Impact of IPO Review Inquiry Intensity on Prospectus Information Disclosure Updates

jing li; ying li; yanchun zhu; wei zhang; yuxia zhao

Outline

Open Access

Research article

Impact of IPO Review Inquiry Intensity on Prospectus Information Disclosure Updates

jing li¹

,

ying li²

,

yanchun zhu¹^*

,

wei zhang³

,

yuxia zhao⁴

¹

Business School, Beijing Normal University, 100875 Beijing, China

²

Office of Financial Affairs, Central University of Finance and Economics, 100081 Beijing, China

³

School of Information, Central University of Finance and Economics, 100081 Beijing, China

⁴

COFCO Grain Co., Ltd, 100020 Beijing, China

Journal of Accounting, Finance and Auditing Studies

|

Volume 10, Issue 4, 2024

|

Pages 205-225

https://doi.org/10.56578/jafas100403

Received: 09-03-2024,

Revised: 11-01-2024,

Accepted: 11-11-2024,

Available online: 11-20-2024

View Full Article|

Download PDF

Abstract:

The prospectus, as the primary vehicle for issuers to disclose information to the public, plays a crucial role in protecting investors’ rights. Review inquiries serve as an important tool to ensure the quality of the prospectus, as the inquiry and feedback mechanism helps to identify potential risks and enhance the quality of information disclosure. This paper, based on the theory of responsive regulation and the attention-based view, takes companies applying for Initial Public Offering (IPO) on the Science and Technology Innovation Board (STAR) Market and ChiNext Board between 2019 and 2023 as the research samples. Using text analysis methods such as the Latent Dirichlet Allocation (LDA) topic model and dictionary-based methods, this study measures the intensity of review inquiries and the extent of information disclosure. It examines the impact of inquiry topics on the disclosure of corresponding information in the prospectus and explores the moderating effects of company ownership structure, sponsor reputation, and auditor reputation on these relationships. Empirical results indicate that: (1) an increase in the formality of review inquiries enhances the optimization of information disclosure in the prospectus; (2) the focus of review inquiries on specific topics has a significant positive impact on the update of relevant information disclosure in the prospectus; and (3) at the ownership structure level, state-owned enterprises dampen the positive influence of review inquiries on the textual features of the prospectus.

Keywords: Review inquiry, Inquiry intensity, Prospectus, Information disclosure, Topic model

JEL Classification: D81; G34; M41

1. Introduction

The prospectus is the main vehicle for information disclosure during the stock issuance phase under the registration-based system. It contains key information about the company’s current and future business operations, financial condition, issuance details, and potential development prospects, showcasing the company’s development trend. It serves as the basis for investors to make value judgments and investment decisions. However, during the pilot phase of the registration system reform, there have been issues such as low readability, insufficient relevance to investment decision-making, and lack of targeted information disclosure in the prospectus (Liu & Li, 2022; Zhou & Zhou, 2020).

Under the registration-based system, regulatory authorities promote the issuer to disclose specific company information that investors are concerned about through review and inquiry. This helps improve the quality of corporate information disclosure and ultimately reduces market risk. In studies on IPO review inquiries, scholars have mainly focused on their impact on pricing efficiency, IPO market performance, and other economic outcomes, but there has been little research on the mechanism through which they affect the update of information disclosure. Regarding the measurement of review inquiry intensity and information disclosure extent, scholars mostly approach this from the perspective of form or text characteristics (Hu & Wang, 2021), using an indicator system and manual scoring methods (Jiang & Zhang, 2021), which are difficult to replicate. Additionally, with regard to the impact of review inquiries on information disclosure, most scholars in the Chinese context have focused on the STAR Market, lacking samples from the ChiNext Board, and have only explored the first round of inquiries without including the characteristics of subsequent inquiry rounds in the measurement. The mechanism of the impact of review inquiries on information disclosure behavior requires further examination.

Therefore, based on responsive regulation theory and the attention-based view theory, this paper attempts to extract review inquiry intensity and information disclosure characteristics, construct quantitative indicators, and explore the mechanism by which the intensity of IPO review inquiries affects the update of information disclosure in the prospectus, providing decision-making basis and references for exchanges to improve the IPO review inquiry system.

2. Related Research

2.1 Review Inquiries

Scholars have explored the economic consequences of review inquiries from the perspective of their impact on IPO outcomes, IPO pricing efficiency (IPO pricing, stock price synchronicity), and IPO market performance (initial day underpricing, liquidity, volatility, returns).

For example, Lowry et al. (2020) found that inquiries related to revenue recognition were more frequent, and companies with less information disclosure were more likely to suspend the issuance and listing review process. Regarding IPO pricing efficiency, Li & Liu (2017) investigated the mechanism through which Securities and Exchange Commission (SEC) review inquiries influence the formation of IPO prices, suggesting that the U.S. SEC’s regulation reduced speculation, leading companies to be more cautious about pricing. Liu et al. (2024) pointed out that the SEC’s review process reduced information asymmetry during mergers and acquisitions, with transactions receiving inquiries more likely to modify the transaction price and complete the deal. Yang (2023) found that more severe inquiries lead to a more negative market response and increased likelihood of management voluntarily canceling the transaction.

