SlideShare a Scribd company logo
Estimating	
  Beta	
  For	
  A	
  Portfolio	
  Of	
  Stocks	
  
Geoffery	
  Mullings	
  
March	
  29,	
  2016	
  
Executive	
  Summary:	
  
For	
  some	
  a	
  stock's	
  beta	
  can	
  be	
  used	
  as	
  a	
  proxy	
  for	
  the	
  asset's	
  systemic	
  risk	
  -­‐	
  that	
  is,	
  
how	
  vulnerable	
  the	
  stock	
  is	
  to	
  movements	
  in	
  the	
  market.	
  This	
  analysis	
  estimates	
  
beta	
  coefficients	
  for	
  a	
  portfolio	
  of	
  stocks	
  while	
  interpreting	
  and	
  comparing	
  those	
  
values	
  to	
  more	
  standard	
  measures	
  of	
  risk	
  such	
  as	
  variance	
  and	
  correlation.	
  While	
  to	
  
a	
  certain	
  extent	
  beta	
  represents	
  systemic	
  risk	
  its	
  ability	
  to	
  proxy	
  for	
  risk	
  is	
  quite	
  
limited.	
  The	
  difference	
  between	
  upside	
  and	
  downside	
  risk	
  is	
  difficult	
  to	
  attain	
  from	
  
beta,	
  while	
  covariance	
  offers	
  much	
  more	
  flexibility	
  in	
  portfolio	
  risk	
  managment.	
  
Introduction:	
  
A	
  stock's	
  beta	
  can	
  be	
  a	
  convenient	
  proxy	
  for	
  risk	
  depending	
  on	
  many	
  factors.	
  Beta	
  is	
  
a	
  measure	
  of	
  comovement	
  with	
  the	
  market,	
  often	
  interpreted	
  as	
  a	
  measure	
  of	
  
systemic	
  risk.	
  If	
  a	
  stock	
  moves	
  more	
  than	
  the	
  market	
  it	
  will	
  have	
  a	
  higher	
  beta,	
  while	
  
the	
  reverse	
  is	
  true	
  for	
  a	
  lower	
  beta.	
  The	
  risk-­‐reward	
  tradeoff	
  framework	
  suggests	
  
that	
  stocks	
  with	
  lower	
  betas	
  may	
  be	
  safer	
  but	
  should	
  provide	
  less	
  opportunity	
  for	
  
large	
  gains.	
  That	
  will	
  be	
  tested	
  in	
  the	
  analysis	
  below.	
  
Beta	
  can	
  be	
  estimated	
  theoretically	
  through	
  a	
  regression	
  model,	
  perhaps	
  even	
  a	
  
linear	
  regression	
  depending	
  on	
  the	
  movement	
  of	
  the	
  stock.	
  It's	
  also	
  possible	
  to	
  
observe	
  historical	
  data	
  and	
  assess	
  whether	
  beta	
  accurately	
  reflects	
  our	
  normal	
  
understanding	
  of	
  "risk."	
  
Data	
  Loading	
  and	
  Munging:	
  
require(tseries)	
  
require(zoo)	
  
require(fBasics)	
  
#	
  Retrieving	
  the	
  Market	
  Factor.	
  Using	
  the	
  S&P	
  500	
  as	
  a	
  proxy	
  for	
  the	
  
market's	
  movements.	
  
sp	
  <-­‐	
  get.hist.quote("^GSPC",	
  start='1989-­‐12-­‐01',	
  quote="AdjClose",	
  
compression="m",	
  quiet=TRUE)	
  
	
  
mytick	
  =	
  c('ATI','TSN','GME','LB','PAYX')	
  #	
  Storing	
  all	
  of	
  the	
  stocks	
  
in	
  a	
  variable	
  for	
  looping.	
  
	
  
data	
  <-­‐	
  get.hist.quote(mytick[1],	
  quote="AdjClose",	
  start="1989-­‐12-­‐01",	
  
compression="m",	
  quiet=TRUE,	
  retclass="zoo")	
  
 
for	
  (i	
  in	
  2:length(mytick))	
  
	
  	
  	
  
{	
  
	
  	
  	
  
	
  	
  temp	
  =	
  get.hist.quote(mytick[i],	
  quote="AdjClose",	
  start="1989-­‐12-­‐
01",	
  compression="m",	
  quiet=TRUE,	
  retclass="zoo")	
  
	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  
	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  data	
  =	
  cbind(data,	
  temp)	
  
	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  
}	
  
	
  
colnames(data)	
  =	
  mytick	
  
	
  
#	
  Transforming	
  into	
  logarithmic	
  returns.	
  Log	
  returns	
  are	
  easier	
  to	
  
compare	
  over	
  time,	
  and	
  accurate	
  as	
  long	
  as	
  returns	
  aren't	
  too	
  large.	
  
ret	
  =	
  100	
  *	
  diff(log(data))	
  
sp	
  =	
  100	
  *	
  diff(log(sp))	
  
colnames(sp)	
  =	
  "SP.500"	
  
StocksMarketRet	
  =	
  cbind(ret,sp)	
  
	
  
sum(is.na(StocksMarketRet))	
  #	
  Any	
  missing	
  return	
  values?	
  
##	
  [1]	
  295	
  
There	
  are	
  some	
  NA	
  values	
  in	
  the	
  historical	
  data	
  because	
  each	
  stock	
  had	
  its	
  IPO	
  after	
  
1990.	
  To	
  reduce	
  bias	
  in	
  our	
  analysis	
  we'll	
  eliminate	
  all	
  observations	
  preceding	
  the	
  
IPO	
  of	
  the	
  youngest	
  stock	
  in	
  the	
  portfolio,	
  GME.	
  
