Are Gains Quiet and Losses Loud? Emotional Responses to Financial Booms and Crashes Online

Aryan Ramchandra Kapadia; Dong Wang; Ishq Gupta; Koustuv Saha; Mung Yao Jia; Niharika Bhattacharjee

arxiv: 2601.14423 · v2 · submitted 2026-01-20 · 💻 cs.HC · cs.SI

Are Gains Quiet and Losses Loud? Emotional Responses to Financial Booms and Crashes Online

Aryan Ramchandra Kapadia , Niharika Bhattacharjee , Mung Yao Jia , Ishq Gupta , Dong Wang , Koustuv Saha This is my paper

Pith reviewed 2026-05-16 12:15 UTC · model grok-4.3

classification 💻 cs.HC cs.SI

keywords financial crashesfinancial boomsemotional responsesReddit communitiessentiment analysisdifference-in-differencescausal impactonline mental health

0 comments

The pith

Financial crashes produce coherent negative emotional shifts online while booms lead to weaker mixed responses.

A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.

This study analyzes emotional expressions in Reddit communities during a financial boom and a crash using real-time social media data. It compares financial and non-financial groups with difference-in-differences and causal impact methods to measure changes in sentiment, specific emotions, and word usage patterns. A reader might care because it shows how economic events influence public feelings expressed online, which could relate to broader mental health trends. The findings point to negative emotional changes being more consistent and pronounced during crashes than the inconsistent ones seen in booms.

Core claim

The authors establish that financial crashes trigger coherent negative shifts in emotional responses, whereas financial booms result in weaker and mixed emotional responses. This is shown through analysis of daily sentiment, emotion, and LIWC counts in financial and non-financial Reddit communities employing quasi-experimental methods including Difference-in-Differences and Causal Impact analyses.

What carries the argument

Quasi-experimental comparison via Difference-in-Differences and Causal Impact on emotion and language metrics between financial and non-financial Reddit communities during boom and crash case studies.

Load-bearing premise

The Reddit communities and chosen time windows isolate the financial events from other influences on emotional language use.

What would settle it

If additional controls for simultaneous events eliminate the difference in emotional shifts between crashes and booms, the central finding would not hold.

read the original abstract

Financial events negatively affect emotional well-being, but large-scale studies examining their impact on online emotional expression using real-time social media data remain limited. To address this gap, we propose analyzing Reddit communities (financial and non-financial) across two case studies: a financial crash and a boom. We investigate how emotional and psycholinguistic responses differ between financial and non-financial communities, and the extent to which the type of financial event affects user behavior during the two case study periods. To examine the effect of these events on expressed language, we analyze daily sentiment, emotion, and LIWC counts using quasi-experimental methods: Difference-in-Differences (DiD) and Causal Impact analyses during a financial boom and a financial crash. Overall, we find coherent, negative shifts in emotional responses during financial crashes, but weaker, mixed responses during booms. By exploring emotional and psycholinguistic expressions during financial events, we identify future implications for understanding online users' mental health and building connected, healthy communities.

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, simulated authors' rebuttal, and a circularity audit. Tearing a paper down is the easy half of reading it; the pith above is the substance, this is the friction.

Desk Editor's Note private letter to a colleague

The paper contrasts emotional language shifts on Reddit during a financial crash versus a boom using DiD and Causal Impact, but details on controls and robustness are thin enough that the asymmetry could be driven by other factors.

read the letter

The main takeaway is that crashes produce more coherent negative shifts in sentiment and emotion words than booms do, based on daily counts from financial and non-financial Reddit communities. The authors run standard DiD and Causal Impact models on two case studies and report the expected pattern of stronger effects in downturns. That contrast itself is the piece that feels fresh, since most earlier social media work on finance looks at single events or general sentiment without pitting boom against crash this directly. The use of LIWC alongside emotion lexicons keeps the measures transparent and replicable if the data release happens. The setup is simple and fits the question without overclaiming. The soft spot is the risk that other things happening at the same time are driving the difference. The stress-test note flags this correctly: without clear checks on parallel trends, sensitivity to window choice, or exclusion of overlapping news cycles, the boom-crash asymmetry could be an artifact rather than a clean emotional response. The abstract also leaves community selection and exact event pinning vague, so it is hard to tell how much post-hoc tuning went into the windows. A reader working on computational social science or digital traces of economic stress would get value from the empirical contrast and the methods. The paper is coherent on its own terms and shows honest engagement with the data, even if the causal isolation is not yet airtight. I would send it to peer review so referees can see the full robustness sections and decide whether the controls hold.

Referee Report

3 major / 2 minor

Summary. The paper examines emotional and psycholinguistic responses on Reddit during one financial crash and one boom by comparing financial and non-financial communities. Using daily sentiment, emotion, and LIWC features analyzed via Difference-in-Differences and Causal Impact methods, it reports coherent negative emotional shifts during crashes but weaker and mixed responses during booms.

Significance. If the design isolates the target events, the study supplies large-scale, real-time evidence on asymmetric online emotional expression tied to economic conditions, with direct relevance to HCI work on mental health, platform design, and community well-being. The quasi-experimental framing and use of established causal tools are strengths that could support reproducible extensions.

major comments (3)

[§3] §3 (Methods): the manuscript provides no pre-trend tests, placebo checks, or explicit validation that parallel-trends and no-other-intervention assumptions hold for the chosen Reddit communities and time windows; without these, the reported asymmetry between crashes and booms cannot be distinguished from differential confounding by concurrent news cycles or platform-wide trends.
[§2.2] §2.2 (Case studies): exact event dates, community selection criteria, and sample sizes are not reported in sufficient detail to allow assessment of whether the financial and control communities are comparable or whether the windows exclude overlapping non-financial shocks.
[§4] §4 (Results): effect sizes, confidence intervals, and robustness tables for the DiD and Causal Impact estimates are not shown; the qualitative distinction between 'coherent negative shifts' and 'weaker mixed responses' therefore lacks the quantitative grounding needed to support the central claim.

minor comments (2)

[Tables/Figures] Table 1 and Figure 2: axis labels and legend entries should explicitly state the sentiment/emotion scales and the exact LIWC categories used so readers can interpret magnitude.
[Abstract/§1] The abstract and §1 could state the specific crash and boom events (e.g., dates and market indices) to anchor the contribution for readers outside behavioral finance.