Regarding the impact of review inquiries on market performance, scholars suggest that as a non-punitive form of regulation, review inquiries are negatively correlated with market performance. Lowry et al. (2020) indicated that a stronger review inquiry, to some extent, reflects poorer compliance with the issuer’s information disclosure, which is related to subsequent low liquidity, low returns, and other long-term market performance issues. Other scholars argue that review inquiries have a certain information effect and can reduce information asymmetry, leading to better market performance. For example, Hu & Wang (2021) believed that more rounds of inquiries lead to increased underpricing on the first day of listing and increased volatility in the first month. In addition, some scholars have examined the impact of IPO review inquiry responses on IPO pricing efficiency. Xue & Wang (2022) measured the quality of information disclosure in inquiry response letters from both “quality” and “quantity” perspectives, finding that the larger the volume of information, the higher the level of visualization, and the smaller the density of accounting terminology and reverse components, the lower the underpricing on the first issuance.

Compared to the U.S., China’s implementation of the registration-based system came later. Lu et al. (2020) were the first to use the number of inquiry rounds and questions during the issuance and listing review process to represent the intensity of review inquiries. They examined the intensity of inquiries at different stages by distinguishing the inquiries from different review entities. Some scholars have counted the total number of specific types of questions in each round to represent the exchange’s attention to those issues (Yu et al., 2022a), using registration duration and inquiry rounds to measure the intensity of review inquiries (Hu & Wang, 2021).

Additionally, some scholars have tried to analyze the content characteristics of inquiry letters using topic modeling methods, examining the text features of inquiry letters on the STAR Market and exploring the topics that exchanges focus on for IPO applicants, as well as the mechanism of review inquiries (Jiang & Zhang, 2021; Yu et al., 2022b). For instance, Tong & Li (2024) used the LDA topic model to refine the text features of review feedback letters and analyzed the impact of corporate bond issuance review on issuance pricing. They found that exchanges’ review inquiries significantly increased the bond issuance price spread, but issuers’ high-quality responses could effectively alleviate this impact.

2.2 Prospectus Information Disclosure

Scholars have used text analysis methods based on theories such as the winner’s curse hypothesis, signaling theory, and book-building theory to extract text features (text length, comprehensibility, readability), information content (effective information content, technical information content, word count per sentence), and management tone (tone, viewpoints) in the prospectus, analyzing how these features impact IPO pricing, market performance, and other economic outcomes (Hu & Wang, 2021).

Regarding text features, Zhou & Zhou (2020) developed a readability index for financial texts and analyzed the mechanism by which prospectus readability affects IPO underpricing. They argue that improving the readability of the prospectus and more accurately conveying company information helps create a better corporate image, reduces information asymmetry between investors and issuers, and thus reduces IPO underpricing. In terms of information content, Hanley & Hoberg (2010) divided the prospectus into two parts: standard and informational, and found that a higher content of information led to more accurate pricing and less underpricing. Building on this approach, Li et al. (2023) explored the relationship between review inquiries, IPO performance, and stock price behavior using Tibet Duo Rui Pharmaceutical as a case study. They found that review inquiries had a certain negative impact on the company’s post-IPO performance and stock price behavior, suggesting that companies receiving more inquiries tended to perform worse in terms of profitability and growth after the IPO. Review inquiries might raise investor concerns about the company, thereby affecting stock price performance.

In terms of management tone, Ferris et al. (2013) used the proportion of negative tone words to measure the conservatism of the prospectus, finding that more conservative tones in the prospectus led to underpricing. Brau et al. (2016) argued that an overuse of positive words could also result in underpricing. Loughran & McDonald (2013) found that IPOs with highly uncertain tones in the text had higher first-day returns, larger absolute offer price revisions, and higher post-IPO volatility.

2.3 Impact of Review Inquiries on Prospectus Information Disclosure

Scholars have mainly studied the impact of IPO review inquiries on the quality of information disclosure from the perspectives of specific topic attention and improvements in the expression features of the prospectus. For example, Lowry et al. (2020) used the LDA topic model to extract the core topics of SEC review inquiries in the U.S. and applied Kullback-Leibler (KL) divergence to match these topics with the related content in the prospectus. They found that listed companies actively added relevant content in response to the topics raised by the SEC in the inquiry. Hu & Wang (2021) focused on the degree of information disclosure updates after a company has been subjected to review inquiries. Their empirical results showed that, after IPO review inquiries, companies tended to enrich the content of their prospectuses on specific topics, significantly increasing word count per sentence while reducing the proportion of technical jargon to improve readability and help investors better understand the information. Xue & Wang (2022) examined the quality of information disclosure in inquiry response letters from the perspectives of both “quality” and “quantity.” They found that higher-quality responses led to increased word count, page numbers, and other measures of information “quantity,” while also improving the “quality” of the information. Jiang & Zhang (2021) approached the issue from the perspective of key issues and established an indicator system to manually score the content of key issues in the prospectus. They found that the more the review inquiry focused on topics such as risk, innovation, foresight, and competition, the more likely it was to guide the issuer’s management to supplement and improve the disclosure of key issues.

In research related to IPO review inquiries, scholars have mainly focused on their impact on pricing efficiency, IPO market performance, and other economic consequences, with relatively little research on the mechanisms by which they affect information disclosure. Most studies have been based on formal characteristics or text content features, using topic models to measure review inquiry intensity and updates in prospectus information disclosure. However, these methods tend to be relatively crude, making it difficult to precisely identify the key topics raised in review inquiries. Alternatively, using manual scoring systems to evaluate these indicators is difficult to replicate (Jiang & Zhang, 2021). Moreover, most research has focused on the STAR Market, with fewer samples from the ChiNext Board. Additionally, research has primarily explored the first round of inquiries, without considering the characteristics of subsequent rounds of inquiry letters. When interpreting the economic implications of extracted inquiry topics, many studies rely on expert experience, which can introduce subjectivity.

