## The Insurer Channel of Monetary Policy Working Paper No. WP/2025/054 ## Source details **Canonical URL:** [The Insurer Channel of Monetary Policy Working Paper No. WP/2025/054](https://www.imf.org/-/media/files/publications/wp/2025/english/wpiea2025054-print-pdf.pdf) ## Other formats - [Markdown version](/-/media/files/publications/wp/2025/english/wpiea2025054-print-pdf.pdf.md) - [Structured JSON version](/-/media/files/publications/wp/2025/english/wpiea2025054-print-pdf.pdf.json) --- ### Role of life insurers and portfolio incentives - Life insurers have very long-term liabilities that shape investment behavior. - Insurers account for close to 20 percent of outstanding corporate bonds and over 50 percent of corporate business debt in the US. - Within insurers’ general account bonds: - Corporate bonds ≈ 60 percent of total value; Treasuries ≈ 13 percent. - Within corporate bonds held by insurers: - More than 90% are investment grade (NAIC categories 1 and 2); NAIC 5 & 6 < 1%. - Key tradeoff for insurers: - Extend asset duration via long-term treasuries (maturities at issuance up to 30 years) versus earn higher yields on medium-term private debt (corporate bonds typically under 15 years). - When discount rates are low, insurers appear to bear negative net duration exposure and may prefer higher-yield, lower-duration corporate bonds over lower-yield, higher-duration treasuries. - Long-term risk-free yields are a crucial reference asset shaping insurers’ demand for risky private debt. ### Primary-market insurer demand: evidence and magnitudes - Data: transaction-level primary market data with close to 400,000 transactions by insurers across more than 3,500 corporate bond issuances (2014–2022). - Main empirical measure: net purchases by insurers within three months of issuance scaled by total supply (“insurer purchase share”). - Baseline bond-level regressions (purchase share in percentage points (pp)): - A 1 pp increase in T30 spread (corporate yield minus 30-year Treasury) → a 5.88 pp increase in insurer purchase share (NAIC 1–3 specification). - Average insurer purchase share ≈ 25%. - Duration-matched credit spreads add no additional explanatory power beyond T30 spreads. - Extensive-margin (number of insurers participating): - A 1 pp increase in T30 spread → 5 more insurers participating. - Average number of insurers per issuance ≈ 20. - Insurer demand increases with bond duration: - A 1 year increase in bond duration → 1 pp increase in insurer purchase share (monetary-policy regression result). - A 5 year increase in bond duration → 3 additional insurers purchasing the bond (extensive-margin results). ### Monetary policy shocks, long-term yields, and insurer behavior - Identification: - Use high-frequency 30-year Treasury yield changes in a one-hour window around FOMC announcements. - Filter to days when bond yields and stock markets move in opposite directions to isolate interest-rate shocks (37 of 71 FOMC announcements in sample). - Quantity responses: - A 10 basis point (bp) contractionary monetary policy shock (∆T30 Yield_MP) → insurers’ purchase shares fall by 8 pp (≈ 33 percent of average purchase share). - Same shock → 8 fewer insurers participate in a corporate bond issuance within three months. - Price responses and risk premia: - Secondary-market sample: >10,000 bonds for 1.8 million bond-weeks (2014–2022). - Risk premia (EBP) defined as yields in excess of duration-matched risk-free benchmarks unexplained by default risk and bond characteristics. - Interaction results: - For each percentage point increase in insurer holding shares (previous quarter), a 10 bp increase in long-term Treasury yields → an additional 0.2 bp increase in excess bond premia (EBP). - A one standard deviation increase in insurer holding shares → a 5 bp larger increase in EBP in response to a 10 bp long-term yield rise. - Monetary-policy-induced short-term yield changes (∆T2) show small or statistically indistinct interactions in risk premia regressions. ### Regulatory reform (RBC C1) and amplification of transmission - 2021 RBC reform timeline: - October 22 2020: ACLI and NAIC public RFP released. - February 1 2021: Moody’s Analytics recommendations delivered. - April 15 2021: ACLI proposed risk weights released. - June 11 2021: Changes formally adopted; revised factors effective with year-end reporting for 2021. - Key features: - NAIC designations increased from 6 to 21; risk weights more granular and more continuous. - Treasuries remained exempt from C1 RBC requirements. - Reform implied substantial variation in magnitude and sign of changes in risk weights across rating categories. - Empirical amplification: - Appendix Table D.5 / main triple-difference: - A one standard deviation increase in risk weights → associated with a 9.7 bp increase in the sensitivity of credit risk premia to a 10 bp variation in the T30 yield (Appendix Table D.5 summary). - Table D.6 (2019–2022, Excl. COVID): - Coefficient on 1{Post} × ∆RBC_r × T30 Spread: All Ratings = 6.58 ∗∗∗ (2.23); NAIC 1–3 = 8.77 ∗∗∗ (2.44). - Table 9 / triple interaction (2019–2022): - 1{Post} × ∆RBC_r × ∆T30 Yield_MP: All Ratings = 1.31 ∗∗ (0.53); NAIC 1–3 = 0.65 ∗∗ (0.28). ### Magnitude comparisons and robustness - Comparative magnitudes: - Palazzo and Yamarthy (2022) estimate that a 10 bp contractionary shock raises credit default spreads by 2 bp; insurer-channel estimates imply substantially larger differential effects when insurers dominate. - Robustness: - Results robust when focusing on all bonds or rating groups with largest insurer exposures. - Excluding the COVID period often strengthens results. - Findings are specific to long-term yield movements (∆T30) and are not found for short-term (∆T2) yield shocks in analogous specifications (see Appendix Tables D.3, D.7). - Similar conclusions hold using a simple credit spread measure instead of the paper’s EBP (Appendix Table D.8). ### Data sources, sample, and summary statistics - Sample period: 2014–2022. - Key data