Why Analysts Herd: Incentives That Drive Wall Street Consensus

Why Do Sell-Side Analysts Herd?

Sell-side analysts appear to be incentivized to provide independent, superior forecasts. Their reputation depends on accuracy; their compensation depends on the assets they attract to their bank; their career trajectory depends on outperforming peers. These incentives seem to promote independent analysis and divergent forecasts. Yet in practice, analysts cluster around consensus with surprising consistency. Earnings forecasts show low dispersion, stock ratings concentrate in bullish ranges, and price targets converge within narrow bands. The herding occurs despite surface-level incentives that should promote divergence.

The explanation lies in the deep structure of analyst careers and compensation. Sell-side analysts do not earn direct commissions on their forecast accuracy. They earn compensation based on research allocation—the assets under management that flow to their bank as a result of the research. Research allocation depends on institutional client preferences, and institutional clients prefer consensus estimates to outlier estimates. Institutional portfolio managers face career risk from underperforming their peers, which means they face greater risk from owning positions that diverge from benchmark allocations. A fund manager holding a stock not in the benchmark and underperforming faces questions about their stock-picking ability; a fund manager holding the same benchmark-constituent stocks as competitors and underperforming faces questions about selection skill within a constrained set. The first seems like a worse outcome, so asset managers prefer analysts who provide consensus estimates that align with benchmark positions.

Analyst herding is therefore not irrational self-deception. It is the rational response to institutional incentive structures that reward clustering over differentiation. Understanding these incentives is crucial to understanding why consensus estimates persist despite their vulnerability to disappointment.

Quick definition: Analyst herding occurs because sell-side analysts' career advancement, compensation, and asset attraction depend more on consensus alignment than on forecast accuracy, creating structural incentives to cluster estimates around consensus rather than publish independent forecasts.

Key takeaways

• Career risk from being notably wrong dominates incentives; an analyst who forecasts divergently and is incorrect risks reputation damage that is asymmetrically worse than consensus correct mistakes
• Institutional asset managers prefer consensus estimates to manage relative performance risk, creating demand for analyst consensus that exceeds demand for independent analysis
• Equity research allocation (the flow of trading commissions and capital to the analyst's bank) depends on institutional client satisfaction, not research accuracy, misaligning incentives from useful analysis to client-pleasing analysis
• Bank compensation structures tie analyst pay to research allocation and trading volumes, not forecast accuracy, removing direct incentives for beating estimates or accurately predicting earnings
• Sell-side analyst time and resources are spent building relationships with institutional clients and company management rather than conducting independent fundamental analysis
• Regulatory constraints (Regulation FD, Chinese walls) limit analyst information access while incentivizing reliance on consensus narratives and management guidance that is available to all analysts equally

The Career Risk of Being Notably Wrong

An analyst who publishes a forecast that diverges significantly from consensus and turns out to be wrong suffers reputational damage that is asymmetrically worse than the cost of being wrong together with the consensus. This asymmetry is the fundamental driver of herding.

Consider an example. Two analysts, Anna and Boris, both work at major investment banks. Both are analyzing Apple's next-year earnings potential. Based on their analysis, both believe Apple will earn $6.50 per share. Consensus among 35 analysts is $6.00 per share. Anna publishes a $6.50 estimate. Boris publishes a $6.00 estimate matching consensus. Actual earnings come in at $5.80.

Anna was more accurate—off by $0.70 versus Boris's $0.20. But Anna's reputation suffers more severely. When results are announced, the headline is "Consensus Misses Again" or "Apple Disappoints Expectations." Anna's notably incorrect outlier forecast appears to have been wrong not because of inherent difficulty in forecasting, but because Anna lacked sound judgment. Boris's consensus estimate is lumped into a broader consensus miss; his error is no individual fault. The market implicitly forgives consensus error—"we were all surprised"—while it explicitly condemns outlier error—"how did Anna miss what others got right?"

