Stock Market Hype: An Empirical Investigation of the Impact of Overconfidence on Meme Stock Valuation

Abstract

This study investigates the relationship between overconfidence and meme stock valuation, drawing on panel data from 28 meme stocks listed from 2019 to 2024. The analysis incorporates key financial indicators, including Tobin’s Q ratio, market capitalization, return on assets, leverage, and volatility. A range of overconfidence proxies is employed, including changes in trading volume, turnover rate, changes in outstanding shares, and alternative measures of excessive trading. We observe a significant positive relationship between overconfidence (as measured by changes in trading volume) and firm valuation, suggesting that investor biases contribute to notable pricing distortions. Leverage has a significant negative relationship with firm valuation. In contrast, market capitalization has a significant positive relationship with firm valuation, implying that meme stock investors respond to both speculative sentiment and traditional firm fundamentals. Robustness checks using alternative proxies reveal that turnover rate and changes in the number of shares are negatively related to valuation. This shows the complex dynamics of meme stocks, where psychological factors intersect with firm-specific indicators. However, results from a dynamic panel model estimated using the Dynamic System Generalized Method of Moments (GMM) show that the turnover rate has a significantly positive relationship with firm valuation. These results offer valuable insights into the pricing behavior of meme stocks, revealing how investor sentiment impacts periodic valuation adjustments in speculative markets.

1. Introduction

The rapid pace of technological innovation in global financial markets has paved a complex pathway for asset valuation, with “meme stocks” emerging as a significant phenomenon in recent years (Chen and Lee 2017; Philander 2023). The rise in meme stocks has been driven by the active participation of retail investors in social media platforms, where trading behaviors often mimic those of influential public figures (Costola et al. 2021; d’Addona and Khanom 2022; Philander 2023; Shternshis and Mazzarisi 2024; Yousaf et al. 2023). This mimicry has given rise to meme stocks, which are financial securities that experience dramatic surges in popularity and trading volume due to their viral appeal within online communities. These stocks have garnered substantial attention and carry significant implications for financial markets (d’Addona and Khanom 2022; Philander 2023). During the 2021 GameStop frenzy, assets such as Tesla, Bitcoin, and Dogecoin exhibited meme-like characteristics, with valuations heavily influenced by internet-driven sentiment (Philander 2023). The GameStop trading episode, in particular, firmly established the meme stock phenomenon, as collective retail action triggered a short squeeze on institutional investors and fundamentally shifted the dynamics of stock valuation (d’Addona and Khanom 2022). This phenomenon, characterized by extreme price fluctuations and spillover effects in global markets, highlights the substantial impact of retail investors’ sentiment. For example, meme stocks such as AMC and GameStop exhibited daily fluctuations of 153% and 301%, respectively, during the frenzy, followed by rapid declines of 50.9% and 56.6% (Yousaf et al. 2023). Such patterns captivated retail investors and extended beyond Reddit forums to institutional investors seeking to hedge risks in volatile option markets. Aloosh et al. (2023) posit that while the surge in popularity of meme stocks is unsurprising, the GameStop episode in 2021 brought the meme concept into sharper focus. Hence, this highlights the significant role of online forums in shaping retail trading behavior.

These dynamics also prompted significant regulatory scrutiny, raising concerns about how trading platforms may have contributed to the turmoil. The losses experienced by hedge funds during this period, estimated at more than USD 68 billion for US-based firms (Yousaf et al. 2023), highlighted the potential for meme stocks to disrupt broader financial markets. Previous studies (Aloosh et al. 2023; d’Addona and Khanom 2022; Yousaf et al. 2023) further suggest that the viral nature of social media-driven trading amplifies investor interest and fosters speculative behaviors that undermine market efficiency. Such behaviors raise critical questions about the underlying factors driving these markets, with overconfidence emerging as a key explanation for the observed patterns.

The literature on overconfidence suggests that investors often overestimate their predictive abilities, leading to excessive risk taking, inflated valuations, and deviations from fundamental asset prices (Aljifri 2023; Bouteska and Regaieg 2020; Tekçe and Yılmaz 2015). The overconfident theory of Daniel et al. (1998) shows that investors tend to overreact to private information signals and underreact to public information signals. This theory provides a theoretical framework for understanding when the security market under- and overreacts due to overconfidence. This is consistent with Graham et al. (2009), who showed the theoretical link between overconfidence and trading frequency. In meme stock markets, overconfident investors tend to overweight small probabilities of extraordinary gains, thereby driving overvaluation (Costola et al. 2021; d’Addona and Khanom 2022; Philander 2023). This behavior has been extensively documented, with studies showing that retail investors in meme markets often engage in high-risk trades, driven by an inflated belief in their market intuition (Aloosh et al. 2023; Shternshis and Mazzarisi 2024).

While prior studies on overconfidence and equity markets have explored diverse financial elements, limited attention has been given to meme stocks and their valuation dynamics. Adebambo and Yan (2018) and Aljifri (2023) showed the significant role overconfidence plays in firm valuation. Trinugroho and Sembel (2011) found that overconfident investors tend to engage in higher levels of trading activity, which contributes to the overvaluation of equities. However, these studies primarily focus on conventional equity markets, overlooking the unique characteristics of meme stocks. This study addresses this gap by examining how overconfidence impacts meme stock valuation, contributing to the broader speculative financial literature. Employing a panel regression framework with 672 quarterly observations of 28 meme stocks from 2019 to 2024, this study incorporates comprehensive measures of firm performance (Tobin’s Q, market capitalization, ROA, leverage, and volatility) and overconfidence (changes in trading volume, turnover rate, and other excess trading measures). The results reveal a positive nexus relationship between overconfidence and meme stock valuation, suggesting investor biases drive price deviations. Interestingly, while leverage has a negative relationship with meme valuation, market capitalization has a positive relationship with meme valuation. The result suggests that meme stock investors do pay attention to the overall financial viability of these stocks and thereby restrict the degree to which their prices deviate from fundamentals. This suggests that a degree of balance exists between the overconfidence and speculative enthusiasm driving the trading activities of these stocks, with traditional fundamental analysis contributing to the determination of the stock’s market price. This observation should therefore alleviate concerns that meme stock prices are completely at the mercy of hype and market excitement.1

This study diverges from the existing literature by focusing on meme stocks, which remain underexplored despite their increasing relevance and implications. The findings broaden our insights on overconfidence’s impact in speculative markets by employing robust dynamic panel modeling. The results have practical implications, suggesting that policymakers and regulators have the scope to improve market efficiency by monitoring and managing social media-inspired trading. However, there is some comfort in the limits this trading has on significantly destabilizing the market efficiency of meme stocks. As such, this study contributes to the literature by exploring the relationship between overconfidence and meme stock valuation, highlighting the nuanced dynamics of this speculative market. We provide valuable insights for investors and regulators, emphasizing the need for balanced regulatory frameworks to address the risks posed by behavioral biases in speculative markets.2

The remainder of this paper is organized as follows: Section 2 discusses the theoretical framework, Section 3 discusses the data and methodology used, Section 4 presents the empirical results, and Section 5 concludes the paper.

