Funding carry factor construction

Abstract#

Traditional portfolio construction relies on equity-centric risk factors — market beta, size, value, momentum — calibrated over decades of equity returns. Crypto markets present fundamentally different risk dynamics: perpetual funding rates act as a continuous cost-of-carry signal, cross-exchange spreads reveal latent liquidity fragmentation, and extreme funding rate dispersion signals regime-dependent tail risk. This paper proposes a three-factor crypto-native framework — funding carry, spread compression, and dispersion momentum — and evaluates its explanatory power using 831,910 funding rate observations and cross-exchange pricing data from Binance, Hyperliquid, and AsterDEX. We find that funding carry factors exhibit strong persistence for a subset of assets (XMR, NEAR, ZRO at 2–5% annualized), but that extreme negative funding on long-tail assets (PROVE, SUPER, ALT at −25% to −44% annualized) is dominated by survivorship and listing effects rather than systematic risk premia. This is a pilot study — sample coverage spans a single market cycle and lacks out-of-sample validation. The framework is proposed as a starting point, not a production-ready allocation model.

Hypotheses#

• H1 (Moderate confidence): Persistent positive funding rates on select mid-cap perpetuals (XMR, NEAR, ZRO) reflect a structural short-demand premium that can be harvested as a carry factor. Confidence: moderate — persistence is observed but the underlying driver (hedging demand vs. speculative sentiment) is unidentified.
• H2 (Low confidence): Cross-exchange spreads between DEX venues (Hyperliquid, AsterDEX) and Binance predict short-term price convergence and can serve as a liquidity-risk factor. Confidence: low — spread data is point-in-time and lacks time-series depth for statistical validation.
• H3 (Exploratory): The dispersion of funding rates across assets (measured as cross-sectional standard deviation) functions as a systemic stress indicator that predicts portfolio drawdown risk. Confidence: exploratory — related work (May 2024 dispersion ratio analysis) showed promise, but this formulation lacks out-of-sample testing.

Data Provenance#

All data is sourced from real-time production systems operated by Vex Capital:

Important caveats:

• Funding rate data covers a single market cycle (not decade-long equity-style data)
• Cross-exchange spreads are point-in-time snapshots, not full time-series
• The sample is biased toward assets with active perpetual listings on Hyperliquid and Binance
• No execution cost, slippage, or fill-rate data is included in spread estimates

Analysis#

Factor 1: Funding Carry#

The funding carry factor captures the expected return from holding positions that earn persistent funding payments. Unlike equity dividends, crypto funding rates are paid continuously (typically every 8 hours) and reflect real-time supply-demand for leverage.

# Funding carry factor construction
import numpy as np

def funding_carry_score(rates, window=3617):
    """
    Score assets by persistent funding carry.
    Positive = shorts pay longs (carry harvest).
    Negative = longs pay shorts (cost of holding).
    """
    avg_ann = np.mean(rates)
    std_ann = np.std(rates)
    # Signal-to-noise ratio for carry quality
    carry_quality = avg_ann / std_ann if std_ann > 0 else 0
    return {
        'avg_annualized': avg_ann,
        'volatility': std_ann,
        'carry_quality': carry_quality,
        'sample_size': len(rates),
        'reliability': 'RELIABLE' if len(rates) >= 500 else 'UNRELIABLE'
    }

Top persistent positive funding (shorts pay longs):

The carry quality metric (mean return / volatility) is highest for XMR and NEAR — these assets exhibit positive funding with relatively contained downside volatility. PURR, despite higher average carry, has extreme range (−606% to +172%) making it unsuitable for risk-managed carry strategies.

Extreme negative funding (longs pay shorts — "crowded long" signal):

These extreme negative rates are dominated by early-listing dynamics — newly listed perpetuals attract one-directional speculative flow, creating funding rates that are structurally non-stationary. Using them as risk factors without controlling for listing age introduces severe survivorship bias.

Factor 2: Cross-Exchange Spread Compression#

Cross-exchange price differentials reveal liquidity fragmentation and execution risk. When spreads widen, portfolio rebalancing costs increase and arbitrage opportunities emerge — but only for capital with cross-venue access.

Top cross-exchange spreads (DEX vs. Binance):

Critical caveat: These annualized yields are calculated by multiplying instantaneous spreads by 365 — they represent potential convergence profit assuming the spread closes in exactly one day. This is a massive assumption. Real execution requires:

• Capital deployment on both venues
• Gas/transaction costs
• Slippage on close
• Counterparty/settlement risk

The annualized figures are not achievable returns. They are signals of mispricing magnitude.

