On the Liquidity of Danish Mortgage Bonds Jesper Lund Department of Finance Copenhagen Business School Joint work-in-progress with: Birgitte Vølund Buchholst, Danish Central Bank Jens Dick-Nielsen, Copenhagen Business School Jacob Gyntelberg, Bank for International Settlements Thomas Sangill, European Central Bank Nykredit Symposium 13th December 2010
Motivation for the paper Increased focus on liquidity after the financial crisis Liquidity reform package proposed by the Basel Committee Regulation of liquidity reserves in financial institutions Strict definition of what constitutes a liquid asset No limits on government bonds, but limits on covered bonds that would pose problems for many Danish financial institutions, and ultimately the Danish mortgage bond market Especially a problem for Denmark where the size of the mortgage bond market is more three times the size of the government bond market (outstanding amount) Basel proposal motivated by the recent financial crisis, but All financial assets were affected by the crisis The real question should be: are Danish mortgage bonds really less liquid than Danish government bonds? If so, how much? Does it apply to all segments of the market? On the Liquidity of Danish Mortgage Bonds 2 / 20
Danish mortgage bonds during the peak of financial crisis During 2008 option adjusted spreads widened significantly Danish mortgage bonds trade at higher yields than government bonds and usually also swap rates (we use option adjusted spreads instead of simple yield spreads to take account of the prepayment option) No different from any other non-government bond market: investors attach a special significance to [most] government bonds and pay a premium for that For that reason covered bond markets attract real money investors as well as leveraged speculators (like hedge funds) Liquidity interpretations of this development Leveraged speculators were forced to liquididate their positions (temporary supply demand pressure on bond prices) Greater premium for liquidity risk in the market On the Liquidity of Danish Mortgage Bonds 3 / 20
Option-adjusted spreads to the swap curve On the Liquidity of Danish Mortgage Bonds 4 / 20
Research framework for this study How to measure liquidity? No single answer Typical measures of liquidity include Bid-ask spread (the cost of portfolio reallocation) The ability to sell a large amount without affecting the price (too much), aka price impact of trades The daily turnover (especially: lack of liquidity if it is low) Main inspiration: Dick-Nielsen, Feldhütter and Lando, Corporate bond liquidity before and after the onset of the subprime crisis, working paper 2010. In that paper, yield spreads for corporate bonds are adjusted for credit risk, and the liquidity premium (remaining part of the yield spread) is studied before and after the crisis We use the same liquidity measures, but do not focus on whether liquidity is priced (yet) Preliminary results (presented here) are from a working paper (2010 70) published by Danmarks Nationalbank On the Liquidity of Danish Mortgage Bonds 5 / 20
Transaction data January 2005 May 2010 All transactions in the Danish bond market Data fields: ISIN, date and timestamp, clean price, nominal amount traded, settlement type (standard vs non-standard), and whether the trade is bank bank or bank client Danish bond market is primarily an OTC ( telephone ) market, but legal requirement that all trades are reported to a central authority Main data source: Nasdaq OMX Copenhagen (all mortgage bonds and government bonds until November 2007) Transactions for Danish goverment bonds since November 2007 were obtained from the Danish Financial Supervisory Authority All repo trades and other trades with non-standard settlement were removed prior to the analysis On the Liquidity of Danish Mortgage Bonds 6 / 20
Subset of data used for this analysis Data selection Trades above 10 mill DKK (focus on professional segment) Bonds with an outstanding amount of at least 1 bn EUR Only bonds with a fixed coupon rate (for now). CIBOR-linked (capped) floaters will be added later First subset: short-term mortgage and government bonds Defined as maturities up to five years The mortgage segment is dominated by one-year bonds used to finance adjustable-rate mortgages Average mortgage maturity is shorter than the average government bond maturity Second subset: long-term mortgage and government bonds Mortgage segment dominated by 30 year callable bonds Lots of small bond series excluded (29 out of 1250 series used!) Large 30-year government bond issue in November 2008 On the Liquidity of Danish Mortgage Bonds 7 / 20