2. Theoretical Framework

Drawing insight from recent studies (Aloosh et al. 2023; d’Addona and Khanom 2022; Philander 2023), it can be deduced that the meme stock phenomenon reflects a fundamental shift toward a behaviorally informed by interpretation of market activity. This stands in contrast to the conventional financial paradigm, which assumes rational investors and markets that are informationally efficient. The empirical anomalies observed in meme stock valuations present an opportunity to explore the role of overconfidence in shaping market outcomes, particularly within speculative settings where traditional valuation anchors appear less relevant.

This investigation is theoretically anchored in Prospect Theory (PT), originally proposed by Kahneman (1979), which challenges the assumption of investor rationality by positing that individuals evaluate outcomes relative to a reference point and exhibit loss-averse preferences. As extended by Barberis et al. (1998), PT provides a framework for understanding how overconfidence may distort investor perceptions and decisions during periods of heightened market attention. In this setting, meme stock traders can be characterized by a tendency to pursue perceived gains despite accumulating losses, especially when such behavior is validated within online communities. This aligns with evidence from Costola et al. (2021); d’Addona and Khanom (2022); Philander (2023), who observe a consistent pattern of excessive trading volume and speculative pricing disconnected from fundamental indicators.

Overconfidence in this context is expressed not only through a miscalibration of predictive ability but also through the overweighting of improbable positive outcomes. Investors engaging with meme stocks often assign inflated probabilities to significant gains and discount the plausibility of negative corrections (d’Addona and Khanom 2022; Philander 2023). This tendency is further supported by the findings of Barber and Odean (1999), who show that retail investors are more likely to engage in high-frequency speculative trading without adequately considering risk exposure. The role of digital technologies has further intensified these behavioral patterns. Philander (2023) argues that technology has amplified speculative momentum by lowering barriers to entry and increasing the speed and scale of information dissemination. Related studies (Aloosh et al. 2023; Costola et al. 2021; Shternshis and Mazzarisi 2024) confirm that online forums serve as vehicles for reinforcing behavioral biases, particularly the illusion of control and herd-driven enthusiasm.

Prospect Theory also emphasizes the significance of loss aversion in influencing investor inertia. Bouteska and Regaieg (2020) show that loss-averse investors are likely to hold declining assets in the hope of future recovery. In the meme stock setting, this behavioral tendency is observed when investors retain loss-making positions under the belief that current valuations will rebound. This conviction is often reinforced by confirmation bias and peer validation, as noted in the work of Adebambo and Yan (2018); Aljifri (2023); Parveen et al. (2020). As a result, trading volumes remain elevated even during downward price adjustments, delaying correction mechanisms typically observed in more efficient markets (Philander 2023; Shternshis and Mazzarisi 2024).

This theoretical lens is consistent with evidence from a growing body of research (Costola et al. 2021; d’Addona and Khanom 2022; Philander 2023; Shternshis and Mazzarisi 2024; Yousaf et al. 2023), which suggests that collective overconfidence in meme stock trading creates feedback loops that sustain speculative activity. The present study also draws on the foundational contributions of Alsabban and Alarfaj (2020); Brown and Cliff (2005); Fang et al. (2009); Trejos et al. (2019); Trinugroho and Sembel (2011), who demonstrate how psychological biases impede the discovery of prices and undermine the empirical relevance of rational expectations models. Within the context of memes, Prospect Theory provides a coherent interpretive framework to explain the observed pricing dynamics, elevated volumes, and trading patterns. These characteristics reflect a market environment in which perceived risk is subdued, confidence in individual trading ability is overstated, and investor sentiment plays a dominant role in valuation.

3. Data and Methodology

3.1. Data

Our target population consists of a comprehensive dataset combining financial market data with proprietary metrics based on social media activity and trading behavior, utilized to rank meme stocks by Quiver Quantitative3. We download quarterly data of 28 listed meme stocks (see

Table 1. List of stocks used in the analysis of this paper.

No.	Ticker	Company Name	No.	Ticker	Company Name
1	ALB	Albemarle Corporation (USA)	16	NVAX	Novavax Inc (USA)
2	AMC	AMC Entertainment Holdings Inc (USA)	17	NTR	Nutrien Ltd. (Canada)
3	AAL	American Airlines Group Inc. (USA)	18	PANW	Palo Alto Networks, Inc. (USA)
4	BB	BlackBerry Limited (Canada)	19	PARA	Paramount Global (USA)
5	BTBT	Bit Digital, Inc. (USA)	20	PHAT	Phathom Pharmaceuticals Inc (USA)
6	BYND	Beyond Meat, Inc. (USA)	21	STX	Seagate Technology Holdings (USA)
7	CGC	Canopy Growth Corporation (Canada)	22	SIRI	Sirius XM Holdings, Inc (USA)
8	CCL	Carnival Corporation (USA)	23	SNDL	Sundial Growers Inc. (Canada)
9	CELH	Celsius Holdings, Inc. (USA)	24	SAVE	Spirit Airlines, Inc. (USA)
10	DKNG	DraftKings Inc. (USA)	25	SBUX	Starbucks Corp (USA)
11	EYEN	Eyenovia, Inc. (USA)	26	SPWR	SunPower Corporation (USA)
12	GME	GameStop Corp. (USA)	27	TLRY	Tilray Brands, Inc. (USA)
13	KRUS	Kura Sushi USA, Inc. (USA)	28	WKHS	Workhorse Group, Inc (USA)
14	MARA	MARA Holdings, Inc. (USA)
15	NKLA	Nikola Corporation (USA)

This table reports the list of stocks used in this study and their respective tickers. The data were downloaded from the Refinitiv Eikon Datastream database. Our quarterly data sample is from 2019 to 2024.

for the company names and tickers) from Refinitiv Eikon Datastream, covering key variables such as Tobin’s Q ratio (Tobin’s Q), market capitalization (MarketCap), return on assets (ROA), leverage ratio (leverage), historical 90-day volatility (vol), and trading volume (TV) (see

Table 2. List of Variables, Definitions, and Data Sources.

Variables	Definition	Source
Tobin’s Q ratio (Tobin’sQ)	Calculated as the ratio of a firm’s market value (DWEV) to its assets’ replacement cost (DWTA), giving an indication of how much the market values a firm relative to its assets.	Refinitiv Eikon Datastream
Market capitalization (MarketCap)	Total market value of a firm’s outstanding shares, which is calculated as the product of share price and the number of shares outstanding.	Refinitiv Eikon Datastream
Return on assets (ROA)	Net profit as a percentage of total assets (%).	Refinitiv Eikon Datastream
Leverage Ratio (Leverage)	Ratio between total debt and total assets.	Refinitiv Eikon Datastream
Historical 90-day volatility	Measures the past 90-day price volatility of the firm.	Refinitiv Eikon Datastream
Investor overconfidence ($\Delta t{v}_{it}$)	Change in trading volume.	Refinitiv Eikon Datastream

This table reports the list of variables, definitions, and data sources used in this paper.

for the list of variables, definitions and sources). The dataset spans the period from 2019 to 2024, with this sample period selected solely based on data availability. A quarterly frequency is used since the parameters of interest are predominantly reported every quarter (Aljifri 2023).

