Lvng arrangement decsons for elderly care n Italy Matteo Lpp Brun, Crstna Ugoln matteo.lppbrun@unbo.t crstna.ugoln@unbo.t Department of Economcs - Unversty of Bologna nd Internatonal Conference on Evdence-based Polcy n Long-term Care 5th - 8th September 0 London
Background The need to respond to the rase n the demand for LTC servces s creatng ncreasng concerns among ctzens and polcymakers n Italy Compared to the other OECD countres, the Italan populaton ranks among the oldest ones as a consequence of: Very low fertlty rate: ncreased lfe expectancy. Most LTC s stll provded by famly caregvers, but households are less lkely to be able to provde care drectly n the future. Several factors contrbute to erode the potental of nformal caregvng wthn the famly: reducton n households sze declne n famly tes ncreased women labour-force partcpaton
Polcy ssues These trends are common to most developed countres and they rase concerns that are recevng ncreasng attenton n the economc lterature They nvolve mportant polcy ssues such as: The assessment of the sustanablty of current welfare systems; The need to contan rsks of early hosptalsaton; problems n ensurng approprate and effectve settngs for carng actvtes; Interplay between publc acton and prvate support (crowdng out?); substtutablty/ complementarty between formal and nformal care
The focus of the paper One of the areas of nvestgaton concerns the determnants of famly decsons regardng lvng arrangements of elderly people affected by lmted autonomy or dsabltes. Man focus of our work s on the role of pad caregvng when elderly people are asssted at home. Formal vs. Informal care typcally grounds on the assumpton that formal care provded by pad helpers occurs exclusvely when the elderly person resdes n a lvng asssted faclty..e. n the tradtonal approach, nformal and home care tend to concde. Clam: when a dependent person s kept at home, the choce between nformal and pad care s an ncreasngly relevant ssue.
Bref outlne of the lterature The theoretcal lterature vares along several dmensons: common preferences [Kotlkoff and Morrs, NBER 988; Hoerger et al., Rev Econ Stat 996] or famly barganng [Stern, J Hum Resour 995; Pezzn and Schone Am Econ Rev 997; Engers and Stern, Int Econ Rev 00]; the type of care (formal or nformal) or lvng arrangements consdered (shared housng, lve ndependently, nursng home), role of chldren n the decson process, etc. The emprcal lterature s as vared as the theoretcal one. It has been conducted wth a varety of econometrc methods, but dsplays remarkably consstent results. The majorty of works relates to the Unted States but a recent stream of research has developed also cross country comparsons across Europe.
The Italan context Snce the late 80 s Italy experenced sgnfcant mgraton flows from Eastern Europe and Afrca, mostly undocumented and llegal, wth a hgh female component that joned the nformal labour market. Two thrd of these women are engaged n domestc or personal care. For many Italan famles, the possblty to delegate home caregvng by buyng assstance at a lower prce wth respect to professonal servces reduced sgnfcantly admsson n nursng homes. n most of cases affordablty was acheved explotng the black market; Socal norms about flal responsblty stll tend to attach a consstent amount of socal stgma to the nsttutonalsaton of the elderly.
The Survey Data are drawn from a cross-sectonal survey carred out n year 00 on a representatve sample of 405 famles of the Italan regon Emla Romagna (around 4 mllons nhabtants). The man purpose of the questonnare was to elct WTP for coverng LTC expendture rsk. Publc vs prvate nsurance (Brau. Lpp Brun PInna 00, Appled Economcs, Brau Lpp Brun, 008 Health Economcs) For the present analyss we extract nformaton from a specfc secton of the questonnare devoted to regster the presence of a dsabled person aged 50 or more n the household (ether co-resdent or not).