To address these shortcomings, this study includes samples from both the STAR Market and ChiNext Board and uses text analysis methods such as the LDA topic model and dictionary method to measure the intensity of review inquiries and the level of information disclosure updates in the prospectus. It aims to explore the incentives behind how review inquiry intensity affects information disclosure updates.

3. Theoretical Foundation and Research Hypotheses

3.1 Theoretical Foundation

3.1.1 Responsive regulation theory

Responsive regulation theory aims to select appropriate and targeted regulatory strategies or measures based on the specific context of the regulated entities. It emphasizes integrating regulatory resources, closely interacting with the regulated entities, and fostering their self-regulatory awareness to achieve optimal regulatory outcomes. Its core includes four aspects: response, shaping, collaboration, and relationality (Ayres & Braithwaite, 1995).

Responsive regulation theory provides the theoretical foundation for the registration-based system. Under the registration system, review inquiries are key to ensuring information quality, reflecting the four main elements of responsive regulation theory: response, shaping, collaboration, and relationality. The inquiry requires regulatory authorities to raise targeted questions based on the issuer’s situation, guiding them to improve disclosure compliance while emphasizing cooperation and building close relationships with the issuer and intermediary institutions. After receiving an inquiry letter, the issuer is required to understand and analyze the concerns raised by the exchange and respond to them one by one. Through this approach, regulatory effectiveness and market efficiency are enhanced.

3.1.2 Attention-based view

The attention-based view proposed by Ocasio (1997) regards enterprises as attention allocation systems. It emphasizes that corporate behavior is essentially a reflection of how decision-makers manage and allocate their attention, indicating that the allocation of attention is constrained by the personal traits of decision-makers and is also influenced by external environmental factors (Ocasio, 1997). To fully understand how a company allocates attention, it is necessary to consider not only the individual factors of decision-makers but also their internal and external environments and how they interpret these environments.

Ocasio (1997) defined attention as the process in which decision-makers allocate time and energy to specific issues and their solutions and take related actions. The allocation of attention is a dynamic process that involves three core principles: focusing, contextualizing, and structuring allocation. These principles together reveal six key factors that influence corporate attention allocation: decision-making environment, issues and answers, procedures and communication channels, attention structures, decision-makers, and corporate behavior.

Under the registration-based system, when a company is conducting an IPO, the regulatory bodies responsible for reviewing the issuer and raising questions set the tone for the regulatory environment. In this environment, the issuer’s application team will notice the intensity of the inquiry on specific issues raised by the regulatory authorities. This attention will, in turn, affect how they allocate attention to different disclosure contents, leading to corresponding adjustments in their information disclosure. This suggests that the regulatory environment the issuer faces is key to shaping how management allocates attention and provides theoretical support for the “review inquiry → management attention allocation → information disclosure behavior adjustment” mechanism proposed in this paper.

3.2 Research Hypotheses

Based on responsive regulation theory, attention-based view, and other theories, scholars have found that IPO review inquiries strengthen the role of exchanges as non-governmental institutions in regulating the capital market. Exchanges and other review bodies apply professional knowledge to inquire into the issuer’s information disclosure from the perspectives of regulators and investors, prompting the issuer to supplement and improve their disclosures (Yu et al., 2022a). Secondly, the issuance of an audit inquiry letter has a certain deterrent effect. This regulatory measure forces issuers to focus on the integrity and accuracy of their disclosures to avoid facing stricter regulatory reviews and potential market losses. Finally, review inquiries have a market governance effect. The public disclosure of the inquiry and response content allows the issuer’s information disclosure to be monitored by the market, increasing transparency (Benveniste & Spindt, 1989).

In 2023, the China Securities Regulatory Commission issued guidelines that state “the prospectus should be easy for investors to read, clear, concise, and as simple as possible, using charts, images, or other intuitive disclosure methods to enhance readability and comprehensibility.” Based on this requirement, this study hypothesizes that as the intensity of review inquiries increases, the readability and comprehensibility of the prospectus improve, with more visual and quantitative information, and the language becoming clearer. Furthermore, if a large portion of the inquiry concerns a specific disclosure issue or wording, it will draw the issuer’s attention to that issue, causing the issuer to actively revise and improve the relevant disclosures and expressions. Therefore, the following hypothesis is proposed:

H1: Other things being equal, the greater the intensity of the exchange’s review inquiries to the company, the higher the level of updates in the company’s prospectus information disclosure.

Empirical studies have shown that political connections between companies and the government can influence the effectiveness of punitive regulation. Companies with close government ties face lighter penalties for violations, and these penalties may be difficult to enforce effectively (Anderson, 1999). There is a time lag effect in the regulatory authorities’ handling of violations by politically connected companies, and political connections weaken the enforcement efficiency of legal protection for small and medium-sized investors (Xu et al., 2013). In the case of non-punitive regulation, review inquiries are less effective for state-owned enterprises. Non-state-owned enterprises are more closely watched by future investors during the IPO process, and their inquiry letters are more likely to influence investor confidence, increasing their financing challenges and leading to economic losses. Therefore, these companies may be forced to disclose more high-quality information to alleviate investors’ concerns (Chen et al., 2019). However, existing literature has seldom explored whether the mechanism of review inquiries and their regulatory effectiveness in updating prospectus information disclosure differs depending on the nature of the company’s ownership. Based on this, this study uses the proportion of state-owned shares to measure the company’s ownership nature and examines the moderating effect of ownership nature, proposing the following hypothesis:

H2: Other things being equal, the higher the proportion of state-owned shares in the company, the weaker the positive impact of review inquiry intensity on the level of updates in the prospectus information disclosure.