This study utilizes Tobin’s Q ratio (Tobin’s Q) as the dependent variable to analyze the valuation of a selected list of meme stocks. Tobin’s Q is a widely recognized metric for evaluating firm performance and has been frequently applied in previous research (see, for example, Aljifri 2023; Buchanan et al. 2018; Fang et al. 2009; Gompers et al. 2003). Additionally, we include various firm characteristics related to valuation proxies, such as market capitalization (MarketCap), return on assets (ROA), leverage, and historical 90-day volatility. Consistent with prior studies, we incorporate these variables in our analysis: market capitalization (Aljifri 2023; Chen and Lee 2017; Feng and Wu 2018; Yung and Jian 2017), ROA (Aljifri 2023; Al-Najjar and Anfimiadou 2012; Ghoul et al. 2017; Hassan 2018), leverage (Aljifri 2023; Al-Najjar and Anfimiadou 2012; Chen and Lee 2017; Feng and Wu 2018; Yung and Jian 2017), and 90-day historical volatility (Bag 2022; Karolyi 1993). To test the null hypothesis ($H_0$) that overconfidence bias positively impacts meme stock valuation, we also include a proxy for overconfidence in our model.

3.2. Investor Overconfidence Metrics

This study investigates the impact of overconfidence on meme stock valuation. Following Aljifri (2023), we focus on identifying suitable proxies for overconfidence, a challenging task given the variable’s intangible nature. Zahera and Bansal (2018) states that finding a consistent proxy of investor overconfidence is complex and contingent on various factors. Moreover, Glaser et al. (2013) caution that overconfidence measures require careful consideration, especially when periodic intervals are applied, as not all proposed metrics are equally effective. Despite these challenges, previous studies (Aljifri 2023; Bouteska and Regaieg 2020; Huang et al. 2022; Parveen et al. 2020; Statman et al. 2006; Strahilevitz et al. 2011; Trejos et al. 2019; Trinugroho and Sembel 2011; Xia et al. 2014) have made significant contribution in establishing reliable proxies for overconfidence. Notably, some studies (Glaser et al. 2013; Mushinada and Veluri 2018; Parhi and Pal 2022) have adopted cross-sectional methods, even though the longitudinal nature of overconfidence traits would typically suggest a time-series approach.

Following an examination of these alternative measures, we align with Aljifri (2023) in selecting overconfidence proxies due to their robustness and comprehensive coverage. This study incorporates six proxies for investor overconfidence, which have been validated in previous literature (Bouteska and Regaieg 2020; Huang et al. 2022; Parveen et al. 2020; Trejos et al. 2019). These proxies are: (1) changes in trading volume ($\Delta t{v}_{it}$), (2) turnover rate ($turnove{r}_{it}$), and (3) changes in outstanding shares ($is{o}_{it}$). Additionally, we assess excessive trading behavior using multiple proxies: $e{t}_{it}$, which captures general increases in trading volume; $et{m}_{it}$, which measures trading volume increase above the mean; and $et{d}_{it}$, which captures trading volume increase above the median.

3.2.1. The Variations in Trading Volume

Following Aljifri (2023), trading volume as a measure of overconfidence has been widely discussed in previous literature (Bouteska and Regaieg 2020; Huang et al. 2022; Mushinada and Veluri 2018; Statman et al. 2006). Bouteska and Regaieg (2020) highlighted the role of overconfidence in explaining the observed patterns in trading volume. Statman et al. (2006) developed an overconfidence model for financial market investors, demonstrating that investors’ confidence in their ability to assess trading volumes often reflects their level of overconfidence. Mushinada and Veluri (2018) showed that overconfident investors are often driven by high trading volumes and attribute their gains to their own expertise. This is consistent with Huang et al. (2022), who developed an overconfidence model for the Chinese equity market. According to these studies, overconfident investors tend to increase trading volume due to self-attribution, often disregarding alternative perspectives. Strahilevitz et al. (2011) also showed that increased overconfidence correlates with higher trading volumes for both individual and corporate investors. Collectively, these studies show a significant positive relationship between overconfidence and trading volume, reinforcing the use of trading volume as a proxy for overconfidence. Aljifri (2023) similarly provided evidence supporting trading volume as a measure of investor overconfidence in the Saudi Arabian equity market, affirming the robustness and reliability of this proxy across various market settings.

In line with Aljifri (2023), this study estimates the variation (relative increase or decrease) in trading volume using the following equation:

$$\Delta t{v}_{it}={\displaystyle \frac{t{v}_{it}-t{v}_{it-1}}{t{v}_{it-1}}}$$

(1)

Here, $\Delta t{v}_{it}$ represents the change in trading volume for company i at time t, where $t{v}_{it}$ is the trading volume in the current period and $t{v}_{it-1}$ is the trading volume in the previous period.

3.2.2. Measures of Excessive Trading

Similar to trading volume, excessive trading has been widely used as a reliable measure of investor overconfidence in prior studies (Aljifri 2023; Mushinada and Veluri 2018; Trejos et al. 2019). We adopt a robust measure of excessive trading through three key factors that help identify these traits among investors. As outlined in Aljifri (2023), these proxies—$e{t}_{it}$, $et{m}_{it}$, and $et{d}_{it}$—enable a comprehensive assessment of investor overconfidence by examining variations at the firm level and across different periods. Evidence from Trinugroho and Sembel (2011) shows that overconfident investors are more likely to engage in excessive trading, often as a result of self-attribution bias (see Mushinada and Veluri 2018). Excessive trading, in turn, contributes to increased market volatility, with the reliability of this measure justified by Trejos et al. (2019) despite challenges in estimation. In line with Aljifri (2023), this study uses three proxies at different thresholds to define “excessive” trading volume.

To begin, we compute excessive trading ($e{t}_{it}$) by including changes in trading volume only when the change is greater than zero, thus identifying overconfident investors. Overconfidence is defined here by a zero threshold, capturing only positive changes in trading volume. This variable is calculated as follows:

$$e{t}_{it}=\left\{\begin{array}{cc}\Delta t{v}_{it},\hfill & \mathrm{if}\Delta t{v}_{it}>0\hfill \\ \varphi ,\hfill & \mathrm{if}\Delta t{v}_{it}\le 0\hfill \end{array}\right.$$

(2)

where $e{t}_{it}$ represents the increase in trading volume, identified as “excessive trading,” and $\varphi$ denotes a null value. Here, overconfidence is defined as a positive change in trading volume. This proxy was calculated for each company and period, recognizing that firm-specific information could drive significant trading volume increases.

Next, we calculate excessive trading ($et{d}_{it}$) based on changes in trading volume exceeding the median trading volume (${\mu }_{1/2i}$), capturing the dynamics of overconfident investors who trade at levels above the median. This proxy is defined as follows:

$$et{d}_{it}=\left\{\begin{array}{cc}\Delta t{v}_{it},\hfill & \mathrm{if}\Delta t{v}_{it}>{\mu }_{1/2i}\hfill \\ \varphi ,\hfill & \mathrm{if}\Delta t{v}_{it}\le {\mu }_{1/2i}\hfill \end{array}\right.$$

(3)

where ${\mu }_{1/2i}$ is the median change in trading volume for company i.