Factor 3: Funding Rate Dispersion#

The cross-sectional standard deviation of funding rates captures systemic stress. When dispersion spikes, it indicates disagreement across the market about directionality — historically correlated with increased drawdown risk.

def dispersion_factor(funding_rates_by_asset):
    """
    Compute cross-sectional funding rate dispersion.
    High dispersion = systemic stress signal.
    """
    avg_rates = [np.mean(rates) for rates in funding_rates_by_asset.values()]
    return {
        'cross_sectional_std': np.std(avg_rates),
        'mean_rate': np.mean(avg_rates),
        'skew': float(np.mean(avg_rates) - np.median(avg_rates)),
        'n_assets': len(avg_rates)
    }

The current snapshot shows:

• Overall mean funding: −0.169% annualized (slight net-negative, implying longs paying shorts on average)
• Cross-sectional range: −1,096% (SUPER) to +171% (PURR)
• Median asset funding: approximately −2% annualized

The extreme negative skew (long-tail assets with −25% to −44% averages) pulls the mean far below the median, indicating that the average is driven by a small number of distressed/speculative listings rather than systematic market conditions.

Discussion#

Robust Findings#

1. Selective carry persistence is real. XMR, NEAR, and ZRO show consistently positive funding rates across 3,617 observations with manageable volatility. This is the strongest finding in the dataset.

2. Funding rate distributions are not normal. Extreme skew and kurtosis across most assets means mean-variance optimization will systematically misestimate risk. Factor models must account for fat tails.

3. Cross-exchange spreads are material. Spreads of 5–9% between DEX and CEX venues represent genuine pricing inefficiency, though their tradability is unproven without execution data.

Preliminary Findings#

1. The carry factor may be asset-specific, not systematic. Only 7 of 20+ assets show persistent positive funding. Whether this generalizes to a broader universe requires more data.

2. Spread compression timing is unknown. Without time-series spread data, we cannot determine whether wide spreads persist, narrow, or widen further. The spread factor is observational, not predictive.

Not Supported#

1. H3 (dispersion as stress indicator) cannot be evaluated with point-in-time data. This requires a time-series of cross-sectional dispersion with corresponding portfolio P&L, which is not available in the current dataset.

Limitations#

This study has significant limitations that must be stated clearly. First, the funding rate dataset covers a single market cycle — we cannot distinguish cycle-dependent effects from structural risk factors. Traditional factor models require multi-decade data; we have months. Second, cross-exchange spread data is point-in-time rather than time-series, preventing any statistical validation of spread dynamics or mean-reversion behavior. Third, the analysis excludes all execution costs, slippage, gas fees, and counterparty risk — any return-to-risk ratio or performance estimate presented here is gross of these costs and systematically overstates achievable returns. Fourth, the asset universe is limited to perpetual futures with active Hyperliquid and/or Binance listings, introducing selection bias toward liquid, exchange-listed tokens. Fifth, funding rates for recently listed assets (PROVE, SUPER, ALT) exhibit extreme non-stationarity that makes them unsuitable for factor construction without controlling for listing age. Finally, no out-of-sample validation or walk-forward testing has been performed — all findings are in-sample and should be treated as hypothesis-generating rather than confirmatory.

Conclusion#

This pilot study proposes a three-factor crypto-native portfolio framework — funding carry, spread compression, and dispersion momentum — and evaluates its components using 831K+ real funding rate observations. The carry factor shows the most promise: a handful of mid-cap perpetuals (XMR, NEAR, ZRO) exhibit persistent positive funding with manageable volatility, suggesting a structural short-demand premium. The spread factor captures real pricing inefficiency between DEX and CEX venues, but lacks the time-series data needed for statistical validation. The dispersion factor remains untested.

What this paper IS: A hypothesis-generating framework for crypto-native portfolio construction, grounded in real production data, with explicit reliability labels and honest assessment of data limitations.

What this paper is NOT: A production-ready allocation model, a backtested strategy, or a claim that any of these factors generate risk-adjusted alpha net of execution costs.

The natural next step is to collect time-series spread data and walk-forward test the carry factor with proper execution cost modeling.

References#

1. Fama, E.F. & French, K.R. (1993). "Common risk factors in the returns on stocks and bonds." Journal of Financial Economics, 33(1), 3–56.
2. Alexander, C. & Dakos, M. (2020). "A critical investigation of cryptocurrency data." Economic Notes, 49(3), e12161.
3. Harvey, C.R. et al. (2016). "...and the Cross-Section of Expected Returns." Review of Financial Studies, 29(1), 5–68.
4. Liu, Y. et al. (2022). "Risks and Returns of Cryptocurrency." Review of Financial Studies, 35(11), 5157–5217.
5. Venym Labs Research (2026). "Systemic stress indicator: funding rate dispersion ratio." Internal research paper.
6. Venym Labs Research (2026). "Funding Rate Mean Reversion: Testing Ornstein-Uhlenbeck Dynamics in Perpetual Futures Markets." Internal research paper.