AXT Inc. Securities Litigation

By /

An initial economic assessment of the AXT Inc. Securities Litigation

Introduction

This post contains an early case assessment of the AXT Inc. Securities Case (3:24-cv-02778-MMC). The article discusses the potential damages in the case, the market efficiency indicators for the company, various loss causation issues, and potential settlement ranges.

This article is part of a series of standardized preliminary economic analyses of recently filed cases. In separate publication I described the potential benefits from standardizing some preliminary case assessments. The opinions expressed are those of the author only and do not necessarily reflect the views of their company, employer, or its clients. This article is for general information purposes and is not intended to be and should not be taken as legal advice. All tables and charts sourced from SCA iPortal are published with permission.

Summary of the AXT Inc. Case

AXT Inc. (“AXTI) is a materials science company that develops and produces high-performance compound and single element semiconductor substrates, also known as wafers. While AXTI is a US company, its operations are based in China. The complaint alleges that the company, via its officers, issues misleading statements or made omissions about various problems with its operations during the class period from March 24, 2021 to April 3, 2024. See Figure 1. The misstatements and omissions were detailed for the first time in a report by J Capital Research on April 4, 2024, which is the only alleged disclosure date in the complaint. Specifically, the disclosures assert that:

  • AXTI’s sales and production in China had significantly declined, with production at no more than 50% of capacity since early 2023.
  • The company’s attempted listing of its subsidiary, Tongmei, on the Shanghai stock market was apparently blocked by Chinese regulators.
  • Tongmei had faced environmental problems, including arsenic contamination in groundwater, and has been fined multiple times for non-compliance with safety regulations.
  • The company had been accused of IP infringement and has faced lawsuits for defaulting on wages to employees.
  • AXTI’s physical infrastructure, including buildings and land holdings, had been called into question, with discrepancies between disclosed values and Chinese government records.

According to the complaint, AXTI’s stock fell by $1.73 per share, or 34.94%, on April 4, 2024, and an additional $0.11 per share, or 3.4%, on April 5, 2024.

Figure 1: AXT Inc. class period (SCA iPortal)

Aggregate Damages

Figure 2: Plaintiff-Style Aggregate Damages (SCA iPortal)

The plaintiff style aggregate damages are computed subject to a number of assumptions. See the Appendix: Damages Assumptions below.

Damages Discussion

Damages Estimates

The aggregate damages “Without 90-day Lookback Cap” ($57.1 million, see Figure 2) are roughly equal to the product of 32 million AXTI public float shares and the $1.78 “peak inflation.” The peak inflation is the damage per share on every share purchased from March 24, 2021, through April 3, 2024. The peak inflation is computed as the combined excess price decline on the two-day disclosure event April 4, 2024, and April 5, 2024. See Figure 3.

The aggregate damages “With 90-day Lookback Cap” ($30.6 million, see Figure 2) applies an upper limit to the $1.78 damage per share as stipulated in the PSLRA. [1] The upper limit is the difference between the purchase price and the average price for up to 90 days after the end of the class period.

Figure 3; AXT Inc. Excess Return Computation over April 4-5, 2024 (SCA iPortal)

The effect of a two-day disclosure event

The complaint alleges that the declines on both April 4, 2024, and April 5, 2024, are associated with the disclosures contained in J Capital Research report issued on April 4, 2024. The “plaintiff-style” analysis applied in this article gives maximum benefit to the plaintiff’s allegations, including the adoption of a 2-day disclosure period.

If the analysis assumes that the disclosure impact is limited to the decline on April 4, 2024, only, the peak inflation would be reduced from $1.78 to $1.64. See Figure 13. The aggregate damages would be reduced to the range of $28.9 million (with 90-day Lookback Cap) to $52.3 million (without 90-day Lookback Cap).

There are at least two reasons defense experts can bring up to justify limiting the disclosure impact to April 4, 2024, only.