Finally, we define excessive trading ($et{m}_{it}$) as changes in trading volume that exceed the mean trading volume (${\mu }_{ik}$), capturing the behavior of overconfident investors relative to an average level. Here, the mean level of trading volume for $\Delta t{v}_{it}$ is used to assess increases above the average. The mean-based excessive trading (${\mu }_{ik}$) is calculated as follows:

$${\mu }_{ik}={\displaystyle \frac{1}{n}}\sum _t\Delta t{v}_{it}$$

(4)

where ${\mu }_{ik}$ represents the mean trading volume for period k for company i, k is a fixed window, and n is the number of observations within k. In this study, k is equal to one year, covering four quarters, with non-overlapping periods created by shifting k forward by n observations. This variable proxy is defined as follows:

$$et{m}_{it}=\left\{\begin{array}{cc}\Delta t{v}_{it},\hfill & \mathrm{if}\Delta t{v}_{it}>{\mu }_{ik},\forall t\in k\hfill \\ \varphi ,\hfill & \mathrm{if}\Delta t{v}_{it}\le {\mu }_{ik}\hfill \end{array}\right.$$

(5)

3.2.3. The Turnover Rate

While trading volume and excessive trading are commonly used to measure investor overconfidence, several studies have shown that the turnover rate is an equally reliable proxy (Bouteska and Regaieg 2020; Ho 2013; Statman et al. 2006; Tekçe and Yılmaz 2015; Trejos et al. 2019). Aljifri (2023) supports this perspective and show the applicability of turnover rate in capturing overconfidence. Notably, the overconfidence model by Statman et al. (2006) extends beyond trading volume, emphasizing the robustness of turnover rate, especially for small-cap stocks. In the case of Ho (2013), investor overconfidence is driven by access to private information that significantly affects turnover, with further implications on trading volume and return volatility. Similarly, Tekçe and Yılmaz (2015) showed how excessive trading behavior among overconfident investors leads to higher turnover, and Trejos et al. (2019) showed that overconfident investors often trade at levels beyond rational thresholds. Bouteska and Regaieg (2020) further justified turnover as a suitable proxy for overconfidence in US-based equities.

In this study, the turnover rate ($turnove{r}_{it}$) is calculated as the ratio of the number of shares traded to the number of shares outstanding. Given that meme-based stocks often exhibit unique trading patterns in response to information, we calculate this measure on a period- and firm-specific basis to account for these variations. The turnover rate is defined as follows:

$$turnove{r}_{it}={\displaystyle \frac{t{v}_{it}}{s{o}_{it}}}$$

(6)

where $s{o}_{it}$ represents the number of shares outstanding for firm i at time t.

3.2.4. Increase in the Number of Outstanding Shares

To provide additional insights, this study incorporates the increase in the quantity of firm-specific outstanding shares ($is{o}_{it}$) as a complementary measure of investor overconfidence. The inclusion of this proxy is motivated by prior studies (Aljifri 2023; Alsabban and Alarfaj 2020; Bouteska and Regaieg 2020; Statman et al. 2006; Tekçe and Yılmaz 2015), which highlight the relationship between overconfidence and share capitalization. Bouteska and Regaieg (2020) showed that capitalization, as reflected in the outstanding shares, can signal investor overconfidence. While none of these studies have relied solely on this measure, it has shown significance when used in conjunction with other indicators of overconfidence. Following Aljifri (2023), this study acknowledges the significance of outstanding shares as a complementary proxy for investor overconfidence. The increase in the number of shares outstanding is measured as follows:

$$is{o}_{it}=\left\{\begin{array}{cc}\Delta s{o}_{it},\hfill & \mathrm{if}\Delta s{o}_{it}>0\hfill \\ \varphi ,\hfill & \mathrm{if}\Delta s{o}_{it}\le 0\hfill \end{array}\right.$$

(7)

where $is{o}_{it}$ represents the increase in the number of shares outstanding and $\Delta s{o}_{it}$ denotes the change in the number of shares outstanding. An increase in $s{o}_{it}$ over the previous period $s{o}_{it-1}$ is considered indicative of investor overconfidence.

3.3. Methodology

Following Aljifri (2023), we employ a panel regression model to capture both cross-sectional and time-series variations, enabling us to account for unobserved heterogeneity across firms. A pooled panel analysis, however, fails to address firm-level heterogeneity adequately (Aljifri 2023; Gujarati and Porter 2009). The homogeneity test (F-value = 7.92, p-value = 0.0000) indicates that a pooled regression model (pooled OLS) is unsuitable due to significant heterogeneity across firms, which aligns with the findings of Aljifri (2023). To determine the most suitable model, we conduct a Hausman test, comparing random effects and fixed effects estimators for consistency and bias. The test results yielded a chi-square (${\chi }^2$) value of 15.54 with a p-value of 0.0000, leading us to reject the null hypothesis. This result indicates that the random effects model is inconsistent, making the fixed effects model the appropriate choice for our analysis. We use robust standard errors to address issues of heteroskedasticity and serial correlation.

The null hypothesis ($H_0$) is that overconfidence bias positively impacts meme stock valuations. The model used to test this effect is specified as follows:

$$Tobi{n}^{\prime }s{Q}_{it}={\alpha }_i+\beta X_{it}+{\delta }_t+{\mu }_i+{ϵ}_{it}$$

(8)

where the subscripts i and t denote entities $i=1,2,\dots ,n$ and time $t=1,2,\dots ,T$, ${ϵ}_{it}$ is the overall error term, Tobin’s Q is the outcome variable for firm i at time t, X is a vector of predictors (thus MarketCap, ROA, Leverage, Volatility, Overconfidence), $\alpha$ is the unknown intercept, ${\delta }_t$ is the unknown coefficient for the time regressors $\left(t\right)$, ${\mu }_i$ is the within-entity error term, and $\beta$ represents a common effect across firms controlling for individual and time heterogeneity.

4. Results

4.1. Descriptive Statistics and Correlations

This section presents the descriptive statistics and pairwise correlations of the variables used in the current analysis.

Table 3. Descriptive statistics and correlations.