StocksMarketRetAdj	
  =	
  StocksMarketRet[151:318,]	
  #	
  Eliminating	
  the	
  first	
  
149	
  observations	
  from	
  all	
  assets	
  to	
  reduce	
  bias.	
  
	
  
sum(is.na(StocksMarketRetAdj))	
  
##	
  [1]	
  0	
  
Let's	
  first	
  analyze	
  some	
  of	
  the	
  basic	
  statistics	
  about	
  this	
  portfolio's	
  and	
  the	
  S&P	
  
500's	
  logarithmic	
  return.	
  
Analysis:	
  
basicStats(StocksMarketRetAdj)	
  
##	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  ATI	
  	
  	
  	
  	
  	
  	
  	
  TSN	
  	
  	
  	
  	
  	
  	
  	
  GME	
  	
  	
  	
  	
  	
  	
  	
  	
  LB	
  	
  	
  	
  	
  	
  	
  PAYX	
  
##	
  nobs	
  	
  	
  	
  	
  	
  	
  	
  168.000000	
  168.000000	
  168.000000	
  168.000000	
  168.000000	
  
##	
  NAs	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  0.000000	
  	
  	
  0.000000	
  	
  	
  0.000000	
  	
  	
  0.000000	
  	
  	
  0.000000	
  
##	
  Minimum	
  	
  	
  	
  	
  -­‐50.573656	
  -­‐31.198382	
  -­‐64.077922	
  -­‐36.862335	
  -­‐16.913253	
  
##	
  Maximum	
  	
  	
  	
  	
  	
  47.759670	
  	
  26.659710	
  	
  24.203372	
  	
  27.204306	
  	
  17.582045	
  
##	
  1.	
  Quartile	
  	
  -­‐8.765500	
  	
  -­‐4.404103	
  	
  -­‐5.359523	
  	
  -­‐3.520060	
  	
  -­‐3.848341	
  
##	
  3.	
  Quartile	
  	
  	
  9.622867	
  	
  	
  6.204392	
  	
  	
  8.764377	
  	
  	
  7.757846	
  	
  	
  4.966828	
  
##	
  Mean	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  0.160533	
  	
  	
  1.092622	
  	
  	
  0.737814	
  	
  	
  1.363810	
  	
  	
  0.436373	
  
##	
  Median	
  	
  	
  	
  	
  	
  	
  -­‐0.441562	
  	
  	
  1.976514	
  	
  	
  1.494889	
  	
  	
  1.851768	
  	
  	
  1.036216	
  
##	
  Sum	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  26.969589	
  183.560521	
  123.952835	
  229.120095	
  	
  73.310595	
  
##	
  SE	
  Mean	
  	
  	
  	
  	
  	
  	
  1.267847	
  	
  	
  0.714102	
  	
  	
  0.954303	
  	
  	
  0.711278	
  	
  	
  0.498472	
  
##	
  LCL	
  Mean	
  	
  	
  	
  	
  -­‐2.342541	
  	
  -­‐0.317208	
  	
  -­‐1.146238	
  	
  -­‐0.040446	
  	
  -­‐0.547746	
  
##	
  UCL	
  Mean	
  	
  	
  	
  	
  	
  2.663608	
  	
  	
  2.502453	
  	
  	
  2.621867	
  	
  	
  2.768066	
  	
  	
  1.420491	
  
##	
  Variance	
  	
  	
  	
  270.049434	
  	
  85.670170	
  152.996535	
  	
  84.994034	
  	
  41.743632	
  
##	
  Stdev	
  	
  	
  	
  	
  	
  	
  	
  16.433181	
  	
  	
  9.255818	
  	
  12.369177	
  	
  	
  9.219221	
  	
  	
  6.460931	
  
##	
  Skewness	
  	
  	
  	
  	
  	
  0.011003	
  	
  -­‐0.400221	
  	
  -­‐1.015992	
  	
  -­‐0.601945	
  	
  -­‐0.147011	
  
##	
  Kurtosis	
  	
  	
  	
  	
  	
  1.055338	
  	
  	
  1.152953	
  	
  	
  3.521931	
  	
  	
  1.594190	
  	
  -­‐0.326907	
  
##	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  SP.500	
  
##	
  nobs	
  	
  	
  	
  	
  	
  	
  	
  168.000000	
  
##	
  NAs	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  0.000000	
  
##	
  Minimum	
  	
  	
  	
  	
  -­‐18.563647	
  
##	
  Maximum	
  	
  	
  	
  	
  	
  10.230659	
  
##	
  1.	
  Quartile	
  	
  -­‐1.773925	
  
##	
  3.	
  Quartile	
  	
  	
  2.984705	
  
##	
  Mean	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  0.341674	
  
##	
  Median	
  	
  	
  	
  	
  	
  	
  	
  1.018362	
  
##	
  Sum	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  57.401290	
  
##	
  SE	
  Mean	
  	
  	
  	
  	
  	
  	
  0.331765	
  
##	
  LCL	
  Mean	
  	
  	
  	
  	
  -­‐0.313320	
  
##	
  UCL	
  Mean	
  	
  	
  	
  	
  	
  0.996669	
  
##	
  Variance	
  	
  	
  	
  	
  18.491439	
  
##	
  Stdev	
  	
  	
  	
  	
  	
  	
  	
  	
  4.300167	
  
##	
  Skewness	
  	
  	
  	
  	
  -­‐0.868752	
  
##	
  Kurtosis	
  	
  	
  	
  	
  	
  1.923297	
  
There's	
  some	
  upside	
  risk	
  in	
  ATI	
  (a	
  small	
  cap	
  metals	
  and	
  mining	
  firm)	
  based	
  on	
  the	
  
mean,	
  median,	
  and	
  high	
  variation,	
  if	
  not	
  the	
  maximum	
  observation.	
  This	
  could	
  be	
  
due	
  to	
  its	
  relative	
  youth	
  on	
  the	
  market,	
  similar	
  to	
  GME	
  (the	
  infamous	
  small	
  cap	
  
video	
  game	
  retailer,	
  GameStop),	
  which	
  also	
  exhibits	
  a	
  high	
  variance	
  and	
  many	
  low	
  
returns,	
  evidenced	
  by	
  a	
  large	
  negative	
  skew	
  and	
  minimum	
  value.	
  High	
  variance	
  is	
  
also	
  not	
  particularly	
  uncommon	
  for	
  technology	
  stocks.	
  