Intermediaries are the first gatekeepers of the prospectus, undertaking substantive review work. To enhance or maintain their reputation rankings, intermediaries are motivated to encourage listed companies to improve their information disclosure quality (Mao et al., 2022). According to the law of diminishing marginal returns, accounting firms with lower reputation rankings have a stronger governance effect with financial report inquiry letters (Fu & Zeng, 2022). Furthermore, because high-reputation intermediaries provide higher quality information disclosures, while low-reputation intermediaries offer lower quality disclosures, review inquiries tend to have a more significant governance effect for companies hiring low-reputation intermediaries than for those hiring high-reputation ones. Since law firms are less involved in writing the prospectus, this study mainly focuses on the moderating role of the reputation of the sponsor and accounting firms in the IPO process.

Thus, the following hypotheses are proposed:

H3: Other things being equal, for companies hiring low-reputation sponsor firms, the positive impact of review inquiry intensity on the level of updates in the prospectus information disclosure is stronger.

H4: Other things being equal, for companies hiring low-reputation accounting firms, the positive impact of review inquiry intensity on the level of updates in the prospectus information disclosure is stronger.

4. Research Design

4.1 Sample Selection and Data Sources

The study selects companies that applied for issuance and listing on the STAR Market and the ChiNext Board between June 2019 and October 2023 as the research subjects. IPO company reply documents for each round of review inquiries were crawled from the official information disclosure websites of the Shenzhen Stock Exchange and Shanghai Stock Exchange, resulting in a total of 7,380 review inquiry reply documents. The study checks whether these correspond to the draft and registration versions of the prospectus, and after excluding those, 5,165 review inquiry reply documents remain. In these documents, there are instances where a company submits multiple replies to a single round of inquiry letters, so only the first response is retained for each inquiry letter. After cross-referencing with the IPO database samples, 2,658 review inquiry reply documents remain, corresponding to 1,094 companies. After excluding companies with missing control variable data, the study finally includes 964 companies as valid samples, corresponding to 2,353 review inquiry letters, 964 prospectus drafts, and 964 prospectus registration versions. The financial data of the companies involved in the study comes from Guotai An (CSAMR), and the ranking of sponsor institutions comes from the China Securities Regulatory Commission, manually organized.

4.2 Text Acquisition and Preprocessing

(1) Text acquisition

Review inquiry letters: The questions in the review inquiry letters were collected from the IPO database and split. These were merged at the “company-round” dimension to form a textual corpus for each company’s review inquiry letters by round.

Prospectus: Prospectus documents in PDF format were downloaded from the official websites. Python was used to extract text from the PDFs and save it in .txt files.

(2) Text preprocessing

The company names, geographic names, and common but economically meaningless professional terms were added to the stopword list. At the same time, the company names, geographic names, accounting and financial terms, professional terms, etc., were de-duplicated and compiled into a retained word list. The Jieba library in Python was used to perform Chinese word segmentation on both types of text corpora. Following the approach of Dyer et al. (2017), word frequencies for the review inquiry letters and prospectuses were calculated after segmentation, and the top 100 (1,000) most frequent words were manually reviewed. Words deemed meaningless were added to the stopword list and the texts were re-segmented.

(3) LDA topic model training

The LDA topic model aims to fit a topic distribution for the given text data, and the results depend heavily on the training corpus used (Omar et al., 2015). Referring to Lowry et al. (2020), the topics of the inquiry letters and the prospectuses were used to train the LDA model separately. A lower perplexity range for the number of topics was found to be between 10-13, and after comparing the consistency of models with different numbers of topics, 12 topics were determined for the review inquiry letters, based on which the LDA topic model for the inquiry letters was trained. Similarly, the number of topics for the prospectus was set to 30, and the LDA topic model for the prospectus was trained.

(4) Matching topics between review inquiry letters and prospectus

KL divergence was used to match the topics between the prospectus and the inquiry letters (Lowry et al., 2020). First, the intersection of the topic word distributions of the inquiry letters and the prospectus was calculated to form a topic vector. Then, the topic vectors for the prospectus and the inquiry letters were reconstructed. Finally, for each of the 12 review inquiry letter topics, the KL distance with the 30 prospectus topics was calculated. The prospectus topic closest to each inquiry letter topic i was chosen as its corresponding topic.

(5) Economic meaning of topics

The study selected project titles that appeared in at least 300 different letters for research purposes. The previously trained LDA topic model was applied to the question paragraphs corresponding to these project titles to identify the topic most closely matched with each paragraph. The topic most representative of the text for each project title was then used to label that project title. Among all the question paragraph samples, 413 paragraphs belonged to risk disclosure, with 323 paragraphs (78.2%) classified as Topic 1. Thus, Topic 1 was labeled as the “Risk Disclosure” topic. In the end, the 12 topics were labeled as follows: “Risk Disclosure,” “Production and Business,” “Profit and Loss Situation,” “Core Technology,” “Customers and Suppliers,” “Stock-based Payments,” “Gross Profit Margin,” “Revenue Recognition,” “Sales,” “Board of Directors/Supervisors/Executives,” and “Assets and Liabilities.”