Panel A: Descriptive statistics
Variable	N	Mean	Median	Sth	Min	Max	Skewness	Kurtosis
Tobin’sQ	672	2.798	1.653	3.873	−0.034	29.645	3.809	20.337
MarketCap (M)	672	14.946	3.405	28.210	0.002	140.300	2.834	10.620
ROA	672	−0.167	−0.068	0.332	−1.952	0.263	−1.764	7.608
Leverage	672	0.345	0.281	0.305	0.000	1.854	1.552	6.743
Volatility	672	0.760	0.670	0.476	0.023	5.641	2.808	22.122
$\Delta tv$	672	−0.075	−0.003	0.649	−5.472	0.974	−3.409	23.178
et	672	0.243	0.189	0.194	0.001	0.974	1.742	6.014
etd	672	0.247	0.197	0.193	−0.065	0.974	1.704	5.991
etm	672	0.226	0.181	0.208	−0.386	0.974	1.355	5.711
turnover	672	2.556	0.990	7.280	0.002	158.732	15.516	319.333
iso	672	0.179	0.002	1.110	0	19.794	11.941	176.364
Panel B: Pairwise correlations
	Tobin’sQ	MarketCap (M)	ROA	Leverage	Volatility
Tobin’sQ	1
MarketCap	0.0603	1
ROA	0.0006	0.1139 ***	1
Leverage	−0.1054 ***	0.0427	0.0318	1
Volatility	0.0906 **	−0.2878 ***	−0.2763 ***	−0.1383 ***	1
Panel C: Correlation of overconfidence proxies
	$\Delta tv$	et	etd	etm	turnover	iso
$\Delta tv$	1
et	0.3534 ***	1
etd	0.3601 ***	0.9778 ***	1
etm	0.3745 ***	0.904 ***	0.909 ***	1
turnover	0.0799 **	0.2097 ***	0.2132 ***	0.1955 ***	1
iso	−0.0816 **	0.0802 **	0.0862 **	−0.0395	−0.0432	1

This table reports the description and correlation matrix of the variables. Panel A summarizes descriptive statistics for the key variables, including the mean, median, standard deviation (std), minimum (min), maximum (max), skewness, and kurtosis and number of 672 observations (n), (M) is million. Panel B shows the pairwise correlation matrix for Tobin’s Q, market capitalization, return on assets (ROA), leverage, and volatility, highlighting the relationships between these financial metrics. Panel C shows the correlation matrix for investor overconfidence proxies. *** and ** denote statistical significance at the 1% and 5% levels, respectively.

is divided into three panels: Panel A displays the descriptive statistics for all variables, including the number of observations, mean, median, standard deviation, minimum and maximum values, skewness, and kurtosis, Panel B shows the pairwise correlations for the key variables in the main analysis, while Panel C reports the correlations among the six investor overconfidence proxies. The dataset comprises 672 quarterly observations for 28 meme stocks from 2019 to 2024. The variables included in this study are Tobin’s Q, market capitalization (MarketCap), return on assets (ROA), leverage, volatility, and investor overconfidence proxies ($\Delta tv$, et, etd, etm, turnover, and iso).

In Panel A of Table 3, we observe that the average Tobin’s Q ratio is 2.798, which is consistent with prior research that typically finds Tobin’s Q ratios for many firms to fall between two and three (see Adebambo and Yan 2018; Aljifri 2023; Villalonga and Amit 2006, for more details). The mean market capitalization in our sample is USD 14.946 million, with a median of USD 3.405 million, highlighting the presence of both large and smaller meme stocks. The firms show an average ROA of −0.167, indicating that, on average, meme stocks are struggling to generate positive returns on their assets. The mean leverage ratio is 0.345, reflecting moderate use of debt financing, while the average volatility is 0.760, capturing the risk or variability in stock prices for these stocks. The main overconfidence variable ($\Delta tv$) has a mean of −0.075, with a median of −0.003 and a standard deviation of 0.649, suggesting considerable variation in trading volume changes across the sample. Other overconfidence proxies, such as (et, etd, etm), display similar patterns, while turnover and iso exhibit more significant variability, as indicated by their higher standard deviations and kurtosis.

Table 3, Panel B provides the pairwise correlation analysis for the variables included in this study, including Tobin’s Q, market capitalization, ROA, leverage, and volatility. The results show that Tobin’s Q is positively correlated with volatility (0.0906), which is significant at the 5% level, indicating that firms with higher valuations tend to experience more significant stock price fluctuations. However, market capitalization has a negative relationship with volatility (−0.2878), significant at the 1% level, suggesting that larger firms typically exhibit more stable stock prices. Additionally, ROA is negatively correlated with volatility (−0.2763), also significant at the 1% level, meaning that more profitable firms tend to have less volatile stock prices. Leverage shows a negative correlation with Tobin’s Q (−0.1054), significant at the 1% level, suggesting that firms with higher market valuations relative to their assets tend to rely less on debt. As expected, Tobin’s Q is positively correlated with volatility and negatively with leverage. The low correlations among the variables suggest no multicollinearity issues and justify the regression analysis employed.

In Panel C of Table 3, we present the pairwise correlations among the six investor overconfidence proxies: $\Delta tv$, et, etd, etm, turnover, and iso. The correlation matrix reveals that $\Delta tv$ (change in trading volume) has a positive and significant relationship with et (0.3534), etd (0.3601), and etm (0.3745), all at the 1% significance level. This implies that fluctuations in trading volume are strongly associated with these overconfidence measures. Aside from iso, turnover significantly correlates with the other overconfidence proxies. It is also notable that iso has weaker correlations with most proxies, except for a modest positive correlation with etd (0.0862). Interestingly, the significant negative correlation between iso and $\Delta tv$ (−0.0816) suggests that investor sentiment does not always correspond with changes in trading volume. These proxies represent different aspects of investor behavior and show how these measures are correlated. This divergence indicates that while trading volume may fluctuate, it does not always reflect shifts in investor sentiment, emphasizing the importance of using multiple proxies to capture the complexity of overconfidence and market dynamics.

We test our data for multicollinearity using a Variance Inflation Factor (VIF) test. The estimation results, available upon request, show that the VIF values are very low, with the highest value of 1.23 observed for volatility and the lowest value of 1.03 for leverage. The mean VIF of 1.10 is well below the commonly accepted threshold of 10, suggesting that multicollinearity is not a concern in our model. These results ensure the reliability of the coefficient estimates and enhance the robustness of our findings. The low to moderate correlations among the predictors further confirm the absence of significant multicollinearity, providing a solid foundation for subsequent regression analysis.

Figure 1 reports the average Tobin’s Q (in blue) and average Overconfidence (in orange). Tobin’s Q ratio is calculated as the ratio of a firm’s market value (DWEV) to its assets’ replacement cost (DWTA), which indicates how much the market values a firm relative to its assets. A value greater than one suggests that the market perceives the firm as having valuable growth opportunities, while value less than one implies that the market views the firm as underperforming relative to its asset base. From the graph, we note that firms such as Beyond Meat, Inc. (BYND) and Celsius Holdings, Inc. (CELH) have very high values of Tobin’s Q, indicating that the market assigns a substantial premium over their assets, signaling high growth potential or significant investor enthusiasm. On the other hand, some firms like American Airlines Group Inc. (AAL) and Spirit Airlines, Inc. (SAVE) have lower Tobin’s Q values, indicating that the market is assigning a lower valuation relative to their assets.

Changes in trading volume represent investor overconfidence. A positive overconfidence value suggests that investors engage in excessive trading, while a negative value indicates more conservative or lower trading volume changes. We note that the average overconfidence is dominated by negative values for most companies, except a few (i.e., Celsius Holdings, Inc. (CELH), DraftKings Inc. (DKNG), Nikola Corporation (NKLA), and SunPower Corporation (SPWR)), which have positive overconfidence. Interestingly, CELH stands out with the highest Tobin’s Q and positive overconfidence, suggesting that the market highly values it and experiences speculative or confident trading behavior. Despite its popularity, AMC Entertainment Holdings Inc (AMC) has a low Tobin’s Q and low overconfidence values, and this suggests that investors have become more cautious or pessimistic about the company’s future despite its past hype.

