GIF89a̙fffff33333!,I8ͻ`(dihlp,tmx|pH,Ȥrl:ШtJZجvzxL.zn|N~@8V8 a%(T@  bĆJhȱcGnXH20dĉDTp2Lqi@I,*D$̣.Qit!|*4QB^TP@.jD FIBLmĺUHBک8jv&ֹC[C[v* $5zG[kpc<[W{.)HWњ> Jmםi!L/ V5&s mjdO ʓh(л˅"ݥ!Ye% QS 2xĀ{,4aUb$$'Ԅh\t C}enpc!z%g"E,=!IE% 85 gSS9VtS7UDԌSquImyC ȇ{]6g 1*l\3tfYۭPCXxOB=>RcP0Xפ^'9t]*\]O ,$1ex}{m9` :s[kfMhfL/'mmsQ){ PV@GQ.Ap[ws,:% oq^ f8!!|AE&Ƒx:ǿd)0 <]g:٭!PD۔z䁾du% FXv HDE%^xaULkF䈧 ix#c22L#ǶaF??T! :#{A!MPcJ>*rk$b١~:" ƞy/dF(HPNJ,SDK\eiL&ʯ ⢣0uD fy@i:N>#!xKeS@Y4,shyl6g&IRT,S πiAFz0ldP2iQxdk֣aw}O dW&}`GO̯ڔDYD {G"פW&)~w_T3A|_ΰv>Vmo{Q^{~|I:cFDa~Ok+D(y<ͨU/D m-,ؗi5"'vb#noҀ  &#7a(6؁ "8$X&x(*,؂.0284X6x8:<؃>@B8DXFxHJp6p00P(pSx^(U\(^XaQ؆ 0rZx`^hH=pȇ}Xp pH"`؈0i8px(`((؈g(Xx[؋V8~NxxxhfHH`xfHhx8X#؍"pe(Ũ ؈0˘XxqobvX昏:Hن)YI!iH&*#و'yh647lXؑs3)6)hF ؆<[M%ЈM\Kۚ ّ=f ]Cz9:0vXݜ <x+Aܾ,-? ڵ-:ߺm>=-]^Pι=ކKiн=a|}%omͬ\}MdMވ͝؋X|۬>n 1ۼߝ:C=xIĽ2( KN]D-Q-T>\l3,:ʪ ͈zm7+iI@ʚ=ms]\F N$톍౎ A^*$>W~(~n,ݞN|띭'!Rm>{=6@~X؍ˎ@ڶ՜ +©̎ŽH" pWmΉN ϑ=5(m︭⯞㘻Nϔu@~_=n>ۍo\OI~亮D-kI4 `ޮ'|+Т`-+9_1);ɲ.\Mi_}/,+)b1|Ժ:M}WH|_?*ƋǍ>m隖/yw?nv* SӦ_'UȌ-Wyc{ޏ# /Sdž*˯/U f&_5@ INYR}:6qԺ".q``shP>: {N? 0 *Ԣ 78KQ0ACVk 1;"Bh"1̏(A(r4Dɯ[rdQ̚oBb)ѡ(o''2qчNLL(۠.\Ql,Ϝ <(O<LgEHb*4\Q`)DosV0U^etawTbZH6ZijsRtjlUT;-JaD#hs=3ZST^>Mw= |?WWlZ}V Wށ8a?9dG&dOF9eWfe_9fgfo9gwg:h&hF:ifi:jj:kk;l&lF;mfm߆;n离n;oox'xG>ygy矇>z駧z>{{?|'|G?}g}߇?~秿~?`8@ЀD`@6Ё`%8A VЂ`5Av(0 $"$@@(B@ @ &< BЃP@81ЁbBD"BD* 07V;TA*E,cи hшM<D3N'v3:t#S%‐Cǒ)RrDr1͑!XI@VQȣK"҉A$!p3~MDDv24m#YfRԘ8g8w~-.QMRG)Ѕ ?SCbPhyKG5XcP I.$pAI7B&3Ae$,EVke!L+6-3@Z-}0+TA8i(3ظ46e+c!63Pb@%ZZlD;ϴ"(%#)Y.+)8gdyT"#QEj 1߸MҀ$EyY $f9UG/nLmiMs=x}d _R%;̡u7[qKevmkdʒ:fiΙOΖh`m5=4yl2(@>$$%kbav~,=JчEu[m2oWߑerOg|Nh7S;us[5x q-j|3tі)r~y}[d#%Qƣ6v?>9ͩzm^<.Δ:ͫtO9oAӛX~ziF}~ܙ޷:7;s.y1v:vz$Gi'z3oUu y'Qݸ j?>ͯg}y_COϙ⛞_~@oG ZjOlyk'Ĩӿ3Y-c@#k:;{@D#[=Ż@_A2A?18C>9R83LӴփ7+)ۺKKq#B=J"@9$%\t!/8BBк/$@8A1-ϻ71ԂC l:,@@6t>k>*8,ړ맢BBH۩LӲJC0 $yK695 C#@Ck6B;I~*3N0F~:*,HBԵ#d/L;4EL*$lt9>$dEm>AD>WdU26ƛk<|GRܤbBo$@I@ GLJ%FC|ES0)):/iCoy\0CI|=w,I:s5W򥉫!lVşܙ2cJ& pJ*X9Yȧ<Z jİc8 G QT_;&k%Bh$BKCdozsCAƍ!qDǿ4\DADBry!ċGDȫ R˛K:2̉Qj0Œ~J4*Z&RIpL̏ 0͌۲dNB2̽5T-M@A#ܜ@HrMT<C#:[-+ν:NKW#ALLl&+PMDTPL̻9:$OF,DzO&-JS O 8ǤIѾf#>J# , 'M;"e3'<p*όPSZ/EC\UD*}EP[P8mQnPěBtS[m2XiVGq ˬM!L$@V]%H 'vC]ҖDЀeTA=Y#3T#Wye0(}-P|6.mS'MXdY&eMZ:0|%,yRZE"*Z#dNZژų$mIXE@Dm(8;YBK[O3S*uCGY`F&`ۼہfIdm#luB[l[T+\@QR$\[|ۇ5Eř)/Sݱ]Rr%U5b݈֕U%]kN]U\*YE:1K8%^٥ŞFpMQ$SPp]WAU_ *_z0BE_&6FVfv `‚_`C`=,EY!J{ݶ\LC ̚]¥PKX9VdŋeRJٱS1ӫIf ε+Ël LDmE#;؄j!$cݤUUM`7<6c,cqx$*3?K3^KO@. / I~G#E~hp t deLJݨv1K1h&GiR,vឭ\)3r,.⠤VUSbB";ElCc4FӰ:m&nE4ޮWmc>Lǫi T{Y0Qdv\L:BX.B=Xf\,*Eh/E\lo~, dTZ.FqjL..BDŽLklmײiMXiudR'SrRvXKaoŇk";e WUk˽r*vJQo*K޺>j-Jԋ%beFDM@(T̝b8Ϟqad$Q1n ĝї>"&n*\'\[eO#[VyNcH^y}<0/vzNfvUg=T ^R) NrVgbS%8C;_^sOUl%ORiK?TxaGuM&)-Z ˾V'zXtumO(> [HO*s&K2fWe+oVtBrIoovPz.VTOV5[ˣfs"~rfog3#mz{s->,p_%/VĔMj-u{2 婗ժOgeXI[dYJKaΨ++,@^B[F_@[`B^O% 'hhJ# Mœ ٭$ WT,)-Ei*ȳ١0ͩ`$qq`rr/H+pyh=z*\fB$R< 4җ"A(7x&#L^ `@Z E0lo&q@Р4zqG"ǒóL,KkPcTb^(KZW;Lx5MTyԔW3+~~MDYvIk5%bM-m ,͏%j\IM&\8퉊ZO^JVc":L@tn:Hc3 0 pF: Tkn08̱r̫(Kh"`A4h.+:4gyWC4 g_ШUĠ#[_}$a~<_HA]gbK{!WhWPgݍEb4rߎFj,JatɜW#{0E;\D$j t%Ywya -0\Xdz/963N>9qډ(7]Y7dYmhMlyjt*pW|#%諆nmXX!;Ԓ֧)~AhyİjaxrDh8Ydm**!81j)L##ZçWnhH,jZ[(*`)+畳{q5AO 8DEK ;,ԹA !&l*{šq O$|A ",BpO紤\hHDK9Iclء>]c1joq6mfku ν$M6yҦ"Gi'5;N"ۏK>9[~9kNy{9衋>:饋ҹ驫:뭻:(>;~;˞;;<<<;<1o[=={0?>>髿>>?????(< 2| #( R 3 r C(&!(!F<"%P2N|"()RV"-r^"(1f<#Ө5n|#(9ұv#=~# )A/;
Wednesday, October 30th 2013

