The effect of cash transfers on subjective well-being and mental health

This is a cross-post that sum­marises a new work­ing pa­per from the Hap­pier Lives In­sti­tute, The im­pact of cash trans­fers on sub­jec­tive well-be­ing and men­tal health in low- and mid­dle-in­come coun­tries: A sys­tem­atic re­view and meta-anal­y­sis, by Joel McGuire, An­ders Malthe Bach-Mortensen and Cas­par Kaiser.

We know cash trans­fers re­duce poverty, im­prove health and en­hance ed­u­ca­tion but what im­pact do they have on how peo­ple feel and think about their lives? Or, to ask a fa­mil­iar ques­tion: does money make peo­ple happy? The liter­a­ture on the link be­tween in­come and sub­jec­tive well-be­ing has long lacked causal ev­i­dence—there is lots of cor­re­la­tional re­search. Luck­ily, there has been re­cent re­search us­ing cash trans­fers in low and mid­dle-in­come coun­tries. We re­viewed this ev­i­dence in a meta-anal­y­sis. We find, in short, that cash trans­fers have a small, pos­i­tive effect on sub­jec­tive well-be­ing, one that lasts for sev­eral years.

At HLI, we search for the most cost-effec­tive ways to im­prove global well-be­ing. Cash trans­fers to those in low-in­come set­tings are a nat­u­ral place to look: they are a sim­ple and scal­able in­ter­ven­tion and (in part be­cause of this) one of the most ex­ten­sively stud­ied and im­ple­mented in­ter­ven­tions in low- and mid­dle-in­come coun­tries. This makes them a use­ful bench­mark to com­pare the effec­tive­ness of other well-be­ing in­creas­ing op­por­tu­ni­ties against.

In this work­ing pa­per, we sys­tem­at­i­cally re­viewed all the available liter­a­ture on the effect of cash trans­fers on sub­jec­tive well-be­ing[1] in low- and mid­dle-in­come coun­tries.[2] We in­cluded var­i­ous types of cash trans­fers, such as un­con­di­tional or con­di­tional trans­fers, and trans­fers paid out in one go or monthly.

In or­der to as­sess cost-effec­tive­ness, we need to es­ti­mate the to­tal effect on sub­jec­tive well-be­ing. In the fol­low­ing sum­mary, we briefly de­scribe the main re­sults in the pa­per that en­able this anal­y­sis, and in­clude an es­ti­mate of the to­tal effect in terms of Well-be­ing Ad­justed Life-Years, or WELLBYs, which are ex­plained fur­ther be­low.

Our search method re­vealed 1,147 po­ten­tially rele­vant stud­ies; 38 of these were deemed rele­vant for our in­ves­ti­ga­tions and in­cluded in the anal­y­sis. We col­lected their re­sults, stan­dard­ized the effects and took their weighted av­er­age. The av­er­age effect was 0.10 stan­dard de­vi­a­tions (95% con­fi­dence in­ter­val: 0.08-0.12), mea­sured af­ter an av­er­age of two years. This is shown as the black di­a­mond at the bot­tom of the plot be­low. By the stan­dards of so­cial sci­ence, this is a small effect size,[3] but it’s worth not­ing that this re­sult is found af­ter two years, which is an un­usu­ally long fol­low up pe­riod. This sug­gests cash trans­fers do make peo­ple hap­pier and do so for a sub­stan­tial length of time. The pre­dicted in­ter­val, rep­re­sented by the dashed line on ei­ther side of the black di­a­mond, shows that 95% of fu­ture stud­ies are ex­pected to find cash trans­fers have a pos­i­tive effect on sub­jec­tive well-be­ing.

Note: ‘Mo. af­ter start’ is the av­er­age num­ber of months since the cash trans­fer be­gan. ‘$PPP monthly’ is the av­er­age monthly value of a CT in pur­chas­ing power par­ity ad­justed US 2010 dol­lars. Lump sum cash trans­fers were con­verted to monthly value by di­vid­ing the mean fol­low-up time by 24 months.

The cash trans­fers vary in sev­eral ways, in­clud­ing their size and whether they were paid out in one go (lump sum) or at reg­u­lar in­ter­vals (stream). We at­tempted to con­trol for fac­tors such as these by con­duct­ing a meta-re­gres­sion. This is a way to es­ti­mate how cer­tain vari­ables (such as the size of cash trans­fer) mod­er­ate the effect size. A meta-re­gres­sion be­haves similarly to a stan­dard re­gres­sion.[4] We next de­scribe the main re­sults from the meta-re­gres­sion, which can be seen in Table 2 of the work­ing pa­per. We also illus­trate some of the re­la­tion­ships in figures be­low; note that the re­gres­sions on the figures do not con­trol for other vari­ables, un­like the re­sults from the meta-re­gres­sion (Table 2).