Conversely, when actual earnings are $6.80 (higher than both forecasts), Anna's $6.50 is closer than Boris's $6.00, and Anna receives credit for superior insight. But the credit is asymmetrically smaller than the blame would have been if Anna had missed. Institutional clients note that Anna was closer to correct, which is good, but they do not systematically reward outlier correctness with increased research allocation. The risk-reward for analysts is asymmetric: large downside from outlier-wrong, modest upside from outlier-right. The rational response is clustering.

Empirical research on analyst careers shows this dynamic clearly. Analysts who are persistently slightly wrong (part of consensus error) retain and advance their careers. Analysts who are occasionally notably wrong (despite sometimes being right) face career damage and reduced research allocation. The market rewards accuracy-relative-to-consensus more than it rewards absolute accuracy, driving clustering.

Institutional Client Preferences for Consensus

Institutional asset managers strongly prefer consensus estimates to help manage their relative performance risk. This preference is not always explicit, but it is consequential. A fund manager who deviates from consensus allocations faces two sources of risk: the risk that their different allocation underperforms (stock-picking risk), and the risk that their different allocation matches different benchmarks than their peers (benchmark-tracking risk).

When a portfolio manager is allocated to beat the S&P 500 index, the benchmark is fixed. But the manager's performance is evaluated relative to peers. If 100 growth fund managers are competing for the same capital, and 90 of them hold similar portfolios with consensus estimates informing allocations, the 10 outlier managers are taking on risk that their divergent analysis provides alpha (true edge). If the outliers are wrong, they underperform dramatically. If they are right, they outperform dramatically. The median fund manager avoids this risk by clustering closer to the 90.

This preference translates directly to sell-side analyst demand. Institutional managers demand research that helps them understand the consensus (to know what they are being evaluated against) and research that provides consensus estimates (to know how their allocation differs from peer allocations). Analysts who provide consensus-challenging estimates are seen as less useful for this purpose. Banks that employ such analysts find that their research is used less frequently by institutional clients, reducing their research allocation and thus reducing analyst compensation.

The preference manifests in analyst rankings. Every year, institutional investors vote on which analysts are most useful. The vote correlates strongly with analyst consensus alignment and popularity, not with forecast accuracy. Analysts recognized as leading analysts in their field are often those who provide consensus estimates aligned with institutional preferences, not those who have made the most accurate independent forecasts. This reinforces analyst herding, as young analysts observe that career success correlates with consensus alignment rather than with beating consensus.

Compensation Structures Misaligned with Accuracy

Sell-side analyst compensation is directly tied to research allocation—the amount of trading volume that institutional clients route to the analyst's bank because they value the research. An analyst who generates $2 million in annual research allocation earns substantially more than one who generates $200,000, regardless of forecast accuracy.

Research allocation depends on institutional client satisfaction, which depends on how useful the analyst's research is for the client's portfolio management. Usefulness includes accurate forecasts, but also includes consensus validation (confirming what clients thought), narrative support (providing intellectual cover for existing positions), and consensus explanation (helping clients understand why consensus has moved).

An analyst can be highly compensated without ever beating consensus on a forecast. In fact, analysts who consistently beat consensus often face pressure to align more closely with consensus, because their divergent forecasts create performance friction with institutional clients who prefer consensus alignment. A bank might tell an analyst: "Your forecast accuracy is strong, but your research allocation is declining. Consider aligning more closely with consensus to improve client satisfaction."

This creates a perverse incentive: analyst compensation decreases forecast accuracy because accuracy that diverges from consensus reduces research allocation and therefore reduces compensation. An analyst optimizing for lifetime earnings should cluster toward consensus, even if they believe their independent analysis is more accurate.

The misalignment between compensation and accuracy is further amplified by the nature of research allocation. Allocation is often tied to trading volumes rather than to institutional asset management directly. A bank earns research allocation when its sales traders execute trades for institutional clients. If an analyst's research drives trading volume, the bank profits and the analyst is compensated. If an analyst's research is accurate but discourages trading because clients are confident in the forecast, the bank earns less allocation. This inverts incentives further: analysts are rewarded for research that generates trading activity, which is often generated by recommendations that diverge from consensus (whether correct or not).