4.2. Discussion of Empirical Results Related to Overconfidence and Meme Stock Valuation

In this section, we investigate the impact of overconfidence on meme stock valuation as hypothesized. The regression results in

Table 4. Investor Overconfidence and Firm Valuation.

Variable	Model
MarketCap	0.0111 ***
	(0.00334)
ROA	0.357
	(0.249)
Leverage	−1.292 ***
	(0.306)
Volatility	0.677
	(0.446)
Overconfidence	0.534 ***
	(0.146)
Constant	2.663 ***
	(0.368)
Observations	672
$R^2$	0.089
${\chi }^2$	15.54 ***
F-value	14.93 ***
Stocks	28

This table reports the panel regression results for investor overconfidence and firm valuation. Robust standard errors are in parentheses. Significance levels: ***: 0.01.

report the effect of investor overconfidence and other key predictors on firm valuation within the context of meme stocks. We employed a panel regression model, focusing on a sample period of 672 quarterly observations across 28 stocks. The fixed effects model accounts for firm-specific characteristics and includes robust standard errors to address heteroskedasticity and potential serial correlation issues. The significance of the model, indicated by an F-value of 14.93, which is significant at a 1% level, highlights the collective predictive power of the predictors on meme stock valuation.

Behavioral finance theory suggests that investor overconfidence should positively impact stock valuation, especially in speculative markets such as those driven by memes. The coefficient for overconfidence in our model is 0.534 and highly significant at a 1% significance level, which aligns with empirical expectations that overconfidence can lead to overvaluation (Aljifri 2023). Overconfidence in this context can be viewed as the tendency of investors to overestimate their ability to predict stock movements, often leading to higher trading volumes and increased demand for certain stocks without regard to underlying fundamentals (Chin 2012; Qadri and Shabbir 2014). This behavior is consistent with prior findings (Adebambo and Yan 2018; Aljifri 2023) of a similar positive relationship between overconfidence and firm valuation.

We note that market capitalization has a positive relationship with firm valuation, with a coefficient of 0.0111 significant at the 1% level. This shows that as firms increase in market size, their valuations increase. This result is consistent with the findings of Brown and Cliff (2005), who stated that larger firms are more susceptible to sentiment-driven overvaluation. In the meme stock context, companies with large market capitalization may attract even more speculative interest as they are perceived to be more stable and prominent within the market. This finding supports the view that meme stock investors are not solely focused on smaller, high-risk firms but may also drive up valuations for more prominent companies based on perceived market momentum.

Conversely, leverage has a statistically significant negative relationship with firm valuation with a coefficient of −1.292. This finding aligns with traditional finance theory, which posits that high levels of debt increase financial risk, potentially making firms less attractive to investors (Diamond and He 2014; Flannery 1994). This negative relationship may suggest that even speculative investors are cautious of highly leveraged firms, possibly due to concerns over financial instability (Bhojraj et al. 2021). This contrasts with the unrealistic notion often associated with meme stocks, indicating that investor overconfidence does not entirely overshadow considerations of financial health.

Interestingly, ROA and volatility have a positive insignificant relationship with the firm valuation. ROA’s insignificant coefficient suggests that meme stock investors may place less emphasis on traditional profitability measures when valuing these stocks. This finding is consistent with the speculative nature of meme stock markets, where fundamental indicators are often overlooked in favor of perceived potential or momentum (Smith 2024). Volatility’s insignificance is somewhat surprising, as one might expect meme stock investors to be attracted to highly volatile stocks. However, the lack of significance here suggests that although volatility might initially capture investor interest, it does not play a fundamental role in driving valuation, particularly over low frequency, such as quarterly periods.

Our findings contribute to the understanding of meme stock valuation by highlighting the significant role of overconfidence. The positive relationship between overconfidence and firm valuation supports existing behavioral finance literature, which suggests that investor biases can lead to price distortions (Aljifri 2023). However, the mixed responses to leverage and market capitalization show that meme stock investors may not disregard all traditional financial indicators and tend to balance overconfidence and speculative enthusiasm with some attention to risk factors. This implies that meme stock valuations may deviate from fundamentals due to behavioral biases, but these deviations are not entirely devoid of rational considerations.

4.3. Robustness and Sensitivity Analysis

To verify the robustness of the results, we ran two robustness checks. First, we re-evaluated the impact of overconfidence on meme stock valuation using five alternative measures of investor overconfidence. Second, we addressed potential endogeneity issues by employing the dynamic panel data model (system GMM) (Arellano and Bover 1995; Arellano and Bond 1991; Blundell and Bond 1998).

4.3.1. Alternative Overconfidence Proxies and Firm Valuation

In this section, we examine the robustness of our findings on the relationship between investor overconfidence and meme stock valuation by using five alternative proxies for overconfidence: et, etd, etm, turnover, and iso.

Table 5. Alternative overconfidence proxies.

Variables	(1)	(2)	(3)	(4)	(5)
MarketCap	0.0110 ***	0.0110 ***	0.0113 ***	0.0112 ***	0.0134 ***
	(0.00351)	(0.00351)	(0.00334)	(0.00336)	(0.00304)
ROA	0.469 *	0.477 *	0.486 *	0.460	0.352
	(0.265)	(0.270)	(0.254)	(0.308)	(0.246)
Leverage	−1.188 ***	−1.183 ***	−1.185 ***	−1.210 ***	−1.309 ***
	(0.266)	(0.265)	(0.279)	(0.293)	(0.303)
Volatility	0.719	0.709	0.629	0.846	0.818 *
	(0.544)	(0.553)	(0.491)	(0.570)	(0.457)
Alt. Overconfidence:
et	0.483
	(1.373)
etd		0.529
		(1.398)
etm			1.013
			(0.937)
turnover				−0.00507
				(0.0255)
iso					−0.0342 **
					(0.0142)
Constant	2.459 ***	2.453 ***	2.412 ***	2.496 ***	2.574 ***
	(0.330)	(0.326)	(0.323)	(0.375)	(0.349)
Observations	672	672	672	672	672
$R^2$	0.083	0.083	0.084	0.082	0.086
${\chi }^2$	15.36 ***	15.12 ***	15.95 ***	96.08 ***	29.89 ***
F-value	18.03 ***	18.29 ***	18.41 ***	21.85 ***	21.79 ***
Stocks	28	28	28	28	28

This table reports the panel regression results for alternative measures of investor overconfidence (et, etd, etm, turnover, and iso) and firm valuation. Robust standard errors are in parentheses. Significance levels: *: 0.1, **: 0.05, ***: 0.01.

reports the effect of the alternative overconfidence proxies on firm valuation alongside other predictors such as market capitalization, return on assets (ROA), leverage, and volatility.

Market capitalization consistently has a positive and statistically significant relationship with firm valuation across all models, with coefficients ranging from 0.0110 to 0.0134, all significant at the 1% level. This positive relationship suggests that larger firms likely attract greater visibility and investor interest and experience higher valuations in the meme stock market. These findings align with sentiment-driven investment theories, which posit that larger firms can be more susceptible to overvaluation due to increased investor attention (Brown and Cliff 2005).

