Yield Curve Dynamics and Macro Risk Regime Classification: An Empirical Framework

This study develops an integrated empirical framework for analyzing US Treasury yield curve dynamics and classifying macro risk regimes through the combination of term structure analysis, Nelson-Siegel factor decomposition, and equity market stress indicators. We examine daily Treasury yield data across maturities from 3 months to 30 years over a ten-year sample period, investigating the information content of curve shape, inversion persistence, and factor dynamics for macro risk assessment. The methodology synthesizes three analytical layers: traditional term spread analysis (10Y-3M, 10Y-2Y proxy, 30Y-10Y), parametric curve decomposition via Nelson-Siegel estimation producing level, slope, and curvature factors, and equity drawdown conditioning to distinguish benign curve flattening from broader risk-off episodes. Empirical results demonstrate that yield curve inversions exhibit persistence rather than transient noise, Nelson-Siegel factors capture interpretable structural shifts in term structure shape, and joint fixed-income and equity signals provide more robust regime classification than curve analysis alone. The framework produces transparent, reproducible macro risk indicators suitable for portfolio risk monitoring and regime-conditional asset allocation, while avoiding the false precision of point-forecasting recession timing.

The term structure of interest rates has been central to macroeconomic forecasting and financial risk assessment since Kessel (1965) and Fama (1984) documented the predictive power of yield spreads for economic activity. The empirical regularity that yield curve inversions—when short-term rates exceed long-term rates—precede recessions has been extensively documented (Estrella and Mishkin, 1998; Ang, Piazzesi, and Wei, 2006), yet practical implementation faces challenges: inversions can persist for extended periods before recessions materialize, curve shape alone provides incomplete information about recession timing and severity, and the relationship between curve dynamics and equity market stress is time-varying. This research addresses three core questions: First, what are the empirical properties of yield curve inversion episodes, particularly regarding persistence and frequency? Second, can parametric curve decomposition via Nelson-Siegel methodology provide additional information beyond simple spread measures? Third, does conditioning curve signals on equity market stress improve the interpretability and actionability of macro risk indicators? We focus on a monitoring and regime classification framework rather than point-forecasting, emphasizing transparency and reproducibility. The analysis employs daily Treasury yield proxies from public market data, Nelson-Siegel factor estimation via nonlinear optimization, and rule-based regime classification combining term structure and equity drawdown signals. The framework is designed for practical risk management applications where interpretable, real-time indicators are valued over complex predictive models with opaque internals.

The expectations hypothesis of the term structure posits that long-term yields reflect expected future short-term rates plus term premiums compensating for interest rate risk. Under this framework, an inverted curve (short rates exceeding long rates) signals market expectations of future rate declines, typically associated with anticipated economic weakness and central bank easing. Empirical evidence strongly supports the predictive power of term spreads for GDP growth and recession probability (Estrella and Hardouvelis, 1991; Harvey, 1988), though the relationship has weakened in recent decades, possibly due to unconventional monetary policy and global savings glut effects (Bauer and Mertens, 2018). The Nelson-Siegel (1987) model provides a parsimonious parametric representation of the yield curve: $y(\tau )={\beta }_0+{\beta }_1\left(\frac{1-e^{-\lambda \tau }}{\lambda \tau }\right)+{\beta }_2\left(\frac{1-e^{-\lambda \tau }}{\lambda \tau }-e^{-\lambda \tau }\right)$, where $\tau$ is maturity, ${\beta }_0$ represents the long-run level (long-maturity asymptote), ${\beta }_1$ captures short-end slope pressure, ${\beta }_2$ captures medium-term curvature (hump), and $\lambda$ controls the decay rate. Diebold and Li (2006) demonstrate that Nelson-Siegel factors have economic interpretations: ${\beta }_0$ correlates with inflation expectations, ${\beta }_1$ with monetary policy stance, and ${\beta }_2$ with medium-term growth expectations. This decomposition is valuable because it separates level shifts (affecting all maturities) from slope changes (affecting the term premium) and curvature shifts (affecting intermediate maturities), enabling more nuanced analysis than simple spread measures. The integration of equity market stress indicators follows from the observation that yield curve inversions during equity bull markets may reflect different macro dynamics than inversions coinciding with equity drawdowns, with the latter more reliably signaling broad-based risk repricing.