All	
  but	
  ATI	
  and	
  PAYX	
  (a	
  large	
  cap	
  IT	
  firm)	
  seem	
  negatively	
  skewed	
  based	
  on	
  mean	
  
and	
  median	
  values,	
  with	
  the	
  largest	
  tails	
  present	
  on	
  GME	
  (which	
  also	
  happens	
  to	
  be	
  
the	
  second-­‐highest	
  stock	
  in	
  variance)	
  based	
  on	
  its	
  kurtosis	
  value.	
  
LB	
  (a	
  large	
  cap	
  retailer	
  of	
  women's	
  underwear)	
  and	
  TSN	
  (the	
  large	
  cap	
  food	
  
producer	
  Tyson	
  Foods)	
  share	
  with	
  the	
  other	
  stocks	
  high	
  variances	
  relative	
  to	
  the	
  
market,	
  accompanied	
  by	
  higher	
  mean	
  returns	
  -­‐	
  except	
  for	
  ATI.	
  
Every	
  stock	
  in	
  the	
  portfolio	
  has	
  a	
  higher	
  return	
  variance	
  than	
  the	
  market.	
  Alongside	
  
variance,	
  comovement	
  between	
  assets	
  and	
  the	
  market	
  should	
  be	
  assessed	
  on	
  the	
  
road	
  to	
  estimating	
  beta.	
  
plot(as.zoo(StocksMarketRetAdj),	
  main="Portfolio	
  Stock	
  Market	
  Returns")	
  
 
plot(as.data.frame(StocksMarketRetAdj),	
  main="Portfolio	
  Stock	
  Market	
  
Returns	
  Plotted	
  Against	
  Each	
  Other")	
  
 
Visually,	
  there's	
  evidence	
  of	
  comovement	
  between	
  all	
  the	
  stocks	
  and	
  the	
  market	
  
factor.	
  Obviously	
  some	
  stock	
  return	
  data	
  exhibits	
  more	
  varied	
  comovement	
  than	
  
others,	
  and	
  finding	
  these	
  differences	
  between	
  the	
  asset	
  and	
  the	
  market	
  factor's	
  
comovement	
  will	
  give	
  us	
  the	
  sought-­‐after	
  estimate	
  of	
  beta.	
  
cor(StocksMarketRetAdj)	
  #	
  Would	
  have	
  included	
  use="pairwise.complete"	
  
if	
  I	
  hadn't	
  eliminated	
  the	
  missing	
  observations	
  already.	
  
##	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  ATI	
  	
  	
  	
  	
  	
  	
  TSN	
  	
  	
  	
  	
  	
  	
  GME	
  	
  	
  	
  	
  	
  	
  	
  LB	
  	
  	
  	
  	
  	
  PAYX	
  	
  	
  	
  SP.500	
  
##	
  ATI	
  	
  	
  	
  1.0000000	
  0.2801193	
  0.2450891	
  0.3920166	
  0.2263508	
  0.5653850	
  
##	
  TSN	
  	
  	
  	
  0.2801193	
  1.0000000	
  0.1623045	
  0.3337894	
  0.2221864	
  0.3987431	
  
##	
  GME	
  	
  	
  	
  0.2450891	
  0.1623045	
  1.0000000	
  0.3515588	
  0.2799321	
  0.3822975	
  
##	
  LB	
  	
  	
  	
  	
  0.3920166	
  0.3337894	
  0.3515588	
  1.0000000	
  0.4319816	
  0.6131435	
  
##	
  PAYX	
  	
  	
  0.2263508	
  0.2221864	
  0.2799321	
  0.4319816	
  1.0000000	
  0.5867857	
  
##	
  SP.500	
  0.5653850	
  0.3987431	
  0.3822975	
  0.6131435	
  0.5867857	
  1.0000000	
  
fit	
  =	
  lm(StocksMarketRetAdj[,mytick]	
  ~	
  StocksMarketRetAdj$SP.500)	
  
coef(summary(fit))	
  
##	
  Response	
  ATI	
  :	
  
##	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  Estimate	
  Std.	
  Error	
  	
  	
  	
  t	
  value	
  	
  	
  	
  	
  
Pr(>|t|)	
  
##	
  (Intercept)	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  -­‐0.5776988	
  	
  	
  1.052226	
  -­‐0.5490256	
  
5.837256e-­‐01	
  
##	
  StocksMarketRetAdj$SP.500	
  	
  2.1606307	
  	
  	
  0.244650	
  	
  8.8315179	
  
1.411880e-­‐15	
  
##	
  	
  
##	
  Response	
  TSN	
  :	
  
##	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  Estimate	
  Std.	
  Error	
  	
  t	
  value	
  	
  	
  	
  	
  Pr(>|t|)	
  
##	
  (Intercept)	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  0.7993742	
  	
  0.6589283	
  1.213143	
  2.267986e-­‐01	
  
##	
  StocksMarketRetAdj$SP.500	
  0.8582674	
  	
  0.1532055	
  5.602066	
  8.630383e-­‐08	
  
##	
  	
  
##	
  Response	
  GME	
  :	
  
##	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  Estimate	
  Std.	
  Error	
  	
  t	
  value	
  	
  	
  	