The coefficient for ROA is positive across all models, with statistical significance at the 10% level in Models 1–3. This finding suggests that meme stock investors may still factor in profitability, albeit with less emphasis than traditional financial metrics. However, the lower significance levels for ROA show that profitability may not be a primary driver of valuation in this highly speculative market. This is consistent with our findings in Table 4.

Leverage has a significant negative relationship with firm valuation in all models, with coefficients ranging from −1.183 to −1.309, which is significant at the 1% level. This negative relationship suggests that investors tend to be cautious towards highly leveraged firms, possibly due to the perceived financial risk associated with high debt levels. Despite the speculative nature of meme stocks, the consistent response to leverage suggests that even overconfident investors are wary of firms carrying substantial debt, aligning with traditional finance principles that link high leverage with greater risk (Diamond and He 2014; Flannery 1994).

The relationship between volatility and firm valuation is less consistent across models. Volatility is positive but only significant in Model 5, where iso is used as the overconfidence proxy. The coefficient of 0.818 is significant at the 10% level, indicating that while volatility may initially attract investor interest, it does not consistently influence valuation in meme stocks. This result suggests that meme stock investors may prioritize other indicators, such as trading volume changes and firm size, over volatility when making valuation decisions.

The alternative overconfidence proxies reveal varying degrees of impact on firm valuation. The coefficients for et, etd, and etm are all positive, though none is statistically significant. These proxies are designed to capture excessive trading behavior based on trading volume changes above certain thresholds. The results show that increased trading alone may not fully account for overconfidence effects in meme stock valuation.

The turnover proxy, representing the ratio of change in trading volume to shares outstanding, shows a slight negative but statistically insignificant coefficient in Model 4. This finding suggests that trading frequency relative to firm size does not have a substantial impact on meme stock valuation in this context, despite its common association with investor sentiment in other markets (Aljifri 2023; Tekçe and Yılmaz 2015).

Finally, the iso proxy, which measures changes in shares outstanding, has a significant negative coefficient of −0.0342 at the 5% level in Model 5. This result implies that an increase in shares outstanding, possibly driven by market demand, may reduce valuation, which suggests a dilution effect as new shares enter the market. The significance of iso highlights the complexity of investor behavior in speculative markets. At the same time, overconfidence can drive up initial demand, and increases in shares outstanding may lead to a reassessment of firm valuation as investors react to the potential dilution of stock value.

In summary, the alternative proxies for overconfidence confirm the complex nature of meme stock valuation, while market capitalization and leverage play consistent roles in influencing valuation. These results align with behavioral finance literature, suggesting that while meme stock valuations are influenced by overconfident trading behavior, certain traditional financial indicators, like firm size and debt levels, continue to shape investor perceptions (Adebambo and Yan 2018; Trinugroho and Sembel 2011). The nuanced direction and significance of different overconfidence proxies signify the complexity of meme stock valuation and investing.

4.3.2. Dynamic System Generalized Method of Moments (GMM) Regression

In this section, we examine the impact of investor overconfidence on meme stock valuation using a dynamic system, the Generalized Method of Moments (GMM) approach. The regression model aims to address potential endogeneity and autocorrelation issues (see Arellano and Bover 1995; Arellano and Bond 1991; Blundell and Bond 1998). The model specification is as follows:4

$$\Delta \mathrm{Tobin}’\mathrm{s}{Q}_{it}={\alpha }_0+\delta ·L.\Delta \mathrm{Tobin}’\mathrm{s}{Q}_{it}+B_i·X_{it}+{ϵ}_{it}$$

(9)

where $\Delta \mathrm{Tobin}’\mathrm{s}{Q}_{it}$ represents the differenced dependent variable, $L.\Delta \mathrm{Tobin}’\mathrm{s}{Q}_{it}$ captures the lagged difference of Tobin’s Q, $\delta$ is the lag time coefficient for $L\Delta \mathrm{Tobin}’\mathrm{s}Q$, and ${ϵ}_{it}$ is the error term. The vector $B_i·X_{it}$ encompasses the predictors:

$$B_i·X_{it}={\beta }_1\Delta {\mathrm{MarketCap}}_{it}+{\beta }_2\Delta {\mathrm{ROA}}_{it}+{\beta }_3\Delta {\mathrm{Leverage}}_{it}+{\beta }_4\Delta {\mathrm{Volatility}}_{it}+{\beta }_5\Delta {\mathrm{Overconfidence}}_{it}$$

This dynamic GMM framework allows us to obtain robust estimates for the relationship between overconfidence and meme stock valuation. In

Table 6. Dynamic system generalized Method of Moments (GMM) regression.

Variables	(1)	(2)	(3)	(4)	(5)
LΔ.Tobin’s Q	−0.135	−0.153	−0.140	−0.138	−0.128	−0.119
	(0.107)	(0.102)	(0.101)	(0.099)	(0.096)	(0.098)
$\Delta$.MarketCap	0.033	0.033	0.028	0.027	0.027	0.020
	(0.035)	(0.032)	(0.036)	(0.025)	(0.032)	(0.031)
$\Delta$.ROA	3.071	2.490	2.990	2.940	3.584	3.876
	(2.038)	(2.431)	(1.957)	(2.040)	(2.152)	(2.366)
$\Delta$.Leverage	0.716	0.419	0.698	0.539	1.210	1.286
	(1.793)	(1.470)	(1.443)	(1.120)	(2.630)	(1.774)
$\Delta$.Volatility	0.756	0.748	0.680	0.719	0.683	0.716
	(0.496)	(0.575)	(0.592)	(0.448)	(0.442)	(0.545)
Overconfidence:
$\Delta$.tv	0.174
	(0.271)
$\Delta$.et		0.565
		(0.956)
$\Delta$.etd			0.461
			(1.120)
$\Delta$.etm				0.320
				(0.828)
$\Delta$.Turnover					0.033 **
					(0.012)
$\Delta$.iso						−0.001
						(0.002)
Constant	−0.102 **	−0.095 **	−0.099 **	−0.092 *	−0.101 **	−0.097 **
	(0.043)	(0.042)	(0.045)	(0.045)	(0.045)	(0.043)
Observations	616	616	616	616	616	616
Number of stocks	28	28	28	28	28	28
F-value	3.931 ***	5.914 ***	4.818 ***	3.400 ***	27.78 ***	4.034 ***
AR(1) test (p-value)	0.0582	0.0650	0.0607	0.0585	0.0571	0.0526
AR(2) test (p-value)	0.630	0.537	0.588	0.569	0.649	0.671
Hansen test (p-value)	0.577	0.702	0.666	0.691	0.588	0.709
Sargan test (p-value)	0.149	0.109	0.106	0.167	0.262	0.207

This table reports the result for the dynamic system generalized Method of Moments (GMM) regression. Significance levels: *: 0.1, **: 0.05, ***: 0.01.

, the coefficient on the lagged difference of Tobin’s Q, $L\Delta \mathrm{Tobin}’\mathrm{s}Q$, is consistently negative across all model specifications. This result suggests a potential mean-reverting pattern, implying that corrective adjustments may follow prior increases in valuation. While insignificant, the consistency of the negative sign aligns with the volatile behavior often observed in speculative markets, where hype tends to be followed by market corrections.