The choce model We assume common preferences among famly members and lvng arrangement decsons taken once-and-for-all. We record a total of 79 households nvolved n the assstance of an elderly dependent, 3 ndvduals lve at home, 79 receve nformal home care, 5 receve pad home care 48 are nsttutonalsed. The decson process of the household can be seen as: smultaneous; sequental:
THE DECISION TREE- Smultaneous decsons Household The decson process can be seen as smultaneous. The household has three alternatves: Resdental Care Famly Home Care Pad Home Care. to nsttutonalse the elderly n a lvng asssted faclty (Resdental Care, RC) ;. to provde care at home through nformal famly support (Famly Home Care, FHC); 3. to hre an external caregver to provde care at home (Pad Home Care, PHC).
THE DECISION TREE- Smultaneous decsons Accordng to the smultaneous decson scheme, each household faces three non-ordered alternatves. For each alternatve, ndrect utlty s composed by a determnstc component (xb) and a stochastc error term e. The underlyng utlty assocated wth each alternatve s not observable but we can estmate the probablty of choosng a gven alternatve by modellng choce process as follows: We explot the observed choce of a specfc lvng arrangement to estmate the set of relevant parameters b
Multnomal Logt and IIA The MultNomal Logt (MNL) specfcaton s typcally used to estmate the model prevously descrbed. Its man lmtaton s that t reles on the IIA assumpton for dentfcaton of the assocated parameter vector. IIA s potentally questonable n ths context, where two alternatves (famly care and pad home care) dsplay larger smlartes compared to the thrd one (nursng home) We estmate also a Multnomal Probt specfcaton, that, despte beng computatonally cumbersome, t allows to relax the IIA assumpton The var/cov matrx s no longer restrcted to be dagonal
THE DECISION TREE Sequental decsons Household The decson process can be seen as artculated n two steps. Resdental Care Home Care The household decdes FIRST whether to nsttutonalse the elderly n a lvng asssted faclty (Resdental Care, RC) or to provde care at home (Home Care, HC). Famly home care Pad home care For those who stay at home, the household decdes whether to provde care drectly (Famly home care, FHC) or to hre a external caregver (pad home care, PHC).
Bv-probt wth sample selecton (e.g. van de Ven, van Pragg, 98 Journal of Econometrcs) =.N t=, j = RC vs HC; PHC vs.ihc ) ( ) ( ) ( ) ( * * IHC PHC IHC PHC RC HC RC HC v X X y v X X y e e b b e e b b jt t jt jt X U e b st Stage y = f Home Care ; y =0 f Resdental Care nd Stage y = f Pad H C; y =0 f Informal H C mssng nformaton f y =0 0 0,, log,, log,, log y y y y y x x x x x
The Data DE characterstcs Age, Sex, Lved alone = f the DE lved alone before becomng dsable; LTC spell Spell of dsablty n years Num ADL Number of ADLs and IADLs the DE s not able to perform Meal = f the DE s unable to prepare meals Heavy help = Publc support for > 40 days (dentfcaton varable) Famly characterstcs House = f the famly owns the house Household ncome Household ncome n Euro (HH+ HH spouse, f present) Eld_rato members >65 / total number of household members Resdence = f the prmary determnant of the resdence choce was the wll to lve close to other members of the famly (famly tes) Head_Age Age of the head of the household Head_Chronc = f the head of the household suffers of chronc condtons
Multnomal probt / RESIDENTIAL CARE(vs.Pad HC) Coef. Std. Err. Age DE.000.00 Sex DE -.936.3939 LTC spell.09.054 Lved alone.485.406 Heavy help.0646.50** Num ADL.6.0737** Cookng meals.0869.509** HH Age -.067.066 HH Chronc -.95.396 Elderly rato.749.057 Famly sze.0538.585 Famly ncome.0000.000 House ownershp -.4057.4689 Town>5000 -.9043.6985*** Town<5000.4696.7050** Unversal access.74.4087 Need-based access -.4375.36 Resdence choce -.40.456 Cash care.0864.3777 Cash care -.357.4499 Constant -.0884.065