  	
  Pr(>|t|)	
  
##	
  (Intercept)	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  0.3620902	
  	
  0.8872695	
  0.408095	
  6.837298e-­‐01	
  
##	
  StocksMarketRetAdj$SP.500	
  1.0996562	
  	
  0.2062965	
  5.330465	
  3.158340e-­‐07	
  
##	
  	
  
##	
  Response	
  LB	
  :	
  
##	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  Estimate	
  Std.	
  Error	
  	
  	
  t	
  value	
  	
  	
  	
  	
  
Pr(>|t|)	
  
##	
  (Intercept)	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  0.9146684	
  	
  0.5653664	
  	
  1.617833	
  1.075972e-­‐
01	
  
##	
  StocksMarketRetAdj$SP.500	
  1.3145315	
  	
  0.1314517	
  10.000109	
  1.006882e-­‐
18	
  
##	
  	
  
##	
  Response	
  PAYX	
  :	
  
##	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  Estimate	
  Std.	
  Error	
  	
  	
  t	
  value	
  	
  	
  	
  	
  
Pr(>|t|)	
  
##	
  (Intercept)	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  	
  0.1351402	
  0.40613107	
  0.3327502	
  7.397427e-­‐
01	
  
##	
  StocksMarketRetAdj$SP.500	
  0.8816360	
  0.09442837	
  9.3365581	
  6.361038e-­‐
17	
  
The	
  p	
  and	
  t-­‐values	
  of	
  all	
  the	
  coefficients	
  suggests	
  that	
  each	
  is	
  statistically	
  significant	
  
and	
  most	
  likely	
  a	
  reliable	
  estimate	
  of	
  their	
  respective	
  betas.	
  
Notably	
  even	
  though	
  TSN	
  and	
  GME	
  have	
  some	
  of	
  the	
  weakest	
  correlations	
  with	
  the	
  
market	
  factor,	
  they	
  have	
  betas	
  that	
  are	
  very	
  close	
  to	
  1.	
  While	
  correlation	
  shows	
  the	
  
strength	
  of	
  return	
  similarities	
  between	
  the	
  stock	
  and	
  the	
  market,	
  beta	
  really	
  
estimates	
  the	
  effect	
  of	
  a	
  market	
  movement	
  on	
  each	
  asset.	
  
Conclusion:	
  
The	
  preceding	
  has	
  hopefully	
  made	
  it	
  clear	
  that	
  beta	
  provides	
  some	
  utility	
  
accompanied	
  by	
  many	
  limits.	
  Beta	
  is	
  much	
  better	
  at	
  capturing	
  all	
  around	
  variance	
  
within	
  a	
  stock's	
  returns,	
  relative	
  to	
  the	
  market,	
  than	
  specifically	
  downside	
  risk.	
  Take	
  
for	
  example	
  that	
  ATI	
  had	
  the	
  highest	
  upside	
  risk	
  as	
  measured	
  by	
  its	
  maximum	
  
observation	
  and	
  other	
  variables,	
  but	
  has	
  a	
  far	
  higher	
  estimated	
  beta	
  than	
  GME	
  -­‐	
  the	
  
stock	
  with	
  the	
  highest	
  downside	
  risk	
  in	
  our	
  portfolio	
  based	
  on	
  the	
  same	
  measures.	
  
Similarly,	
  beta	
  does	
  not	
  communicate	
  well	
  correlation	
  or	
  covariances,	
  measures	
  
with	
  the	
  potential	
  to	
  control	
  portfolio	
  risk	
  much	
  more	
  effectively.	
  
It's	
  a	
  relatively	
  easy	
  to	
  estimate	
  the	
  measure,	
  but	
  beta	
  hardly	
  scratches	
  the	
  surface	
  
in	
  relaibility	
  when	
  it	
  comes	
  to	
  assessing	
  portfolio	
  risk.	
  

More Related Content

Viewers also liked

Improving Returns from the Markowitz Model using GA- AnEmpirical Validation o...
Improving Returns from the Markowitz Model using GA- AnEmpirical Validation o...Improving Returns from the Markowitz Model using GA- AnEmpirical Validation o...
Improving Returns from the Markowitz Model using GA- AnEmpirical Validation o...
idescitation
 
PorfolioReport
PorfolioReportPorfolioReport
PorfolioReportAlbert Chu
 
Group_5_stream_80.docx (1)
Group_5_stream_80.docx (1)Group_5_stream_80.docx (1)
Group_5_stream_80.docx (1)Arthur Qiu
 
Is capm a good predictor of stock return in the nigerian banking stocks
Is capm a good predictor of stock return in the nigerian banking stocksIs capm a good predictor of stock return in the nigerian banking stocks
Is capm a good predictor of stock return in the nigerian banking stocks
Alexander Decker
 
Polyu Reference letter
Polyu Reference letterPolyu Reference letter
Polyu Reference letterKelvin Lam
 
Nduati Michelle Wanjiku Undergraduate Project
Nduati Michelle Wanjiku Undergraduate ProjectNduati Michelle Wanjiku Undergraduate Project
Nduati Michelle Wanjiku Undergraduate ProjectMichelle Nduati
 
Portfolio Optimization Project Report
Portfolio Optimization Project ReportPortfolio Optimization Project Report
Portfolio Optimization Project ReportJohn Cui
 
ma574-final-paper
ma574-final-paperma574-final-paper
ma574-final-paperXuan Ning
 
CAPM Fama French
CAPM Fama FrenchCAPM Fama French
CAPM Fama FrenchAhmedSaba
 
Application of Graphic LASSO in Portfolio Optimization_Yixuan Chen & Mengxi J...
Application of Graphic LASSO in Portfolio Optimization_Yixuan Chen & Mengxi J...Application of Graphic LASSO in Portfolio Optimization_Yixuan Chen & Mengxi J...
Application of Graphic LASSO in Portfolio Optimization_Yixuan Chen & Mengxi J...Mengxi Jiang
 

Viewers also liked (15)

Case 2
Case 2Case 2
Case 2
 
Improving Returns from the Markowitz Model using GA- AnEmpirical Validation o...
Improving Returns from the Markowitz Model using GA- AnEmpirical Validation o...Improving Returns from the Markowitz Model using GA- AnEmpirical Validation o...
Improving Returns from the Markowitz Model using GA- AnEmpirical Validation o...
 
