4.1.6 Modeling Non-Parallel Term Structure Shift and Hedging

6. Modeling Non-Parallel Term Structure Shift and Hedging

6.1 Principal Components Analysis

6.1.1 Weakness of One-Factor Risk Metrics

One-factor risk metric focuses on parallel shift in the interest rate term structure. In practice, rate movements in different regions are not perfectly correlated and multi-factor risk metrics addresses this non-parallel shit pattern.

6.1.2 Principal Components Analysis

PCA is a statistical technique that can be used to understand term structure changes in historical data.

PCA can analyze the effects of multiple factors and estimate their relative importance in describing movements in the term structure.

This technique looks at the daily changes in the interest rates corresponding to various maturities and identifies factors that have the following characteristics:

• These factors are uncorrelated.
• Daily changes in term structure are linear combinations of the factors.
• The first two or three factors account for the majority of the observed daily movements.

$$\Delta r_f=\sum _{i=1}^8{a}_i{f}_{ij}=a_1{f}_{1,j}+a_2{f}_{2,j}+\cdots +a_8{f}_{8,j}$$

• Factor loading($f_{i,j}$): the amount by which rate $j$ moves when there is one-unit change of the $i_{th}$ factor.
• Factor score($a_i$): the number of units change of the $i_{th}$ factor in a daily term structure change.

Rate Maturity	Factor 1	Factor 2	Factor 3	Factor 4	Factor 5	Factor 6	Factor 7	Factor 8
1 year	$-0.129$	$0.384$	$0.778$	$-0.478$	$-0.009$	$0.003$	$-0.017$	$-0.007$
2 year	$-0.258$	$0.490$	$0.081$	$0.583$	$0.588$	$0.000$	$0.025$	$0.001$
3 year	$-0.326$	$0.426$	$-0.119$	$0.241$	$-0.727$	$-0.292$	$0.150$	$0.062$
5 year	$-0.410$	$0.203$	$-0.317$	$-0.216$	$-0.068$	$0.574$	$-0.548$	$-0.096$
7 year	$-0.434$	$-0.008$	$-0.294$	$-0.397$	$0.221$	$-0.707$	$-0.338$	$-0.284$
10 year	$-0.414$	$-0.193$	$-0.104$	$-0.204$	$0.211$	$-0.707$	$-0.338$	$-0.284$
20 year	$-0.385$	$-0.397$	$0.234$	$0.161$	$-0.023$	$0.019$	$-0.169$	$0.764$
30 year	$-0.367$	$-0.440$	$0.348$	$0.317$	$-0.164$	$0.275$	$0.193$	$-0.557$

The importance of the factors are measured by the standard deviation of the factor scores.

Factor	1	2	3	4	5	6	7	8
Standard deviation	$14.15$	$4.91$	$2.44$	$1.59$	$1.09$	$0.85$	$0.78$	$0.68$

For our data, the total variance is:
$$14.15^2+4.91^2+\cdots +0.68^2=235.77$$

The first three factors account for $97,66\%$ of the variation of the observed rate daily movements.
$$\frac{14.15^2+4.91^2+2.44^2}{235.77}=97.66\%$$

Factor 1 ($85\%$) is roughly a parallel shift in the term structure.

Factor 2 ($10\%$) corresponds to a steepening or flattening of the term structure.

Factor 3 ($3\%$) is a bowing(曲度变化) of the term structure.

6.2 Key Rate 01s

6.2.1 Key Rate Shifts

Key rate shifts are non-parallel shifts and the change in the term structure is determined by the change in several selected key rates. It assumes that the impact of the key rate shifts declines linearly and reaches zero at the term of the adjacent key rates.

2.2 Key Rate 01s

Key Rate 01s is the change in a portfolio’s value from a $1$ bp change of the key rate shifts.

