Introduction
Arora, Merali and Ouellet Leblanc (2018) document the growing role that Canadian corporate bond funds (CCBFs) play in channelling savings to investments through capital markets. In this note, we analyze the liquidity-management decisions of CCBFs: how do fund managers meet the redemption requests of investors? Liquidity-management decisions are important because fixed-income funds offer investors the right to daily redemptions, but fund managers invest in relatively less-liquid assets.
Our analysis produces one core finding: liquidity-management decisions depend on market conditions. We find that CCBFs tend to use liquid holdings (cash and government bonds) to meet investor redemptions when volatility is low. When volatility is high, however, CCBFs continue to use liquid holdings but also sell less-liquid assets to maintain the liquidity of their portfolios.
On the one hand, using liquid holdings or selling assets to meet redemptions reflects normal day-to-day liquidity-management decisions by fund managers. On the other hand, the sale of less-liquid assets (corporate bonds) in times of stress could exacerbate market volatility and illiquidity if redemptions are large. When volatility rises, liquidity providers are more reluctant to supply liquidity and the cost of selling less-liquid assets increases (Dick-Nielsen, Feldhütter and Lando 2012; Nagel 2012).
Our results are consistent with the existing concern that less-liquid mutual funds can amplify shocks in the financial system (IMF 2015). This concern is relevant for Canada because Arora, Merali and Ouellet Leblanc (2018) show that CCBFs have become larger and increased their holdings of riskier assets since 2007. The larger size of CCBFs suggests that both the potential size of redemptions and sales of corporate bonds have increased, with higher potential impact on the liquidity of fixed-income markets. Holding riskier assets also suggests that the likelihood of large redemptions has increased, which we discuss below.
Strategies fund manager use to meet investor redemptions
When investors redeem their fund shares for cash, fund managers can choose whether to draw on their liquid holdings or sell less-liquid assets.1 Using liquid holdings can be described as slicing the portfolio horizontally. Alternatively, selling assets proportional to their investment allocation can be described as slicing the portfolio vertically. Figure 1 illustrates horizontal and vertical slicing in a hypothetical portfolio.
Horizontal slicing lowers the share of cash and government bonds held by a fund in the subsequent period but raises the share of corporate bonds (Figure 1a). This strategy reduces the liquidity of the fund portfolio. However, vertical slicing maintains the asset allocation and the liquidity of the portfolio (Figure 1b).
Figure 1: Fund managers can adopt different strategies to meet investor redemptions
Figure 1a: Horizontal slicing strategy
Figure 1b: Vertical slicing strategy
Note: We assume a fund facing 15 per cent of redemptions at time t. The dashed bar represents the amount sold for a given asset class.
CCBFs’ liquidity-management decisions between horizontal and vertical slicing reflect a trade-off. Using cash to meet redemptions reduces transaction costs and implies that assets remaining in the portfolio have higher expected returns. However, using cash holdings increases the liquidity mismatch between assets and liabilities, strengthening the first-mover advantage for investors who remain with the funds (Chen, Goldstein and Jiang 2010). This strengthening of the first-mover advantage increases the risk of redemptions (Arora forthcoming). If investors expect costly sales of assets to meet redemptions, this can also increase the vulnerability of mutual funds to “bank run.”
Whether funds favour horizontal slicing to meet redemptions remains highly debated, with empirical support for both horizontal and vertical slicing. Chernenko and Sunderam (2016) find that US equity funds tend to use horizontal slicing to meet redemptions. Jiang, Li and Wang (2016) validate this result for US corporate bond funds but find vertical slicing during periods of high market uncertainty. Morris, Shim and Shin (2017) find that global bond funds sell a greater share of less-liquid assets both to meet redemptions and to increase cash holdings (cash-hoarding strategy).
CCBFs are more likely to use horizontal slicing
In this section, we examine whether CCBFs, on average, choose a horizontal or vertical-slicing strategy to meet investor redemptions. Following Jiang, Li and Wang (2016), we estimate how fund holdings change after redemption claims and flows. The appendix contains more details about the data and estimation.
Our results indicate that, on average, CCBF managers first use liquid holdings (horizontal slicing) to meet investor redemptions. Chart 1 shows that the share of government bonds decreases by 31 basis points (bps) following an increase of one standard deviation in redemptions, but the share of corporate bonds increases by 26 bps.