  • First, the defense experts are likely to argue that in an efficient market, i.e. a market in which all value-relevant information is incorporated into the stock price quickly, the disclosure effect cannot be assumed to take two full days. Remember that the J Capital Research report was released before the market open on April 4, 2024. The market efficiency literature has documented that most information is incorporated within minutes of the release, certainly within the day.[2] Plaintiff experts cannot reject the “efficient market” argument itself because market efficiency is a pre-requisity to obtain class certification. However, as I highlight below, they could argue about the apprpriate speed of information incorporation in the case of AXTI as I discuss below.
  • Second, the defense experts could highlight the fact that the AXTI return on April 5, 2024, when tested on its own, is not statistically significant. See Figure 13. The combined April 4-5, 2024, return is statistically significant because the return on April 4, 2024, is large enough to make the combined return significant as well.

On the other side, plaintiff experts can advance arguments supporting the inclusion of April 5, 2024, in the damage computations.

  • First, plaintiff experts can argue that a short report that questions the viability of the entire company, particularly a small and lesser-known company which operates exclusively in an opaque market like China, could take longer than usual to get incorporated into the stock price. Even simple tasks, like confirming underlying facts and documents from China-based sources, could conceivably take more than a day to complete.
  • Second, plaintiff experts may argue that some of the information released on April 5, 2024, is new. The analysis included in the Loss Caustion section provides some information in that regard. In other words, an argument could be made that the decline on April 5, 2024, is not in response to information released on April 4, 2024, but in response to new information released on April 5, 2024. A potential shortcoming of such an argument would be that it calls for statistical testing of April 5, 2024, on its own. As I mentioned above, and document in the Loss Causation section, the return on this date is not statistically significant. Plaintiff experts would have to make an argument that the ‘totality’ of the information releases on both April 4 and 5, 2024 should be considered as a disclosure rather than each individual day’s news.

Class Certification – Market Efficiency

Defendants usually face an uphill battle in fighting market efficiency on US exchange listed securities. AXTI is not an exception to that rule but, as I show below, it does present some opportunities for defense experts. In the discussion below, I cite the test and the corresponding numeric values, and compare those value to benchmarks provided in a paper by Bhole, Surana and Torchio (2020).[3]

Note that the authors of the paper propose, without supporting analysis, that a factor which is above the 10th percentile for all stocks in the direction required for efficiency should be considered to be indicative of efficiency. For example, if the Cammer turnover statistic for AXTI is higher than the turnover of 10% of the US-listed securities, then Bhole, Surana and Torchio (2020) argue that it should be considered supportive of market efficiency.

Cammer Factors:

Cammer 1: Weekly Volume Turnover

Purpose: According to Cammer v Bloom “[t]he reason the existence of an actively traded market, as evidenced by a large weekly volume of stock trades, suggests there is an efficient market is because it implies significant investor interest in the company.” The decision states that “turnover measured by average weekly trading of 2% or more of the outstanding shares would justify a strong presumption that the market for the security is an efficient one; 1% would justify a substantial presumption.”[4]

Result: Average ratio of weekly trading volume to securities outstanding over the class period is 5.20%. See Figure 4. Excluding the extraordinary spike in turnover during the week of February 26, 2024, the average turnover is 3.5%. Both of these estimated, according to Cammer, justify “a strong presumption that the market for the security is an efficient one.” Both values are also over the 95th percentile among US listed stocks.

Figure 4: Weekly Trading Volume (SCA iPortal)

Cammer 2: Number of Analysts Following the Stock

Purpose: According to Cammer v Bloom, “it would be persuasive to allege a significant number of securities analysts followed and reported on a company’s stock during the class period. The existence of such analysts would imply, for example, the [company’s] reports were closely reviewed by investment professionals, who would in turn make buy/sell recommendations to client investors.”[5]

Result: Average number of analysts covering the company over the class period is 5. See Figure 5. This is below the 50th percentile (aka median) of US listed firms.

Figure 5: Number of Securities Analysts (SCA iPortal)

Cammer 3: Number of Market Makers

Purpose: According to Cammer v Bloom, “[t]he existence of market makers and arbitrageurs *1287 would ensure completion of the market mechanism; these individuals would react swiftly to company news and reported financial results by buying or selling stock and driving it to a changed price level.”

However, the “Market makers count” was criticized for its misleading nature as early as 5 years after the Cammer decision. Barber, Griffin and Lev (1993) concluded that “apparently, market makers just “make a market” in the stock, namely match buy and sell orders, without contributing to the information available about the stock.”[6] The definition of a “market maker” also differs across market structures, e.g., NYSE vs NASDAQ. As a result, economic experts have used ‘proxies’ for what the Cammer decision envisioned are “individuals would react swiftly to company news and reported financial results by buying or selling stock and driving it to a changed price level.”[7] The most common such ‘proxies’ are the percentage of institutional investors holding a security and the short interest as a fraction of the shares outstanding.

