High-Frequency Trading Market Impact: 2023-2024 Findings

# High-Frequency Trading Market Impact: 2023-2024 Findings

Abstract

This study examines the microstructural effects of high-frequency trading (HFT) activity across major equity markets during 2023-2024, with particular emphasis on liquidity provision, price discovery efficiency, and adverse selection costs. Through analysis of order book data from NYSE, NASDAQ, and CME futures markets, we document significant changes in market quality metrics coinciding with increased algorithmic participation rates exceeding 65% of total volume. Our findings reveal that HFT market makers reduced bid-ask spreads by 23% while simultaneously increasing quote instability during volatility regimes.

Introduction

High-frequency trading has fundamentally altered equity market microstructure since the mid-2000s. While regulatory scrutiny peaked following the 2010 Flash Crash, the subsequent decade witnessed continued evolution in HFT strategies and market architecture. This research provides updated empirical evidence on HFT's role in contemporary markets, addressing three primary questions:

1. How has HFT activity influenced bid-ask spreads and effective spread costs in post-pandemic markets?

2. What is the relationship between algorithmic trading intensity and price discovery efficiency?

3. Do high-frequency market makers contribute to or mitigate intraday volatility spikes?

The period 2023-2024 represents a critical juncture for algorithmic trading analysis. Macroeconomic uncertainty, elevated interest rates, and persistent volatility created an environment where HFT strategies faced significant stress testing. Unlike the low-volatility regime of 2017-2019, recent markets exhibited higher volatility clustering and regime changes that challenged traditional market making algorithms.

Literature Review

Academic research on high-frequency trading has evolved considerably since early studies by Hendershott, Jones, and Menkveld (2011) documented the positive relationship between algorithmic trading and liquidity provision. Subsequent work by Brogaard, Hendershott, and Riordan (2014) found that HFT firms contributed to price discovery while potentially exacerbating short-term volatility.

More recent investigations have focused on the strategic behavior of HFT participants during stress periods. Kirilenko et al. (2017) analyzed HFT activity during the Flash Crash, revealing how liquidity provision evaporated precisely when markets needed it most. This research extends that framework to examine whether structural changes in market architecture and regulation have altered HFT behavior during volatility events.

The debate over HFT's net impact remains contentious. Proponents emphasize reduced transaction costs and enhanced liquidity depth. Critics highlight concerns about market fragility, fleeting liquidity, and the technological arms race that disadvantages slower market participants. Our analysis contributes empirical evidence to this ongoing discussion using granular order book data unavailable in earlier studies.

Data and Methodology

Our dataset comprises tick-by-tick order book snapshots from three primary venues: NYSE (equities), NASDAQ (equities), and CME (E-mini S&P 500 futures). The sample period spans January 2023 through October 2024, encompassing approximately 18 billion individual order book events across 500 frequently traded securities.

HFT activity identification follows the methodology established by SEC Rule 15c3-5 disclosures and exchange-provided flags for algorithmic order flow. We classify orders as HFT-originated based on:

- Order cancellation rates exceeding 90%

- Order-to-trade ratios above 20:1

- Message traffic concentrated in sub-millisecond intervals

- Co-location facility participation flags

Market quality metrics include:

**Effective Spread:** Measured as twice the absolute difference between transaction price and mid-quote at execution time, expressed in basis points.

**Quoted Spread:** The difference between best bid and best offer, normalized by mid-quote.

**Depth:** Total quantity available within 10 basis points of the mid-quote on each side of the order book.

**Price Impact:** The permanent price movement following large trades, measured at 1-minute, 5-minute, and 15-minute horizons.

**Volatility Measures:** Realized volatility computed using 5-minute returns, along with intraday range-based volatility estimators.

Results: Spread Dynamics and Transaction Costs

Analysis of quoted spreads reveals substantial heterogeneity across market capitalization segments. Large-cap securities (market cap >$10B) experienced average quoted spreads of 1.2 basis points in 2023-2024, representing a 23% decline from 2021-2022 levels. This compression occurred despite higher absolute volatility, suggesting increased competition among HFT market makers.

Mid-cap securities ($2B-$10B market cap) showed more modest spread improvements, averaging 3.8 basis points compared to 4.3 basis points in the prior period. Small-cap securities (<$2B) exhibited minimal change, with average spreads remaining near 12 basis points.

The relationship between HFT participation and spread levels demonstrates non-linearity. Securities with HFT market share between 40-60% exhibited the tightest spreads, while those with either very low (<20%) or very high (>80%) HFT participation showed wider spreads. This suggests an optimal range of algorithmic trading intensity for market quality.

Effective spreads, which capture price impact and execution quality, painted a more nuanced picture. While quoted spreads compressed, the ratio of effective spread to quoted spread increased from 1.15 to 1.28 during high-volatility periods. This divergence indicates that rapid quote updates by HFT firms sometimes created execution difficulty for slower market participants attempting to access displayed liquidity.

