SSRN

@ssrn.bsky.social

3141 Followers 594 Following

SSRN’s mission is to rapidly share early-stage research and empower global scholarship to help shape a better future. You can submit your research to SSRN online at www.ssrn.com.

Figure 2: Channels linking sustainability disclosure to analyst and investor
forecast accuracy

Figure1 Comparison of Winrate Matrices

Figure 3. The BTC case: The potential excess loss at various trading frequencies.

Figure 1. The inverted development curve.

Figure 2. Where the 17-to-18 hours go. The solid black bars are the components that routine automation reduces to minutes; the diagonally hatched bars are the ones that turn on human judgement. The compressible component follows LeBlanc et al.’s (2019) median, read broadly to include basic submission preparation as well as reformatting, whereas the judgement component is my own estimate.

Figure 1: Nash equilibrium bias across experimental conditions. (a) LLMs exhibit consistently higher Nash ratios (Equation 1) than humans, with limited sensitivity to contextual framing. (b) Eliciting explicit reasoning further increases Nash-consistent behavior, indicating that reasoning reinforces equilibrium-oriented strategies rather than diversifying them. Significance levels: *** p < 0.001, ** p < 0.01, * p < 0.05. ns denotes not statistically significant.

Figure 1: Time-dependent parameters of the Heston model used in the numerical example.

This figure shows the cumulative sum of SDF returns for the linear attention and nonlinear transformer models and benchmark models. The left panel shows the full out-of-sample period beginning in 1968, and the right panel begins in 2003.

Figure 1: Share of series where each model achieves a lower RMSFE than AR(1), by forecast horizon—6-month delayed actuals. Dotted line at 50%; values above indicate that the majority of series beat the benchmark. Forecasts targeting Q1–Q2 2020 excluded

Birthright Citizenship, Taxation, and the Draft

Figure 1: Spatial distribution of cumulative climate hazard exposure across U.S. public school districts during 2013-2022. Maps show total district-days of exposure to extreme heat, ozone exceedances (MDA8 ozone ≥140 μg/m3), wildfire smoke (wildfire-attributable PM2.5 >0 μg/m3), wildfire burn exposure, and tropical storms. Supplementary analyses evaluate alternative thresholds including hurricane-force winds (≥33 m/s) and wildfire smoke PM2.5 >5 μg/m3

Figure 2: Examples of spatially separate collections of markings on bone produced
by non-anthropogenic agents, potentially insects, for non-utilitarian purposes
(unpublished data by L.A. Courtenay obtained from fossils from Konso-Gardula,
Ethiopia, dated at ca. 1.7 Ma).

Figure 1: Next-month returns by momentum decile, classic momentum 𝑀 versus the reversal-tilted factor 𝐵 (equal-weighted, U.S., 1927–2025). The winner deciles (right) coincide almost exactly; the factors diverge at the loser end (left), where 𝐵’s bottom decile is deeper. The difference between 𝐵 and 𝑀 is a loser-side phenomenon.

Figure 8. Estimated racial/ethnic composition of party-list seats won by Democrats under a Mixed-Member Proportional (MMP) system (based on the 2024 U.S. House elections). Party-list seats distributed under the assumption of ethno-racial bloc voting under STV (Droop) candidate preferential votes

Confronting AI Poverty in the Digital Age

Figure 1: Average Effective Tariff Rate on All U.S. Imports
This figure displays average effective tariff rate on all imports from 1821 to 2025. We
provide four reference dates for the convenience of understanding the time-series variation in the average effective tariff rate. In 1930, the Smoot-Hawley Tariff Act, signed by President Herbert Hoover in June, raised tariffs on more than 20,000 imported goods. In 1947, the General Agreement on Tariffs and Trade (GATT) was signed by 23 countries in October with the goal of reducing trade barriers. In January of 1994, the North American Free Trade Agreement came into effect to, again, promote multilateral trade by reducing tariffs. In the beginning of 2018 and 2025, the Trump administration imposed tariffs on various US imports. The time-series data is from the Tax Foundation

Figure 1: Real-time smoothing of the VIX around the global nancial crisis. The raw VIX is shown in black, the proposed edge-preserving lters in solid color, and the benchmark smoothers in dashed lines. Perona-Malik retains the crisis spike more tightly than the other lters while still attenuating short-run variation

Distributing Risk in an Age of AI: Procedural Bad Faith and AI Claims Handling

Figure 1: A Country Life Cycle

Exhibit 4. A few states capture most announced data center investment (active pipeline, June 16, 2026)

This image shows an image that saYS Call for applicants: Scientific Editor Role.