Firstly, and un­sur­pris­ingly, the size of a cash trans­fer mat­ters! It is a statis­ti­cally sig­nifi­cant pre­dic­tor of the effect size, both for ab­solute (the dol­lar value) and rel­a­tive size (the pro­por­tion of pre­vi­ous con­sump­tion). Dou­bling in­come for a year leads to a 0.11 stan­dard de­vi­a­tion in­crease in our sub­jec­tive well-be­ing in­dex[5] and a trans­fer equiv­a­lent to a monthly value of $100 PPP re­ceived over two years[6] leads to an in­crease in SWB of 0.10 stan­dard de­vi­a­tions (all else held equal).[7] Note that these find­ings may not ap­ply to high-in­come coun­tries.

Note: effect size com­pared to ab­solute cash trans­fer size (left, in US dol­lars per month) and rel­a­tive size (right, as a pro­por­tion of pre­vi­ous in­come). The blue lines are sim­ple re­gres­sions of y against x; see Table 2 in the pa­per for the full re­sults.

Se­condly, we in­ves­ti­gated the du­ra­tion of the effects. The figure be­low shows the data for lump sum and stream cash trans­fers in yel­low and pur­ple, re­spec­tively, and a sim­ple re­gres­sion of effect size against time for each. The vast ma­jor­ity of data was col­lected less than three years af­ter the trans­fers be­gan. Only one study has a fol­low-up af­ter five years. We find that there is a de­cay through time, whereby the av­er­age effect would reach zero af­ter ap­prox­i­mately seven years have passed since the trans­fer started, as­sum­ing a lin­ear de­cay with time (see Models 3 & 4 in Table 2 in the pa­per).[8] We also con­duct the analy­ses on lump sum and streams cash trans­fers sep­a­rately; in these cases we do not de­tect a statis­ti­cally sig­nifi­cant de­cay, which may be due to low power. More data of the long-run effect of cash trans­fers would im­prove this anal­y­sis.

Thirdly, as the figure be­low sug­gests and our anal­y­sis con­firms (see Model 1, Table 2), cash trans­fers have a greater effect (0.04 stan­dard de­vi­a­tions, or nearly half the av­er­age effect size) on life satis­fac­tion than de­pres­sion. Re­sults for hap­piness are less clear, but ap­pear to be in­ter­me­di­ate be­tween de­pres­sion and life satis­fac­tion.

Note: This figure shows the effect size for each study as cir­cles, grouped by type of out­come. A box plot for each out­come shows the val­ues of the me­dian, first and third quar­tiles. The sur­round­ing vi­o­lin plots illus­trate the den­sity of data points on the y-axis.

We also tested if the study’s de­sign, con­ti­nent, or con­di­tion­al­ity in­fluenced the mag­ni­tude of the effect. We only found that con­di­tion­al­ity had a sig­nifi­cant effect: con­di­tional cash trans­fers had a con­sid­er­ably smaller im­pact on SWB (-0.040 SDs).

Fi­nally, we do not de­tect a sig­nifi­cant spillover effect, that is an effect on non-re­cip­i­ents in other house­holds in the same com­mu­nity. This re­sult is im­por­tant as a po­ten­tial con­cern about cash trans­fers, one em­pha­sised by Plant (2019, p230), is that trans­fers might in­crease re­cip­i­ents’ well-be­ing at the cost of mak­ing non-re­cip­i­ents worse off, thereby re­duc­ing their over­all effec­tive­ness. How­ever, we note only four of the stud­ies in­cluded a mea­sure of spillovers to the com­mu­nity.

What does all this im­ply for our un­der­stand­ing of the to­tal im­pact a $1,000 cash trans­fer, given to some­one in global poverty, would have on their well-be­ing? We use the unit of a WELLBY to an­swer this, defined as an in­crease of one sub­jec­tive well-be­ing point (on a 0-10 scale) for one year.[9] We es­ti­mate that the to­tal effect over time is 0.38 stan­dard de­vi­a­tions, or roughly 0.87 WELLBYs.[10][11]

We don’t know of any re­views of a similar in­ter­ven­tion (i.e. a ‘micro’-in­ter­ven­tion in low-in­come coun­tries) that al­lows us to com­pare the to­tal im­pact on sub­jec­tive well-be­ing. How­ever, our fu­ture work aims to fill this gap, by re­view­ing ev­i­dence on the effects from in­ter­ven­tions such as cataract surgery and group ther­apy.