Suppose a portfolio consists of a $ $1$ million investment in 1-year, 3-year, 5-year, 9-year and 15-year zero coupon bond. The decrease in the portfolio’s value for a one-basis-point increase in the relevant spot rates is shown below:

Spot Rate Maturity	1 year	3 year	5 year	9 year	15 year
Portfolio Value Decrease	$95.62$	$270.26$	$424.35$	$677.93$	$944.75$

The calculation of KR01 are listed below:

1 bp shift in the following key rate	Change in bp of the Spot Rate(Maturities)
	1 year	3 year	5 year	9 year	15 year
2 year	1	0.667	0	0	0
5 year	0	0.333	1	0.2	0
10 year	0	0	0	0.8	1

Key Rate	Change in $ value fro the zero bonds in the portfolio					Total
	1 year	3 year	5 year	9 year	15 year
KR01-2yr	95.6	180.3	0.0	0.0	0.0	275.9
KR01-5yr	0.0	90.0	424.4	135.6	0.0	649.9
KR01-10yr	0.0	0.0	0.0	542.3	944.8	1487.1

Relationship between KR01 and DV01:

DV01 is the result of shifting all rates by 1 bp while KR01 is a result of shifts in key rates.

$DV01=KR01_{2yr}+KR01_{5yr}+KR01_{10yr}=2412.9$

6.2.3 Key Rate Hedging

For a portfolio with $N$ key rate exposures, we would need $N$ assets to hedge for the key rate risk.

What positions do we need for the 3 hedging instruments to hedge the key rate risk for the portfolio?

Data for Hedging Using KR01s
KR01 Measure	Portfolio	Hedging Instrument
		A	B	C
KR01-2yr	126	20	3	3
KR01-5yr	238	2	22	4
KR01-10yr	385	1	4	25

If $x_1$, $x_2$ and $x_3$ are the positions in the three hedging instruments, the equations are:
$$\{\begin{array}{c}126+20x_1+3x_2+3x_3=0\\ 238+2x_1+22x_2+4x_3=0\\ 385+x_1+4x_2+25x_3=0\end{array}\to \begin{array}{c}{x}_1=-3\\ x_2=-8\\ x_3=-14\end{array}$$

So the portfolio can be hedged with short positions of 3,8 and 14 in the three hedging instruments.

6.2.4 Portfolio Risk

Bank regulators require banks to analyze the risks in their portfolios by considering ten different $KR01s$. The standard deviation of the change in value of the portfolio in one day is:

$${\sigma }_p=\sqrt{\sum _{i=1}^{10}\sum _{j=1}^{10}{\rho }_{i,j}{\sigma }_i{\sigma }_{j}KR01_{i}KR01_j}$$

• ${\sigma }_i$ is the standard deviation of the daily movement in rate $i$.
• ${\rho }_{i,j}$ is the correlation between the daily movement in rate $i$ and rate $j$.

We can generalize the idea so that any set of term structure movement is considered:

$${\sigma }_p=\sqrt{\sum _{i=1}^n\sum _{j=1}^n{\rho }_{i,j}{\sigma }_i{\sigma }_j{w}_i{w}_j}$$

• $w_i$ is the exposure to the $i_{th}$ term structure movement.
• ${\sigma }_i$ is the standard deviation of the $i_{th}$ term structure movement.
• ${\rho }_{i,j}$ is the correlation between the $i_{th}$ and $j_{th}$ structure movement.

It is particularly easy to apply Principal Components Analysis for the calculation of the portfolio risk:

$${\sigma }_p=\sqrt{{\sigma }_1^2{f}_1^2+{\sigma }_2^2{f}_2^2+{\sigma }_3^2{f}_3^2}$$

• ${\sigma }_i$ is the standard deviation of the factor score for $i_{th}$ factor

• $f_i$ is the change in the value of the portfolio when there is a movement in the term structure corresponding to one unit of the $i_{th}$ factor.

• The first three factors as which account for about $98\%$ of the variance in the rate movement.

• The correlation efficient can be discarded in the formula as the factors are uncorrelated with each other.

Suppose a portfolio has an exposure of USD $50$ to a $1$ bp increase in the 5-year rate and USD $-100$ to a $1$ bp in the 10-year rate in the following table and no other exposures.What is the portfolio’s exposure to the first two factors? Assuming the daily standard deviation for Factor 1 and Factor 2 are 14.15 units and 4.91 units respectively. What is the portfolio’s daily standard deviation?