PorfolioReport
PorfolioReportPorfolioReport
PorfolioReport
 
Saha, Final Project
Saha, Final ProjectSaha, Final Project
Saha, Final Project
 
Group_5_stream_80.docx (1)
Group_5_stream_80.docx (1)Group_5_stream_80.docx (1)
Group_5_stream_80.docx (1)
 
Is capm a good predictor of stock return in the nigerian banking stocks
Is capm a good predictor of stock return in the nigerian banking stocksIs capm a good predictor of stock return in the nigerian banking stocks
Is capm a good predictor of stock return in the nigerian banking stocks
 
Polyu Reference letter
Polyu Reference letterPolyu Reference letter
Polyu Reference letter
 
Project Report
Project ReportProject Report
Project Report
 
Project final
Project finalProject final
Project final
 
Portfolio Analysis
Portfolio AnalysisPortfolio Analysis
Portfolio Analysis
 
Nduati Michelle Wanjiku Undergraduate Project
Nduati Michelle Wanjiku Undergraduate ProjectNduati Michelle Wanjiku Undergraduate Project
Nduati Michelle Wanjiku Undergraduate Project
 
Portfolio Optimization Project Report
Portfolio Optimization Project ReportPortfolio Optimization Project Report
Portfolio Optimization Project Report
 
ma574-final-paper
ma574-final-paperma574-final-paper
ma574-final-paper
 
CAPM Fama French
CAPM Fama FrenchCAPM Fama French
CAPM Fama French
 
Application of Graphic LASSO in Portfolio Optimization_Yixuan Chen & Mengxi J...
Application of Graphic LASSO in Portfolio Optimization_Yixuan Chen & Mengxi J...Application of Graphic LASSO in Portfolio Optimization_Yixuan Chen & Mengxi J...
Application of Graphic LASSO in Portfolio Optimization_Yixuan Chen & Mengxi J...
 

Similar to Estimating Beta For A Portfolio Of Stocks

Analyzing_ETF_Financial_Data_In_R
Analyzing_ETF_Financial_Data_In_RAnalyzing_ETF_Financial_Data_In_R
Analyzing_ETF_Financial_Data_In_RGeoffery Mullings
 
Strategy Part 1,2 & 3
 Strategy Part 1,2 & 3 Strategy Part 1,2 & 3
Strategy Part 1,2 & 3demarcog
 
Presentation on Bad Beta, Good Beta
Presentation on Bad Beta, Good BetaPresentation on Bad Beta, Good Beta
Presentation on Bad Beta, Good Beta
Michael-Paul James
 
8_borri.pdf
8_borri.pdf8_borri.pdf
8_borri.pdf
ClaudioTebaldi2
 
The Original Draft Copy of VaR and VaR Derivatives
The Original Draft Copy of VaR and VaR DerivativesThe Original Draft Copy of VaR and VaR Derivatives
The Original Draft Copy of VaR and VaR Derivatives
Ralph 刘冶民 Liu
 
Bayesian Dynamic Linear Models for Strategic Asset Allocation
Bayesian Dynamic Linear Models for Strategic Asset AllocationBayesian Dynamic Linear Models for Strategic Asset Allocation
Bayesian Dynamic Linear Models for Strategic Asset Allocation
max chen
 
QNBFS Daily Technical Trader - Qatar for October 03, 2017
QNBFS Daily Technical Trader - Qatar for October 03, 2017QNBFS Daily Technical Trader - Qatar for October 03, 2017
QNBFS Daily Technical Trader - Qatar for October 03, 2017
QNB Group
 
Fixed Factors, Terms of Trade, and Growth in Unbalanced Productive Structures
Fixed Factors, Terms of Trade, and Growth in Unbalanced Productive StructuresFixed Factors, Terms of Trade, and Growth in Unbalanced Productive Structures
Fixed Factors, Terms of Trade, and Growth in Unbalanced Productive Structures
pkconference
 
QNBFS Daily Technical Trader - Qatar for September 19, 2017
QNBFS Daily Technical Trader - Qatar for September 19, 2017QNBFS Daily Technical Trader - Qatar for September 19, 2017
QNBFS Daily Technical Trader - Qatar for September 19, 2017
QNB Group
 
Global Equity Market Neutral Factsheet_End of April 2020
Global Equity Market Neutral  Factsheet_End of April 2020Global Equity Market Neutral  Factsheet_End of April 2020
Global Equity Market Neutral Factsheet_End of April 2020
Giuseppe Piazzolla
 
QNBFS Daily Technical Trader - Qatar for January 03, 2018
QNBFS Daily Technical Trader - Qatar for January 03, 2018QNBFS Daily Technical Trader - Qatar for January 03, 2018
QNBFS Daily Technical Trader - Qatar for January 03, 2018
QNB Group
 
QNBFS Daily Technical Trader - Qatar for October 04, 2017
QNBFS Daily Technical Trader - Qatar for October 04, 2017QNBFS Daily Technical Trader - Qatar for October 04, 2017
QNBFS Daily Technical Trader - Qatar for October 04, 2017
QNB Group
 