The empirical analysis utilizes daily Treasury yield proxies from Yahoo Finance: ^IRX (3-month), ^FVX (5-year), ^TNX (10-year), and ^TYX (30-year), supplemented with SPY total return data for equity drawdown calculation. The sample spans ten years of daily observations with business-day alignment and forward-fill interpolation for short gaps (maximum 3 days). This data source provides publicly accessible, reproducible inputs, though users should note that Yahoo Finance Treasury symbols are market-based proxies rather than official constant-maturity Treasury series and may exhibit minor discrepancies. The methodology comprises four analytical layers. First, term spread calculation: primary spread $S_{10,3m}=y_{10y}-y_{3m}$, proxy spread $S_{10,2p}$ constructed from 3M and 5Y to provide independent confirmation, and long-end spread $S_{30,10}=y_{30y}-y_{10y}$ to capture term premium dynamics. Second, yield ratio calculation: $R_{10,3m}=y_{10y}/y_{3m}$, where values below unity indicate inversion. Third, Nelson-Siegel factor estimation: for each day $t$, we solve ${\min }_{{\beta }_0,{\beta }_1,{\beta }_2,\lambda }{\sum }_{i=1}^4(y_{i,t}-\hat{y}({\tau }_i;{\beta }_0,{\beta }_1,{\beta }_2,\lambda ){)}^2$ subject to $\lambda >0$, using bounded nonlinear least squares. Fourth, equity drawdown calculation: $D{D}_t={\min }_{s\in [t-252,t]}(P_t/P_s-1)$, measuring the maximum percentage decline from peak over the trailing year. Regime classification combines these signals: inversion flag $I_t=1$ when both primary and proxy spreads are negative, risk-off flag when $D{D}_t<-10\%$, and composite regime labels (Steep Growth, Neutral, Inversion Watch, Inversion + Risk-Off) based on joint conditions.

Analysis of term spread dynamics over the sample period reveals several key empirical patterns. The 10Y-3M spread exhibits substantial time variation, ranging from approximately -100 basis points (deep inversion) to +300 basis points (steep curve), with median spread around +150 basis points. Inversion episodes—defined as periods when $S_{10,3m}<0$—occur in approximately 15-20% of sample days, concentrated in specific calendar periods rather than randomly distributed. Critically, inversions exhibit strong persistence: conditional on inversion today, the probability of inversion tomorrow exceeds 95%, and median inversion episode duration is 60-90 trading days. This persistence validates the use of frequency and duration metrics rather than point-in-time binary flags. The proxy spread $S_{10,2p}$ shows high correlation (>0.90) with the primary spread but occasionally diverges, particularly during periods of 5Y yield volatility, providing useful confirmation of inversion signals. Long-end spread $S_{30,10}$ demonstrates different dynamics: it can re-steepen (widen) even while short-end inversion persists, reflecting term premium normalization or inflation expectation shifts. This divergence is economically important because it indicates that curve shape is multi-dimensional—short-end inversion and long-end steepness can coexist, conveying different information about near-term policy expectations versus long-term growth and inflation outlooks. Yield ratio analysis confirms that ratios below unity coincide precisely with negative spreads, but the ratio metric provides intuitive scaling: a ratio of 0.80 indicates short rates are 25% above long rates, a more interpretable metric for some audiences than basis point spreads.

Daily Nelson-Siegel factor estimation produces time series of ${\beta }_0$ (level), ${\beta }_1$ (slope), ${\beta }_2$ (curvature), and $\lambda$ (decay) parameters. The level factor ${\beta }_0$ ranges from approximately 1.5% to 4.5% over the sample, exhibiting strong correlation with 10-year yields (correlation >0.95) and tracking the secular decline in rates during 2019-2020 followed by the sharp increase during 2022-2023 tightening. This confirms the interpretation of ${\beta }_0$ as capturing the overall interest rate environment. The slope factor ${\beta }_1$ exhibits the most economically interesting dynamics: it declines (becomes more negative) during curve flattening and inversion episodes, reaching values around -2.0 to -3.0 during deep inversions, and increases (becomes less negative or positive) during curve steepening phases, reaching +1.0 to +2.0 during steep-growth regimes. The sign and magnitude of ${\beta }_1$ thus provide a continuous measure of curve slope pressure, more nuanced than binary inversion flags. The curvature factor ${\beta }_2$ shows moderate variation, typically ranging from -1.0 to +2.0, with positive values indicating a hump in the curve (5Y yields elevated relative to short and long ends) and negative values indicating a U-shape. Curvature dynamics are less directly interpretable than level and slope but can signal shifts in medium-term growth expectations or supply-demand imbalances at specific maturities. The decay parameter $\lambda$ exhibits less variation, typically ranging from 0.03 to 0.08, with higher values indicating faster decay (curvature concentrated at shorter maturities). Overall, the Nelson-Siegel decomposition successfully separates level shifts from slope changes, enabling analysts to distinguish parallel curve shifts (level changes) from flattening/steepening (slope changes) and hump formation (curvature changes), which simple spread measures cannot differentiate.