Result 1: The average percentage of AXTI stock held by institutional investors during the class period was 57.6%. See Figure 6. This is above the 25th percentile but below the median of US listed securities.

Result 2: The average short interest of AXTI stock as fraction of securities outstanding during the class period was 2.29%. See Figure 6. This is above the 25th percentile but below the median of US listed stocks.

Figure 6: Sophisticated Investors (SCA iPortal)

Cammer 4: Eligibility to file SEC Form S-3

Purpose: According to Cammer v Bloom, “it would be helpful to allege the Company was entitled to file an S-3 Registration Statement in connection with public offerings or, if ineligible, such ineligibility was only because of timing factors rather than because the minimum stock requirements set forth in the instructions to Form S-3 were not met.” [8]

Result: AXTI actually filed 7 forms S-3 as early as December 7, 2021, and as late as May 5, 2022. AXTI naturally satisfied the main requirements for filing form S-3. AXTI’s market capitalization was above the minimum required ($75 million) for filing form S-3 throughout the class period. The stock had traded on a US exchange for longer than 12 months and the company was up to date with its financial filings.

Cammer 5: Cause-and-Effect Relationship Between Material News and Stock Returns

Purpose: The Cause-and-Effect (aka Fifth Cammer) Factor is considered, by most experts and courts, the most direct test of market efficiency. Cammer v Bloom states “it would be helpful to a plaintiff seeking to allege an efficient market to allege empirical facts showing a cause-and-effect relationship between unexpected corporate events or financial releases and an immediate response in the stock price. This, after all, is the essence of an efficient market and the foundation for the fraud on the market theory.”[9]

There are a number of tests that can be applied to this factor. Two standardized tests are presented below:

Result 1: Percent of earnings announcements with statistically significant price reaction during the class period: 50% (6 out of 12). See Figure 7.

This is a result of an event study on the days of earnings releases by AXTI during the class period. Earnings days are the most commonly used dates with potentially material new information. It should be noted, however, that the use of earnings announcements needs to be reviewed further by economic experts to determine whether the lack of statistically significant reaction to any of them is scientifically expected or surprising. Analysis needs to be conducted to determine if the earnings announcements were in line with expectations, contained earnings surprises or material new information such as future guidance. This analysis will involves significant amount of expert judgment.

Figure 7: Earnings Dates Event Study (SCA iPortal)

Result 2: The correlation between volume and security returns IS statistically significant. See Figure 8.

This test assumes that volume is a proxy for news flow.[10] Thus, a test of the correlation between returns and volume is arguably a test of the cause-and-effect between information flow and returns. According to Bhole, Surana and Torchio (2020), over 98% of US listed stocks exhibit statistically significant correlation between daily returns and volumes.

Figure 8: Return – Volume Regression (SCA iPortal)

Krogman Factors

Krogman 1: Market Capitalization

Purpose: According to Krogman v. Sterritt, “[m]arket capitalization, calculated as the number of shares multiplied by the prevailing share price, may be an indicator of market efficiency because there is a greater incentive for stock purchasers to invest in more highly capitalized corporations.”[11]

Results: The average market capitalization of AXTI stock during the class period was $256.4 million, though it varied between $84 million and $523 million. The $256.4 million average is above the 25th percentile but below the median of US listed stocks. See Figure 9.

Figure 9: AXT Inc. Market Capitalization (SCA iPortal)

Krogman 2: Bid-ask Spread

Purpose: According to Krogman v Sterritt, “A large bid-ask spread is indicative of an inefficient market, because it suggests that the stock is too expensive to trade.”[12]

Results: The average Bid-Ask Spread of a AXTI Inc Ordinary Share was, on average, $0.01, or 0.29% of the security price. See Figure 10. An average bid-ask spread of 0.29% is higher than the 50th percentile of bid-ask spreads of US listed companies. (Note: For Bid-Ask spreads LOWER is better)

Figure 10: AXT Inc. Bid-Ask Spread (SCA iPortal)

Krogman 3: Public float

Purpose: According to Krogman v Sterritt, “Because insiders may have private information that is not yet reflected in stock prices, the prices of stocks that have greater holdings by insiders are less likely to accurately reflect all available information about the security.”[13]

Result: The average Public Float of AXTI Inc Ordinary Shares was, on average, 40.4 million, or 93.6% of the outstanding securities. See Figure 11. There is no benchmark for public float shares of outstanding stock of which I am aware.