sources: - Insurer holdings and transactions: S&P Capital IQ Pro (holdings quarterly; transactions daily). - Primary market corporate and Treasury characteristics: Mergent Fixed Income Securities Database and TreasuryDirect. - Secondary market corporate bonds: TRACE. - High-frequency long-term yields and stock returns: Bloomberg (30-year Treasury yields around FOMC). - Default risk and option-adjusted duration: Moody’s CreditEdge. - Insurer stock price data: CRSP. - Selected primary-market summary statistics (Table 1): - Insurer purchase share (net purchases within three months as fraction of issuance): 2014–2018 averages between 0.23 and 0.29 by year; 2019–2022 means: 2019 0.25, 2020 0.18, 2021 0.24, 2022 0.24. - Number of insurers per issuance: yearly means range ≈ 17.0 to 22.6 (annual panel values provided). - T30 spread (percentage points): yearly means range 0.44 to 1.37 across 2014–2022. - Average bond duration purchased by insurers: yearly means close to 9–11 years (e.g., 2014 9.8; 2022 9.6). - Distance to default (Dtdefault) fell in 2020 (2020 mean 5.2) and rose thereafter. - Offering amounts (BN): yearly means reported (e.g., 2014 0.66 BN; 2020 0.99 BN; 2022 0.82 BN). - Insurer holdings composition (Table D.1 / Table 2 summary): - Corporate shares by par value: 2014 57.2% → 2019 58.4% → 2022 57.0%. - Government bonds: 2014 15.8% → 2019 13.2% → 2022 11.1%. - Private MBS and other categories reported year-by-year (see Table D.1 values). - NAIC rating composition (Table D.2): - NAIC-1 + NAIC-2 together comprise the vast majority of corporate holdings (e.g., 2014 NAIC-1 43.7% and NAIC-2 47.9%; 2019 NAIC-1 44.4% and NAIC-2 50.0%; 2022 NAIC-1 42.7% and NAIC-2 52.3%). ### Appendix highlights: construction and additional regressions - Risk premia construction (Appendix A): - Model: ln S_{c,i,t} = β EDF_{c,i,t} + γ′ Z_{c,i,t} + ε_{c,i,t}. - Predicted spread: ˆS_{c,i,t} = exp[ˆβ EDF_{c,i,t} + ˆγ′ Z_{c,i,t} + ˆσ^2 / 2 ]. - Excess bond premium: EBP_{c,i,t} = S_{c,i,t} − ˆS_{c,i,t}. - Uses option-adjusted duration and bond-specific default-risk measures accounting for seniority. - Insurers are short-duration (Appendix B): - Two-factor rolling regressions (weekly, two-year windows): R_{i,t} = α + β R_{m,t} + γ R_{10,t} + ε_{i,t}. - One percentage point increase in 10-year Treasury returns associated with one percentage point decline in insurance portfolio returns in sample periods; estimated duration gap implied ~10 years. - Insurers sometimes add asset duration via interest-rate derivatives equivalent to nearly $200 BN in bonds. - Key tables with exact coefficients and standard errors: - Table 3 (primary-market purchase-share regressions): T30 Spread coeff ~5.35–5.87 with standard errors reported; Duration coeff ~0.70–0.84. - Table 4 (extensive margin): T30 Spread coeff ~4.81–4.99 with SEs reported; Ln(Offering Amount) large positive effect on number of insurers. - Table 6–7 (monetary-policy regressions): ∆T30 Yield_MP coefficients range from −0.64 ∗∗∗ (0.18) to −0.86 ∗∗∗ (0.20) depending on sample. - Table 8 (risk premia interaction): Insurer Exposure × ∆T30 Yield_MP coeff = 0.03 ∗∗ (0.01) for All Ratings (2014–2022), Observations = 108,053. - Appendix tables D.3–D.8 provide robustness and alternative specifications reported in the text. ### Conclusions and implications - Long-term risk-free rates (the long end of the yield curve) are a crucial benchmark shaping insurers’ demand for private debt. - Conventional monetary policy that moves long-term risk-free yields affects insurers’ relative demand for privately issued debt and, through insurers’ large market footprint, alters risk premia. - Transmission from short-term policy rate changes to risk premia is mediated by the response of long-term yields because insurers compare privately issued yields to long-term treasuries (T30 spreads). - The importance of yield differentials to long-term treasuries produces pronounced cross-sectional heterogeneity in monetary transmission across corporate bonds. _Source: wpiea2025054-print-pdf_ ### 1. Introduction ### 1. Introduction ### Role of life insurers in monetary transmission - Financial intermediaries are central to the transmission of monetary policy to the real economy; the literature has focused on banks but given the growing importance of non-bank financial intermediaries (NBFIs), life insurers merit attention. - Key features of life insurers: - Very long-term liabilities that shape investment behavior (Koijen and Yogo,2023; Scharfstein,2018). - Account for over 20 percent of corporate bond holdings in the US. - The sector’s allocation skews more towards medium-term corporate bonds than to long-term treasuries. - Long-term risk-free yields are a crucial reference asset for insurers and shape their demand for risky private debt. ### Insurer tradeoff: duration versus yield - Life insurers face a tradeoff between: - Extending asset duration by investing in long-term treasuries (maturities at issuance up to 30 years), and - Earning higher yields on some medium-term private debt (corporate bonds with maturities at issuance typically under 15 years) due to credit risk. - When discount rates are low, life insurers appear to bear negative net duration exposure. - As a result, insurers may prefer higher-yield, lower-duration corporate bonds over lower-yield, higher-duration treasuries. ### Evidence on insurer demand in primary markets - Data used: transaction-level primary market data with close to 400,000 transactions by insurers across more than 3,500 corporate bond issuances from 2014-22. - Main empirical measure: net purchases by insurers close to issuance—scaled by total supply at issuance—(“insurer purchase share”) to isolate insurer demand relative to other institutional investors. - Key estimates: - For each percentage point increase in the spread between yields on privately issued bonds and long-term risk-free rates, insurers’ share of purchases rises by 5.9 percentage points (25 percent of their average purchase