4.3 Variable Definitions

4.3.1 Review inquiry intensity

Based on existing studies, this paper measures the intensity of review inquiries using three parameters: the number of inquiry rounds (LNum), the number of questions in the first round of inquiry (FQNum), and the total number of questions in the inquiry (AQNum). Additionally, the intensity of review inquiries on a specific topic is measured using the LDA topic model (LTopici) (Omar et al., 2015; Yu et al., 2022b). The calculation formula is as follows:

${LTopic}_i={ln}(\text{Probability of Topic}\ i \% \times \text{Total character count of review inquiry letter} )$

(1)

4.3.2 Prospectus information disclosure

Drawing from existing literature, the following six aspects are used to construct textual feature indicators for the prospectus:

(1) Text length: The total number of characters in the prospectus is used to represent the text length, which measures the overall information content of the prospectus.

(2) Text readability: The readability of the text is measured by dividing the total number of words in the prospectus by the number of sentence-ending and pause marks. A higher word count per punctuation mark indicates higher text complexity and lower readability.

(3) Text understandability: The percentage of accounting terms in the text is used to measure text understandability. The greater the density of financial and accounting terminology in the annual report, the higher the complexity and the lower the understandability of the text (Wang et al., 2018).

(4) Text quantitative information: Referring to Huang et al. (2018), quantitative information in the text is measured by counting the occurrences of symbols like “%”, “￥”, “$”, and other numeric characters.

(5) Text visual information: According to Xue & Wang (2022), the degree of visualization in the prospectus is measured by the number of lines occupied by images and tables in the document.

(6) Text ambiguity: Based on the Chinese financial sentiment dictionary developed by Zhang & Zhou (2020), uncertainty and negative terms are used to measure the level of ambiguity in the prospectus. Furthermore, from the issuer’s response perspective, textual indicators for information disclosure on specific topics in the prospectus are constructed (Yu et al., 2022a), measured by the following formula:

${PTopic}_i={ln}\left(\frac{\text { Probability of Topic } i \% \times \text { Prospectus registration draft total character count }}{\text { Probability of Topic } i \% \times \text { Prospectus draft total character count }}\right)$

(2)

4.3.3 Moderating variables

State-owned property: Referring to previous studies, the percentage of shares held by state-owned shareholders in the listed company is used to measure state-owned property.

Auditor reputation: Auditor reputation is evaluated using business income rankings. The top ten accounting firms in terms of business income for the year are classified as high-reputation firms, while the remaining firms are classified as low-reputation firms.

Sponsor reputation: According to the China Securities Regulatory Commission’s classification supervision regulations, the rating results of brokers each year are used to assign a reputation score to underwriters. When the broker’s rating is AA or higher, the underwriter is classified as high-reputation, while others are classified as low-reputation.

4.3.4 Control variables

Based on literature (Hu & Wang, 2021; Jiang & Zhang, 2021; Lowry et al., 2020; Yu et al., 2022a), the following variables are selected as control variables: Ownership Concentration (OC), Return on Equity (ROE), Debt-to-Asset Ratio (Lev), Company Size (CSize), Company Age (LnAge), Venture Capital (PEVC), R&D Investment Ratio (R&D), Issuance Size (Offer_size), Draft Text Length (V1_length). To control for industry and year effects on the regression results, industry effects (Ind), year effects (Year), and board segment types (Seg) are included as dummy variables in the regression. The variable definitions are shown in Table 1.

Table 1. Variable definitions

Variable Type	Variable Name	Variable Symbol	Variable Description
Prospectus Information Disclosure Update Indicators	Text Length Update	LenUp	ln(Character count of the registration draft/Character count of the filing draft)
	Text Readability Update	SenUP	Ratio of sentence-readability word count in the registered draft to that in the draft
	Text Understandability Update	UnderstandUp	Ratio of accounting professional vocabulary in the registered draft to that in the draft
	Text Quantitative Information Update	QuantityUP	Ratio of numeric characters and units in the registered draft to those in the draft
	Text Visualization Information Update	VisualUP	Ratio of image and table rows in the registered draft to those in the draft
	Text Ambiguity Update	FuzzyUP	Ratio of uncertain and negative words in the registered draft to those in the draft
	Topic i Information Disclosure Update	PTopic_i	Derived from formula (2), the higher the value, the higher the information disclosure update for Topic i in the prospectus
Review Inquiry Intensity Indicators	Review Inquiry Rounds	LNum	Total number of review inquiries from IPO pre-disclosure to meeting stage
	First Round Inquiry Questions	FQNum	Number of questions in the first round of review inquiry
	Total Number of Questions	AQNum	Total number of questions in all review inquiries received during IPO
	Topic i Inquiry Intensity	LTopic_i	Derived from formula (1), the higher the value, the greater the inquiry intensity on Topic i
Moderating Variables	State-Owned Property	Nature	Sum of shares held by state-owned shareholders
	Auditor Reputation	Account_repu	Set to 1 if business income is in the top ten, otherwise 0
	Sponsor Reputation	Sponsor_repu	Set to 1 if the rating is AA or higher, otherwise 0
Control Variables	Ownership Concentration	OC	Proportion of shares held by the largest shareholder
	Return on Equity	ROE	Ratio of net profit to net assets
	Debt-to-Asset Ratio	Lev	Ratio of total debt to total assets
	Company Size	CSize	Natural logarithm of total assets
	Venture Capital	PEVC	Set to 1 if venture capital is present, otherwise 0
	Company Age	LnAge	Natural logarithm of the company’s age
	R&D Investment Ratio	R&D	Ratio of R&D investment to operating income
	Issuance Size	Offer_size	Natural logarithm of the raised funds from the IPO
	Draft Text Length	V1_length	Natural logarithm of the character count of the main text in the prospectus draft
	Segment Type	Seg	Set to 1 for the STAR Market, 0 for the ChiNext Board
	Year Effect	Year	Year of the company’s IPO application
	Industry Effect	Ind	Industry of the company