Short-term bonds: summary statistics for the data On the Liquidity of Danish Mortgage Bonds 8 / 20
Long-term bonds: summary statistics for the data On the Liquidity of Danish Mortgage Bonds 9 / 20
Measures of liquidity used in the empirical analysis Median trade size for each month Roll (1984) implied bid-ask spread Roll t = 2 Cov( P i, P i 1 ) (1) This works because bid-ask bounces introduce negative serial correlation of price changes between adjacent trades Amihud (2002) price impact of a trade Amihud t = 1 N t N t P j P j 1 P j 1 Q j j=1 = 1 N t N t j=1 r j Q j (2) where Q j the the size of the j th trade Assumption: price impact is a linear function of trade size We report the median Roll and Amihud measure for each month (aggregated over all bonds in the data subset) On the Liquidity of Danish Mortgage Bonds 10 / 20
Empirical results for short-term bonds Median trade size has increased slightly for mortgage bonds since 2008 Careful with interpretation because of spikes during refinancing auctions in November December and the new smaller ones in March and September Before the crisis, the trade size was smaller for mortgage bonds, but by 2010 there is no discernible difference Implied bid-ask spreads (Roll) Sharp increase during the financial crisis, but much larger increase for government bonds Seems to have stabilized at a higher level compared to pre-crisis Price impact of a trade Substantial increase i 2008, but post-crisis levels are comparable to pre-crisis Drop in liquidity during 2008 seems to have an equal effect on mortgage bonds and government bonds On the Liquidity of Danish Mortgage Bonds 11 / 20
Short-term bonds: median trade size On the Liquidity of Danish Mortgage Bonds 12 / 20
Short-term bonds: implied bid-ask spread (Roll) On the Liquidity of Danish Mortgage Bonds 13 / 20
Short-term bonds: price impact of a trade (Amihud) On the Liquidity of Danish Mortgage Bonds 14 / 20
Empirical results for long-term bonds Median trade size greater for government bonds Not surprising given the much larger number of mortgage bonds (only one 4% DGB 2019, but multiple 4% 2041 bonds) The effect of the crisis was greater for government bonds! Some (most?) of the variation in median trade size for mortgage bonds can probably be explained by issuance patterns and refinancing waves (missing here) Implied bid-ask spreads (Roll) Sharp increase during the financial crisis, with the same effect on mortgage and government bonds Post-crisis levels are higher than pre-crisis, but the effect is actually greater for government bonds Likely explanation: introduction of 30-year bond in late 2008 Price impact of a trade is greater for mortgage bonds Substantial increase i 2008: Greater effect on mortgage bonds Post-crisis impact measure has stabilized at a higher level On the Liquidity of Danish Mortgage Bonds 15 / 20
Long-term bonds: median trade size On the Liquidity of Danish Mortgage Bonds 16 / 20
Long-term bonds: implied bid-ask spread (Roll) On the Liquidity of Danish Mortgage Bonds 17 / 20
Long-term bonds: price impact of a trade (Amihud) On the Liquidity of Danish Mortgage Bonds 18 / 20
Future extensions of the empirical analysis Include CIBOR-linked mortgage bonds in the analysis Nonparametric estimation of price impact as function of the trade size (before, during and after the crisis) Use ownership data from VP to estimate the share of leveraged speculative investors vs real-money investors Study individual bonds: a lot of potentially interesting results might be washed out in our averages (medians) Can investor composition explain variation in the liquidity measures across different bonds? Investigate whether liquidity is priced in the market, e.g. by using OAS in leui of yield spreads On the Liquidity of Danish Mortgage Bonds 19 / 20
Conclusions Disclaimer: preliminary conclusions... Overall, the financial crisis had an equal effect on the liquidity of mortgage bonds and government bonds The temporary increase in the price impact of a trade (Amihud measure) during the peak of the crisis was slightly higher for long-term mortgage bonds than government bonds, though Mortgage bonds are not just mortgage bonds Liquidity measures of short-term bonds were affected by the financial crisis to a lesser degree than long-term bonds Not surprising: less price exposure to spread widening (maturity effect), and mortgage bullet bonds do not have the complexity of the prepayment option On the Liquidity of Danish Mortgage Bonds 20 / 20