## wpiea2022202-print-pdf

## Source details

**Canonical URL:** [wpiea2022202-print-pdf](https://www.imf.org/-/media/files/publications/wp/2022/english/wpiea2022202-print-pdf.pdf)

## Other formats

- [Markdown version](/-/media/files/publications/wp/2022/english/wpiea2022202-print-pdf.pdf.md)
- [Structured JSON version](/-/media/files/publications/wp/2022/english/wpiea2022202-print-pdf.pdf.json)

---

### I. Motivation
- European policy rates were dramatically slashed in late 2008 and again between 2011 and 2014, leading to historically ultra-low policy rates of less than 1 percent.
- Central banks implemented unconventional monetary policies, notably large acquisitions of sovereign and corporate bonds, producing a “low-for-long” era.
- Common perception: worsening quality of insurers’ bond portfolios during low-for-long was driven principally by a more aggressive search for yield.
- Alternative/additional factors highlighted:
  - A multi-year surge in corporate and sovereign rating downgrades following the global financial and European debt crises.
  - Mechanical upward revaluation of bond holdings as interest rates fell.
  - Fundamental transformation of the regulatory framework and incentives around 2014 with preparations for Solvency II (implementation date 1 January 2016) that may have affected portfolio decisions prospectively.
- Paper’s main contribution: track insurers’ bond investments from before the low-for-long until recently and quantify the impact of rating migrations, bond revaluations, and hunt for yield/regulatory incentives to assess whether life insurers became more risk seeking.

### II. The changing credit quality of insurers’ bond portfolios, 2005–2021
- Sample composition:
  - 56 European life insurance firms from Belgium (1), Finland (1), France (3), Germany (9), Italy (4), Netherlands (2), Norway (5), Portugal (2), Spain (2), Sweden (5), Switzerland (6), United Kingdom (15).
  - Sample covers around ¾ of the total European life insurance sector.
- Key rating-allocation changes:
  - Share of AAA bonds: dropped from 43 percent of total bond holdings in 2005 to 15 percent in 2021.
  - Holdings of BBB paper: rose from around 5 to 6 percent in 2005-08 to 22 percent of the entire bond portfolio in recent years.
  - Deterioration in credit quality proceeded slowly in 2006-2011, accelerated markedly in 2012 and 2013, and stabilized thereafter.
  - Variation of AA, A, and sub-BBB and unrated bonds was much smaller.
- Main question: what drove substitution of AAA for BBB bonds — net purchases (hunt for yield), rating migrations, or bond revaluations under falling yields and changing regulatory incentives (Solvency II)?

### Rating downgrades: evidence and magnitude
- Downgrades increased in the aftermath of the global financial and European debt crises in 2009–2012.
- Mechanical effect: downgrades increase BBB holdings for buy-and-hold investors.
- Transition-rate comparisons (summary):
  - For 1981-2009, downgrades from AAA to BBB were non-existent.
  - For 2010-2014, downgrades from AAA to BBB averaged 1.0 percent of all AAA-rated bonds per year.
  - Downgrades from AA to BBB were 0.8 p.p. higher in 2010-2014 than in 1981-2009.
  - Downgrades from A to BBB were 2.5 p.p. higher in 2010-2014 compared to 1981-2009.
- Sovereign downgrades to BBB during 2009–2013 included Greece (2009), Ireland (2011), Cyprus (2011), Portugal (2011), Italy (2012), and Spain (2012).
- Insurer behavior evidence:
  - A 2019 EIOPA survey: almost 80 percent of European life insurers self-declared as buy-and-hold investors.
  - EIOPA (2017): 40 percent of respondents claimed decrease in average investment grade stemmed from rating changes.
  - EIOPA (2020): insurers sold downgraded bonds in 2016Q1 to 2020Q2 at a quarterly average of 3.7 percent of downgraded bonds—sales “remained largely contained.”

### Bond revaluations due to falling yields
- Falling interest rates mechanically increase the market value of bond holdings; the increase is larger the larger the drop in yields.
- For 2009-2020 cumulative yield declines were larger for BBB bonds than for AAA, AA, and A bonds.
- Bond yield definition: yield-to-maturity of the euro-denominated Bank of America/Merrill Lynch bond index for each rating.