4.4 Model Construction

To test H1, the following regression Models 1 and 2 are established:

${InfoUp} =\beta_0+\beta_1 {Letters} +\beta_2 {Controls} +\varepsilon$

(3)

${PTopic}_i=\beta_0+\beta_1 {LTopic}_i+\beta_2 {Controls} +\varepsilon$

(4)

In Formula (3), InfoUp refers to the update indicators for six prospectus text features: LenUp, SenUP, UnderstandUp, QuantityUP, VisualUP, and FuzzyUP. These indicators are individually regressed in Model 1 as the dependent variables.

Letters refers to the three inquiry intensity indicators: LNum, FQNum, and AQNum. These indicators are regressed one at a time as independent variables in Model 1.

In Formula (4), PTopic_i refers to the disclosure update of topic i in the registration draft and formal draft of the prospectus, while Ltopic_i refers to the inquiry intensity of topic i in the inquiry letter. Controls represent control variables.

To test H2, the following regression models 3 and 4 are established:

${InfoUp}=\beta_0+\beta_1 {Letters} +\beta_2 {Nature} +\beta_3 {Letters} \times {Nature} +\beta_4 {Controls} +\varepsilon$

(5)

${PTopic}_i=\beta_0+\beta_1 {LTopic}_i+\beta_2 {Nature} + \beta_3 {LTopic}_i \times {Nature} +\beta_2 {Controls} +\varepsilon$

(6)

Here, Letters × Nature and LTopic_i × Nature represent the moderating effect of state-owned property.

To test H3 and H4, using Formulas (3) and (4), group regression can be conducted based on the reputation level of the intermediary organizations (high reputation vs. low reputation), the significance and magnitude of the regression coefficients can be observed, and a test of the difference in regression coefficients can be conducted accordingly.

5. Empirical Analysis

5.1 Descriptive Statistics

Descriptive statistics for the variables in the sample are presented in Table 2.

Table 2. Descriptive statistics of key variables

Variable	Sample Size	Minimum	Maximum	Mean	Standard Deviation	Median
LNum	964	1	6	2.441	0.765	2
FQNum	964	5	91	27.998	11.042	27
AQNum	964	5	116	42.5	17.466	40
LenUp	964	-0.304	0.928	0.295	0.235	0.244
SenUP	964	-0.098	0.132	0.012	0.028	0.01
UnderstandUp	964	-0.275	0.135	-0.075	0.065	-0.067
QuantityUP	964	-0.219	2.317	0.468	0.446	0.351
VisualUP	964	-0.341	2.231	0.348	0.374	0.243
FuzzyUP	964	-0.194	0.286	-0.002	0.064	-0.006
LTopic₁	964	-0.87	9.007	4.137	2.626	4.663
LTopic₂	964	-0.866	9.117	4.963	2.519	5.403
LTopic₃	964	-0.854	9.02	4.464	2.18	4.749
LTopic₄	964	-0.852	9.411	4.901	2.458	5.116
LTopic₅	964	-0.866	8.98	3.995	2.583	4.142
LTopic₆	964	-0.862	9.485	3.24	2.773	3.324
LTopic₇	964	-0.886	9.001	4.405	2.435	4.555
LTopic₈	964	-0.831	9.443	5.627	2.21	5.64
LTopic₉	964	-0.875	9.311	5.266	2.332	5.477
LTopic₁₀	964	-0.813	9.359	5.729	1.979	5.942
LTopic₁₁	964	1.482	9.364	7.591	1.304	7.908
LTopic₁₂	964	-0.875	9.047	3.792	2.564	3.99
PTopic₁	964	-0.553	1.375	0.301	0.335	0.278
PTopic₂	964	-0.546	0.969	0.17	0.276	0.158
PTopic₃	964	-0.546	0.918	0.11	0.256	0.102
PTopic₄	964	-0.678	0.999	0.185	0.262	0.185
PTopic₅	964	-0.724	0.891	0.079	0.272	0.069
PTopic₆	964	-0.604	1.02	0.105	0.284	0.097
PTopic₇	964	-0.44	2.435	0.745	0.554	0.674
PTopic₈	964	-0.897	1.614	0.173	0.38	0.128
PTopic₉	964	-0.649	1.542	0.317	0.394	0.275
PTopic₁₀	964	-0.809	1.068	0.094	0.319	0.089
PTopic₁₁	964	-0.679	1.454	0.307	0.362	0.285
PTopic₁₂	964	-0.462	1.192	0.287	0.291	0.256
Nature	964	0	0.778	0.068	0.157	0.01
OC	964	0.12	0.982	0.433	0.188	0.41
ROE	964	-0.602	0.667	0.129	0.128	0.124
Lev	964	0.055	0.95	0.38	0.177	0.362
CSize	964	9.622	14.92	11.307	0.866	11.179
LnAge	964	1.792	3.62	2.835	0.343	2.89
R&D	964	0.001	75.256	0.518	5.36	0.058
Offer_size	964	0.89	4.684	1.911	0.659	1.791
V1_length	964	12.143	13.327	12.607	0.191	12.596