Decomposition of effective spreads into adverse selection and order processing components revealed that adverse selection costs rose during 2023-2024. The adverse selection component increased from 32% to 41% of total effective spread, suggesting that informed trading became more concentrated or that HFT firms improved their ability to detect and avoid toxic order flow.

Results: Liquidity Provision and Order Book Depth

Order book depth analysis produced counterintuitive findings. Total displayed liquidity within 10 basis points of mid-quote increased by 18% on average, but depth volatility also rose substantially. The coefficient of variation for quoted depth increased from 0.42 to 0.67, indicating that while average depth improved, liquidity became less stable.

This instability manifested most clearly during volatility events. Using a threshold of realized volatility exceeding the 90th percentile, we identified 147 distinct high-volatility episodes during the sample period. During these episodes, quoted depth declined by an average of 54% within the first two minutes of volatility onset, compared to 38% in comparable 2021-2022 events.

The recovery pattern of liquidity following volatility spikes also changed. In 2023-2024, depth recovery took an average of 12.3 minutes to return to pre-spike levels, versus 8.7 minutes in earlier periods. This slower recovery suggests that HFT algorithms have become more conservative in their liquidity provision following volatility events, possibly due to enhanced risk management protocols implemented after recent market disruptions.

Depth resilience varied significantly by venue. NYSE displayed more stable depth patterns than NASDAQ during stress periods, with depth declining by 47% versus 61% respectively during high-volatility events. This venue difference persisted even after controlling for differences in listed security characteristics, suggesting that market structure and maker-taker fee schedules influence HFT behavior during stress.

Results: Price Discovery and Information Efficiency

Price discovery efficiency was assessed using vector autoregression models examining the lead-lag relationships between venues and the contribution of each market to price innovation. The information share metric, which quantifies each venue's contribution to permanent price movements, revealed that HFT-dominated venues contributed disproportionately to price discovery.

CME E-mini S&P 500 futures, where HFT participation exceeds 75%, accounted for 68% of total price discovery in the equity index complex, up from 62% in 2021-2022. This increased dominance suggests that HFT activity has become more informative or that these algorithms have improved at detecting and trading on information signals.

However, the speed of price discovery also increased volatility transmission across assets. Cross-asset correlation in 5-minute returns rose from 0.34 to 0.48 during our sample period, with the increase concentrated in periods of high HFT activity. This finding indicates that while HFT improves information aggregation, it may also amplify correlation during stress periods, potentially reducing diversification benefits.

Variance ratio tests examining return predictability at high frequencies showed mixed results. At 1-second intervals, returns exhibited significant negative autocorrelation (-0.08) during high HFT activity periods, consistent with temporary price pressure from large HFT trades. However, at 1-minute intervals, autocorrelation approached zero, suggesting rapid price correction and efficient incorporation of information.

The permanence of HFT-initiated price movements was examined by tracking the price impact of identified HFT trades. Large HFT buy orders (>$500k notional) showed 73% permanent price impact at the 15-minute horizon, comparable to non-HFT institutional orders at 76%. This similarity suggests that HFT trades contain genuine information rather than representing pure noise trading.

Results: Volatility Dynamics and Market Stability

Intraday volatility patterns showed distinct signatures associated with HFT activity. High-frequency volatility (5-minute realized volatility) increased by 31% during 2023-2024 compared to 2021-2022, while lower-frequency volatility (daily) increased by only 17%. This divergence indicates that HFT may contribute to short-term volatility amplification even as it provides liquidity on average.

Volatility clustering at high frequencies intensified during the sample period. The autocorrelation of 5-minute absolute returns at lag 1 increased from 0.23 to 0.31, suggesting that short-term volatility shocks persist longer than previously observed. This persistence pattern aligns with HFT risk management systems that reduce liquidity provision following volatility spikes, creating a feedback loop where initial volatility leads to reduced liquidity, which amplifies subsequent volatility.

Analysis of volatility around macroeconomic announcements revealed changes in HFT behavior. Prior to major announcements (FOMC decisions, employment reports), HFT participation declined by 15% in the five minutes before release, indicating strategic withdrawal by algorithmic traders facing elevated uncertainty. This withdrawal coincided with spread widening of 28% pre-announcement, suggesting that HFT liquidity provision is state-dependent and may not be available precisely when market stress is anticipated.

The relationship between HFT activity and tail risk events was examined using extreme value theory. Days with HFT participation exceeding the 90th percentile showed a 34% higher frequency of 5-minute returns exceeding 3 standard deviations. However, the magnitude of these extreme moves did not differ significantly, suggesting HFT increases the frequency but not severity of tail events.

