Yuliya Baranova, Louisa Chen and Nicholas Vause.
Many investors report recent declines in market liquidity, suggesting dealers have become less willing to trade corporate bonds and other fixed-income securities due to additional costs of holding them on their balance sheets. Some fear that if asset managers began to sell these securities then prices could fall sharply. Focusing on high-yield corporate bonds, we use an econometric model to investigate whether the typical responses of dealer inventories and market prices to falls in asset manager demand have changed in recent years. We find that dealer holdings act less as a shock absorber than they did around a decade ago. Instead, bond spreads rise more. We also find that greater declines in issuance now follow these shocks.
Shrinking corporate bond dealer inventories – should we be concerned?
The market for a security is liquid if investors can trade it – even in large size – at low cost, including because their orders have little effect on its price. This matters because any decline in the liquidity of securities could make it more expensive for borrowers to issue them in the first place, and this could hurt investment and economic growth.
Several indicators of market liquidity remain healthy by historical standards. These include bid-ask spreads, price changes relative to trading volumes and price differences between frequently and infrequently traded securities or economically equivalent derivatives. Of course, each of these indicators relates to observed market activity.
However, many market participants fear that prices could move sharply under larger order flows. A particular concern is that dealers may have limited capacity to intermediate large sales of corporate bonds by asset managers, which could therefore generate sharp price falls. Assets under management of corporate bond funds have more than doubled since the 2008 financial crisis, and they may be vulnerable to a wave of redemptions if rising interest rates started to weigh on prices. Furthermore, dealers have cut their inventories of fixed-income securities by more than 75% during the same period, suggesting a significant withdrawal from market making activity.
That said, other factors may also help explain the sharp decline in dealer inventories. First, reflecting the dramatic decline in the securitisation industry since the crisis, dealers are warehousing fewer securities for repackaging. Second, as a response to the regulatory reforms, they are holding fewer securities as proprietary investments. Although these reforms apply equally to securities held as a result of market making activities, these are not necessarily kept on balance sheets for long. Hence, the level of dealer inventories could be a poor proxy for their responsiveness to market events, which is what matters for liquidity.
Modelling the impact of a fall in demand from asset managers on corporate bond market
In this blog we use a structural vector auto-regression (SVAR) model to investigate whether the typical responses of dealer inventories and market prices to changes in asset manager demand for high-yield corporate bonds have changed in recent years. We also use this model to investigate whether the response of issuance has changed. The model is
yt is a vector of the model’s four variables, which are weekly changes in corporate bond spreads and holdings of high-yield corporate bonds by asset managers, dealers and other investors. B0, B1, B2 and c are matrices of coefficients, which we need to calibrate. Finally, ut is a vector of different types of market news that affect the variables in the model. Given how we calibrate the coefficients (see below), we can identify one of the elements in this vector as news about asset manager demand for bonds. We can then simulate significant falls in asset manager demand and see how the variables in the model respond. To compare responses over time, we calibrate the model to two periods: the past few years (since 2012) and a period of similar length before the global financial crisis (2004-2006).
To help calibrate the model in each of these periods, we require that candidate sets of coefficients are consistent with certain sign patterns in the way that the model’s variables respond to news. For example, and as reflected in Figure 1, we require that a fall in asset manager demand for bonds (DAM) leads to a higher spread (s), a reduction in asset manager holdings (HAM) and an increase in the holdings of both dealers (HD) and other investors (HOI). Our full set of sign requirements is reported in Table 1. We discard any coefficient sets that do not fit these restrictions, retaining the others as valid model calibrations. Estimation of SVAR models based on sign restrictions is described more fully here and, in the context of another application, here.
The choice of variables for the SVAR was guided by the quality and availability of data as well as the interpretability of the model’s outputs. Note, first, that we focus on high-yield corporate bonds as we are confident that the high-yield funds in our data set, as opposed to investment-grade funds, invest in little else. Second, we use changes in US primary dealer inventories as a proxy for global dealer inventories since the latter are not available. Moreover, these are inventories of corporate fixed-income securities, as holdings of high-yield corporate bonds are not available for the required time period. And third, because the corporate bond market has grown over time, we scale changes in the holdings of asset managers, dealers and other investors by outstanding volumes, making our results easier to interpret.
Has dealers’ capacity to intermediate the corporate bond market changed post-crisis?
The four SVAR variables are plotted in Figure 2. From this we can see that spreads were more volatile in the post-crisis period than in the pre-crisis period, whereas the opposite is true for dealer holdings. This is consistent with dealers’ intermediation capacity having fallen, leading to greater volatility in market prices.
Our impulse response analysis reaches the same conclusion. Figure 3 shows projected responses of spreads and dealer holdings to a negative shock to asset manager demand for high-yield corporate bonds, based both on our pre-crisis and post-crisis calibrations of the SVAR. These are medians of responses generated by the SVAR calibrations that are consistent with our sign restrictions. The magnitude of the shock generating the responses is equal to one standard deviation of the shocks in the pre-crisis period. To facilitate comparisons, we use the same size shock in the post-crisis period. As the SVAR is a linear model, however, the effects of larger shocks could be studied simply by scaling up the impulses responses in the figure. The illustrated responses suggest that in the pre-crisis period dealers would have increased their bond holdings by 1.5 basis points of outstanding market volume in the same week as the asset manager demand shock and that they would have partially reversed that change within three weeks. In the post-crisis period, however, dealers would increase their holdings by much less in response to the shock – just 0.2 basis points of outstanding market value, and would reverse most of that within three weeks. In contrast, bond spreads respond by more in the post-crisis period (17.3 basis points) than in the pre-crisis period (8.5 basis points). Both spread responses come in the same week as the shock, with minimal effects thereafter.
We also find that net issuance is more affected by negative asset manager demand shocks in the post-crisis period than it was pre-crisis. The impulse responses for net issuance, which we computed by summing the changes in holdings of all market participants, are shown in the right-hand panel of Figure 3. In the post-crisis calibration, this falls by as much as 6.7 basis points of outstanding market value in the third week after the shock. This compares with a maximum decline of 5.4 basis points in the pre-crisis calibration, which is also in the third week. In both cases, net issuance continues to decline for a few more weeks, with effects finally dying down around eight weeks after the shock.
In summary, based on a SVAR model, we find that dealer holdings respond less and spreads respond more to a decline in asset manager demand for high-yield corporate bonds now than they did around a decade ago. These findings support the claim that the market-making capacity of dealers has fallen in recent years, reducing secondary market liquidity. Reflecting the larger increases in spreads that now follow falls in asset manager demand, we also find a greater decline in issuance resulting from these shocks.
Yuliya Baranova, Louisa Chen and Nicholas Vause work in the Bank’s Capital Markets Division.
Bank Underground is a blog for Bank of England staff to share views that challenge – or support – prevailing policy orthodoxies. The views expressed here are those of the authors, and are not necessarily those of the Bank of England, or its policy committees.
If you want to get in touch, please email us at bankunderground@bankofengland.co.uk