5.2 Correlation Analysis

The Pearson correlation coefficients between the key variables are calculated and presented in Table 3 and Table 4. In Table 3, the correlation coefficients in Model 1 show that LNum is positively correlated with LenUp, QuantityUP, and VisualUP. It is negatively correlated with UnderstandUp, which is consistent with expectations. That is, as the number of review rounds, first-round questions, and total number of questions increase, the text length increases, the proportion of accounting terminology decreases, and the text becomes easier to understand, with more quantitative and visual information.

In Table 4, except for Topic 1 and Topic 12, the correlation coefficients between LTopic_i and PTopic_i are positive, which is also in line with expectations. This suggests that the greater the inquiry intensity from the review agency on a particular topic, the greater the update in the disclosure of the corresponding topic in the prospectus. The Pearson correlation coefficients between the key variables are all below 0.8, indicating that the variables selected in this study are reasonable and do not cause multicollinearity, which would affect the signs of the regression coefficients.

Table 3. Pearson correlation coefficients for Model 1 variables

	*LNum*	*FQNum*	*AQNum*	*LenUp*	*SenUP*	*UnderstandUp*
LNum	1
FQNum	0.439	1
AQNum	0.671	0.911	1
LenUp	0.304	0.265	0.387	1
SenUP	0.151	0.178	0.226	0.44	1
UnderstandUp	-0.156	-0.03	-0.118	-0.372	-0.252	1
QuantityUP	0.262	0.195	0.311	0.931	0.39	-0.287
VisualUP	0.243	0.183	0.307	0.853	0.436	-0.358
FuzzyUP	0.031	0.196	0.164	-0.048	-0.22	0.363
OC	-0.087	-0.125	-0.098	0.082	0.009	0.032
ROE	0.044	-0.061	-0.019	0.136	0.091	-0.107
Lev	-0.027	-0.007	-0.006	-0.013	0.007	0.068
CSize	-0.035	0.087	0.047	-0.1	-0.157	0.172
LnAge	0.004	0.072	0.066	0.092	-0.008	0.065
R&D	-0.022	-0.009	-0.032	-0.051	0.011	0.012
Offer_size	-0.09	0.037	-0.035	-0.199	-0.131	0.105
V1_length	-0.145	-0.108	-0.133	-0.326	-0.198	0.158
	*QuantityUP*	*VisualUP*	*FuzzyUP*	OC	*ROE*	*Lev*
uantityUP	1
VisualUP	0.854	1
FuzzyUP	-0.076	-0.16	1
OC	0.107	0.106	-0.029	1
ROE	0.15	0.144	-0.153	0.131	1
Lev	-0.026	-0.004	0.052	0.083	-0.229	1
CSize	-0.108	-0.099	0.202	0.117	-0.422	0.416
LnAge	0.094	0.083	0.039	0.093	0.096	0.006
R&D	-0.059	-0.036	0.058	-0.055	-0.357	-0.061
Offer_size	-0.199	-0.167	0.109	0.048	-0.234	0.168
V1_length	-0.322	-0.27	0.136	0.04	-0.28	0.202
	*CSize*	*LnAge*	*R&D*	*Offer_Size*	*V1_Length*
CSize	1
LnAge	-0.027	1
R&D	0.03	-0.067	1
Offer_size	0.668	-0.206	0.147	1
V1_length	0.491	-0.014	0.069	0.382	1

