What Do Investors Learn in Private Meetings? Evidence from 4,700 Encounters with Portfolio Firms

Private meetings between institutional investors and the boards and management of their portfolio firms have grown substantially worldwide as investors comply with stewardship codes and engage in active ownership. At the same time, these meetings are frequently undisclosed, raising concerns about fair disclosure and insider trading. While prior research has studied investor engagement through the lens of activist campaigns or global activism patterns, little is known about the day-to-day private interactions between mainstream asset managers and their portfolio companies. In our paper, The Benefits of Access: Evidence from Private Meetings with Portfolio Firms, forthcoming in the Journal of Finance, we use large language models (LLMs) to analyze the content of 4,700 private meetings between a large active asset manager and its portfolio firms and examine how the information obtained in these meetings affects trading decisions and performance.

We find that private meetings convey predominantly soft information — qualitative, judgment-laden assessments that require interpretation — rather than hard, quantitative facts. This soft information significantly influences fund managers’ trading decisions, with meeting-informed portfolios generating substantial outperformance. Even after controlling for the presence of hard information, soft information remains the primary driver of trading around meetings. Our results show that the value of private meetings lies in the accumulation of soft, contextual insight — the kind of information that is difficult to transmit and hard to interpret. In a regulatory environment with strict enforcement against trading on material nonpublic information, stewardship engagement generates value through the gathering and interpretation of soft information. Our evidence informs the debate as to the balance between transparency and the need for private, candid interactions between asset managers and portfolio firms.

The Data

Our data come from the equity investments of the world’s 30th largest active asset manager in 2017, Standard Life Investments (now Aberdeen Investments). The data cover the period 2007 to 2015 and include detailed meeting notes, daily fund-level holdings totaling 10.45 million positions, internal analyst recommendations, and shareholder votes. We observe two types of private meetings. Fund manager (FM) meetings are convened by the analyst responsible for a given sector and typically involve senior management of the portfolio firm, including the CEO and CFO. There are 3,410 such meetings, with an average portfolio firm market capitalization of US$7.7 billion. Governance specialist (GS) meetings focus on environmental, social, and governance (ESG) risks and are typically held with the Chairperson, nonexecutive directors, and chairs of key board committees. There are 1,285 GS meetings, allocated to more important stock positions and often triggered by an active governance health warning.

Analyzing Meeting Content with Large Language Models

The meeting notes in our sample are long, averaging up to 2,200 words, with hundreds of meetings taking place each year. Extracting complex text-based information at this scale has been beyond the reach of researchers until recently, with the advent of large language models. We combine LLM processing with traditional natural language processing and human labeling to analyze the content of these meetings. We assign the LLM the role of a finance expert and parse all meeting notes through its API.

The LLM classifies information as hard (quantitative, easily transmissible) or soft (qualitative, requiring interpretation). For the full sample, 67% of meeting information is soft and only 33% is hard. The dominance of soft information is even more pronounced in the conclusions that fund managers draw from meetings, where soft information accounts for 79% of the content. We further decompose soft information by subject and find that 72% relates to the firm, 19% to the industry, and 9% to the market. Meeting content differs markedly between the two meeting types: FM meetings focus on business models, cost and efficiency, and growth, while GS meetings are dominated by audit and compliance, sustainability and corporate responsibility, and remuneration.

To illustrate the LLM’s capabilities, consider a December 2015 meeting with the Chairman of Carillion, a multinational construction firm. The meeting note describes the Chairman as appearing to be a “light touch” who “had just returned from Lesotho by way of a spa in Thailand” while admitting that the company “hadn’t really made any progress” on strategy. The LLM correctly interprets the Chairman’s being “busy” as a negative signal about monitoring, classifying the topics as “leadership dynamics,” “CEO mentoring,” and “governance reporting.” Two weeks later, the internal analyst downgraded the firm from “Hold” to “Sell.” Carillion subsequently collapsed in January 2018.

We also use the LLM to perform cluster analysis, identifying 17,452 unique meeting topics in a first pass, which are then iteratively aggregated into 53 subject categories. Using supervised machine learning, we confirm that the content of meetings differs systematically depending on who convenes the meeting. GS meetings are best identified by words like “governance,” “board,” “chairman,” and “remuneration,” while FM meetings are best discriminated by “broker,” “growth,” “margin,” “cost,” and “buy.”

