COVID search anxiety tracked weaker market performance during the pandemic
A COVID-19 search sentiment index built from U.S. household Google queries moved in the opposite direction from stock market returns. Daily data from May 1, 2020, through July 30, 2021, showed that increased searches for terms such as “coronavirus pandemic,” “coronavirus epidemic,” and “coronavirus outbreak” were associated with lower equity returns and later return reversals.
The finding is straightforward. When households searched more intensely for pandemic-related terms, markets tended to weaken.
The study framed those searches as a behavioral signal of fear, uncertainty, and attention spillover rather than as a direct medical measure of disease burden. In financial terms, the index captured how pandemic worry moved from public consciousness into asset pricing. During a public health crisis, elevated attention can act as a proxy for perceived risk, and that risk perception can alter trading behavior, expected cash flows, and the discount rates investors apply to future earnings.
Household search behavior captured a distinct pandemic sentiment signal
The analysis used an alternative measure called the SCARES index, short for Scared COVID-19 Attitude Revealed by Eager Search. Unlike broad market volatility measures, this index was designed from household search volume and focused on pandemic phrasing that reflected ordinary public concern. The aim was not to model infection dynamics, but to quantify how the public reacted to the pandemic in real time.
That distinction matters for interpretation. Search behavior may rise before official policy shifts, before full epidemiologic clarity, or even before economic indicators catch up. In that sense, the index functions as an early behavioral marker of anxiety. The market response then becomes a downstream effect of collective attention, not simply a reaction to news headlines.
Daily observations across more than a year allowed the analysis to examine how the relationship changed as the pandemic evolved. Early periods of sharp uncertainty did not behave the same way as later periods when people had already adapted to repeated waves of disruption. As search intensity changed, so did the market’s sensitivity to it.
Smaller and weaker firms showed the clearest return reversals
Decile regressions on characteristics-sorted portfolio returns showed that the SCARES index predicted return reversals most clearly among firms that were small, less profitable, and characterized by low investment. Those are the kinds of firms that usually have less financial cushion when macroeconomic stress rises. They can be more exposed to liquidity pressure, demand contraction, and abrupt repricing when investors reassess risk.
This pattern fits the basic mechanics of market stress. Smaller firms often have thinner buffers and greater vulnerability to changes in sentiment, especially when the shock is broad-based and difficult to hedge. Less profitable companies also have less room to absorb a sudden deterioration in expectations. Low-investment firms may be priced differently because markets assign them a particular growth profile, making them more sensitive when uncertainty spikes.
The return reversals are especially important. They suggest that sentiment shocks did not merely depress prices in a static way; they altered the direction of subsequent returns, implying instability in how investors processed pandemic information over time.
The search shock did not move most Fama-French factors
COVID-19 household search shocks did not significantly predict any of the Fama-French five factors except SMB, the small-minus-big factor. In other words, the index was not a broad mechanical driver of all standard risk factors. Its strongest and clearest connection was to firm size, which aligns with the portfolio results showing greater vulnerability among smaller companies.
That selective association is informative. It suggests the shock was not simply a generic market-wide macro variable. Rather, it operated through a narrower channel tied to firm characteristics and investor behavior. In behavioral finance terms, pandemic worry appears to have had a concentrated effect on the segment of the market least able to absorb uncertainty.
Structural breaks changed the signal in high-volatility regimes
The analysis also used a two-state Markov switching model to test how structural breaks associated with pandemic phases altered the relationship between search sentiment and other indicators. In the high-volatility regime, the SCARES index became positively related to measures of Twitter-based uncertainty, the volatility index, economic policy uncertainty, and business condition indicators.
This is where the pandemic period becomes especially revealing. When volatility was elevated, public search anxiety moved in step with other established uncertainty gauges. Search behavior did not sit apart from financial stress; it intensified alongside broader uncertainty in the economy and markets. In calmer regimes, the relationship was weaker, which suggests the index became more informative when conditions were most unstable.
The model captures a familiar crisis pattern. During acute shocks, sentiment indicators can synchronize, each one reflecting the same underlying fear but through different channels: household search, social media, policy uncertainty, and market volatility. When that synchronization appears, the market is responding to a shared information environment rather than to one isolated signal.
Pandemic attention grew slower and less reactive over time
Sub-period analysis showed that people gradually became less responsive to the COVID-19 search keywords in the stock market context. Early in the pandemic, search intensity likely carried more informational weight because uncertainty was raw and the economic consequences were less understood. Later, repeated waves of news, policy responses, and market adaptation may have reduced the shock value of each new round of pandemic-related searches.
That habituation effect is plausible. In prolonged crises, public attention often becomes more selective, and the market may discount the same signal more quickly once it has seen it repeatedly. What once moved prices sharply can later look routine, even if the underlying public concern remains elevated.
The result has practical implications for analysts who track sentiment data. A search spike does not always mean the same thing across time. Its predictive value depends on the phase of the shock, the surrounding volatility regime, and the type of firms being evaluated.
Why this matters for market participants during public health shocks
The study suggests that household pandemic searches can function as a real-time behavioral indicator of asset return pressure during crisis periods. For investors and other market stakeholders, the key message is not simply that COVID-19 sentiment matters, but that it matters unevenly. Small, less profitable, and low-investment firms appear more exposed to the sentiment channel, while broad market factors respond less consistently.
In practical terms, that means pandemic-related attention can help explain why prices move even when no single earnings report or policy change seems sufficient to justify the shift. Public fear can travel quickly through expectations, and expectations are central to pricing. When households collectively search for signs of a worsening outbreak, markets may start to discount future cash flows before fundamental data fully arrives.
The dataset ends in July 2021, so the evidence reflects a defined pandemic window rather than the full arc of COVID-19 or later economic normalization. The model also speaks to associations, not causal proof. Even so, the study gives a clear picture of how search-based pandemic sentiment aligned with equity weakness, return reversals, and regime-specific volatility during one of the most disruptive public health shocks in recent memory.