### Regulatory incentives and Solvency II
- Preparations for Solvency II began in 2014; Solvency II came into effect on 1 January 2016 but rules and requirements were prospectively applied well before 2016.
- Under Solvency II, capital charges for bond investments depend mainly on bond rating and duration; insurers compare bond yield with capital cost when deciding holdings.
- Key empirical observation:
  - Before 2014, the ratio ‘yield/capital cost’ was notably larger (more favorable) for highly rated bonds than for BBB-rated bonds.
  - From 2014, the difference effectively disappeared, suggesting that under ultra-low interest rates capital savings from holding higher-rated bonds did not compensate for lower yields of safer assets.
- Calculation notes:
  - Bond yields: yield-to-maturity of the euro-denominated Bank of America/Merrill Lynch corporate bond index for bonds with 5 to 10 year maturity.
  - Solvency II capital charges: average for bonds with maturities between 5 to 10 years, by rating class.

### III. Quantification of the portfolio drivers
- Law of motion for bond holdings (descriptive):
  - Holdings H_t evolve from rating migrations T_{t-1}, bond valuation changes (function of duration D_{t-1} and interest-rate change Δi_t), and net purchases P_t.
  - Notation includes h_{t}^a, τ_{a,b,t-1}, δ_{t-1}^a, (i_{t}^a − i_{t-1}^a).
- Purpose: quantify impact of rating migrations and valuation changes separately from net purchases to identify contribution of hunt for yield/regulatory incentives.

### IV. Data and key assumptions
- Insurers’ bond investments by rating: S&P Global Capital IQ.
- Corporate vs sovereign split: S&P Capital IQ does not provide the split; EIOPA data used for 2016-2020.
  - EIOPA 2016-2020: sovereign/corporate split oscillated within a narrow band: 51-54 percent sovereign and 49-46 percent corporate for insurers in sample.
  - Industry discussions indicated split constant through time; authors extrapolated the average split for 2016-2020 to 2005-2015.
- Annual rating transition matrices: corporate bonds from Standard & Poor’s; sovereign transitions from Fitch Ratings.
- Duration data: available only for EIOPA (2014) and EIOPA (2018); average applied to all years:
  - Corporate bond duration: 5.7 years
  - Sovereign bond duration: 8.0 years
- Interest-rate changes by rating category:
  - Corporate bond changes from ICE’s corporate bond indices (maturity bucket chosen to match assumed durations).
  - Sovereign yield changes from iBoxx euro-denominated sovereign indices (maturity bucket chosen to match assumed durations).
- Section V presents robustness checks for the duration and other assumptions.

### V. Decomposition results: drivers of BBB bond holdings (2009–2020)
- Portfolio dynamics arise from multiple drivers; net bond purchases are only one element and not always dominant (years where purchases were not dominant include 2009, 2011, 2014, 2015, 2020).
- Example year 2010:
  - Life insurers sold BBB bonds equivalent to 1.8 p.p. of their bond portfolio.
  - Rating migrations added around 2.1 p.p. of BBB bonds.
  - Net annual change in BBB holdings in 2010: +0.5 p.p.
- Rating migrations and downgrades:
  - Rating downgrades added about 2 to 3 p.p. of BBB paper per year during 2009-2012.
  - Cumulative net accumulation due to downgrades by 2013: 10.9 p.p.
  - Insurers reacted by selling some of these bonds in 2009-2011 to maintain allocations.
- Rating cycle reversal:
  - From 2014 onwards, upgrades overcompensated downgrades, especially for sovereigns.
  - Sovereign rating cycle fully closed by 2018; corporate cycle still unwinding in 2020.
  - Insurers responded by net-buying BBB bonds, especially in 2016-2018.
- Net purchases and “low-for-long”:
  - Net purchases of BBB bonds took place from 2012 onwards.
  - Annual net purchases during 2012-2018 oscillated between -1.5 and 1.8 p.p.
  - Cumulative net acquisition of BBB bonds by 2018: 6.9 p.p. of the bond portfolio.
- Revaluations:
  - Upward bond revaluations were an important driver in 2012-2015, especially for sovereign bonds as yields on BBB paper fell from 2011 highs.
- Peak accumulation and contributions by 2017:
  - Peak accumulation of BBB bonds was in 2017; total cumulative increase by 2017: 17.2 p.p.
  - Contributions to the 17.2 p.p. cumulative increase by 2017:
    - Net purchases: 5.2 p.p. (30 percent)
    - Rating migrations: 5.8 p.p. (34 percent)
    - Revaluations: 6.2 p.p. (36 percent)