Chart 1: Liquid assets tend to be the primary liquidity tool for CCBFs to manage redemptions
Note: *** indicates statistical significance at 1%
Sources: Bloomberg, Morningstar, Thomson Reuters and Bank of Canada calculations
Last observation: 2016Q4
This result is in line with the work of Jiang, Li and Wang (2016) on US corporate bond funds. On average, CCBFs favour horizontal slicing to meet redemptions. But the CCBFs’ strategy for meeting redemptions may change when fund managers assign greater probability to a loss in asset values. In this case, horizontal slicing strengthens the first-mover advantage and increases the risk of self-fulfilling redemptions. Greater downside risks to asset prices could shift the trade-off toward vertical slicing, which maintains the liquidity of the portfolio.
CCBFs’ liquidity-management decisions are influenced by market conditions
In this section, we examine whether market conditions influence the decision of CCBFs to use horizontal or vertical slicing to meet redemptions. We re-estimate our baseline model when volatility is low or high, respectively (equation 2 in the appendix). We use the volatility index (VIX) to proxy for downside risk. VIX is the volatility implied in the prices of equity options.
Our results indicate that, on average, CCBFs choose horizontal slicing when volatility is low. However, we find that, on average, CCBFs tend to use vertical slicing when volatility is high. Similar to our previous results, Chart 2 shows that when volatility is low the share of cash holdings declines by 22 bps and the share of corporate bonds increases by 27 bps. However, the share of changes to holdings shows no significant difference when volatility is high, suggesting that fund managers tend to sell assets proportional to their initial holdings (vertical slicing).
Chart 2: Liquid assets tend to be the primary liquidity tool for CCBFs to manage redemptions
Note: *** indicates statistical significance at 1%. High volatility results are not statistically significant.
Sources: Bloomberg, Morningstar, Thomson Reuters and Bank of Canada calculations
Last observation: 2016Q4
This shift by CCBFs from horizontal to vertical slicing suggests that the trade-off changes under different market conditions. Redemption risk increases when volatility rises because there is a higher likelihood of negative fund returns. In this case, managers of CCBFs prefer vertical slicing to maintain the liquidity of their portfolios. This finding is in line with the conclusions of Jiang, Li and Wang (2016) on US corporate bond funds and survey results of UK asset managers (BoE 2015), which concluded that fund managers also sell less-liquid assets to meet redemptions in times of stress.
Conclusion
In this note, we find that CCBFs tend to use liquid holdings to meet investor redemptions during periods of low volatility. However, we find that during periods of high volatility, CCBFs continue to use liquid holdings but also sell less-liquid assets to maintain the liquidity of their portfolios.
On the one hand, using liquid holdings or selling assets to meet redemptions reflect normal day-to-day liquidity management decisions by fund managers. On the other hand, our results have implications for the stability of bond markets. The sale of less-liquid assets (corporate bonds) in times of stress could exacerbate volatility and illiquidity if redemptions are large. Our findings are particularly relevant for Canada given the recent growth in both size and riskiness of CCBF exposures (Arora, Merali and Ouellet Leblanc 2018). Overall, this note is an important contribution to building stress tests that quantify the likelihood and potential impact of CCBF asset sales on the financial system.
Arora, Fan and Ouellet Leblanc (forthcoming) will further investigate the key drivers of the liquidity-management decisions to meet redemptions, assessing the benefits of decision trees and random forest algorithms against those of the traditional models used in this note.
Appendix
This note uses both the definition and sample of CCBFs presented in Arora, Merali and Ouellet Leblanc (2018). Table A1 describes the variables included in our empirical analysis.