Big Data & Insurance

I like reading Mark Stephens (to separate from others who also use the Cringely persona). Every once in a while he strays into the intersection of his world and health care. In a recent post the argument seems to be that big data has allowed the underwriters to hyperspecifically target to deny or price out individuals at risk.

[I]n the 1990s something happened: the cost of computing came down to the point where it was cost-effective to calculate likely health outcomes on an individual basis. This moved the health insurance business from being based on setting rates to denying coverage. In the U.S. the health insurance business model switched from covering as many people as possible to covering as few people as possible — selling insurance only to healthy people who didn’t much need the healthcare system.

It’s a nice theory but I don’t think this has played a large role in the rise of the uninsured in America. I think its small potatoes in the factors that are driving such.

Many states have specific restrictions on the use of individual data in underwriting. New York and Vermont have essentially pure community rating (at least for some insurers). Many other states have other, less restrictive but still important limitations on the use of individual data in underwriting or on how it can be used to determine premiums. These include Maine, Rhode Island, Massachusetts, Connecticut, New Jersey, Washington, Oregon, Colorado, Minnesota, North Dakota, South Dakota, Iowa, New Hampshire, Montana, Nevada, Utah, Kentucky, Idaho, Louisiana. All of these restrictions put some brakes on the influence of big data in moving underwriting much. Not that such methods aren’t nowadays used by health insurance actuaries but the influence may be overstated.

The fact is that the majority of Americans continue to receive their health insurance through group plans, primarily through their employers. And most of those people get their insurance through large group plans where such individual underwriting has less to no influence on premiums.

The rise in the percentage of uninsured correlates nicely with the decline in employer provided insurance and that reflects the growing costs of premiums. Indeed the individuals buying insurance on the open market, those you would imagine most hurt by complex underwriting techniques based on big data, has remained relatively stable.


Source: U.S. Census Data

There are plenty of factors behind the rise in group premiums which have driven employers to drop health insurance as a benefit but I don’t think the use of big data by actuaries is a major one of them.

Share/Bookmark