The coefficient of $\Delta \mathrm{MarketCap}$ is positive across all models, ranging from 0.020 to 0.033, but remains statistically insignificant throughout. This negative coefficient suggests that rapid increases in market capitalization may trigger cautious reactions from the market, especially in speculative stocks. Unlike the positive relationship typically observed in more stable contexts Aljifri (2023); Brown and Cliff (2005), the results here may suggest that meme stock valuations are more sentiment-driven and less anchored by firm size. We observe that $\Delta \mathrm{ROA}$ consistently yields positive but statistically insignificant coefficients across all models. This suggests that meme stock investors may not prioritize profitability when evaluating firm value; this aligns with previous results discussed in Table 4. The lack of a significant relationship between $\Delta \mathrm{ROA}$ and firm valuation aligns with the expected speculative nature of meme stocks, where fundamental performance indicators like ROA may often be overshadowed by trading momentum and strong investor sentiment. The coefficients of $\Delta \mathrm{Leverage}$ also remain statistically insignificant across all models, with positive signs but no consistent economic interpretation. This implies that changes in leverage are not systematically incorporated into the valuation of meme stocks. Traditional finance theory suggests that high leverage is a risk factor, but in speculative environments, such signals may be disregarded, particularly when short-term trading gains dominate investment motives. Likewise, $\Delta \mathrm{Volatility}$ exhibits a consistently positive but insignificant relationship with firm valuation. This suggests that volatility is neither a reward nor a penalty in the pricing of meme stocks. Investors may view high volatility as an inherent characteristic of these assets rather than as a risk or an opportunity factor.

The results for alternative overconfidence proxies provide further insights. In Model 1, the trading volume change proxy, $\Delta \mathrm{tv}$, is not statistically significant, suggesting that raw shifts in trading activity alone may not capture overconfidence in this setting. Similarly, the excessive trading proxy $\Delta \mathrm{et}$ in Model 2 yields a positive but insignificant coefficient (0.565), reinforcing the notion that trading frequency alone may not adequately reflect investor exuberance. Models 3 and 4 include deviation-based proxies: $\Delta \mathrm{etd}$ and $\Delta \mathrm{etm}$, reflecting deviations from median and mean trading volumes, respectively. Both coefficients are positive but statistically insignificant, further indicating that meme stock valuations do not systematically respond to excess trading defined relative to central tendencies. Model 5 presents the most compelling evidence. Here, the turnover proxy exhibits a positive and statistically significant coefficient of 0.033 at the 5% level. This result suggests that heightened turnover is closely associated with elevated valuations, consistent with overconfidence-driven trading activity. Among all proxies considered, turnover emerges as the most reliable indicator of speculative investor behavior and its effect on firm value. This aligns with behavioral finance arguments that overconfident investors drive frequent trading and inflate asset prices in speculative markets. Finally, Model 6 incorporates the shares outstanding proxy, $\Delta \mathrm{iso}$, which yields a negative and statistically insignificant coefficient. While changes in share supply might signal demand in some contexts, the result here suggests that such effects are negligible, possibly due to offsetting concerns about dilution or a general disregard for fundamental share structure in highly speculative environments.

In summary, the GMM regression results indicate that among several proxies for overconfidence, turnover emerges as a significant predictor of meme stock valuation. Other proxies, while directionally consistent with theoretical expectations, lack statistical power in this context. The findings highlight the behavioral underpinnings of meme stock pricing, where overconfidence and frequent trading, rather than fundamentals, drive firm valuation (Adebambo and Yan 2018; Trinugroho and Sembel 2011). The evidence reveals the interplay between sentiment-driven trading and corrective forces characteristic of speculative markets.

5. Conclusions

This study investigates the relationship between investor overconfidence and meme stock valuation, employing 28 listed meme stocks downloaded from Refinitiv Eikon Datastream, spanning 2019 to 2024. We estimate several key variables, including Tobin’s Q ratio, market capitalization, return on assets, leverage ratio, historical 90-day volatility, and the primary proxy for overconfidence defined as the changes in trading volume ($\Delta$).

We observe a positive relationship between overconfidence and firm valuation, which aligns with behavioral finance theories. This implies that investor biases can lead to substantial price distortions. However, traditional financial indicators such as leverage and market capitalization show differing effects on firm value. Leverage has a significant negative impact, while market capitalization exhibits a significant positive effect. These results suggest that meme stock investors consistently balance overconfidence and speculative enthusiasm with traditional fundamental analysis. Rather than focusing exclusively on hype and excitement, their behavior reflects a blend of speculative sentiment and rational consideration. Although meme stock valuations often deviate from intrinsic values due to behavioral biases, these deviations appear to incorporate elements of fundamental reasoning.

To ensure the robustness of our results, we conduct two robustness tests. First, we evaluate five alternative measures of investor overconfidence and confirm the consistency of a positive, though not statistically significant, relationship between overconfidence proxies ($et$, $etm$, and $etd$) and meme stock valuation. Interestingly, $turnover$ and $iso$ have a negative relationship with firm valuation, with $iso$ having statistical significance at 5% level, highlighting the unique dynamics of meme stocks where psychological factors intersect with financial fundamentals. Second, we employ a dynamic panel data model (system GMM) to address potential endogeneity issues. The dynamic panel results confirm the significant impact of overconfidence (turnover) on meme stock valuations. The varying responses to different overconfidence proxies reveal that meme stock values are shaped by sentiment-driven trading and the cyclical corrections characteristic of speculative markets.

The limitations of this study include (i) the selection of individual meme stocks, which was constrained by data availability. This study focuses on 28 listed meme stocks identified on Quiver Quant and sourced from the Refinitiv Eikon Datastream, representing a sample of popular meme stocks during the study period. While this dataset provides meaningful insights into meme stock valuation, it may not capture the full diversity of firms that may exhibit similar speculative behavior. As such, emerging or lesser known meme stocks could not be included in this study due to data unavailability, which may limit the generalizability of our findings. We believe that future research could expand this analysis by incorporating a broader dataset that includes a wider range of meme stocks, allowing for a more granular understanding of sentiment-driven valuation dynamics in this market. (ii) The duration covered by this study (2019–2024) reflects the availability of consistent, high-quality data. While this period includes significant market events and the peak of the meme stock phenomenon, it may not capture earlier or future phases of speculative trading in similar assets. The relatively short timeframe restricts our ability to analyze long-term trends or structural shifts in meme stock valuation driven by investor overconfidence. Extending the study period or including additional historical data could provide deeper insights into the cyclical nature of sentiment and its impact on valuation. (iii) The study relies on overconfidence proxies such as changes in trading volume ($\Delta tv$), turnover rate ($turnover$), and changes in outstanding shares ($iso$), among others. While these proxies are well suited to capturing key aspects of investor behavior, they may not fully account for the complexity of psychological biases influencing trading activity. Perhaps future research could develop alternative behavioral metrics to better disentangle sentiment-driven trading from other market forces, potentially refining the understanding of the effects of overconfidence in speculative markets. Despite these limitations, the findings of this study form the foundation for exploring the multifaceted role of investor psychology in meme stock valuation and contribute to the speculative financial literature.