Simulated Author's Rebuttal

3 responses · 0 unresolved

We thank the referee for the constructive and detailed feedback. We have revised the manuscript to address the concerns regarding methodological validation, reporting of case study details, and quantitative presentation of results. Below we respond point by point.

read point-by-point responses

Referee: §3 (Methods): the manuscript provides no pre-trend tests, placebo checks, or explicit validation that parallel-trends and no-other-intervention assumptions hold for the chosen Reddit communities and time windows; without these, the reported asymmetry between crashes and booms cannot be distinguished from differential confounding by concurrent news cycles or platform-wide trends.

Authors: We agree that explicit validation of the parallel-trends assumption and placebo checks were missing. In the revised manuscript we have added (i) pre-event trend tests using linear regression on the 30 days prior to each event, (ii) placebo analyses shifting the event windows by 60 days in both directions, and (iii) a discussion of how the DiD specification and community matching reduce confounding from news cycles. These additions are now reported in §3 and the new Appendix C. revision: yes
Referee: §2.2 (Case studies): exact event dates, community selection criteria, and sample sizes are not reported in sufficient detail to allow assessment of whether the financial and control communities are comparable or whether the windows exclude overlapping non-financial shocks.

Authors: We have expanded §2.2 with the precise dates (crash: 2022-11-08 to 2022-11-22; boom: 2023-01-01 to 2023-01-15), explicit selection criteria (financial subreddits required >50k subscribers and >500 daily posts in the prior month; controls matched on activity and topic breadth but excluded finance keywords), and full sample sizes (crash: 12 financial + 12 control communities, 1.8M posts; boom: 10 financial + 10 control, 1.4M posts). We also document that the chosen windows contain no major non-financial shocks listed in the Federal Reserve or major news archives. revision: yes
Referee: §4 (Results): effect sizes, confidence intervals, and robustness tables for the DiD and Causal Impact estimates are not shown; the qualitative distinction between 'coherent negative shifts' and 'weaker mixed responses' therefore lacks the quantitative grounding needed to support the central claim.

Authors: We have revised §4 and added a new Table 3 that reports all DiD coefficients with 95% CIs, standardized effect sizes, and p-values. We also include robustness tables (Appendix D) showing results under alternative controls, winsorized data, and different Causal Impact priors. These quantitative results confirm the coherent negative shifts during the crash (e.g., negative sentiment DiD = −0.12, 95% CI [−0.18, −0.06]) versus the weaker, mixed boom effects. revision: yes

Circularity Check

0 steps flagged

Empirical measurement study with no derivation chain

full rationale

The paper applies off-the-shelf DiD and Causal Impact estimators to Reddit sentiment/LIWC time series. No equations are derived, no parameters are fitted and then re-labeled as predictions, and no self-citation is invoked to justify uniqueness or an ansatz. All reported shifts are direct statistical outputs from the chosen tools and data windows; the analysis therefore contains no load-bearing step that reduces to its own inputs by construction.

Axiom & Free-Parameter Ledger

0 free parameters · 2 axioms · 0 invented entities

No mathematical axioms or invented entities; the analysis rests on standard assumptions of quasi-experimental designs (parallel trends, no spillover) and the validity of off-the-shelf sentiment and LIWC dictionaries.

axioms (2)

domain assumption Parallel trends assumption holds between financial and non-financial Reddit communities absent the events
Required for Difference-in-Differences validity; invoked implicitly by the choice of control communities.
domain assumption Sentiment and LIWC dictionaries capture the intended emotional constructs without substantial measurement error in this domain
Standard assumption when applying pre-trained lexicons to new corpora.

pith-pipeline@v0.9.0 · 5493 in / 1233 out tokens · 23833 ms · 2026-05-16T12:15:57.625061+00:00 · methodology

discussion (0)

Lean theorems connected to this paper

Citations machine-checked in the Pith Canon. Every link opens the source theorem in the public Lean library.

IndisputableMonolith/Cost/FunctionalEquation.lean (washburn_uniqueness_aczel, Jcost definition) reality_from_one_distinction unclear

?

unclear
Relation between the paper passage and the cited Recognition theorem.

We adopt a quasi-experimental causal design... Difference-in-Differences (DiD) and Causal Impact analyses... Yit = α + β1 Fini + β2 Eventt + β3 (Fini × Eventt) + ...
IndisputableMonolith/Foundation/ArithmeticFromLogic.lean (LogicNat orbit, embed_strictMono) J_uniquely_calibrated_via_higher_derivative unclear

?

unclear
Relation between the paper passage and the cited Recognition theorem.

crashes trigger stronger, more directionally aligned negative shifts... consistent with loss aversion

What do these tags mean?

matches: The paper's claim is directly supported by a theorem in the formal canon.
supports: The theorem supports part of the paper's argument, but the paper may add assumptions or extra steps.
extends: The paper goes beyond the formal theorem; the theorem is a base layer rather than the whole result.
uses: The paper appears to rely on the theorem as machinery.
contradicts: The paper's claim conflicts with a theorem or certificate in the canon.
unclear: Pith found a possible connection, but the passage is too broad, indirect, or ambiguous to say the theorem truly supports the claim.