Figure 3: Case CH < C ∗. The equilibrium level of security φ∗, liquidation fee z∗, and socially optimal level of security φ F B are represented by the black, red, and gray curves, respectively. Over and under enforcement regions corresponds to the blue-green and yellow filled areas, respectively

Figure 1: Location of Categories of Individuals under the Two Citizenship
Clauses

From Microeconomics to AI Research: A Guide for Economists

Figure 4: Out-of-sample forecasting evaluation for the PCE ination. Figure shows the forecasting accuracy gains in % relative to the baseline M0 model (BVAR-CSV) with respect to point (RMSE) and density(CRPS) forecasting accuracy metrics. RMSE and CRPS denote the root mean squared error and continuous ranked probability score metrics, respectively. The competing models, M1 - M7, are described in Table 2.

To be or not to be by-object

Figure 2: The moving manifold (Section 3.3). With state-dependent sensitivities B(z) the tangent space TxMt rotates as the driver state Z flows. The intertemporal term hedges this predictable rotation of the geometry; in the fixed-geometry case the span does not move and only the premium does

This image shows different colored lights on a dark blue background.

Fig. 1. Professions and compliance in the context of data protection process-based law

Figure 1. The locus of judgement determines the locus of scholarly answerability. Left: the author brings the question, structures the argument, verifies outputs, and stands behind the claims; the LLM is an instrument. Right: the model produces, the named human approves without verifying the output or standing behind it; answerability becomes nominal, same software, opposite outcomes.

Figure 2: Term frequencies in the SpaceX IPO prospectus. Counts are word-boundary
matches in the text of the SpaceX Form S-1 (SEC accession 0001628280-26-036936), the registration statement carrying the full preliminary prospectus including the financial-statement notes. Matching is case-insensitive except for “Mars” and “Martian,” which are counted with the planet capitalized; the Moon count includes “lunar” variants. Multi-word concepts combine the variants listed in the replication code (for example, “AI” and “artificial intelligence”). Counts include the compensation-disclosure notes, where the Mars Colony Milestone is defined

Figure 4 ECUs: industry – production quantities on example of cement industry, and CTS – number of employees

FIGURE 1: The Centralization Reform

Figure 2: Effect of changing liquidity premium and leverage on PE activity. The figure illustrates the effect of changes in the illiquidity premium and/or leverage on the economic viability of potential PE targets. The y-axis measures the excess return on assets generated during PE ownership, αA,for a given firm. The x-axis measures the d istribution of potential buyout targets, sorted from the firms with the highest to the lowest αA. The horizontal lines show the minimum αA, denoted ˆαA, that is required to cover the costs of private equity CPE = ϕ + λ, where ϕ represents the costs of
delegated portfolio management (fees and carry) and λ is the illiquidity premium required by LPs. Given a leverage level L, we have ˆαA = (1 −L)×CPE. The top horizontal line represents “cold” market conditions when λ is high and/or L is low, whereas the bottom line represents “hot” market conditions when λ is low and/or L is high.

Figure 1: Domination of the variable X_AGR by the variable Fx.

Productivity and Living Standards Are Not Everything

Making Teenagers Read Newspapers. A Nationwide Experiment in French High Schools

Figure 1: Wavelet coherence and DCC-GARCH plots: Financials and Gold

Figure 1: Illustrating autoregressive generation.
The input prompt is “Mr. and Mrs. Durs,” for which the (autoregressive) LLM produces a distribution over the next tokens in its vocabulary (all possible tokens it can generate). At each generation step, the decoding procedure illustrated here selects the highest-probability token as the one to generate. In this example, the top-ranked token has enormous probability at each step. The output after several steps is verbatim text from Harry Potter and the Sorcerer’s Stone. Figure produced by the authors, based on results for running this prompt on Llama 3.1 70B with greedy decoding..

Optimising for Happiness in EU Technology Policy to Repay Public Subsidies for AI

Automating the Safety Net

Figure 3: Estimated average treatment effect on ad x-outs within out-of-bag
CATE-quantile bins. Construction identical to Figure 2. The dashed horizontal line marks the AIPW estimate of the population ATE (+0.1 x-outs). Only the bottom bin shows a negative AIPW estimate with a confidence interval strictly below zero; the absolute magnitude of the most negative bin (−0.07) is small relative to the most positive bin
(+0.93).

Exp. 1: Productivity, Average Judge Score, and Total Time.

Figure 2: Distribution of Predicted Fraud Probabilities

This image shows a figure titled Quarterly Event-Study Estimates of the Golden Visa Reform on Transaction Values for years 2021-2024.

Reconciling AI Innovation With Copyright Values – Why Reliance on Individual Licences is not Enough

Figure 2: Time series comparing median market land sale prices and statutory parcel valuations per square metre between 2012 and 2025. The interquartile range (IQR) of the sales price is shaded in grey. The trends exhibit strong temporal co-movement, confirming that statutory valuations track underlying market dynamics.

This image shows a figure titled Outcomes from regression.

Strategy's Bitcoin Treasury Model: Corporate Omphaloskepsis, Polypharmacy of Risk, and Shareholder and Societal Welfare

Figure 2:Extended FFNN for DDL

Figure 1: PE Ownership in the Insurance Sector
(a) Number of Insurers

Fig. 1: Number of funds, per month.