Our study has sev­eral limi­ta­tions, most no­tably that the ev­i­dence is thin (with only four stud­ies) on spillovers to other mem­bers of the com­mu­nity, and in the long run (greater than five years). There is no data on the effect on other mem­bers of the re­cip­i­ent’s house­hold. Some gen­eral caveats are (1) that in­signifi­cant re­sults do not mean an effect does not ex­ist—it could be too small to de­tect given our sam­ple size and (2) this effect only per­tains to similar pop­u­la­tions (the very poor in low- and mid­dle-in­come coun­tries).

This re­view pro­vides strong ev­i­dence that cash trans­fers im­prove the sub­jec­tive well-be­ing of peo­ple in low-in­come con­texts. Our es­ti­mate of the to­tal effect on well-be­ing is novel, and we en­courage fu­ture stud­ies to make similar es­ti­mates in or­der to com­pare the cost-effec­tive­ness of differ­ent health and de­vel­op­ment in­ter­ven­tions.

  1. We re­garded ques­tions that asked about life satis­fac­tion, hap­piness or cheer­ful­ness as sub­jec­tive well-be­ing. Sub­jec­tive well-be­ing mea­sures tend to as­sess how some­one is do­ing over­all, some­times in­clud­ing sep­a­rate mea­sures of pos­i­tive and nega­tive men­tal states. We there­fore con­sid­ered men­tal health ques­tion­naires which asked about af­fec­tive or mood di­s­or­ders as prox­ies for the nega­tive af­fect as­pect of sub­jec­tive well-be­ing. The “mea­sures” column of Table A4 in the pa­per in­cludes all in­cluded mea­sures. ↩︎

  2. We in­cluded ex­per­i­men­tal and quasi-ex­per­i­men­tal study de­signs, pub­lished since the year 2000. ↩︎

  3. Co­hen (1992) es­tab­lished the con­ven­tion that small = 0.2, mod­er­ate = 0.5 and large = 0.8. ↩︎

  4. One way that a meta-re­gres­sion is differ­ent to a stan­dard re­gres­sion is that in­di­vi­d­ual stud­ies are used as data points, rather than in­di­vi­d­ual peo­ple. Another differ­ence is that stud­ies are weighted by their pre­ci­sion, so gen­er­ally, larger stud­ies have a greater in­fluence than smaller stud­ies. Meta-re­gres­sions also al­low for within study and in the case of ran­dom effects meta-re­gres­sion, be­tween study vari­abil­ity. ↩︎

  5. As men­tioned in Foot­note 1, we in­clude many out­come met­rics in the over­all effect size in­dex, in­clud­ing life satis­fac­tion, hap­piness and men­tal health out­comes that mea­sure af­fec­tive states. ↩︎

  6. Stream CTs were re­ceived for an av­er­age of two years, and this is the pe­riod we as­sumed a lump CT was con­sumed in. ↩︎

  7. A $120 monthly value would dou­ble house­hold in­come for the year and lead to a 0.12 stan­dard de­vi­a­tion effect on SWB. ↩︎

  8. This is con­sis­tent with the re­sult even when re­mov­ing the fol­low-up at nine years, or us­ing the de­cay for stud­ies with mul­ti­ple fol­low-ups. ↩︎

  9. As noted pre­vi­ously, this in­dex in­cludes life satis­fac­tion, hap­piness, and mea­sures of mood di­s­or­ders. ↩︎

  10. Us­ing a stan­dard de­vi­a­tion of 2.3, as we did pre­vi­ously (see cell 4 in our pre­vi­ous model. ↩︎

  11. This is no­tably less than our ear­lier es­ti­mate of ~1.6 WELLBYs for the in­di­vi­d­ual, where WELLBYs were in terms of life satis­fac­tion points. Given the sig­nifi­cant differ­ence in effect for life satis­fac­tion and de­pres­sion out­comes, our es­ti­mate here is likely an un­der­es­ti­mate, as com­pared to our pre­vi­ous es­ti­mate. We plan to in­ves­ti­gate the effect of differ­ent sub­jec­tive well-be­ing met­rics in due course. Fur­ther, pre­vi­ously we con­sid­ered GiveDirectly cash trans­fers only, but, for ex­am­ple, here we in­clude con­di­tional as well as un­con­di­tional cash trans­fers, which also sug­gests this is an un­der­es­ti­mate. We ex­pect to up­date our ear­lier es­ti­mate in fu­ture, based on the in­for­ma­tion from this meta-anal­y­sis. ↩︎