share), with 5 additional insurers participating. - Yield differentials to duration-matched risk free benchmark rates offer no additional information about insurers’ relative demand. ### Monetary policy shocks, long-term yields, and insurer behavior - Identification strategy: - Use high-frequency variation in 30 year treasury yields in a one-hour window around monetary policy announcements to isolate impacts on long-term yields. - Following Jarociński and Karadi (2020), restrict to interest-rate shocks (announcements where bond yields and stock markets move in opposite directions) to separate interest-rate shocks from information shocks. - Stylized fact documented: long-term yields respond systematically to interest-rate shocks but not to information shocks. - Quantity responses: - Contractionary monetary policy shocks induce insurers to substitute away from privately issued debt. - A 10 basis point contractionary monetary policy shock reduces insurers’ purchase shares by 8 percentage points (33 percent of insurers’ average purchase share), with 8 fewer insurers participating. - Price responses and risk premia: - Use secondary market pricing with a weekly frequency (to allow for illiquidity) covering more than 10,000 bonds for 1.8 million bond-weeks over 2014-22. - Define risk premia as yields in excess of duration-matched risk-free benchmark rates that cannot be explained by default risk and other bond-specific characteristics (following Gilchrist and Zakrajšek,2012), with improvements: - Use option-adjusted measures of bond duration. - Use bond-specific measures of default risk accounting for seniority across claims. - Two exercises linking insurers to transmission of monetary shocks to risk premia: - Risk premia respond more strongly to monetary policy shocks when insurers are a more important component of the investor base: a one standard deviation increase in insurer holding shares in the previous quarter is associated with an additional increase of 5 basis points in risk premia in response to a 10 basis point contractionary monetary shock. - Regulatory shifts that changed insurers’ relative attractiveness of corporate bonds versus treasuries altered the sensitivity of risk premia: a one percentage point increase in risk weights for insurers leads to an additional increase of 9.7 basis points in risk premia given a 10 basis point contractionary monetary policy shock. - Comparative magnitude: Palazzo and Yamarthy (2022) estimate that a 10 basis point contractionary shock raises credit default spreads by 2 basis points. ### Contributions and implications - Two main contributions: 1. Clarify how insurers’ long-term liabilities make long-term risk-free debt a crucial reference asset shaping insurer demand relative to other investors. 2. Show that the impact of conventional monetary policy on the long end of the yield curve and on risk premia are causally connected through insurers’ investment behavior; isolate the price of risk from quantity shifts using improved measures. - Implications: - Transmission of monetary policy targeting short-term rates to risk premia depends on the reaction of long-term rates because these shape insurers’ investment behavior. - Transmission to risk premia is highly heterogeneous across privately issued debt; yield differentials relative to long-term risk free rates are a critical determinant across bonds with varying risk and maturities. ### Data overview - Sample period across datasets: 2014-2022. - Data sources and constructs: - Insurer holdings and transactions: S&P Capital IQ Pro (holdings quarterly; transactions daily). - Primary market corporate and treasury bond characteristics: Mergent Fixed Income Securities Database and TreasuryDirect. - Secondary market corporate bond information: TRACE; median yield and duration across transactions on the last active day in each bond-week. - High-frequency long-term bond yields and stock returns: Bloomberg (30 year treasury yields in a one-hour window around FOMC announcements). - Bond default risk and duration-matched treasury spreads: Moody’s CreditEdge (bond-week level measures that account for seniority differences and option-adjusted spreads). - Insurer stock price data: Center for Research in Security Prices (CRSP) to estimate net interest-rate exposure. - Summary statistics (Table 1 highlights): - Insurer purchase share (net purchases within three months of issuance as a fraction of total issuance): on average between 20 to 25% across years. - Average number of insurers investing in a corporate bond within three months of issuance: close to 20. - T30 spread (spread between corporate bond yields and 30 year Treasury yields, in percentage points): ranged from 0.5 to 1.5 percent over time. - Average duration of corporate bonds purchased by insurers: close to 10 years. - Distance to default decreased significantly in 2020 due to higher default risk during COVID; robustness checks document results with and without the COVID period. _1. Introduction — wpiea2025054-print-pdf_ ### 3. Institutional background ### 3. Institutional background ### Role of life insurers in US corporate bond markets - Life insurers are the largest institutional holders of corporate bonds in the US. - Insurers hold close to 20 percent of outstanding corporate bonds. - Corporate bonds constitute more than 50 percent of corporate business debt in the US. - Within life insurers’ general account bonds: - Corporate bonds make up close to 60 percent of the total value. - Treasuries account for about 13 percent. - Within corporate bonds held by insurers: - More than 90% are investment grade (NAIC categories 1 and 2). - The riskiest categories (NAIC 5 & 6) make up less than 1% of all corporate bonds held by insurers. - Appendix references (D.1, D.2) indicate this composition has remained stable over time. ### Insurer capital regulation reform in 2021 (RBC C1) - Life insurers are subject to risk-based capital (RBC) requirements; charges applied to bonds based on credit