Information Access Asymmetries and Consensus Reliance

Sell-side analysts operate under regulatory constraints (Regulation Fair Disclosure, or Reg FD) that limit their access to non-public information from company management. Before Reg FD was implemented in 2000, analysts could obtain detailed non-public guidance directly from executives, creating information asymmetries where well-connected analysts had superior forecasts. After Reg FD, all analysts receive the same official guidance, and non-public information is limited.

This regulatory change inadvertently increased analyst herding. When all analysts have access to the same public information and company guidance, their independent analyses converge more naturally. An analyst cannot differentiate through superior information access; differentiation requires superior analytical skill. But the market does not systematically reward superior analysis if it diverges from consensus. So analysts cluster around the common information (guidance) and reduce independent analysis.

Additionally, analyst time is finite. With thousands of stocks to cover, analysts face severe time constraints. An analyst covering 30 technology stocks cannot conduct exhaustive independent analysis on each company each quarter. Practical necessity drives reliance on management guidance, sell-side consensus, and published industry data. The analyst produces a consensus estimate in a reasonable timeframe by adjusting the consensus by a small amount based on incremental analysis. This is rational time management but drives herding.

Management Guidance as a Clustering Anchor

Company management provides earnings guidance that serves as a focal point for analyst estimates. Once management has guided for specific earnings, analyst estimates cluster around that guidance. Management guidance is often derived from company expectations for business performance, but it is also often influenced by desires to manage expectations downward (to ensure beat potential) or to support current stock prices (to provide confidence to investors).

Management guidance has become tighter and more conservative in recent decades, reflecting companies' desires to ensure beat potential and manage expectations. An analyst who deviates significantly from management guidance risks being wrong relative to management's own expectations, which is a reputational risk. This anchoring effect drives analyst clustering around guidance, even when analysts might independently forecast different numbers.

The tightening of guidance ranges has paradoxically increased analyst clustering. When management guidance was broad (often 20-30% ranges), analysts had room to diverge while still staying within guidance. As guidance has tightened (often 5-10% ranges), analyst estimates naturally cluster within those tight ranges. This makes analyst estimate dispersion decline structurally, even if analyst skill or analytical independence remains constant.

The Influence of Relative Performance Evaluation

Equity research departments at investment banks are often evaluated on relative performance—how well their research allocation grows compared to competitors. This creates tournament dynamics where analysts compete to grow allocation, not to improve accuracy. Tournament dynamics favor clustering toward the median, because taking risk (diverging from consensus) can reduce expected allocation even if it increases expected accuracy.

A research director at a bank observes that her technology analysts have underperformed in research allocation relative to competitors. She might tell them: "Align more closely with consensus. We are losing business because clients find our divergent forecasts risky." The implicit message is clear: accuracy without consensus alignment is less valuable than consensus accuracy.

This creates herding at the institutional level, beyond individual analyst incentives. Banks as organizations herd toward consensus because research allocation depends on client satisfaction, and clients prefer consensus. Individual analysts then inherit these institutional herding pressures as baseline expectations for their performance.

The Role of Sell-Side Competition

Competition among sell-side research providers would logically drive differentiation and reduce herding. In practice, competition reinforces herding. Banks compete for research allocation by hiring the analysts ranked highest by clients. Ranked analysts are those aligned with consensus. Banks then employ these consensus-oriented analysts, reducing the population of independent analysts competing in the marketplace. The pool of available top talent is filtered toward consensus-alignment, creating structural herding across the industry.

Additionally, competition creates pressure for rapid consensus formation. If an analyst takes a contrarian view and is proven correct, the analyst can claim superior insight and differentiate. But the outcome is uncertain for months or years. During that period, the analyst's research allocation may decline if clients perceive the view as contrarian risk rather than contrarian insight. The competitive incentive is to align quickly with emerging consensus, not to maintain contrarian views long enough to validate them.