QNBFS Daily Technical Trader - Qatar for October 05, 2017
QNBFS Daily Technical Trader - Qatar for October 05, 2017QNBFS Daily Technical Trader - Qatar for October 05, 2017
QNBFS Daily Technical Trader - Qatar for October 05, 2017
QNB Group
 
Daily news letter equity 14 jan2013
Daily news letter equity 14 jan2013Daily news letter equity 14 jan2013
Daily news letter equity 14 jan2013TheEquicom Advisory
 
Daily sgx report by epic research 26 august 2016
Daily sgx report by epic research 26 august 2016Daily sgx report by epic research 26 august 2016
Daily sgx report by epic research 26 august 2016
epicresearchsgmy
 
QNBFS Daily Technical Trader - Qatar June 29, 2016
QNBFS Daily Technical Trader - Qatar June 29, 2016QNBFS Daily Technical Trader - Qatar June 29, 2016
QNBFS Daily Technical Trader - Qatar June 29, 2016
QNB Group
 

Similar to Estimating Beta For A Portfolio Of Stocks (20)

Analyzing_ETF_Financial_Data_In_R
Analyzing_ETF_Financial_Data_In_RAnalyzing_ETF_Financial_Data_In_R
Analyzing_ETF_Financial_Data_In_R
 
Strategy Part 1,2 & 3
 Strategy Part 1,2 & 3 Strategy Part 1,2 & 3
Strategy Part 1,2 & 3
 
Final Paper
Final PaperFinal Paper
Final Paper
 
Presentation on Bad Beta, Good Beta
Presentation on Bad Beta, Good BetaPresentation on Bad Beta, Good Beta
Presentation on Bad Beta, Good Beta
 
Trading on SPY
Trading on SPYTrading on SPY
Trading on SPY
 
8_borri.pdf
8_borri.pdf8_borri.pdf
8_borri.pdf
 
Gsk
GskGsk
Gsk
 
Case Studies
Case StudiesCase Studies
Case Studies
 
The Original Draft Copy of VaR and VaR Derivatives
The Original Draft Copy of VaR and VaR DerivativesThe Original Draft Copy of VaR and VaR Derivatives
The Original Draft Copy of VaR and VaR Derivatives
 
Bayesian Dynamic Linear Models for Strategic Asset Allocation
Bayesian Dynamic Linear Models for Strategic Asset AllocationBayesian Dynamic Linear Models for Strategic Asset Allocation
Bayesian Dynamic Linear Models for Strategic Asset Allocation
 
QNBFS Daily Technical Trader - Qatar for October 03, 2017
QNBFS Daily Technical Trader - Qatar for October 03, 2017QNBFS Daily Technical Trader - Qatar for October 03, 2017
QNBFS Daily Technical Trader - Qatar for October 03, 2017
 
Fixed Factors, Terms of Trade, and Growth in Unbalanced Productive Structures
Fixed Factors, Terms of Trade, and Growth in Unbalanced Productive StructuresFixed Factors, Terms of Trade, and Growth in Unbalanced Productive Structures
Fixed Factors, Terms of Trade, and Growth in Unbalanced Productive Structures
 
QNBFS Daily Technical Trader - Qatar for September 19, 2017
QNBFS Daily Technical Trader - Qatar for September 19, 2017QNBFS Daily Technical Trader - Qatar for September 19, 2017
QNBFS Daily Technical Trader - Qatar for September 19, 2017
 
Global Equity Market Neutral Factsheet_End of April 2020
Global Equity Market Neutral  Factsheet_End of April 2020Global Equity Market Neutral  Factsheet_End of April 2020
Global Equity Market Neutral Factsheet_End of April 2020
 
QNBFS Daily Technical Trader - Qatar for January 03, 2018
QNBFS Daily Technical Trader - Qatar for January 03, 2018QNBFS Daily Technical Trader - Qatar for January 03, 2018
QNBFS Daily Technical Trader - Qatar for January 03, 2018
 
QNBFS Daily Technical Trader - Qatar for October 04, 2017
QNBFS Daily Technical Trader - Qatar for October 04, 2017QNBFS Daily Technical Trader - Qatar for October 04, 2017
QNBFS Daily Technical Trader - Qatar for October 04, 2017
 
QNBFS Daily Technical Trader - Qatar for October 05, 2017
QNBFS Daily Technical Trader - Qatar for October 05, 2017QNBFS Daily Technical Trader - Qatar for October 05, 2017
QNBFS Daily Technical Trader - Qatar for October 05, 2017
 
Daily news letter equity 14 jan2013
Daily news letter equity 14 jan2013Daily news letter equity 14 jan2013
Daily news letter equity 14 jan2013
 
Daily sgx report by epic research 26 august 2016
Daily sgx report by epic research 26 august 2016Daily sgx report by epic research 26 august 2016
Daily sgx report by epic research 26 august 2016
 
QNBFS Daily Technical Trader - Qatar June 29, 2016
QNBFS Daily Technical Trader - Qatar June 29, 2016QNBFS Daily Technical Trader - Qatar June 29, 2016
QNBFS Daily Technical Trader - Qatar June 29, 2016
 