risk are captured under C1 RBC requirements. - Treasuries remained exempt from C1 RBC requirements. - Prior to 2021: - NAIC classified bonds into six designations (NAIC-1 through NAIC-6), mapped to Moody’s rating ranges (e.g., NAIC-1 = Aaa to A3; NAIC-2 = Baa1 to Baa3; NAIC-3 = Ba1 to Ba3; NAIC-4 = B1 to B3; NAIC-5 = Caa1 to Caa3; NAIC-6 = the rest). - Reform timeline and process: - American Council of Life Insurers (ACLI) and NAIC released a public request for proposal to assess new RBC C1 Bond Factors on October 22 2020. - Moody’s Analytics delivered a report with recommendations on February 1 2021. - ACLI released proposed risk weights on April 15 2021. - Changes were formally adopted by the Life Risk-Based Capital Working Group on June 11 2021. - The revised factors came into effect with year-end reporting for 2021. - Key features of the 2021 reform: - Increased granularity of risk weights: number of NAIC designations rose from six to 21, aligning more closely with private rating-agency granularities. - Risk weights became more continuous compared to pre-2021 discontinuities. - The reform implied substantial variation in both the magnitude and sign of changes in risk weights across rating categories. - Empirical relevance: - Insurers held substantial amounts in rating categories subject to significant increases in risk weights, enabling sharp identification of insurer responses to changes in risk weights and to monetary policy via spreads to long-term treasuries. ### Insurers and the tradeoff between asset duration and investment yields - Insurers skew allocations toward bonds with longer maturities: - Insurers purchase a significantly larger share of longer-maturity bonds in the primary market, across corporate bonds and treasuries. - Insurer purchase shares are larger for corporate bonds than for treasuries. - Despite extending asset maturities relative to other institutional investors, insurers appear to be short duration (net negative duration exposure). - Equity returns are negatively correlated with returns on long-term bonds (see Appendix B). - Insurers face a tradeoff: higher yields on medium-term corporate debt than on long-term risk-free debt lead insurers to balance investment income against further extending asset duration. - Long-term treasuries serve as a crucial reference asset for insurers—corporate bonds with lower yields than long-term treasuries offer no tradeoff between duration and investment income. ### Insurer relative demand driven by yield differentials to long-term treasuries (T30 spread) - Data and scope: - Transaction-level dataset: close to 400,000 transactions by insurers across more than 3,500 corporate bonds issued from 2014-22. - Key empirical findings on primary-market demand: - Insurers’ relative demand for corporate bonds falls sharply when bond yields are close to or below the 30 year treasury yield (T30 spread). - Insurer holdings jump once the T30 spread is positive, and insurers continue to focus on T30 spreads after issuance. - Bond-level regression evidence (baseline focuses on NAIC 1-3; purchase shares and T30 spread expressed in percentage points (pp)): - A 1 pp increase in T30 spread is associated with a 5.88 pp increase in insurer purchase share. - The average insurer purchase share is close to 25%. - Duration-matched credit spreads add no additional explanatory power beyond T30 spreads: insurers view long-term treasuries as the relevant benchmark risk-free asset regardless of bond maturity. - Extensive-margin evidence (number of insurers participating per issuance): - A 1 pp increase in T30 spread is associated with 5 more insurers participating in a bond issuance. - The average number of insurers participating in issuance of corporate bonds is close to 20. - Insurer demand strongly increases with bond duration: - An increase in bond duration is associated with higher relative demand from insurers, consistent with insurers seeking to extend asset duration when possible. - Prior literature suggests insurer behavior is driven more by incentives to match duration than by reaching for yield. ### Transmission of monetary policy to risk premia via life insurers - Conceptual channel: - Conventional monetary policy that raises long-term risk-free yields makes long-term treasuries more attractive relative to privately issued debt, reducing insurers’ demand for private debt and potentially increasing risk premia given insurers’ large market footprint. - Insurer demand for long-term treasuries: - Regression using treasuries with maturities greater than or equal to 10 years shows insurers increase relative demand for long-term treasuries when treasury offering yields rise. - Appendix Figure C.1 shows a strong positive correlation between insurers’ relative demand for treasuries and treasury yields. - Identification of monetary policy shocks: - Focus on narrow windows around FOMC announcements to isolate monetary-policy-induced movements in long-term yields. - Baseline includes only FOMC announcement days when bond yields and the S&P 500 move in opposite directions in a one-hour window, following Jarociński and Karadi (2020), to separate interest rate shocks from information shocks. - Of 71 FOMC announcements in the sample, 37 are included as monetary policy shocks using this filter. - Figure 5 shows a strong correlation between short- and long-end yield variation on these filtered FOMC days. - Effect of monetary-policy-induced long-term yield changes on insurer demand for corporate bonds: - Specification: quarterly change in insurer purchase share by rating category regressed on high-frequency monetary-policy-induced changes in the 30 year treasury yield aggregated to quarterly frequency. - Monetary-policy-induced increases in long-term rates lower insurers’ relative demand for corporate bonds: - A 10 bp increase in long-term risk-free rates is associated with an 8 pp fall in insurer purchase shares. - A 1 year increase in bond duration is associated with a 1 pp increase