Real-world examples

2008 Financial Crisis Analyst Clustering: In 2006-2007, sell-side analysts covering mortgage-backed securities and housing stocks overwhelmingly maintained positive estimates and ratings. Analyst earnings estimates for Countrywide and other mortgage lenders clustered around bullish assumptions about housing price stability. Despite growing evidence of mortgage delinquencies and structural problems, analyst consensus remained positive through 2007. The herding was not due to analyst incompetence; it was due to institutional incentives. Banks profited from mortgage-backed securities and mortgage lender stock underwriting. Research allocation rewarded analysts who maintained bullish narratives aligned with the market consensus that housing was a stable, low-risk asset class. Analysts diverging from this consensus faced institutional pressure. The result was clustering around a fundamentally flawed consensus that was invalidated catastrophically in 2008.

Technology Stock Collapse of 2000: In 1999-2000, technology stock analyst estimates clustered around extremely bullish growth assumptions. Internet companies expected to achieve extraordinary revenue growth and profitability despite lacking profitable business models. Analyst estimates incorporated these assumptions, and earnings forecasts for technology stocks were dramatically optimistic. As the bubble burst and companies failed, analyst estimates were revised downward repeatedly. Analysts had clustered around a consensus that technology would transform the world (true) and that current stock prices were justified by this transformation (false). The herding was driven by client preferences (technology fund managers wanted bullish analyst coverage) and bank incentives (banks profited from technology underwriting).

Energy Stocks Post-2010: After the 2008 financial crisis, energy stocks underperformed due to oil price volatility and negative sentiment about fossil fuels. Analyst estimates for energy companies clustered around conservative growth assumptions. For years, analyst estimates missed on the upside as energy companies demonstrated resilience and dividend sustainability. The herding toward conservative consensus cost institutional clients who preferred consensus alignment upside potential. Only when sentiment shifted in 2021 did analyst estimates begin moving upward. The clustering had persisted for a decade despite evidence that conservative estimates were systematically missing.

FAANG Clustering of 2012-2018: The "FAANG" technology stocks (Facebook, Apple, Amazon, Netflix, Google) experienced analyst consensus clustering throughout 2012-2018. Analyst estimates and ratings converged on bullish assumptions about these companies' market dominance and growth. Institutional managers, heavily weighted to FAANG due to their large index weights, preferred analyst research validating their positions. Analysts who expressed concern about FAANG valuations or market saturation found their research allocation declining. The clustering persisted despite concerns about valuation extremity, which turned out to be justified when growth decelerated and valuations compressed in 2022.

Common mistakes

Assuming Analyst Estimates Reflect Best Collective Wisdom: Investors often treat analyst consensus as the market's best estimate of future earnings, assuming that the aggregation of many analysts produces a good forecast. In reality, consensus reflects institutional preferences and incentive alignment as much as analytical skill. Consensus can be systematically biased (too bullish during bull markets, too bearish during bear markets) for years.

Trusting Analyst Recommendations Without Checking Dispersion: When 90% of analysts recommend a stock, the high recommendation concentration should trigger caution about herding, not confidence in the thesis. A stock with 70% buy ratings and 20% hold ratings and 10% sell ratings shows healthy analytical diversity; a stock with 90% buy and 10% hold shows consensus clustering.

Following Analyst Revisions as Short-Term Trading Signals: Analyst estimate revisions appear to provide short-term trading signals, but many revision patterns are herding-driven rather than information-driven. An analyst revising earnings higher following an initial analyst revision is often herding rather than conducting independent analysis. Sophisticated traders recognize that early revisions are genuine information but later revisions are often herding, reducing their alpha potential.

Underweighting Analyst Divergence as a Positive Signal: When analyst estimates show high dispersion, it is tempting to interpret this as analytical weakness (lack of clarity). In reality, high dispersion indicates that analysts disagree about future outcomes, which often means the stock has real genuine upside and downside potential. High dispersion is sometimes a sign of genuine opportunities where outliers have superior insight.

Assuming Consensus Accuracy Improves Over Time: In many fields, aggregate forecasting improves over time as information accumulates. In equity markets, consensus does not systematically improve because the incentives driving herding remain constant. Consensus four weeks before earnings is often as herded as consensus eight weeks before, despite the additional information that has accumulated. Time does not reduce herding pressure; it usually reinforces it.

FAQ

How much of analyst herding is driven by incentives versus behavioral bias?