Rate Maturity	Factor 1	Factor 2	Factor 3	Factor 4	Factor 5	Factor 6	Factor 7	Factor 8
1 year	$-0.13$	$0.38$	$0.78$	$-0.48$	$-0.01$	$0.00$	$-0.02$	$-0.01$
2 year	$-0.26$	$0.49$	$0.08$	$0.58$	$0.59$	$0.00$	$0.03$	$0.00$
3 year	$-0.33$	$0.43$	$-0.12$	$0.24$	$-0.73$	$-0.29$	$0.15$	$0.06$
5 year	$-0.41$	$0.20$	$-0.32$	$-0.22$	$-0.07$	$0.57$	$-0.55$	$-0.10$
7 year	$-0.43$	$-0.01$	$-0.29$	$-0.40$	$0.22$	$0.10$	$0.71$	$0.11$
10 year	$-0.41$	$-0.19$	$-0.10$	$-0.20$	$0.21$	$-0.71$	$-0.34$	$-0.28$
20 year	$-0.39$	$-0.40$	$0.23$	$0.16$	$-0.02$	$0.02$	$-0.17$	$0.76$
30 year	$-0.37$	$-0.44$	$0.35$	$0.32$	$-0.16$	$0.28$	$0.19$	$0.56$

The exposure to 1 unit of the first factor is:
$50\times (-0.41)+(-100)\times (-0.41)=20.5$

The exposure to 1 unit of the second factor is:
$50\times 0.20+(-100)\times (-0.19)=29.0$

The daily standard deviation of portfolio is
${\sigma }_p=\sqrt{14.15^2\times 20.5^2+4.91^2\times 29^2}=323.14$

6.2.5 The Use of Par Yields

Besides using spot rates to define key rate shift, a more practical way is to use par yields because:

Par yield government bonds are much more actively traded instruments and we can immediately calculate the position necessary to hedge a portfolio once we have calculated the exposure of the portfolio to the key rate shifts.

6.3 Forward Bucket 01

6.3.1 Bucketing Approach

Bucketing approach is to divide the interest rate term into segments referred to as buckets, and then calculate the dollar impact of changing all the rates in a bucket by one basis point on the value of a portfolio.

Relationship between Bucket shifts and DV01:

$DV01=B_{0-2yr}01+B_{2-10yr}01+B_{10-30yr}01$

6.3.2 Forward Bucket 01

Apply the bucketing approach to the forward curve as well:

There are three buckets for forward curve: 0-2 years, 2-10 years and 10-30years,

• The decrease in portfolio value after applying $1$ bp shift to forward curve in the first bucket. (The forward rates for the 6-months periods beginning in 0, 6, 12 and 18 months increase by $1$ bp.)
• Similarly, we can define Forward $bucke{t}_{2-10yr}$, and Forward $bucke{t}_{10-30yr}$.

Consider a simple portfolio consisting of a two-year bond with a face value of USD $100$ and a coupon of $6\%$ per year. Assume the term structure is flat at $4\%$ with semiannual compounding. Suppose there are two buckets; 0-1 years and 1-2 years. Calculate the forward bucket 01 for each bucket?

The price of the bond:

$$\frac{3}{(1+2\%)}+\frac{3}{(1+2\%{)}^2}+\frac{3}{(1+2\%{)}^3}+\frac{103}{(1+2\%{)}^4}=103.8077$$

Forward rates in the 0-1 years bucket are increased by $1bp$

$$\frac{3}{(1+2.005\%)}+\frac{3}{(1+2.005\%{)}^2}+\frac{3}{(1+2.005\%{)}^2(1+2\%)}+\frac{103}{(1+2.005\%{)}^2(1+2\%{)}^2}=103.7977$$

Forward Bucket $01_{0-1yr}=103.8077-103.7977=0.01$

Forward rates in the 1-2 years bucket are increase by $1bp$

$$\frac{3}{(1+2\%)}+\frac{3}{(1+2\%{)}^2}+\frac{3}{(1+2\%{)}^2(1+2.005\%)}+\frac{103}{(1+2\%{)}^2(1+2.005\%{)}^2}=103.7983$$

Forward Bucket $01_{1-2yr}=103.8077-103.7983=0.0094$

6.3.3 Duration Measures

We know the relationship between DV01 and duration is given as followed:
$$\text{DV01}=\frac{\text{DollarDuration}}{10,000}=\frac{\text{Duration}\times \text{Valueof Portfolio}}{10,000}$$

Similarly, we can convert any of the 01 measure to its corresponding duration measure:

$$\text{01Measure}=\frac{\text{DurationMeasure}\times \text{Valueof Portfolio}}{10,000}$$

$$\to \text{DurationMeasure}=\frac{\text{01Measure}\times 10,000}{\text{Valueof Portfolio}}$$