### Robustness checks — corporate vs sovereign bond split
- Baseline assumption: insurers allocated around 52 percent of their bond portfolio to sovereigns throughout 2016-2020; same allocation assumed for 2005-2015.
- Extreme distributions analyzed for 2005-2015:
  - Sovereign bias distribution: 70-30 percent (sovereign-corporate).
  - Corporate bias distribution: 30-70 percent (sovereign-corporate).
- Findings under extremes:
  - Sovereign bias: stronger impact of sovereign downgrades and revaluations on BBB holdings; bond purchases account for only one-fifth of the cumulative change by 2017.
  - Corporate bias: net purchases account for slightly over two-fifths of the cumulative change by 2017; corporate re-ratings and revaluations are next most important.
- Non-uniform distribution scenario (sovereign concentrated in higher ratings; corporate concentrated in lower ratings) using Table 3 distribution (Sov % / Corp %):
  - AAA: 80% / 20%
  - AA: 80% / 20%
  - A: 20% / 80%
  - BBB: 20% / 80%
  - NR: 0% / 100%
  - Under this scenario, net purchases account for 38 percent of the total accumulation until 2017 (i.e., 6.5 p.p. out of 17.2 p.p.).
- Aggregate robustness outcome:
  - Across distributions, net purchases would account for at most one-half of the total change and, in the most extreme corporate-bias case, up to 44 percent of the cumulative change by 2017.
  - Passive drivers (rating migrations and revaluations) together account for one-half or more of the change, consistent with a “buy-and-hold” investor behavior.

### Robustness checks — bond duration assumption
- Alternative to the constant-average-duration assumption: a linear extrapolation from the two available duration years (2014 and 2018) was considered.
- Result: results using the linear extrapolation are very close to those using the average-duration assumption; no substantive change to main findings.

### Concluding remarks and policy implications
- Portfolio rebalancing between 2005 and 2017:
  - AAA share fell from 43 percent to about 18 percent.
  - BBB share rose from about 5 percent to 23 percent.
  - The rebalancing has remained largely constant since 2017.
- Interpretation of insurers’ behavior:
  - Insurers actively increased BBB holdings via net purchases starting in 2012, but rating migrations and bond revaluations were equally important—each accounting for about 1/3 of the cumulative change between 2008 and 2017.
  - Even in extreme scenarios, net purchases would account for at most two-fifths of the total deterioration in credit quality during this period.
  - Passive drivers (rating migrations and revaluations) were equally responsible for the changing allocation.
- Financial stability perspective:
  - Insurers largely retained a “buy-and-hold” investment stance despite ultra-low yields and widespread downgrades.
  - By not actively rebalancing portfolios en masse, insurers avoided aggravating fire-sale dynamics and additional stress for borrowers.
  - Evidence of contained net-selling of downgraded bonds during the COVID-19 crisis corroborates insurers’ relative stability (EIOPA, 2021).
- Regulatory question:
  - Are the capital costs embedded in Solvency II strong enough to promote safer investment allocations through the cycle?
  - Findings suggest:
    - Before 2014, the yield–capital-charge relationship favored investments in higher-quality assets.
    - From 2014 on, the return/capital-cost trade-off was not different across ratings, making higher-yielding assets more attractive.

*Source: IMF Working Paper — Section V (Results) and subsequent robustness checks and concluding remarks.*

### References .............................................................................................................