Multnomal probt / INFORM H-CARE(vs.Pad HC) Coef. Std. Err. Age DE -.0449.064*** Sex DE -.3675.3363 LTC spell -.00.03 Lved alone -.795.3465** Heavy help -.43.4884 Num ADL -.055.065 Cookng meals -.60.3545* HH Age -.048.047 HH chronc -.6350.3480* Elderly rato.433.77** Famly sze.53.44* Famly ncome -.0006.000*** House ownershp.0586.3784 Town>5000 -.0665.4836 Town<5000.49.689** Unversal access -.9055.330*** Need-based access.575.37* Resdence choce.9400.369** Cash care.395.335 Cash care -.399.4435*** Constant 6.5933.5457***
Probt model wth sample selecton / HOME CARE(vs. RESID CARE) Coef. Std. Err. Age DE -.036.07* Sex DE.65.656 LTC spell -.048.0084*** Lved alone -.7946.95*** Heavy help -.0504.39*** Num ADL -.96.05** Cookng meals -.74.3967*** Age HH.030.0089 Chronc HH.0634.797 Elderly rato.0398.609 Famly sze.0.0980 Famly ncome -.0003.000** House ownershp.337.836 Town>5000.7060.5975*** Town<5000 -.3677.4494 Unversal access -.6607.898** Need-based access.668.600** Resdence choce.8983.849*** Cash care.354.35 Cash care -.067.339 Constant 3.6498.353***
PAID H-CARE(vsINF CARE) Probt model wth sample selecton / Coef. Std. Err. Age DE.033.07*** Sex DE.065.40 LTC spell.0033.009 Lved alone.454.73* Num ADL -.046.0468 Cookng meals.580.505 Age HH.09.0080 Chronc HH.479.44* Elderly rato -.80.5088** Famly sze -.55.090* Famly ncome.0004.000*** House ownershp -.0545.946 Town>5000.937.3449 Town<5000-7896.4398* Unversal access.6699.784** Need-based access -.3335.349 Resdence choce -.394.447* Cash care -.96.04 Cash care.4745.834* Constant -4.758.0866***
Econometrc ssues SIMULTANEOUS MODEL Although prevous lterature (e.g. Borsch-Supan et al. 990) support the dea that home based solutons are more strongly correlated compared to resdental care, there s no strkng evdence aganst the IIA hypothess (Hausman test). MNL and MNProbt specfcatons provde farly smlar results. SEQUENTIAL MODEL If H 0 : =0 Separate estmatons generate unbased coeffcents probt RC vs HC on the whole sample probt PHC vs IHC on the subsample where y= Our emprcal evdence does not allow to reject the Independence hypothess, but the result s not robust. Hence we fnd advsable to keep the jont model.
Emprcal Results Severty related varables ncrease the probablty of nsttutonalsaton but have lmted mpact effect on the decson of hrng an external caregver Age s an excepton n that t nfluences the second but not the frst stage decson Lved Alone ncreases the probablty of hrng an external caregver Income does not nfluence the choce between pad home care and resdental care, but low ncome household are more lkely to provde nformal care. Household composton has lmted nfluence on the decson process Poor health condtons of the head of the household ncrease the propensty to recur to external help. Very frequent publc support captures extremely severe cases (hghly nsttutonalsed) Strong famly tes ncrease the probablty of choosng nformal care.
Conclusons Assstng elderly dependent people by means of formal care s an ncreasngly followed strategy also when famles opt for a home care soluton. The determnants of formal care dffer substantally f one consders PHC vs FHC nstead of the more tradtonal choce between RC vs HC. Specfc nvestgaton of the ssue s needed, n partcular today that publc polces strongly encourage home care and that the opportunty cost of nformal care rses. Resdental care becomes the preferred alternatve when health deterorates, a smlar trend does not hold for pad home care. Lvng arrangement decsons are strongly nfluenced by the economc motvatons (low ncome households more frequently opt for nformal care).