3

I am trying to get some advice or direction (brainstorm) as to the best way to summarise/cluster/etc. the equity fund universe (which for my purposes consists of about 150 funds).

Some of my ideas at current:

-I have access to a Value and Growth Index so could perhaps try segregate funds into value/momentum/growth groups by looking at their betas to these indices. Could also do the same to estimate the funds' exposures to certain sectors: Large Cap, Mid Cap, Small Cap; Resources, Financials, Industrials; etc.

-regression on index to see which funds are closet benchmark trackers (or look at tracking error)

-maybe PCA on the equity fund universe to see which funds are more dissimilar to universe...

I am really just looking for several ways of analysing the funds and try to get some groupings or measures of similarity between the funds. Any pointers of what to read up on or suggestions would be appreciated

PS: I would be using R for this work so if there are any R-specific libraries to look at then please notify

NickF
  • 85
  • 3
  • 1
    perhaps look into k-means clustering? – bcf Nov 17 '15 at 14:33
  • 2
    The "BEST" way will strongly depend on what you want to achieve by this. You can analyse funds by cost, by past performance by active share by fund management company and so on and so on. So what are your goals when clustering? – g g Jun 15 '16 at 13:01

2 Answers2

1

Capital IQ has an existing classifications by asset class, country, family, method, region, sector, size, and style. These definitely are not clever classifications, but form the baseline for fund taxonomy.

  • AssetClass 
    Alternative   ALTERN
Commodities   COMMOD
Currencies    CURR
Equity    EQUITY
Fixed Income  FIXINC
Mixed Assets  MIXASST
  • country follows ISO standards

  • Family

    BLDRS BLDRS
Claymore  CLAYMORE
Currency Shares (Rydex)   CURRENCY
Direxion  DIREXION
Exchange Traded Notes ETNS
First Trust   FTRUST
FocusShares   FOCUS
HealthShares  HEALTHSH
HOLDRS    HOLDRS
iShares   ISHARES
Macro Shares  MACROSH
Market Vectors    MVECTORS
Miscellaneous MISCL
NETS (Northern Trust) NETS
PowerShares   POWERSH
ProShares PROSH
Realty Funds  REALTY
RevenueShares REVSH
Rydex RYDEX
SPA   SPA
SPDR  SPDR
TDX Independence  TDX
United States Trust   USTRUST
Vanguard  VANGUARD
Wisdom Tree   WISDOM

  • Method

Hedged    HEDGED
Leveraged Long    LEVLONG
Leveraged Short   LEVSHORT
Quant Model   QUANT
Special Weights   SPWEIGHTS
Standard Long STANLONG
Standard Short    STANSHORT
  • Region
Asia  ASIA
BRIC-Chindia  BRIC
Developed DEVELOP
Emerging  EMERG
Europe    EUROPE
Global    GLOBAL
Global Ex US  GLOBALXUS
Latin America LATIN
MidEast-Africa    MIDEAST
North America NAMERICA
Pacific Ex Japan  PACIFIC
  • Sector
Agriculture   AGRIC
Alternate Energy  ALTENERGY
Consumer  CONSUMER
Energy    ENERGY
Financial FINANCIAL
General   GENSECT
Healthcare    HEALTHCAR
Housing   HOUSING
Industrials   INDUST
Infrastructure    INFRASTR
Materials MATERIALS
Municipal fixed inc   MUNIS
Precious Metals   PRECIOUS
Real Estate   REALEST
Resources (General)   RESOURC
Services  SERVICES
Social    SOCIAL
Special Theme SPECIAL
Taxable Fixed Inc TXFIXINC
Technology    TECHNOL
Telecomm  TELECOMM
Timber    TIMBER
Transportation    TRANSPORT
Utilities UTILITIES
Water WATER
  • Size
General   GENSIZE
Large-Mega    ETF_LARGECAP
Mid   ETF_MIDCAP
Small-Micro   ETF_SMALLCAP
  • Style
Equity Income EQINCOME
General   GENSTYLE
General Fixed Inc GENFIXINC
Growth    GROWTH
High Yield Fixed Inc  HIGHYLD
Intermediate Fixed Inc    INTFIXINC
Long Fixed Inc    LTFIXINC
Short Fixed Inc   STFIXINC
Value VALUE
David Addison
  • 2,995
  • 15
  • 32
0

for model construction purposes, I would suggest factor analysis or PCA. Have a look at EM algorithm

owner
  • 315
  • 1
  • 6
  • 3
    The EM algorithm is rather general. You mean EM-clustering. Would usual clustering (k-means, hyrarchical) do the same? Which distance measure would you propose? – Richi Wa Mar 17 '16 at 08:16