### wpiea2022202-print-pdf - References

### I. Motivation
- European policy rates were dramatically slashed in late 2008 and again between 2011 and 2014, leading to historically ultra-low policy rates of less than 1 percent.
- Central banks implemented unconventional monetary policies, notably large acquisitions of sovereign and corporate bonds, producing a “low-for-long” era.
- Institutional investors, confronted with anemic returns, were suspected to move into riskier and less liquid asset classes; observers documented increases in insurers’ investments into less liquid assets and riskier bonds and a lengthening of duration (IMF, 2015 and 2017; Domanski et al. 2017).
- Common perception: worsening quality of insurers’ bond portfolios during low-for-long was driven principally by a more aggressive search for yield.
- Alternative/additional factors that could explain portfolio changes:
  - A multi-year surge in corporate and sovereign rating downgrades following the global financial and European debt crises.
  - Mechanical upward revaluation of bond holdings as interest rates fell.
  - Fundamental transformation of the regulatory framework and incentives around 2014 with preparations for Solvency II (implementation date 1 January 2016) that may have affected portfolio decisions prospectively.
- Paper’s main contribution: track insurers’ bond investments from before the low-for-long until recently and quantify the impact of rating migrations, bond revaluations, and hunt for yield/regulatory incentives to assess whether life insurers became more risk seeking.
- Structure: Section II documents changes to credit quality; Section III checks robustness via alternative assumptions; Section IV concludes.

### II. The Changing Credit Quality of Insurers’ Bond Portfolios, 2005-2021
- Sample:
  - 56 European life insurance firms from Belgium (1), Finland (1), France (3), Germany (9), Italy (4), Netherlands (2), Norway (5), Portugal (2), Spain (2), Sweden (5), Switzerland (6), United Kingdom (15).
  - Sample covers around ¾ of the total European life insurance sector.
- Key findings on rating allocation (Figure 1 summary):
  - Share of AAA bonds: dropped from 43 percent of total bond holdings in 2005 to 15 percent in 2021.
  - Holdings of BBB paper: rose from around 5 to 6 percent in 2005-08 to 22 percent of the entire bond portfolio in recent years.
  - Deterioration in credit quality proceeded slowly in 2006-2011, accelerated markedly in 2012 and 2013, and stabilized thereafter.
  - Variation of AA, A, and sub-BBB and unrated bonds was much smaller.
- Question posed: what drove substitution of AAA for BBB bonds?
  - Widely accepted explanation: ultra-low policy rates since 2008 pushed insurers to hunt for higher returns in riskier assets.
  - Other possible contributors: rating downgrades, bond revaluations (due to falling yields), changes to regulatory incentives from Solvency II.

### Rating downgrades (evidence and magnitude)
- Rating downgrades increased in aftermath of global financial and European debt crises in years 2009 to 2012.
- Mechanical effect: downgrades increase BBB holdings for buy-and-hold investors like life insurers.
- Table 1 (summary comparisons of average one-year corporate rating transition rates):
  - For 1981-2009, downgrades from AAA to BBB were non-existent.
  - For 2010-2014, downgrades from AAA to BBB averaged 1.0 percent of all AAA-rated bonds per year.
  - Downgrades from AA to BBB were 0.8 p.p. higher in 2010-2014 than in 1981-2009.
  - Downgrades from A to BBB were 2.5 p.p. higher in 2010-2014 compared to 1981-2009.
- Examples of sovereign downgrades to BBB during 2009-2013 included Greece (in 2009), Ireland (in 2011), Cyprus (in 2011), Portugal (in 2011), Italy (in 2012), and Spain (in 2012).
- Evidence on insurer behavior:
  - A 2019 EIOPA survey: almost 80 percent of European life insurers self-declared as buy-and-hold investors.
  - EIOPA (2017): 40 percent of respondents claimed decrease in average investment grade stemmed from rating changes.
  - EIOPA (2020): insurers sold downgraded bonds in 2016Q1 to 2020Q2 at a quarterly average of 3.7 percent of downgraded bonds—sales “remained largely contained.”

### Bond revaluations due to falling yields
- Falling interest rates mechanically increase the market value of bond holdings.
- The increase in bond values was greater the larger the drop in yields; for 2009-2020 cumulative yield declines were larger for BBB bonds than for AAA, AA, and A bonds (Figure 2: cumulative change in bond yields since end-2008, in basis points, by bond rating).
- Note: bond yields are yield-to-maturity of the euro-denominated Bank of America/Merrill Lynch bond index for each rating.