Discussion: Market Quality Trade-offs

The empirical evidence presents a nuanced view of HFT's market impact. On one hand, algorithmic trading clearly compresses bid-ask spreads and contributes to price discovery, reducing transaction costs for most market participants most of the time. The 23% reduction in large-cap spreads represents substantial savings for institutional investors executing large portfolios.

On the other hand, the increased fragility of liquidity during stress periods raises concerns about market resilience. The 54% depth decline during volatility events, combined with slower recovery times, suggests that HFT liquidity provision may be less reliable than traditional market making during periods when liquidity is most valuable.

This conditional liquidity provision creates a form of liquidity illusion. Displayed depth appears ample during normal market conditions, potentially encouraging larger position sizes and more aggressive trading strategies. However, when volatility materializes, much of this displayed liquidity evaporates, potentially exacerbating price movements and creating conditions for cascading volatility.

The increased adverse selection component of spreads indicates that HFT firms have become more sophisticated at avoiding toxic flow. While this skill benefits HFT profitability, it may disadvantage other market participants who face wider effective spreads when their orders are more likely to contain information. This dynamic could reduce market participation by institutional investors who find execution quality deteriorating despite tighter quoted spreads.

The concentration of price discovery in HFT-dominated venues has implications for market structure. As futures markets increasingly lead cash equity markets, and as HFT dominates futures trading, the effective price formation process becomes concentrated in a narrow segment of market participants. This concentration may be efficient but also creates potential fragility if HFT participation were to decline suddenly.

Policy Implications

These findings inform several ongoing policy debates regarding market structure and HFT regulation:

**Market Access Fees:** The variation in HFT behavior across venues suggests that fee structures influence algorithmic trading strategies. NYSE's more stable depth during stress periods may relate to its different maker-taker fee schedule. Regulators should consider whether current fee structures optimize market quality or primarily benefit specific market participants.

**Minimum Resting Times:** Some jurisdictions have proposed minimum resting times for orders to reduce quote flickering. Our evidence on increased depth volatility supports the argument for such requirements, though implementation must balance reduced quote instability against potential liquidity costs.

**Circuit Breakers:** The faster volatility transmission in 2023-2024 suggests that existing circuit breaker thresholds may need updating. As HFT accelerates both price discovery and volatility propagation, circuit breakers calibrated for slower markets may trigger too late to prevent cascading volatility.

**Market Maker Obligations:** The conditional nature of HFT liquidity provision raises questions about whether algorithmic market makers should face formal obligations similar to traditional designated market makers. Such obligations could ensure more consistent liquidity provision during stress periods.

**Consolidated Audit Trail:** The complexity of HFT strategies and their market impact underscores the importance of comprehensive surveillance capabilities. Full implementation of the Consolidated Audit Trail would enable more detailed analysis of HFT behavior and its market effects.

Limitations and Future Research

This study faces several limitations that future research should address. First, HFT identification based on order characteristics provides reasonable proxies but cannot definitively classify all algorithmic trading. Some institutional algorithms may exhibit HFT-like characteristics, while some HFT strategies may not fit standard classification criteria.

Second, our analysis focuses on market quality metrics but does not directly observe HFT profitability or risk management protocols. Understanding the economic incentives driving HFT behavior requires proprietary data unavailable to academic researchers. Regulatory access to such data would enhance understanding of algorithmic trading dynamics.

Third, the sample period coincides with specific macroeconomic conditions (high inflation, rising interest rates) that may limit generalizability. Replication during different market regimes would strengthen confidence in findings.

Future research directions include:

- Analysis of HFT behavior in less liquid markets and emerging market venues

- Investigation of machine learning techniques used in contemporary HFT strategies

- Examination of HFT impact on options markets and derivatives pricing

- Study of regulatory arbitrage across jurisdictions with different HFT rules

- Assessment of retail investor execution quality in HFT-dominated markets

Conclusion

High-frequency trading continues to reshape equity market microstructure in profound ways. Our analysis of 2023-2024 market data reveals that HFT provides substantial benefits through reduced spreads and enhanced price discovery, but these benefits come with costs in terms of liquidity fragility and increased short-term volatility.

The optimal policy approach likely involves accepting HFT as a permanent feature of modern markets while implementing guardrails to mitigate its destabilizing potential during stress periods. Rather than attempting to eliminate or severely restrict algorithmic trading, regulators should focus on ensuring market structure supports stable liquidity provision and efficient price discovery across all market conditions.

Market participants must also adapt to this reality. Institutional investors should recognize that displayed liquidity may not be accessible during volatility events and should adjust risk management accordingly. Technology investments enabling faster execution and better order routing become increasingly necessary to compete in HFT-influenced markets.

The evolution of HFT continues. As artificial intelligence and machine learning techniques become more sophisticated, algorithmic trading strategies will likely become more complex and potentially more difficult to regulate or monitor. Ongoing research and regulatory attention will be essential to ensure that market structure evolves in ways that benefit all participants while maintaining market integrity and stability.