in insurer purchase shares. - No similar effects are found when focusing on interest-rate shocks affecting short-term treasury yields (Appendix Table D.3). - Changes in number of insurers purchasing newly issued bonds: - A 10 bp increase in long-term treasury yield results in 8 fewer insurers purchasing a corporate bond within three months of issuance. - A 5 year increase in bond duration is associated with 3 additional insurers purchasing the bond. - Transmission to risk premia in secondary markets: - Secondary-market sample: more than 10,000 bonds for 1.8 million bond-weeks over 2014-22. - Risk premia constructed to capture variation in the price of risk separately from changes in the quantity of risk (see Appendix A). - Two key exercises demonstrate insurers’ footprint in transmission: 1. Bonds with larger insurer footprints in the investor base: - Regression interacting insurer exposure share (previous-quarter insurer holdings relative to par outstanding) with monetary-policy-induced ∆T30 yields shows stronger transmission when insurers account for a larger investor share. - Table 8 result: For each percentage point increase in insurer holding shares, a 10 bp increase in long-term treasury yields is associated with a 0.2 bp greater increase in excess bond premia (EBP). - A one standard deviation increase in insurer holding shares translates to a 5 bp larger increase in EBP in response to a 10 bp increase in long-term treasury yields. - Appendix Table D.4 shows no similar effects for monetary-policy-induced changes in short-term treasury yields. 2. Bonds affected by insurer-specific regulatory reform: - The 2021 RBC reform increased insurer sensitivity to T30 spreads for bonds in rating categories with increased risk weights (Appendix Table D.6). - Hypothesis: Monetary-policy-induced increases in long-term treasury yields should have a stronger impact on risk premia for bonds subject to increases in risk weights, because insurers become more sensitive to spreads to long-term treasuries. - A triple-difference regression framework is used to test whether ∆RBC (change in risk weights by rating) amplifies the effect of ∆T30 YieldMP on ∆EBP, especially in the post-reform period. *Source: IMF working paper section "3. Institutional background" (wpiea2025054-print-pdf).* ### Appendix Table D.5. ### Appendix Table D.5. ### Key empirical finding on risk weights and monetary policy transmission - The coefficient of interest is β6. - ∆RBCr denotes the change in risk weights associated with a rating category r after the reform. - ∆T30 YieldMP,w denotes the measure of monetary policy induced changes in 30-year treasury yields. - One standard deviation increase in risk weights is associated with a 9.7 bp increase in the sensitivity of credit risk premia to a 10 bp variation in the T30 yield. - The reform only applied to life insurers, so the change in sensitivity across rating categories can only be explained by a change in the investment behavior of insurers. ### Robustness and comparative magnitudes - Findings are robust when focusing on all bonds and when zooming in on the rating groups in which insurers have the largest exposures. - Excluding the COVID period strengthens the results. - The differential shift in risk premia documented is large relative to prior reduced form estimates: Palazzo and Yamarthy (2022) find that a 10 basis point contractionary shock raises risk premia by about 2 basis points. - The paper does not find similar results if focusing on how monetary policy shocks shift two-year treasury yields rather than long-term treasury yields (see Appendix Table D.7). - Results are similar if a simple measure of credit spreads is used instead of the paper’s measure of risk premia (see Appendix Table D.8). ### Conclusion — insurers and the transmission of monetary policy - Long-term risk-free rates shape insurers’ demand for private debt. - When monetary policy shifts the long end of the yield curve, this shifts insurers’ demand for private debt and therefore changes risk premia. - Across different bonds available as investment options at the same time, yield differentials to long-term risk-free rates are a crucial determinant of demand from insurers relative to other institutional investors. - The importance of yield differentials to long-term treasuries for insurers leads to stark cross-sectional heterogeneity in the impact of monetary policy on risk premia. ### Selected summary statistics (primary market) - Table 1, Panel (a): 2014-2018 (Mean, S.D. by year) - Purchase Share: 2014 Mean 0.28 S.D. 0.24; 2015 Mean 0.29 S.D. 0.25; 2016 Mean 0.26 S.D. 0.24; 2017 Mean 0.23 S.D. 0.22; 2018 Mean 0.25 S.D. 0.22. - Number of Insurers: 2014 Mean 19.9 S.D. 11.7; 2015 Mean 22.6 S.D. 13.8; 2016 Mean 22.1 S.D. 12.7; 2017 Mean 18.6 S.D. 10.8; 2018 Mean 20.5 S.D. 12.3. - T30 Spread: 2014 Mean 0.64 S.D. 1.21; 2015 Mean 1.00 S.D. 1.38; 2016 Mean 0.88 S.D. 1.53; 2017 Mean 0.79 S.D. 1.11; 2018 Mean 1.14 S.D. 0.84. - Yield: 2014 Mean 3.90 S.D. 1.19; 2015 Mean 3.86 S.D. 1.39; 2016 Mean 3.50 S.D. 1.56; 2017 Mean 3.69 S.D. 1.11; 2018 Mean 4.27 S.D. 0.85. - Duration: 2014 Mean 9.8 S.D. 4.7; 2015 Mean 9.6 S.D. 4.8; 2016 Mean 9.4 S.D. 4.9; 2017 Mean 9.1 S.D. 5.0; 2018 Mean 9.4 S.D. 5.1. - Dtdefault: 2014 Mean 11.8 S.D. 4.6; 2015 Mean 10.7 S.D. 4.7; 2016 Mean 9.4 S.D. 4.1; 2017 Mean 11.5 S.D. 5.6; 2018 Mean 9.7 S.D. 4.1. - Offering Amount (BN): 2014 Mean 0.66 S.D. 0.44; 2015 Mean 0.87 S.D. 0.71; 2016 Mean 0.89 S.D. 0.65; 2017 Mean 0.79 S.D. 0.63; 2018 Mean 0.87 S.D. 0.93. - Observations: 2014 408; 2015 467; 2016 431; 2017 440; 2018 363. - Table 1, Panel (b): 2019-2022 (Mean, S.D. by year) - Purchase Share: 2019 Mean 0.25 S.D. 0.20; 2020 Mean 0.18 S.D. 0.15; 2021 Mean 0.24 S.D. 0.20; 2022 Mean 0.24 S.D. 0.23. - Number of Insurers: 2019 Mean 19.4 S.D. 10.6; 2020 Mean 19.0 S.D. 9.4; 2021 Mean 19.1 S.D. 10.6; 2022 Mean 17.0 S.D. 9.6. - T30 Spread: 2019 Mean 1.06 S.D. 0.95; 2020 Mean 1.23 S.D. 1.34; 2021 Mean 0.44 S.D. 0.91; 2022 Mean 1.37 