Table A1: Variables and data sources
Table A1: Variables and data sources | |||
---|---|---|---|
Variables | Description | Source | Computation and Units |
Share of cash holdings | Share of cash and equivalents holdings in book value | Morningstar Holdings and Thomson Reuters fixed-income reference data | In percentage points. Cash and equivalents include cash holdings and certificates of deposit. |
Share of government bond holdings | Share of government bond holdings in book value | Morningstar Holdings and Thomson Reuters fixed-income reference data | All government bonds held by a fund are matched with the Thomson Reuters fixed-income reference data to obtain the book value of each security. We calculate the asset class holdings by aggregating all government bonds securities at each period. We then compute its share of the portfolio in percentage points. |
Share of corporate bond holdings | Share of corporate bond holdings in book value | Morningstar Holdings and Thomson Reuters fixed-income reference data | All corporate bonds held by a fund are matched with the Thomson Reuters fixed-income reference data to obtain the book value of each security. We calculate the asset class holdings by aggregating all corporate bonds securities at each period. We then compute its share of the portfolio in percentage points. |
Inflows | Net inflows as a percentage of total net assets | Morningstar Direct | In percentage points. Inflows are winsorized at the 1 and 99 percentiles. |
Outflows | Net outflows (redemptions) as a percentage of total net assets | Morningstar Direct | In percentage points. Inflows are winsorized at the 1 and 99 percentiles. |
Fund return | Quarterly net return | Morningstar Direct | In percentage points. |
Fund age | Age of the fund since inception | Morningstar Direct | In years. |
Total net assets | Fund-level total net assets | Morningstar Direct | Log of total net assets. |
Term spreads | Yield on 10-year Canadian government bonds minus yield on 3-month Treasury bills | Bloomberg | In percentage points. |
Credit spreads | Option-adjusted spreads of the Bank of America Merrill Lynch Canadian corporate bond index | Bloomberg | In percentage points. |
VIX | Chicago board options exchange volatility index | Bloomberg | In percentage points. |
Low versus high volatility | Dummy to determine market conditions | Bloomberg | Dummy takes a value of 1 if the VIX is below its 2-year historical median (low volatility), and is 0 otherwise (high volatility). |
Table A2 reports the results from equation 1 on how funds’ holdings change following inflows and outflows. We estimate the model on quarterly data from March 2002 to December 2016. Equation 1 is estimated using the Arellano and Bond (1991) generalized method of moments estimator. This estimator uses instrument variables to correct for potential endogeneity bias. Also reported are (i) the results of a Sargan-Hansen test of the validity of the overidentifying restrictions, and (ii) the results of an m2 test to check that there is no second-order autocorrelation in the residual.
Equation 1: Dynamic panel estimation
\( Share \, of \, asset \, holdings(\%)_{i,t} \)\( =α_i\)\( +β_1 Share\,of\,asset \, holdings(\%)_{i,t-1}\)\( +β_2 Inflows_{i,t}\)\( +β_3 Outflows_{i,t}\)\( +γ' Controls_{i,t-1}\)\( +ε_{i,t} \)
The coefficient of interest is \(β_3\) since it captures the effects of investor redemptions on CCBF liquidity management practices. For brevity, the coefficient estimates of the control variables are not shown in Table A2. Because CCBFs can change their asset allocation according to valuation changes, we include lagged (i.e., as at the end of quarter \(t-1\)) fund return, term spreads and credit spreads in the control variable set. Equation 1 also accounts for fund age, fund return and total net assets as well as time effects and fund fixed effects. For robustness, we also estimated a static panel model (not shown in this note) proposed in Jiang, Li and Wang (2016). Results are consistent and robust using both models.
Table A2: Changes in fund holdings following inflows and outflows
Table A2: Changes in fund holdings following inflows and outflows | |||
---|---|---|---|
Cash_{i,t} (%) | Government_{i,t} (%) | Corporate_{i,t} (%) | |
Outflows_{i,t} (β_{3}) |
1.41 (2.99)*** |
2.08 (3.16)*** |
-1.74 (-4.21)*** |
Inflows_{i,t} |
-0.17 (-0.76) |
-0.99 (-2.98)*** |
0.28 (0.51) |
Sargan-Hansen test (p-value) | 0.45 | 0.54 | 0.71 |
m2 test (p-value) | 0.30 | 0.26 | 0.54 |
Standard errors are clustered at the fund level, *** 1% significance, ** 5% significance, *10% significance
Table A3 reports the estimation results from equation 2. We estimate the model on quarterly data from March 2002 to December 2016.