### Regulatory incentives and Solvency II
- Preparations for Solvency II began in 2014; Solvency II came into effect on 1 January 2016 but rules and requirements were prospectively applied well before 2016.
- Under Solvency II, capital charges for bond investments depend mainly on bond rating and duration; insurers compare bond yield with capital cost when deciding holdings.
- Intuitive metric: ratio of bond yields to capital surcharges.
  - Before 2014, the ratio ‘yield/capital cost’ was notably larger (more favorable) for highly rated bonds than for BBB-rated bonds.
  - From 2014, the difference effectively disappeared (Figure 3 insert), suggesting that under ultra-low interest rates capital savings from holding higher-rated bonds did not compensate for lower yields of safer assets.
- Notes on calculation:
  - Bond yields: yield-to-maturity of the euro-denominated Bank of America/Merrill Lynch corporate bond index for bonds with 5 to 10 year maturity (matches average maturity in insurers’ bond portfolios).
  - Solvency II capital charges: calculated as the average for bonds with maturities between 5 to 10 years, for each rating class.

### III. Quantification of the Portfolio Drivers
- Law of motion for bond holdings (descriptive formulation):
  - H_t (N x 1 vector) of holdings at time t equals rating migrations at t-1 (T_t matrix) plus bond valuation changes between t-1 and t (function of duration D_{t-1} and change in interest rates Δi_t) plus net purchases P_t at time t.
  - Notation specified: h_{t}^a denotes amount of bonds rated a at time t; τ_{a,b,t-1} denotes percent migrating from rating a at t-1 to rating b at t; δ_{t-1}^a denotes duration of bonds rated a at t-1; (i_{t}^a − i_{t-1}^a) denotes change in interest rates for bonds rated a between t and t-1.
- Purpose: quantify impact of rating migrations and valuation changes separately from net purchases to identify contribution of hunt for yield/regulatory incentives.

### IV. Data
- Data sources and key data handling choices:
  - Insurers’ bond investments by rating: S&P Global Capital IQ.
  - S&P Capital IQ does not provide corporate vs. sovereign split; EIOPA data used to estimate split (available only for 2016-2020).
  - EIOPA 2016-2020 data show sovereign/corporate split oscillated within a narrow band: 51-54 percent sovereign and 49-46 percent corporate for insurers in sample.
  - Industry discussions indicated split constant through time.
  - Lacking other data, authors extrapolate the average split for 2016-2020 to 2005-2015.

*International Monetary Fund: content as provided in the source PDF.*

### Section V

### Section V

### Data and key assumptions
- Annual rating transition matrices: corporate bonds from Standard & Poor’s; sovereign transitions from Fitch Ratings.
- Duration data: only available from EIOPA (2014) and EIOPA (2018). The average of the two years was computed and applied to all years in the sample:
  - Corporate bond duration: 5.7 years
  - Sovereign bond duration: 8.0 years
- Interest-rate changes by rating category:
  - Corporate bond changes from ICE’s corporate bond indices (maturity bucket chosen to match the assumed durations).
  - Sovereign yield changes from iBoxx euro-denominated sovereign indices (maturity bucket chosen to match the assumed durations).
- Section V presents robustness checks for the duration and other assumptions.

### Decomposition results: drivers of BBB bond holdings (2009–2020)
- Portfolio dynamics arise from multiple drivers; net bond purchases are only one element and not always dominant (examples of years where purchases were not dominant: 2009, 2011, 2014, 2015, 2020).
- Example interaction of drivers in 2010:
  - Life insurers sold BBB bonds equivalent to 1.8 percentage points (p.p.) of their bond portfolio.
  - Rating migrations added around 2.1 p.p. of BBB bonds.
  - Net annual change in BBB holdings in 2010: +0.5 p.p.
- Rating migrations and downgrades:
  - Rating downgrades added about 2 to 3 p.p. of BBB paper per year during 2009-2012.
  - Cumulative net accumulation due to downgrades by 2013: 10.9 p.p.
  - Insurers reacted by selling some of these bonds in 2009-2011 to maintain allocations.
- Rating cycle reversal:
  - From 2014 onwards, upgrades overcompensated downgrades, especially for sovereigns.
  - Cumulative impact of re-ratings on the BBB portfolio started to reverse from 2014.
  - Sovereign rating cycle was fully closed by 2018; corporate cycle still unwinding in 2020.
  - Insurers responded by net-buying BBB bonds, especially in 2016-2018.
- Net purchases and “low-for-long”:
  - Net purchases of BBB bonds took place from 2012 onwards.
  - Annual net purchases during 2012-2018 oscillated between -1.5 and 1.8 p.p.
  - Cumulative net acquisition of BBB bonds by 2018: 6.9 p.p. of the bond portfolio.
- Revaluations:
  - Upward bond revaluations were an important driver in 2012-2015, especially for sovereign bonds as yields on BBB paper fell from 2011 highs.
- Peak accumulation and contributions by 2017:
  - Peak accumulation of BBB bonds was in 2017; total cumulative increase by 2017: 17.2 p.p.
  - Contributions to the 17.2 p.p. cumulative increase by 2017:
    - Net purchases: 5.2 p.p. (30 percent)
    - Rating migrations: 5.8 p.p. (34 percent)
    - Revaluations: 6.2 p.p. (36 percent)