Estimating Beta For A Portfolio Of Stocks

  • 1. Estimating  Beta  For  A  Portfolio  Of  Stocks   Geoffery  Mullings   March  29,  2016   Executive  Summary:   For  some  a  stock's  beta  can  be  used  as  a  proxy  for  the  asset's  systemic  risk  -­‐  that  is,   how  vulnerable  the  stock  is  to  movements  in  the  market.  This  analysis  estimates   beta  coefficients  for  a  portfolio  of  stocks  while  interpreting  and  comparing  those   values  to  more  standard  measures  of  risk  such  as  variance  and  correlation.  While  to   a  certain  extent  beta  represents  systemic  risk  its  ability  to  proxy  for  risk  is  quite   limited.  The  difference  between  upside  and  downside  risk  is  difficult  to  attain  from   beta,  while  covariance  offers  much  more  flexibility  in  portfolio  risk  managment.   Introduction:   A  stock's  beta  can  be  a  convenient  proxy  for  risk  depending  on  many  factors.  Beta  is   a  measure  of  comovement  with  the  market,  often  interpreted  as  a  measure  of   systemic  risk.  If  a  stock  moves  more  than  the  market  it  will  have  a  higher  beta,  while   the  reverse  is  true  for  a  lower  beta.  The  risk-­‐reward  tradeoff  framework  suggests   that  stocks  with  lower  betas  may  be  safer  but  should  provide  less  opportunity  for   large  gains.  That  will  be  tested  in  the  analysis  below.   Beta  can  be  estimated  theoretically  through  a  regression  model,  perhaps  even  a   linear  regression  depending  on  the  movement  of  the  stock.  It's  also  possible  to   observe  historical  data  and  assess  whether  beta  accurately  reflects  our  normal   understanding  of  "risk."   Data  Loading  and  Munging:   require(tseries)   require(zoo)   require(fBasics)   #  Retrieving  the  Market  Factor.  Using  the  S&P  500  as  a  proxy  for  the   market's  movements.   sp  <-­‐  get.hist.quote("^GSPC",  start='1989-­‐12-­‐01',  quote="AdjClose",   compression="m",  quiet=TRUE)     mytick  =  c('ATI','TSN','GME','LB','PAYX')  #  Storing  all  of  the  stocks   in  a  variable  for  looping.     data  <-­‐  get.hist.quote(mytick[1],  quote="AdjClose",  start="1989-­‐12-­‐01",   compression="m",  quiet=TRUE,  retclass="zoo")  
  • 2.   for  (i  in  2:length(mytick))         {            temp  =  get.hist.quote(mytick[i],  quote="AdjClose",  start="1989-­‐12-­‐ 01",  compression="m",  quiet=TRUE,  retclass="zoo")                                                                                                        data  =  cbind(data,  temp)                                                       }     colnames(data)  =  mytick     #  Transforming  into  logarithmic  returns.  Log  returns  are  easier  to   compare  over  time,  and  accurate  as  long  as  returns  aren't  too  large.   ret  =  100  *  diff(log(data))   sp  =  100  *  diff(log(sp))   colnames(sp)  =  "SP.500"   StocksMarketRet  =  cbind(ret,sp)     sum(is.na(StocksMarketRet))  #  Any  missing  return  values?   ##  [1]  295   There  are  some  NA  values  in  the  historical  data  because  each  stock  had  its  IPO  after   1990.  To  reduce  bias  in  our  analysis  we'll  eliminate  all  observations  preceding  the   IPO  of  the  youngest  stock  in  the  portfolio,  GME.   StocksMarketRetAdj  =  StocksMarketRet[151:318,]  #  Eliminating  the  first   149  observations  from  all  assets  to  reduce  bias.     sum(is.na(StocksMarketRetAdj))   ##  [1]  0   Let's  first  analyze  some  of  the  basic  statistics  about  this  portfolio's  and  the  S&P   500's  logarithmic  return.   Analysis:   basicStats(StocksMarketRetAdj)   ##                                        ATI                TSN                GME                  LB              PAYX   ##  nobs                168.000000  168.000000  168.000000  168.000000  168.000000   ##  NAs                      0.000000      0.000000      0.000000      0.000000      0.000000   ##  Minimum          -­‐50.573656  -­‐31.198382  -­‐64.077922  -­‐36.862335  -­‐16.913253   ##  Maximum            47.759670    26.659710    24.203372    27.204306    17.582045   ##  1.  Quartile    -­‐8.765500    -­‐4.404103    -­‐5.359523    -­‐3.520060    -­‐3.848341   ##  3.  Quartile      9.622867      6.204392      8.764377      7.757846      4.966828  
  • 3. ##  Mean                    0.160533      1.092622      0.737814      1.363810      0.436373   ##  Median              -­‐0.441562      1.976514      1.494889      1.851768      1.036216   ##  Sum                    26.969589  183.560521  123.952835  229.120095    73.310595   ##  SE  Mean              1.267847      0.714102      0.954303      0.711278      0.498472   ##  LCL  Mean          -­‐2.342541    -­‐0.317208    -­‐1.146238    -­‐0.040446    -­‐0.547746   ##  UCL  Mean            2.663608      2.502453      2.621867      2.768066      1.420491   ##  Variance        270.049434    85.670170  152.996535    84.994034    41.743632   ##  Stdev                16.433181      9.255818    12.369177      9.219221      6.460931   ##  Skewness            0.011003    -­‐0.400221    -­‐1.015992    -­‐0.601945    -­‐0.147011   ##  Kurtosis            1.055338      1.152953      3.521931      1.594190    -­‐0.326907   ##                                  SP.500   ##  nobs                168.000000   ##  NAs                      0.000000   ##  Minimum          -­‐18.563647   ##  Maximum            10.230659   ##  1.  Quartile    -­‐1.773925   ##  3.  Quartile      2.984705   ##  Mean                    0.341674   ##  Median                1.018362   ##  Sum                    57.401290   ##  SE  Mean              0.331765   ##  LCL  Mean          -­‐0.313320   ##  UCL  Mean            0.996669   ##  Variance          18.491439   ##  Stdev                  4.300167   ##  Skewness          -­‐0.868752   ##  Kurtosis            1.923297   There's  some  upside  risk  in  ATI  (a  small  cap  metals  and  mining  firm)  based  on  the   mean,  median,  and  high  variation,  if  not  the  maximum  observation.  