Equation 2: Dynamic panel estimation
\( Share \, of \, asset \, holdings(\%)_{i,t} \)\( =α_i\)\( +β_1 Share\,of\,asset \, holdings(\%)_{i,t-1}\)\( +β_2(Inflows_{i,t} * LowVol_{i,t})\)\( +β_3 ( Inflows_{i,t} * HighVol_{i,t})\)\( +β_4 ( Outflows_{i,t} * LowVol_{i,t})\)\( +β_5 ( Outflows_{i,t} * HighVol_{i,t})\)\( +γ' * Controls_{i,t-1}+ε_{i,t} \)
The set of control variables used for the estimation of equation 2 remains the same.
Table A3: Changes to fund holdings following inflowss and outflowss: effect of market conditions
Table A3: Changes to fund holdings following inflowss and outflowss: effect of market conditions | |||
---|---|---|---|
Cash_{i,t} | Government_{i,t} | Corporate_{i,t} | |
Outflows_{i,t} * Low Vol_{i,t} (β_{4}) |
1.46 (2.89)*** |
2.80 (4.34)*** |
-1.85 (-2.57)*** |
Outflows_{i,t }*_{ }High Vol_{i,t} | 1.36 (0.50) |
0.12 (0.79) |
-1.40 (-0.75) |
Inflows_{i,t} * Low Vol_{i,t} | -0.15 (-0.39) |
-0.77 (-0.95) |
0.46 (-1.26) |
Inflows_{i,t }*_{ }High Vol_{i,t} | -0.42 (-1.32) |
-0.42 (-1.32) |
0.03 (0.10) |
Sargan-Hansen test (p-value) | 0.42 | 0.48 | 0.68 |
m2 test (p-value) | 0.28 | 0.25 | 0.51 |
Standard errors are clustered at the fund level, *** 1% significance, ** 5% significance, *10% significance
Acknowledgements
We thank Amy Li and Nadeem Merali for research assistance. We are also thankful to Guillaume Bédard-Pagé, Jean-Sébastien Fontaine, Guillaume Nolin and Jonathan Witmer for helpful comments and suggestions.
References
Arellano, M. and S. Bond. 1991. “Some Tests of Specification for Panel Data: Monte Carlo Evidence and an Application to Employment Equations.” The Review of Economic Studies 58 (2): 277–297.
Arora, R. Forthcoming. “Redemption Run Risk in Bond Mutual Funds.” Bank of Canada Staff Analytical Note.
Arora, Fan and Ouellet Leblanc. Forthcoming. “How do Canadian Corporate Bond Funds Meet Investor Redemptions? – A Machine Learning Approach.” Bank of Canada Staff Analytical Note.
Arora, R., N. Merali and G. Ouellet Leblanc. 2018. “Did Canadian Corporate Bond Funds Increase their Exposures to Risks.” Bank of Canada Staff Analytical Note No. 2018-7. Available at https://www.bankofcanada.ca/2018/03/staff-analytical-note-2018-7/
Bank of England (BoE). 2015. Financial Stability Report (December).
Chen, Q., I. Goldstein and W. Jiang. 2010. “Payoff Complementarities and Financial Fragility: Evidence From Mutual Fund Outflows.” Journal of Financial Economics 97 (2): 239–262.
Chernenko, S. and A. Sunderam. 2016. “Liquidity Transformation in Asset Management: Evidence from the Cash Holdings of Mutual Funds.” European Systemic Risk Board Working Paper No 23.
International Monetary Fund (IMF). 2015. Global Financial Stability Report (April).
Jiang, H., D. Li and A. Wang. 2016. “Dynamic Liquidity Management by Corporate Bond Mutual Funds.” SSRN.
Morris, S., I. Shim and H. S. Shin. 2017. “Redemption Risk and Cash Hoarding by Asset Managers.” Bank for International Settlements Working Paper No. 608.
Nagel, S. 2012. “Evaporating Liquidity.” The Review of Financial Studies 25 (7): 2005–2039.
Dick-Nielsen, J., P. Feldhütter and D. Lando. 2012. “Corporate Bond Liquidity Before and After the Onset of the Subprime Crisis.” Journal of Financial Economics 103 (3): 471–492.
- Fund managers can also engage in a mixed strategy, combining attributes from both horizontal and vertical slicing. Further, fund managers can borrow funds under a line of credit or the repo market, but these strategies are found to be of minor importance in the literature. [←]