstrate this for Albania and do not assert that published work on other countries is affected; we specify the test that would settle it. The dating choice leaves the estimated Okun coefficient itself almost unchanged. That coefficient is negative in every specification, estimator, subsample and phase definition we report, with a normal-times impact estimate of approximately −0.21, smaller in absolute value than the range reported for advanced economies. Its statistical significance is sensitive to the mode of inference: under the null-imposed bootstrap we regard as most appropriate at this sample size, the coefficient is not significant at conventional levels in the specification that carries that estimate. The evidence for the sign of the relationship is considerably stronger than the evidence for its magnitude, and comparisons of level with advanced-economy estimates should be read against that. The three crisis coefficients are large but imprecisely estimated, and restrictions of equality between them are not rejected, so we report them as descriptive magnitudes and do not rank the episodes. We validate the modeled unemployment series against the national Labour Force Survey over the years in which both exist and re-estimate the model on the national series; the crisis coefficients and the crisis-dating pattern are essentially unchanged. We do not detect a structural break at the 1997 pyramid-scheme collapse once dynamics and crisis indicators are jointly modeled, and we do not detect cyclical asymmetry, though the confidence intervals are wide enough that both are failures to detect rather than evidence of stability or symmetry.”

From a paper by Nauro F. Campos, Corrado Macchiarelli, and Fotios Mitropoulos:

“This paper provides new, structural estimates of Okun’s unemployment-output relationship for euro area countries between 1979 and 2019. We show that these structural estimates are stable over time and yet substantially smaller than the reduced-form estimates that tend to characterise the literature. We also find that country specific factors largely shape how output responds to unemployment in both core and periphery economies. Specifically, for the euro periphery we find that product market regulation plays a major role in explaining the significance of Okun’s estimates. Our results are robust, inter alia, to conditioning on diverse institutional set-ups.”

From an article by Daron Acemoglu:

“Artificial intelligence is reshaping the world in front of our eyes. The concerns around AI, however, extend beyond the uncertainty and disruption that accompany any other radical technological change. They stem in part from the direction AI development has taken: many leading companies are engaged in a single-minded pursuit of artificial general intelligence (AGI).

AGI is generally understood as an AI system that can match or exceed human capabilities on most economically valuable tasks. In such a world, the automation of work would not be limited to specific industries or routine jobs. Instead, machines would take over many forms of work, including highly skilled professions. The consequences of this displacement of human labour would be unparalleled.

To draw attention to this urgent public-policy challenge, I joined thousands of economists and AI researchers in signing We Must Act Now, a statement urging policymakers and technology leaders to steer AI towards complementing, rather than replacing, human labour.”

Continue reading here.

From a paper by Mark Zandi, Cristian deRitis, Marisa DiNatale, Dante DeAntonio, Matt Colyar, Shandor Whitcher, Justin Begley, Ilir Hysa, and Gwen Semmens:

“Artificial intelligence stands to become one of the most consequential technologies in generations, if not in the history of humankind, with enormous implications for the economy. However, the specifics of how it will shape the future remain highly uncertain and are the subject of immense debate. Many technologists deeply involved in AI’s development believe it will massively increase productivity, resulting in significant net job loss and much higher unemployment. Conversely, most economists who look to the economic history of past general-purpose technologies tend to be more circumspect, expecting AI to lift productivity but also to diffuse slowly enough through the economy for the job market to adjust more gracefully. There may be bouts of higher unemployment, but any increase will be modest, as jobs lost to AI will be largely offset by new jobs supported by the income and wealth it creates.”

From a paper by Marina da Silva Sanches:

“This study examines the effects of fiscal consolidation on income inequality in OECD countries from 1978 to 2014, using narrative data. Employing local projections methodology and using a Gini decomposition interpretation, we estimate the impact of austerity episodes on disposable income, market income, wage, and functional inequalities. While the literature has primarily focused on the overall effect of fiscal consolidation on disposable income inequality, this paper examines how different dimensions of inequality respond to fiscal consolidation episodes. We find an increase in wage inequality in the short and medium runs, and a decrease in labor’s share of income in the short run. The results also underscore the importance of social protection in the short term. Additionally, we observe a significant increase in earnings inequality, when including the lower end of the distribution. Finally, spending-based austerity measures are, in general, more relevant than tax-based ones. Results are robust to several tests.”