This  could  be   due  to  its  relative  youth  on  the  market,  similar  to  GME  (the  infamous  small  cap   video  game  retailer,  GameStop),  which  also  exhibits  a  high  variance  and  many  low   returns,  evidenced  by  a  large  negative  skew  and  minimum  value.  High  variance  is   also  not  particularly  uncommon  for  technology  stocks.   All  but  ATI  and  PAYX  (a  large  cap  IT  firm)  seem  negatively  skewed  based  on  mean   and  median  values,  with  the  largest  tails  present  on  GME  (which  also  happens  to  be   the  second-­‐highest  stock  in  variance)  based  on  its  kurtosis  value.   LB  (a  large  cap  retailer  of  women's  underwear)  and  TSN  (the  large  cap  food   producer  Tyson  Foods)  share  with  the  other  stocks  high  variances  relative  to  the   market,  accompanied  by  higher  mean  returns  -­‐  except  for  ATI.   Every  stock  in  the  portfolio  has  a  higher  return  variance  than  the  market.  Alongside   variance,  comovement  between  assets  and  the  market  should  be  assessed  on  the   road  to  estimating  beta.   plot(as.zoo(StocksMarketRetAdj),  main="Portfolio  Stock  Market  Returns")  
  • 4.   plot(as.data.frame(StocksMarketRetAdj),  main="Portfolio  Stock  Market   Returns  Plotted  Against  Each  Other")  
  • 5.   Visually,  there's  evidence  of  comovement  between  all  the  stocks  and  the  market   factor.  Obviously  some  stock  return  data  exhibits  more  varied  comovement  than   others,  and  finding  these  differences  between  the  asset  and  the  market  factor's   comovement  will  give  us  the  sought-­‐after  estimate  of  beta.   cor(StocksMarketRetAdj)  #  Would  have  included  use="pairwise.complete"   if  I  hadn't  eliminated  the  missing  observations  already.   ##                            ATI              TSN              GME                LB            PAYX        SP.500   ##  ATI        1.0000000  0.2801193  0.2450891  0.3920166  0.2263508  0.5653850   ##  TSN        0.2801193  1.0000000  0.1623045  0.3337894  0.2221864  0.3987431   ##  GME        0.2450891  0.1623045  1.0000000  0.3515588  0.2799321  0.3822975   ##  LB          0.3920166  0.3337894  0.3515588  1.0000000  0.4319816  0.6131435   ##  PAYX      0.2263508  0.2221864  0.2799321  0.4319816  1.0000000  0.5867857   ##  SP.500  0.5653850  0.3987431  0.3822975  0.6131435  0.5867857  1.0000000   fit  =  lm(StocksMarketRetAdj[,mytick]  ~  StocksMarketRetAdj$SP.500)   coef(summary(fit))   ##  Response  ATI  :   ##                                                          Estimate  Std.  Error        t  value           Pr(>|t|)   ##  (Intercept)                              -­‐0.5776988      1.052226  -­‐0.5490256   5.837256e-­‐01  
  • 6. ##  StocksMarketRetAdj$SP.500    2.1606307      0.244650    8.8315179   1.411880e-­‐15   ##     ##  Response  TSN  :   ##                                                        Estimate  Std.  Error    t  value          Pr(>|t|)   ##  (Intercept)                              0.7993742    0.6589283  1.213143  2.267986e-­‐01   ##  StocksMarketRetAdj$SP.500  0.8582674    0.1532055  5.602066  8.630383e-­‐08   ##     ##  Response  GME  :   ##                                                        Estimate  Std.  Error    t  value          Pr(>|t|)   ##  (Intercept)                              0.3620902    0.8872695  0.408095  6.837298e-­‐01   ##  StocksMarketRetAdj$SP.500  1.0996562    0.2062965  5.330465  3.158340e-­‐07   ##     ##  Response  LB  :   ##                                                        Estimate  Std.  Error      t  value           Pr(>|t|)   ##  (Intercept)                              0.9146684    0.5653664    1.617833  1.075972e-­‐ 01   ##  StocksMarketRetAdj$SP.500  1.3145315    0.1314517  10.000109  1.006882e-­‐ 18   ##     ##  Response  PAYX  :   ##                                                        Estimate  Std.  Error      t  value           Pr(>|t|)   ##  (Intercept)                              0.1351402  0.40613107  0.3327502  7.397427e-­‐ 01   ##  StocksMarketRetAdj$SP.500  0.8816360  0.09442837  9.3365581  6.361038e-­‐ 17   The  p  and  t-­‐values  of  all  the  coefficients  suggests  that  each  is  statistically  significant   and  most  likely  a  reliable  estimate  of  their  respective  betas.   Notably  even  though  TSN  and  GME  have  some  of  the  weakest  correlations  with  the   market  factor,  they  have  betas  that  are  very  close  to  1.  While  correlation  shows  the   strength  of  return  similarities  between  the  stock  and  the  market,  beta  really   estimates  the  effect  of  a  market  movement  on  each  asset.   Conclusion:   The  preceding  has  hopefully  made  it  clear  that  beta  provides  some  utility   accompanied  by  many  limits.  Beta  is  much  better  at  capturing  all  around  variance   within  a  stock's  returns,  relative  to  the  market,  than  specifically  downside  risk.  Take   for  example  that  ATI  had  the  highest  upside  risk  as  measured  by  its  maximum   observation  and  other  variables,  but  has  a  far  higher  estimated  beta  than  GME  -­‐  the   stock  with  the  highest  downside  risk  in  our  portfolio  based  on  the  same  measures.   Similarly,  beta  does  not  communicate  well  correlation  or  covariances,  measures   with  the  potential  to  control  portfolio  risk  much  more  effectively.  
  • 7. It's  a  relatively  easy  to  estimate  the  measure,  but  beta  hardly  scratches  the  surface   in  relaibility  when  it  comes  to  assessing  portfolio  risk.