Private Meetings and Trading

We examine the relation between private meetings and daily trading activity. During the [0, 5]-day window around FM meetings, the average fund that trades increases its position by 3.2% per day, while around GS meetings it decreases its position by 2.3% per day. The opposite trading directions reflect the different information content: FM meetings tend to be interpreted positively, while GS meetings, which often take place against a background of concern, tend to be interpreted negatively. Both the size and the sign of the trading response depend on the information content of meetings, particularly soft information. Our estimates, based on daily data, are orders of magnitude larger than those in prior studies that infer trading from institutional quarterly holdings in 13-F filings. Compared with Bradley, Jame, and Williams (2022), who find that nondeal roadshow meetings increase implied daily trading by about 1.4%, our FM meeting daily trading estimates of 3.2% are more than twice as large.

We show that trading on meeting days is significantly more pronounced for meetings assessed as unusually high or low quality by fund managers, that are high profile with respect to portfolio firm attendees, and that are unusually positive or negative based on meeting sentiment. An interesting feature of our results is that only a minority of positions trade in response to the information from private meetings. Using the LLM, we show that one reason for reduced trading is a lack of consensus: when the information in a meeting is assessed as subject to interpretation, there is significantly less trading.

Soft Information in a Strict Regulatory Environment

A natural question is whether the trading we observe could be driven by material nonpublic information (MNPI) rather than by the soft information content of meetings. To address this, we use the LLM to systematically screen all meeting notes for nonpublic information. We follow the Securities and Exchange Commission and consider as MNPI any information which, if known, could reasonably be expected to move a firm’s stock price.

The LLM assesses 98% of meetings as “not at all likely” to discuss nonpublic information and 1% as “slightly likely.” We read the remaining roughly 1% of meeting notes, since they are LLM-assessed as “moderately likely” or higher, and after adjusting for false positives, we confirm that only 17 meetings, or 0.4% of the total, discuss nonpublic information. These meetings cover topics such as next-day quarterly earnings, CEO departures, M&A deals with competitors, and rights issues. To our knowledge, this is the first sample of MNPI-related private meetings between an institutional investor and its portfolio firms.

We find no evidence of trading around meetings where MNPI is discussed. Event-time estimates show that all trading coefficients around these meetings are statistically indistinguishable from zero. This confirms that the trading patterns we document are driven by the soft information gathered in meetings, not by access to hard, price-sensitive facts. The United Kingdom’s strict “parity of information” regime, which requires listed companies to disclose all MNPI as soon as possible, appears to function as intended: private meetings operate within the boundaries of fair disclosure, and the information advantage they confer stems from the interpretation of soft, contextual insight.

Performance

The large volume of trading around meetings is economically important. We calculate the actual money made by the asset manager through active trading around private meetings. For FM meetings, fund managers earn 15 to 19 basis points per position during a [0, 5]-day window around the meeting, depending on assumptions about intraday trade execution. For GS meetings, the corresponding figure is 7 basis points per position. Interestingly, not all funds trade in response to meetings, which limits the aggregate profits from this channel. To benchmark the potential value of meeting information, we also construct hypothetical long-short portfolios. A meeting-informed portfolio combining both FM and GS meetings generates a significant alpha of 180 basis points per month, compared with 49 basis points for a portfolio without any meetings. The difference suggests that the information conveyed in meetings creates profitable trading opportunities beyond the skill already reflected in non-meeting trading.

Regulatory Implications

Our evidence speaks directly to the regulatory debate around private meetings. We show that private meetings convey mostly soft information, which is used for profitable trading. Where we identify meeting notes that contain MNPI, we find no evidence of trading. Since our analysis relies on the meeting notes, we cannot exclude the possibility that they do not reflect the full conversation. If the regulatory goal is a level playing field, then allowing private meetings seems inconsistent with that goal, since even soft information can provide active asset managers with a competitive edge. At the same time, private meetings satisfy commitments to active ownership and investor stewardship. Our evidence informs the debate as to the balance between transparency and the need for private, candid interactions between asset managers and portfolio firms. Active management, through direct engagement and the gathering of soft information, can create opportunities for profitable trading that passive strategies cannot benefit from, a finding relevant to the ongoing debate about the efficacy and value of active management.

The full paper is available here (Journal of Finance) and here (SSRN).