S.D. 0.64. - Yield: 2019 Mean 3.51 S.D. 1.04; 2020 Mean 2.68 S.D. 1.34; 2021 Mean 2.53 S.D. 0.91; 2022 Mean 4.52 S.D. 0.87. - Duration: 2019 Mean 10.4 S.D. 5.5; 2020 Mean 10.8 S.D. 6.0; 2021 Mean 11.1 S.D. 6.1; 2022 Mean 9.6 S.D. 5.2. - Dtdefault: 2019 Mean 8.6 S.D. 4.2; 2020 Mean 5.2 S.D. 2.9; 2021 Mean 8.9 S.D. 3.6; 2022 Mean 8.4 S.D. 3.3. - Offering Amount (BN): 2019 Mean 0.78 S.D. 0.55; 2020 Mean 0.99 S.D. 0.72; 2021 Mean 0.97 S.D. 0.76; 2022 Mean 0.82 S.D. 0.55. - Observations: 2019 332; 2020 668; 2021 386; 2022 330. ### Composition of insurer bond holdings (selected) - Table 2, Panel (a): Insurer bond holdings by type (Share (Unique CUSIPs), Share (Total Value)) - Corporate Bonds: 19.5% , 58.4% - Government Bonds: 18.6% , 13.2% - Private MBS: 20.5% , 15.8% - Agency MBS: 31.6% , 7.0% - Other: 9.8% , 5.6% - Total: 100% , 100% - Table 2, Panel (b): Insurer corporate bond holdings by NAIC designation (Share (Unique CUSIPs), Share (Total Value)) - NAIC Category 1: 41.9% , 44.4% - NAIC Category 2: 42.0% , 50.0% - NAIC Category 3: 7.4% , 3.7% - NAIC Category 4: 4.9% , 1.3% - NAIC Categories 5 & 6: 3.8% , 0.6% - Total: 100% , 100% *Source: wpiea2025054-print-pdf - Appendix Table D.5.* ### 2019. Government bonds include Federal and Municipal bonds. Panel (b) classifies insurer holdings of corporate bonds a ### wpiea2025054-print-pdf - 2019. Government bonds include Federal and Municipal bonds. Panel (b) classifies insurer holdings of corporate bonds a ### Insurer holdings composition (2019) - Holdings focused on 2019 to illustrate composition prior to COVID in 2020 and an insurer specific regulatory reform in 2021. ### Insurer relative demand driven by yield differentials to long-term risk-free rates (Table 3) - Dependent variable: net purchases by insurers within three months of bond issuance as a fraction of the total issuance amount (percentage points), winsorized at the 1st and 99th percentiles by rating category. - Main independent variable: T30 Spread (spread of corporate bond yields over 30 year Treasury yields, percentage points). - All Ratings (2014-2022): T30 Spread = 5.35 ∗∗ (2.33). Duration-matched Spread = -2.14 (2.38). Duration = 0.84 ∗∗∗ (0.21). Dtdefault = 0.28 (0.21). Log(Offering Amount) = -6.24 ∗∗∗ (1.54). Liquidity = 1.18 (1.54). 1{Callable} = 3.22 (2.58). Observations = 2,610. R2 = 0.68. - NAIC 1-3 (2014-2022): T30 Spread = 5.87 ∗∗ (2.45). Duration-matched Spread = -1.06 (2.41). Duration = 0.70 ∗∗∗ (0.23). Dtdefault = 0.47 ∗∗ (0.20). Log(Offering Amount) = -6.36 ∗∗∗ (1.69). Liquidity = 1.45 (1.56). 1{Callable} = 1.89 (2.15). Observations = 2,277. R2 = 0.69. - Regressions include issuer and rating-time fixed effects; standard errors double-clustered by issuer + time. - Significance notation: ∗ p <0.10, ∗∗ p <0.05, ∗∗∗ p <0.01. ### Insurer relative demand: extensive margin (Table 4) - Dependent variable: number of insurers participating in the bond sale within three months of issuance. - All Ratings (2014-2022): T30 Spread = 4.81 ∗∗∗ (1.21). Duration-matched Spread = -2.86 ∗∗ (1.35). Duration = 0.29 ∗∗∗ (0.09). Dtdefault = 0.16 ∗ (0.08). Ln(Offering Amount) = 10.49 ∗∗∗ (1.07). Liquidity = 5.09 ∗∗∗ (0.75). 1{Callable} = 0.30 (1.40). Observations = 2,716. R2 = 0.59. - NAIC 1-3 (2014-2022): T30 Spread = 4.99 ∗∗∗ (1.25). Duration-matched Spread = -2.37 (1.41). Duration = 0.26 ∗∗ (0.11). Dtdefault = 0.14 (0.08). Ln(Offering Amount) = 10.98 ∗∗∗ (1.23). Liquidity = 5.14 ∗∗∗ (0.85). 1{Callable} = 0.70 (1.55). Observations = 2,370. R2 = 0.60. - Controls: issuer and rating-quarter fixed effects; SE double-clustered by issuer + time. ### Insurer relative demand for long-term treasuries increases with yields (Table 5) - Sample: Treasury bonds with maturities >= ten years issued between 2014 and 2022. - Purchase Share (2014-2022): Offering Yield = 6.57 ∗∗∗ (2.04). Log(Offering Amount) = 7.27 ∗ (4.19). Observations = 135. R2 = 0.64. - Purchase Share (Excl. COVID): Offering Yield = 6.28 ∗∗∗ (2.02). Log(Offering Amount) = 8.72 ∗ (4.59). Observations = 130. R2 = 0.64. - Extensive Margin (2014-2022): Offering Yield = 5.22 ∗∗∗ (1.68). Log(Offering Amount) = 50.17 ∗∗∗ (4.60). Observations = 135. R2 = 0.86. - Extensive Margin (Excl. COVID): Offering Yield = 4.75 ∗∗∗ (1.67). Log(Offering Amount) = 52.47 ∗∗∗ (5.06). Observations = 130. R2 = 0.86. - Regressions include quarter fixed effects; SE clustered by quarter. ### Monetary policy shifts insurers’ relative demand (Tables 6 and 7) - Main independent variable: ∆T30 Yield_MP (change in 30 year Treasury yield in a one hour window around FOMC announcements; MP component identified using days when bond yields move opposite to stock market prices). - Table 6 — Dependent: net purchases of corporate bonds by insurers within three months as fraction of total issuance, averaged by rating: - All Ratings (2014-2022): ∆T30 Yield_MP = -0.64 ∗∗∗ (0.18). ∆Yield = -0.40 (0.29). ∆Duration = 0.83 ∗∗∗ (0.22). ∆Ln(Offering Amount) = -11.46 ∗∗∗ (1.62). Quarter Two = -3.23 ∗∗∗ (0.85). Quarter Three = -3.11 ∗∗ (1.25). Quarter Four = -12.06 ∗∗∗ (1.58). Observations = 400. R2 = 0.694. - NAIC 1-3 (2014-2022): ∆T30 Yield_MP = -0.79 ∗∗∗ (0.26). ∆Yield = 0.58 (0.66). ∆Duration = 0.69 ∗∗ (0.24). ∆Ln(Offering Amount) = -13.22 ∗∗∗ (1.42). Quarter Four = -13.20 ∗∗∗ (1.56). Observations = 329. R2 = 0.746. - Specifications include changes in controls and their lags, inflation, GDP growth, unemployment, VIX, and lagged purchase share; SE double-clustered by rating + quarter. - Table 7 — Dependent: number of insurers participating in sale within three months, averaged by rating: - All Ratings (2014-2022): ∆T30 Yield_MP = -0.59 ∗∗∗ (0.17). ∆Yield = 0.48 (0.44). ∆Dtdefault = -0.32 ∗∗ (0.11). ∆Duration = 0.64 ∗∗∗ (0.15). ∆Ln(Offering Amount) = 10.02 ∗∗∗ (1.31). Quarter effects reported. Observations = 400. R2 = 0.583. - NAIC 1-3 (2014-2022): ∆T30 Yield_MP = -0.86 ∗∗∗ (0.20). ∆Yield = 2.01 ∗∗∗ (0.52). ∆Dtdefault = -0.39 ∗∗∗ (0.11). ∆Duration = 0.52 ∗∗∗ (0.14). ∆Ln(Offering Amount) = 9.10 ∗∗∗ (1.15). Observations = 329. R2 = 0.655. - Controls and identification as in Table 6; SE double-clustered by rating + quarter. ### Risk premia are more responsive when insurers dominate investor base (Table 8) - Dependent variable: change in excess bond premia at the weekly frequency (basis points). - Main interaction: Insurer Exposure_{c,q−1} × ∆T30 Yield_MP. - All Ratings (2014-2022): coefficient = 0.03 ∗∗ (0.01). Observations = 108,053. R2 = 0.233. - NAIC 1-3 (2014-2022): coefficient = 0.01 ∗∗ (0.01). Observations = 100,221. R2 = 0.369. - Insurer Exposure_{c,q−1} is fraction of an outstanding bond held by life insurers (percentage points, quarter preceding FOMC). ∆T30 Yield_MP expressed in basis points. Controls include bond liquidity, inflation, GDP growth, unemployment, VIX; issue and week fixed effects; SE double-clustered by issue + time. ### Risk premia responsiveness increases when higher risk weights raise insurer sensitivity to yield spreads (Table 9) - Sample: 2019-2022. - Main triple interaction: 1{Post} × ∆RBC_r × ∆T30 Yield_MP. - 2019-2022 All Ratings: coefficient = 1.31 ∗∗ (0.53). Observations = 61,320. R2 = 0.24. - 2019-2022 NAIC 1-3: coefficient = 0.65 ∗∗ (0.28). Observations = 58,396. R2 = 0.30. - Notes: Post = 1 for months after August 2021. ∆RBC_r denotes change in risk based capital weights for rating category r. Controls include bond liquidity, inflation, GDP growth, unemployment rate, and VIX; controls interacted with ∆RBC_r and 1{Post}. SE double-clustered by issue + time. ### Appendix A — Construction of risk premia - Model specification: ln S_{c,i,t} = β EDF_{c,i,t} + γ′ Z_{c,i,t} + ε_{c,i,t}. - S = maturity-matched treasury spread; EDF = expected default frequency; Z = controls (bond duration, coupon, face value, indicator for callability). - Predicted spread: ˆS_{c,i,t} = exp[ˆβ EDF_{c,i,t} + ˆγ′ Z_{c,i,t} + ˆσ^2 / 2 ]. - Excess bond premium: EBP_{c,i,t} = S_{c,i,t} − ˆS_{c,i,t}. - Bond-level distance to default and option-adjusted duration used; industry fixed effects included. ### Appendix B — Insurers are short duration - Two-factor model: R_{i,t} = α + β R_{m,t} + γ R_{10,t} + ε_{i,t}, where R_{10,t} is return on 10 year treasury. - Rolling regressions (two years weekly data) indicate life insurers had significant net negative duration exposure when long-term risk-free rates were low. - Quantitative statement: one percentage point increase in 10 year treasury returns associated with one percentage point decline in returns of the insurance portfolio. - Estimated duration gap implied ~10 years (Koijen and Yogo, 2023). Evidence that life insurers use interest-rate derivatives; at times adding asset duration equivalent to nearly $200 BN in bonds with same duration as bond portfolios. - Limited risk-bearing capacity and regulatory frictions may prevent full hedging of liability duration exposure. *Italic line: Source: wpiea2025054-print-pdf - 2019. Government bonds include Federal and Municipal bonds. Panel (b) classifies insurer holdings of corporate bonds a* ### 0.5 percent. ### The Insurer Channel of Monetary Policy Working Paper No. WP/2025/054 ### Life insurers’ interest rate exposure and portfolio duration - Figure B.1 notes: exposure of US life insurers to interest rate risk is estimated as γ in the following 2 year rolling regression at the weekly frequency: R_{i,t} = α + β R_{m,t} + γ R_{10,t} + ε_{i,t}, where R_{i,t} is the return on insurance portfolio in week t, R_{m,t} is the return on value-weighted stock market portfolio in week t, and R_{10,t} is the return on 10 year treasury bond in week t. - The black line in the figure represents the estimated coefficient γ based on rolling weekly regressions using returns for the past two years. - The grey band represents the 95 percent confidence intervals based on heteroskedasticity robust standard errors. - The grey line is the ten year treasury yield over time. - Standalone numeric highlight from page: 0.5 percent. ### Insurers’ demand for long-term Treasury bonds - Figure C.1 notes: purchases share is measured as the net purchases by insurers within three months of bond issuance as a fraction of the total issuance amount, plotted against offering yield. - Purchase share is expressed in percentage points and winsorized at the 1st and 99th percentiles by rating category. - The offering yield is expressed in percentage points. - Sample: all primary issuances of treasuries between 2014 and 2022. ### Insurers’ holdings by asset type (Table D.1) - Notes: classifies insurer holdings of general account bonds based on issuer at end of each year from 2014 to 2022. Government bonds include Federal and Municipal bonds. Shares based on par value of bonds for each issuer category relative to total par value of bonds held. - Holdings by year (percent shares): - 2014: Corporate 57.2% | Govt. 15.8% | Private MBS 15.7% | Agency MBS 8.2% | Other 3.1% - 2015: Corporate 57.7% | Govt. 15.9% | Private MBS 15.2% | Agency MBS 7.7% | Other 3.5% - 2016: Corporate 58.0% | Govt. 16.1% | Private MBS 15.1% | Agency MBS 7.6% | Other 3.2% - 2017: Corporate 58.0% | Govt. 16.1% | Private MBS 14.6% | Agency MBS 7.5% | Other 3.9% - 2018: Corporate 57.8% | Govt. 13.7% | Private MBS 15.3% | Agency MBS 7.4% | Other 5.8% - 2019: Corporate 58.4% | Govt. 13.2% | Private MBS 15.8% | Agency MBS 7.0% | Other 5.6% - 2020: Corporate 58.5% | Govt. 12.4% | Private MBS 16.0% | Agency MBS 5.9% | Other 7.3% - 2021: Corporate 58.9% | Govt. 12.2% | Private MBS 16.5% | Agency MBS 4.9% | Other 7.4% - 2022: Corporate 57.0% | Govt. 11.1% | Private MBS 17.2% | Agency MBS 4.1% | Other 10.7% ### Insurers’ holdings by NAIC rating (Table D.2) - Notes: classifies insurer holdings of corporate bonds 2014–2022 based on NAIC designations (as defined prior to the regulatory change in 2021). Shares based on par value of bonds for each rating category relative to total par value of corporate bonds held. - Holdings by year (percent shares): - 2014: NAIC-1 43.7% | NAIC-2 47.9% | NAIC-3 5.3% | NAIC-4 