Table 4. Pearson correlation coefficients for Model 2 variables

	*LTopic₁*	*LTopic₂*	*LTopic₃*	*LTopic₄*	*LTopic₅*	*LTopic₆*	*LTopic₇*	*LTopic₈*
LTopic₁	1
LTopic₂	0.208	1
LTopic₃	0.409	0.086	1
LTopic₄	0.414	0.41	0.367	1
LTopic₅	0.085	0.384	-0.209	0.158	1
LTopic₆	0.453	-0.274	0.213	-0.012	-0.102	1
LTopic₇	-0.108	0.43	-0.202	0.163	0.456	-0.308	1
LTopic₈	0.212	-0.117	0.537	0.148	-0.423	0.149	-0.354	1
LTopic₉	0.111	0.005	-0.068	0.085	0.059	0.096	0.192	-0.163
LTopic₁₀	0.351	0.018	0.237	0.272	0.201	0.356	-0.008	0.057
LTopic₁₁	-0.157	0.077	0.033	-0.1	0.173	-0.038	0.264	0.04
LTopic₁₂	0.169	0.395	0.203	0.13	0.297	-0.045	0.233	0.023
PTopic₁	-0.138	-0.106	-0.04	-0.177	-0.044	-0.031	0.068	0.058
PTopic₂	-0.002	0.102	0.11	0.102	-0.021	-0.027	0.06	0.064
PTopic₃	-0.025	-0.001	0.058	-0.032	-0.084	-0.017	0.007	0.075
PTopic₄	-0.047	0.108	0.019	0.132	-0.009	-0.108	0.132	-0.04
PTopic₅	-0.026	-0.109	-0.031	-0.156	0.084	0.083	0.079	-0.018
PTopic₆	-0.039	-0.258	-0.022	-0.213	-0.105	0.177	-0.087	0.021
PTopic₇	-0.158	-0.085	-0.033	-0.155	-0.015	-0.055	0.083	0.065
PTopic₈	-0.158	-0.201	0.069	-0.238	-0.168	-0.007	-0.115	0.279
PTopic₉	-0.181	-0.172	-0.021	-0.213	-0.124	-0.049	0.056	0.068
PTopic₁₀	-0.053	-0.047	0.028	-0.074	-0.009	0	0.084	0.001
PTopic₁₁	-0.162	-0.027	-0.033	-0.122	0.01	-0.092	0.122	0.013
PTopic₁₂	-0.197	-0.048	-0.138	-0.091	0.115	-0.089	0.133	-0.066
	*LTopic₉*	*LTopic₁₀*	*LTopic₁₁*	*LTopic₁₂*	*PTopic₁*	*PTopic₂*	*PTopic₃*	*PTopic₄*
LTopic₉	1
LTopic₁₀	0.05	1
LTopic₁₁	0.173	0.284	1
LTopic₁₂	-0.256	0.237	0.239	1
PTopic₁	0.169	0.074	0.33	-0.063	1
PTopic₂	0.027	0.077	0.147	0.053	0.383	1
PTopic₃	0.12	0.065	0.129	0	0.343	0.261	1
PTopic₄	0.094	0.051	0.151	-0.013	0.374	0.374	0.26	1
PTopic₅	0.013	0.161	0.223	0.084	0.415	0.256	0.231	0.254
PTopic₆	0.123	0.051	0.155	-0.157	0.436	0.146	0.207	0.162
PTopic₇	0.169	0.05	0.322	-0.033	0.702	0.453	0.399	0.42
PTopic₈	-0.14	0.019	0.24	-0.014	0.529	0.308	0.291	0.207
PTopic₉	0.229	-0.013	0.283	-0.134	0.671	0.385	0.361	0.38
	*LTopic₉*	*LTopic₁₀*	*LTopic₁₁*	*LTopic₁₂*	*PTopic₁*	*PTopic₂*	*PTopic₃*	*PTopic₄*
PTopic₁₀	0.138	0.121	0.176	0.03	0.444	0.358	0.33	0.252
PTopic₁₁	0.113	0.044	0.29	-0.013	0.653	0.459	0.373	0.443
PTopic₁₂	0.087	0.069	0.285	-0.032	0.54	0.342	0.303	0.373
	*PTopic₅*	*PTopic₆*	*PTopic₇*	*PTopic₈*	*PTopic₉*	*PTopic₁₀*	*PTopic₁₁*	*PTopic₁₂*
PTopic₅	1
PTopic₆	0.317	1
PTopic₇	0.392	0.317	1
PTopic₈	0.397	0.36	0.518	1
PTopic₉	0.391	0.399	0.708	0.524	1
PTopic₁₀	0.337	0.27	0.465	0.366	0.498	1
PTopic₁₁	0.385	0.286	0.747	0.495	0.656	0.5	1
PTopic₁₂	0.382	0.294	0.62	0.397	0.537	0.381	0.579	1

5.3 Multivariate Regression Analysis

5.3.1 The mechanism of inquiry intensity impacting the update of prospectus information disclosure

To test H1, regression analysis was conducted based on Models 1 and 2 using the sample data. After controlling for year, industry, board type, and other control variables, the study examines the impact of inquiry intensity on six textual features of the prospectus and the effect of review inquiry intensity on the information disclosure of the corresponding topics in the prospectus. The regression results are presented in Table 5, Table 6, Table 7, Table 8, and Table 9.

As shown in Table 5, the number of review inquiry rounds is significantly positively correlated with the changes in the length of the prospectus text, text readability, the extent of changes in quantitative information, and the extent of changes in visualization information. Text comprehensibility, which is measured by the proportion of accounting terminology, has a negative regression coefficient, indicating that the more review rounds there are, the lower the proportion of accounting terminology in the revised prospectus, and consequently, the higher the text’s comprehensibility. The expected regression coefficient for text readability was negative; however, the result is positive, suggesting that as the number of inquiry rounds increases, sentence length increases, and the amount of information per sentence increases, but relative readability decreases.

Table 5. Effect of the number of audit enquiry rounds on the textual features of the prospectus

Variable	Dependent Variable: Changes in Prospectus Textual Features
	(1)	(2)	(3)	(4)	(5)	(6)
	Text Length	Text Readability	Text Comprehensibility	Text Quantitative Information	Text Visualization Information	Text Information Ambiguity
Audit Enquiry Rounds	0.059^***(6.668)	0.004^***(2.636)	-0.016^***(-5.080)	0.115^*** (5.852)	0.091^***(5.062)	-0.004 (-1.324)
Intercept	3.922^***(8.116)	0.084^***(4.156)	-0.171^***(-3.944)	0.739^*** (3.441)	0.638^***(3.243)	-0.121^** (-2.287)
Control Variables	Control	Control	Control	Control	Control	Control
Industry Effects	Control	Control	Control	Control	Control	Control
Time Effects	Control	Control	Control	Control	Control	Control
Board Effects	Control	Control	Control	Control	Control	Control
Sample Size	964	964	964	964	964	964
Adjusted R²	0.507	0.135	0.097	0.391	0.333	0.121