This is difficult to measure precisely, but research suggests both factors contribute. Behavioral bias (anchoring to consensus, overconfidence in group decisions, conformity pressure) plays a role, but the largest component is incentive alignment. Analysts who are given explicit incentives to beat consensus (hypothetically) would likely reduce herding substantially. The bias would remain, but the institutional incentives driving herding would be removed.

Can independent research firms (not affiliated with investment banks) reduce herding?

Independent research firms do show lower analyst clustering and higher estimate dispersion than sell-side banks. The reason is that independent analysts' compensation depends more directly on forecast accuracy and less on client satisfaction and trading volumes. This suggests that changing incentives would reduce herding. However, independent research is a small portion of overall research, and institutional clients still prefer consensus alignment. Even independent analysts face pressure toward consensus if clients demand it.

Why don't institutional clients demand more independent analyst research?

Institutional clients claim to want independent research, but their revealed preferences (measured by research allocation) show preference for consensus alignment. The reason is that consensus-aligned research helps portfolio managers manage relative performance risk. A portfolio manager who deviates from consensus based on analyst recommendations assumes the risk of underperformance; it is easier to match consensus and underperform than to diverge and overperform. The institutional system creates demand for consensus validation rather than independent insight.

How has increased transparency in analyst recommendations changed herding?

Ironically, increased transparency has likely increased herding. When analyst recommendations and estimates were less widely available, analysts had more freedom to diverge because divergence was less visible. Modern analyst databases (Bloomberg, FactSet) make all recommendations and estimates immediately visible to all participants, increasing conformity pressure. An analyst publishing an outlier estimate sees the divergence immediately reflected in consensus summaries, creating real-time feedback that drives clustering.

Can fundamental analysts avoid herding if they conduct independent research?

Individual analysts can conduct independent research and maintain independent views, but they still operate within institutional systems that reward clustering. A portfolio manager who conducts excellent independent analysis and identifies a contrarian opportunity still faces relative performance pressure that discourages acting on the independent insight unless confidence is very high. The herding operates at the system level, not just the individual analyst level.

Why don't analyst rankings reward accuracy more explicitly?

Analyst rankings do attempt to reward accuracy, but accuracy is measured relative to consensus rather than in absolute terms. An analyst who is notably wrong in a divergent direction is ranked poorly; an analyst who is notably right in a consensus direction is ranked well. This implicit relative performance evaluation drives clustering even though rankings ostensibly measure accuracy. Changing this would require measuring and rewarding absolute accuracy, which institutional clients currently do not demand.

Related concepts

• Herding Behavior Defined
• Social Proof in Investing
• Herding in Earnings Consensus
• Contrarian Signals From Herding

Summary

Analysts herd not primarily because of incompetence or behavioral bias, but because institutional incentive structures reward clustering toward consensus. Sell-side analyst compensation depends on research allocation, which is determined by institutional client satisfaction. Institutional clients prefer consensus estimates to manage their relative performance risk. This creates a cascading system where analyst compensation rewards consensus alignment, and analyst careers advance through consensus clustering rather than through independent accurate analysis.

Career risk amplifies herding: an analyst who diverges from consensus and is wrong faces asymmetrically worse reputational damage than consensus participants face for being wrong together. The risk-reward for analysts is asymmetric, favoring clustering. Additionally, institutional preferences for consensus are explicit: banks that employ consensus-aligned analysts attract more research allocation; banks that employ independent analysts find their research allocation declining. This creates competitive pressure for consensus across the industry.

Regulatory constraints (Reg FD) and practical time constraints also drive reliance on management guidance and consensus narratives, further reducing independent analysis. Management guidance serves as an anchor that clusters analyst estimates mechanistically. The combination of incentive alignment, relative performance evaluation, and information access constraints creates a system where analyst herding is rational and rational actors should expect it to persist.

Portfolio managers who recognize this structure can adjust their approach by (1) using analyst consensus as one data point rather than truth, (2) paying attention to analyst estimate dispersion as a herding signal, and (3) maintaining healthy skepticism toward consensus narratives that have clustered too tightly.

Next

→ Herding in Mutual Fund Flows