### Robustness checks — corporate vs sovereign bond split
- Baseline assumption: insurers allocated around 52 percent of their bond portfolio to sovereigns throughout 2016-2020; same allocation was assumed for 2005-2015 due to lack of data.
- Extreme distributions analyzed for 2005-2015:
  - Sovereign bias distribution: 70-30 percent (sovereign-corporate).
  - Corporate bias distribution: 30-70 percent (sovereign-corporate).
- Findings under extreme distributions:
  - Sovereign bias: stronger impact of sovereign downgrades and revaluations on BBB holdings; bond purchases account for only one-fifth of the cumulative change by 2017.
  - Corporate bias: net purchases account for slightly over two-fifths of the cumulative change by 2017; corporate re-ratings and revaluations are next most important.
- Non-uniform distribution scenario (sovereign concentrated in higher ratings; corporate concentrated in lower ratings):
  - Table 3 distribution used (Sov % / Corp %):
    - AAA: 80% / 20%
    - AA: 80% / 20%
    - A: 20% / 80%
    - BBB: 20% / 80%
    - NR: 0% / 100%
  - Under this scenario, net purchases account for 38 percent of the total accumulation until 2017 (i.e., 6.5 p.p. out of 17.2 p.p.).
- Aggregate robustness outcome:
  - Across distributions, net purchases would account for at most one-half of the total change and, in the most extreme corporate-bias case, up to 44 percent of the cumulative change by 2017.
  - Passive drivers (rating migrations and revaluations) together account for one-half or more of the change, consistent with a “buy-and-hold” investor behavior.

### Robustness checks — bond duration assumption
- Alternative to the constant-average-duration assumption:
  - A linear extrapolation from the two available duration years (2014 and 2018) was considered.
- Result:
  - The results obtained using the linear extrapolation are very close to those obtained with the average-duration assumption; no substantive change to main findings.

### Concluding remarks and policy implications
- Portfolio rebalancing between 2005 and 2017:
  - AAA share fell from 43 percent to about 18 percent.
  - BBB share rose from about 5 percent to 23 percent.
  - The rebalancing has remained largely constant since 2017.
- Interpretation of insurers’ behavior:
  - While insurers actively increased BBB holdings via net purchases starting in 2012, rating migrations and bond revaluations were equally important—each accounting for about 1/3 of the cumulative change between 2008 and 2017.
  - Even in extreme scenarios, net purchases would account for at most two-fifths of the total deterioration in credit quality during this period.
  - Passive drivers (rating migrations and revaluations) were equally responsible for the changing allocation.
- Financial stability perspective:
  - Insurers largely retained a “buy-and-hold” investment stance despite ultra-low yields and widespread downgrades.
  - By not actively rebalancing portfolios en masse, insurers avoided aggravating fire-sale dynamics and additional stress for borrowers.
  - Evidence of contained net-selling of downgraded bonds during the COVID-19 crisis corroborates insurers’ relative stability (EIOPA, 2021).
- Regulatory question raised:
  - Are the capital costs embedded in Solvency II strong enough to promote safer investment allocations through the cycle?
  - Findings suggest:
    - Before 2014, the yield–capital-charge relationship favored investments in higher-quality assets.
    - From 2014 on, the return/capital-cost trade-off was not different across ratings, making higher-yielding assets more attractive.

*Source: IMF Working Paper — Section V (Results) and subsequent robustness checks and concluding remarks.*

---


_Source: https://www.imf.org/-/media/files/publications/wp/2022/english/wpiea2022202-print-pdf.pdf_