2.3% | NAIC-5/6 0.8% - 2015: NAIC-1 42.7% | NAIC-2 48.9% | NAIC-3 5.5% | NAIC-4 2.2% | NAIC-5/6 0.8% - 2016: NAIC-1 42.9% | NAIC-2 48.5% | NAIC-3 5.5% | NAIC-4 2.2% | NAIC-5/6 0.9% - 2017: NAIC-1 43.8% | NAIC-2 48.8% | NAIC-3 4.7% | NAIC-4 2.0% | NAIC-5/6 0.7% - 2018: NAIC-1 43.1% | NAIC-2 50.8% | NAIC-3 4.0% | NAIC-4 1.5% | NAIC-5/6 0.6% - 2019: NAIC-1 44.4% | NAIC-2 50.0% | NAIC-3 3.7% | NAIC-4 1.3% | NAIC-5/6 0.6% - 2020: NAIC-1 40.8% | NAIC-2 52.3% | NAIC-3 4.9% | NAIC-4 1.4% | NAIC-5/6 0.5% - 2021: NAIC-1 41.3% | NAIC-2 52.8% | NAIC-3 4.3% | NAIC-4 1.2% | NAIC-5/6 0.4% - 2022: NAIC-1 42.7% | NAIC-2 52.3% | NAIC-3 3.5% | NAIC-4 1.0% | NAIC-5/6 0.4% ### Regression evidence on insurer behavior around monetary policy (Tables D.3–D.8) - Table D.3 (Net purchases by insurers within three months as fraction of total issuance; dependent variable averaged for rating category r): - Sample and columns: All Ratings 2014-2022; Excl COVID 2014-2022; All Ratings NAIC 1-3 2014-2022; Excl COVID NAIC 1-3 2014-2022. - Main independent variable: ∆T2 Yield_MP = change in 2 year Treasury yield in a one hour window around FOMC announcements. - Coefficients and standard errors (in parentheses): - ∆T2 Yield_MP: -0.12 * (0.06); -0.12 (0.08); -0.12 (0.11); -0.13 (0.11) - ∆(Yield): -0.29 (0.27); -0.34 (0.30); 0.51 (0.70); 0.25 (0.88) - ∆(Dtdefault): -0.01 (0.17); 0.05 (0.15); -0.16 (0.18); -0.07 (0.18) - ∆(Duration): 0.81 ∗∗∗ (0.22); 1.00 ∗∗∗ (0.18); 0.67 ∗∗ (0.24); 0.87 ∗∗∗ (0.20) - ∆(Ln(Offering Amount)): -11.47 ∗∗∗ (1.60); -11.23 ∗∗∗ (1.60); -13.15 ∗∗∗ (1.39); -12.90 ∗∗∗ (1.44) - ∆Liquidity: -0.46 (0.49); -0.39 (0.45); -0.97 (0.74); -0.74 (0.68) - Quarter Two: -3.59 ∗∗∗ (0.89); -3.84 ∗∗∗ (1.01); -3.46 ∗∗∗ (0.95); -3.77 ∗∗∗ (1.17) - Quarter Three: -2.86 ∗ (1.38); -2.76 ∗ (1.48); -2.59 (1.68); -2.72 (1.80) - Quarter Four: -11.89 ∗∗∗ (1.79); -11.88 ∗∗∗ (2.00); -13.17 ∗∗∗ (1.89); -13.42 ∗∗∗ (2.14) - Observations: 390, 354, 320, 291 - R^2: 0.687, 0.706, 0.738, 0.752 - SE clustering: Rating + Time - Significance notation: ∗ p <0.10, ∗∗ p <0.05, ∗∗∗ p <0.01 - Table D.4 (Change in excess bond premia at weekly frequency, basis points; bond-level regressions 2014–2022): - Key interaction: Insurer Exposure_{c,w−1} × ∆T2 Yield_MP,w — coefficient reported as -0.00 with standard errors (0.00) in all columns. - Observations: 154,751; 132,145; 143,987; 122,452 - R^2: 0.176, 0.192, 0.275, 0.320 - Issue FE: yes; Time FE: yes; Controls: yes; SE Clustering: Issue + Time - Table D.5 (Change in duration matched treasury spread at weekly frequency, basis points; bond-level regressions 2014–2022): - Key interaction: Insurer Exposure_{c,w−1} × ∆T30 Yield_MP,w - Coefficients and standard errors: - 0.03 ∗∗ (0.01) - 0.04 ∗∗ (0.01) - 0.01 ∗∗ (0.01) - 0.02 ∗∗∗ (0.01) - Observations: 108,053; 100,379; 100,221; 92,890 - R^2: 0.232, 0.241, 0.363, 0.379 - Issue FE: yes; Time FE: yes; Controls: yes; SE Clustering: Issue + Time - Table D.6 (Effect of higher risk weights on sensitivity of insurer relative demand to yield spreads to long-term risk-free rates; 2019–2022, Excl. COVID): - Dependent variable: net purchases by insurers within three months of bond issuance as fraction of total offering amount. - Coefficient for 1{Post} × ∆RBC_r × T30 Spread: - All Ratings: 6.58 ∗∗∗ (2.23) - NAIC 1-3: 8.77 ∗∗∗ (2.44) - All Ratings (alternate col): 4.19 (3.99) - NAIC 1-3 (alternate col): 8.23 ∗ (4.32) - Observations: 1,573; 1,132; 1,550; 1,119 - R^2: 0.45 (all columns) - Time FE: yes; Controls: yes; SE Clustering: Issuer + Time - Notes: T30 spread expressed in percentage points; RBC_r expressed in percentage points. - Significance notation: ∗ p <0.10, ∗∗ p <0.05, ∗∗∗ p <0.01 - Table D.7 (Change in excess bond premia at weekly frequency, basis points; bond-level regressions 2019–2022, Excl. COVID): - Coefficient for 1{Post} × ∆RBC_r × ∆T2 Yield_MP: - All Ratings: -0.22 (0.16) - Excl. COVID All Ratings: -0.22 (0.26) - NAIC 1-3: 0.00 (0.07) - Excl. COVID NAIC 1-3: 0.12 (0.27) - Observations: 93,847; 69,717; 89,254; 64,758 - R^2: 0.14; 0.20; 0.24; 0.32 - Issue FE: yes; Time FE: yes; Controls: yes; SE Clustering: Issue + Time - Table D.8 (Change in duration matched treasury spread at weekly frequency, basis points; bond-level regressions 2019–2022, Excl. COVID): - Coefficient for 1{Post} × ∆RBC_r × ∆T30 Yield_MP: - All Ratings: 1.19 ∗ (0.58) - Excl. COVID All Ratings: 1.59 ∗∗ (0.72) - NAIC 1-3: 0.64 ∗∗ (0.30) - Excl. COVID NAIC 1-3: 1.06 ∗∗ (0.35) - Observations: 61,320; 53,119; 58,396; 50,615 - R^2: 0.22; 0.26; 0.27; 0.30 - Issue FE: yes; Time FE: yes; Controls: yes; SE Clustering: Issue + Time - Significance notation: ∗ p <0.10, ∗∗ p <0.05, ∗∗∗ p <0.01 ### Key empirical implications from the presented tables and figures - Life insurers exhibit short duration exposure as summarized by the rolling γ coefficient in the weekly regressions against 10 year Treasury returns. - Insurers’ relative demand for long-term Treasury bonds increases with offering yields (Figure C.1 sample 2014–2022). - Corporate bonds consistently constitute the largest share of insurers’ general account bond holdings (roughly high-50s percent across 2014–2021; 57.0% in 2022). - The share of corporate bonds rated NAIC-1 and NAIC-2 together comprises the vast majority of corporate holdings (e.g., NAIC-1 plus NAIC-2 = 91.6% in 2014; 95.0% in 2022). - Around FOMC announcements, changes in short-term Treasury yields (∆T2 Yield_MP and ∆T30 Yield_MP) and insurer exposure interact with bond-level outcomes: small or statistically indistinct point estimates for ∆T2 interactions in risk premia regressions (Table D.4), but positive and statistically significant interactions for ∆T30 in credit-spread regressions (Table D.5 and D.8). - Regulatory changes in risk-based capital weights (∆RBC_r) are associated with larger insurer sensitivity of relative demand to T30 spreads post-reform (Table D.6). *Source: wpiea2025054-print-pdf* --- _Source: https://www.imf.org/-/media/files/publications/wp/2025/english/wpiea2025054-print-pdf.pdf_