Figure 11: AXT Inc. Public Float (SCA iPortal)

Autocorrelation

Purpose: Autocorrelation is a term used to describe predictability of future returns by current returns. The presence of autocorrelation, also known as serial correlation, is inconsistent with market efficiency because it implies that not all value-relevant information is quickly incorporated into the security price. Therefore, autocorrelation tests have been a staple of market efficiency testing since the introduction of the Efficient Markets Hypothesis in 1970.[14] Courts have also relied on autocorrelation results in establishing market efficiency.[15]

Result: There is evidence of statistically significant autocorrelation, i.e., predictability, in both the AXTI daily raw and excess returns. See Figure 12. According to the benchmark in Bhole, Surana, and Torchio (2020), about 27% of US listed stocks had a statistically significant autocorrelation. This is an important finding and likely to be brought up by defense experts. Additional questions will have to be addressed by experts on both sides. Important questions are whether the statistically significant autocorrelation is driven by some subperiod of the class period and whether an investor aware of this autocorrelation could generate above-market profits based on this information.

Figure 12: Autocorrelation Regression (SCA iPortal)

Loss Causation / Price Impact

Event Study

The event study is a typical starting point of any analysis of loss causation or price impact.

In the AXT securities case, the complaint alleges that the disclosure impact lasted over two days – April 4 and 5, 2024. There are two ways to look at the statistical significance of the disclosure return.

The first option is to test the statistical significance of the combined 2-day return. This is shown in Figure 3. The combined 2-day return is statistically significant.

The second option is to test the statistical significance of the two days individually. This is shown in Figure 13. The analysis shows that the return on April 4, 2024, is statistically significant and the return on April 5, 2024, is not statistically significant.

The implication of the two points above is that the combined April 4-5, 2024, return, shown in Figure 3, is statistically significant because of the large decline on April 4, 2024, by itself.

Figure 13: Disclosure Dates Event Study (SCA iPortal)

As the Damages section noted, the difference in damages between using 1-day or 2-day disclosure impact is between $1 million (with 90-day Lookback Cap) and $5 million (without 90-day Lookback Cap). Total damages are $30.6 to $57.1 million. So, in terms of damage claims, it is far more important to determine the loss causation on April 4, 2024, than whether April 5, 2024, can be added to the disclosure and damages computations.

Some helpful information in terms of loss causation and price impact comes from the intraday and news analysis.

Intraday Price and News Headlines Analysis

Figure 14 shows AXTI’s intraday price movement on April 4, 2024. It is clear that the majority of the impact, as expected, occurred at the market open. A review of the headlines on the same date clearly attribute the decline to the short report by J Capital Research. There are also reports of law firm investigations of securities law violations.

Figure 14: Intraday Price on Apri 4, 2024 (SCA iPortal)

Figure 15 on the other hand, shows the intraday price movement on April 5, 2024. Note that pattern here is more complicated. The AXTI stock price actually opened higher on the day but gradually declined for the remainder of the day. News headlines on the day repeated the already known information in J Capital Research report. However, additional securities law investigations by more law firms were also announced. As I noted above, further analysis is needed to confirm if the stock price decline on April 5, 2024, is driven by continuous validation of the news released by the J Capital Research or by the additional information released on this date.

Figure 15: Intraday Price on April 5, 2024 (SCA iPortal)

Settlement Analysis

The expected settlement in the AXT Inc. Securities Litigation, at the outset of the case, is approximately $3.05 million. See Figure 16. This is approximately 10% of the roughly $30.6 million in estimated damages and consistent with historical experience.[16]

The prediction interval from the 25th to 75th percentiles (50% confidence) ranges from $1.6 to $5.9 million. The prediction interval is computed via an econometric model calibrated on similar settled cases filed since 2005. The prediction interval takes into account both the uncertainty of the calibrated model parameters and the existence of unpredictable factors in the settlement of every case.