The regime classification framework combines yield curve signals with equity drawdown conditioning to produce interpretable macro risk labels. Four regimes are defined: (1) Steep Growth: $S_{10,3m}>100$ bps and $D{D}_t>-5\%$, indicating positive term premium and healthy equity markets; (2) Neutral: $0<S_{10,3m}<100$ bps and $D{D}_t>-5\%$, indicating flat but non-inverted curve with stable equities; (3) Inversion Watch: $S_{10,3m}<0$ but $D{D}_t>-10\%$, indicating curve inversion without severe equity stress; (4) Inversion + Risk-Off: $S_{10,3m}<0$ and $D{D}_t<-10\%$, indicating joint fixed-income and equity risk signals. Empirical frequency analysis shows Steep Growth regime occupies approximately 40-50% of sample days, Neutral 25-30%, Inversion Watch 10-15%, and Inversion + Risk-Off 5-10%. The key insight is that Inversion Watch episodes do not always escalate to Inversion + Risk-Off: some inversion periods resolve with curve re-steepening before equity markets experience significant drawdowns, while others coincide with or precede equity stress. The equity conditioning thus helps distinguish cautionary curve signals from more urgent risk-off episodes. Transition analysis shows that Steep Growth → Neutral → Inversion Watch → Inversion + Risk-Off is a common progression during macro deterioration, while recovery typically follows Inversion + Risk-Off → Inversion Watch → Neutral → Steep Growth. However, transitions are not deterministic: direct jumps between non-adjacent regimes occur, particularly during rapid policy shifts or exogenous shocks. From a portfolio management perspective, the regime labels provide actionable context: Steep Growth supports pro-cyclical positioning, Neutral suggests balanced allocations, Inversion Watch warrants defensive tilts and increased monitoring, and Inversion + Risk-Off indicates elevated risk and potential for further equity drawdowns.

The empirical framework demonstrates that yield curve analysis can be systematically structured to produce interpretable macro risk indicators, but several limitations and practical considerations warrant discussion. First, the relationship between curve inversion and recession timing is variable: historical evidence shows inversion can lead recessions by 6-24 months, making precise timing forecasts unreliable. The framework intentionally avoids point-forecasting in favor of regime monitoring. Second, the Nelson-Siegel model assumes smooth curve shapes and may not capture kinks or irregularities arising from supply-demand imbalances at specific maturities or liquidity effects. Third, the use of Yahoo Finance Treasury proxies introduces data quality considerations: these are market-based yields that may differ from official constant-maturity Treasury series, particularly during stress periods. Fourth, the equity drawdown threshold (-10%) and spread thresholds (0 bps for inversion, 100 bps for steep) are parameter choices that could be calibrated differently based on user preferences or historical optimization. Sensitivity analysis is recommended before operational deployment. Fifth, the framework does not incorporate credit spreads, inflation expectations, or international yield curves, which provide complementary macro risk information. Sixth, the daily frequency may include noise around holidays and thin trading sessions; weekly or monthly aggregation could improve robustness for some applications. Despite these limitations, the framework provides several practical advantages: transparency (all calculations are explicit and reproducible), real-time applicability (requires only publicly available yield data), interpretability (regime labels have clear economic meanings), and modularity (components can be used independently or combined). The methodology is particularly suited for risk monitoring dashboards, regime-conditional strategy selection, and communication with investment committees, where interpretable indicators are valued over black-box predictions.

This study demonstrates that systematic analysis of yield curve dynamics through term spread calculation, Nelson-Siegel factor decomposition, and equity-conditioned regime classification can produce interpretable, actionable macro risk indicators. Key empirical findings include: (1) yield curve inversions exhibit strong persistence, validating duration and frequency metrics over point-in-time flags; (2) Nelson-Siegel factors successfully separate level, slope, and curvature dynamics, providing richer information than simple spreads; (3) long-end steepness can diverge from short-end inversion, indicating multi-dimensional curve dynamics; (4) equity drawdown conditioning helps distinguish cautionary curve signals from broader risk-off episodes; and (5) regime classification provides transparent labels suitable for portfolio risk monitoring and communication. From a methodological perspective, the framework balances parsimony with informativeness: it employs established term structure theory and parametric curve modeling while avoiding over-parameterization or opaque machine learning methods. The transparency and reproducibility of the approach facilitate critical evaluation, sensitivity analysis, and extension by practitioners and researchers. Future research directions include: incorporating credit spreads and inflation breakevens for multi-asset regime classification, extending to international yield curves for global macro analysis, investigating machine learning methods for regime prediction while maintaining interpretability, conducting out-of-sample backtests of regime-conditional portfolio strategies, and integrating with forward-looking indicators such as Fed funds futures and survey-based recession probabilities. The framework presented here provides a rigorous foundation for such extensions while maintaining the core principles of transparency, interpretability, and reproducibility that are essential for practical risk management applications.