Figure 16: Settlement Prediction Interval (SCA iPortal)

Appendix: Damages Assumptions

  1. The Plaintiff-style damages are the product of “damages per share” and “damaged shares”
  2. Damages per share during the class period are computed as the difference in purchase price inflation and sale price inflation.
    1. price inflation is computed using the sum of the disclosure-days’ excess dollar returns.
    2. Price inflation is assumed to be a constant dollar inflation. Constant dollar inflation is a constant dollar amount between the disclosure dates.
    3. Excess dollar returns (i.e., price changes adjusted for market and industry effects) are computed based on a market model regression using S&P 500 Index (market index) and S&P 500 Biotechnology Index (industry index).
  3. Damages per share for shares sold after the end of the class period are based on the following assumptions:
    1. Without the 90-day Lookback Cap: Damages per share are difference between the purchase price inflation and the price inflation at the end of the class period ($0).
    2. With the 90-day Lookback Cap: As stipulated in the regulation, the damage per share is the smaller of (a) and the difference between the purchase price and the average price over a period of at most 90-days following the class period.
  4. Damaged shares are computed via multi-trader model:
    1. The model counts the number of damaged shares, subject to assumptions, from institutional holdings data.
    2. The model applied a 2-trader trading model for the non-institutional holdings.

Footnotes:

  1. “In any private action arising under this chapter in which the plaintiff seeks to establish damages by reference to the market price of a security, if the plaintiff sells or repurchases the subject security prior to the expiration of the 90-day period described in paragraph (1), the plaintiff’s damages shall not exceed the difference between the purchase or sale price paid or received, as appropriate, by the plaintiff for the security and the mean trading price of the security during the period beginning immediately after dissemination of information correcting the misstatement or omission and ending on the date on which the plaintiff sells or repurchases the security.” 15 U.S. Code § 78u–4 – Private securities litigation ↑

  2. Busse J. and T. Clifton Green, “Market efficiency in real time,” Journal of Financial Economics, Volume 65, Issue 3, 2002, Pages 415-437, ↑

  3. Bharat Bhole, Sunita Surana. & Frank Torchio (2020), “Benchmarking Market Efficiency,” 2020 U. Ill. L. Rev. Online 96. While the paper uses data up to 2018, the 3-year measurement periods provide some credibility to the benchmark values. ↑

  4. Cammer v. Bloom, 711 F. Supp. 1264 (D.N.J. 1989). ↑

  5. Cammer v. Bloom, 711 F. Supp. 1264 (D.N.J. 1989). ↑

  6. Barber, Brad M., Paul A. Griffin, and Baruch Lev. “The fraud-on-the-market theory and the indicators of common stocks’ efficiency.” J. Corp. L. 19 (1993): 285. ↑

  7. Cammer v. Bloom, 711 F. Supp. 1264 (D.N.J. 1989). ↑

  8. Cammer v. Bloom, 711 F. Supp. 1264 (D.N.J. 1989). ↑

  9. Cammer v. Bloom, 711 F. Supp. 1264 (D.N.J. 1989). ↑

  10. For review of studies on the topic, see Jonathan M. Karpoff, The Relation Between Price Changes and Trading Volume: A Survey, 22(1) J. FIN. & QUANTITATIVE ANALYSIS 109, 121 (1987). See also Chen, Gong-Meng, Michael Firth, and Oliver M. Rui. 2001. The dynamic relation between stock returns, trading volume, and volatility. Financial Review 36: 153–74. ↑

  11. Krogman v. Sterritt, 202 F.R.D. 467, 474 (N.D. Tex. 2001). ↑

  12. Krogman v. Sterritt, 202 F.R.D. 467, 474 (N.D. Tex. 2001). ↑

  13. Krogman v. Sterritt, 202 F.R.D. 467, 474 (N.D. Tex. 2001). ↑

  14. Fama, Eugene F. “Efficient capital markets.” Journal of finance 25.2 (1970): 383-417. ↑

  15. In re PolyMedica Corp. Sec. Litig., 453 F. Supp. 2d 260, 276-78 (D. Mass. 2006)) ↑

  16. Cornerstone Research, “Securities Class Action Settlements: 2023 Review and Analysis”. ↑

Scroll to Top