Unconditional Cash and Family Investments in Infants: Evidence from a Large-Scale Cash Transfer Experiment in the U.S.

Economists have limited causal evidence on how families receiving unconditional income would spend those funds. We examine financial and time investments in infants among families living in poverty from a large-scale, multi-site randomized controlled study of monthly unconditional cash. We find increased spending on child-specific goods and mothers’ early-learning activities with their infants. The marginal propensity to consume child-focused items from the cash transfer exceeded that from other income, consistent with the behavioral cues in the design. We find no statistically detectable offsets in household earnings or impacts on pre-registered outcomes related to expenditures, labor supply, childcare or subjective well-being.

How people residing in poverty spend their money and time has long 1 animated public and political debate in the U.S. Economists wrestle with 2 competing goals of designing policies that will reduce the consequences and 3 social costs of poverty yet will not reinforce or fuel harmful behaviors or long-4 term dependence on government aid (National Academies of Sciences, other behaviors such as earning money or indirect benefits from consuming more 23 or higher-quality goods is methodologically difficult. 1 Second, other types of 24 studies that aim to identify key mechanisms related to investments in children are 25 not well suited to assess overall impacts on time and money investments in 26 response to income either because they examine specific parental behaviors or 27 parenting skills, such as taking a child to a doctor's appointment, ensuring 28 attendance at school, or interacting directly with a child ( propensity to spend on other pre-registered outcomes, including on alcohol or 85 cigarettes, did not differ between these groups at conventional levels of statistical 86 significance. Aspects of economic hardship such as homelessness, evictions, and 87 missed bill payments did not statistically differ between high-and low-cash-gift 88 families. High-cash-gift mothers did not report less worry over finances or 89 improved subjective well-being in terms of happiness or optimism. Although 90 mothers talked about the "miracle" of the cash gift, they also talked about 91 financial strain as top of mind. demonstrate substantial family heterogeneity in spending, with some prioritizing 169 child-specific goods (e.g., children's clothing, private school tuition) or 170 establishing a savings account for their child and others allocating the EITC 171 importance of kindergarten attendance reduced chronic absenteeism (Robinson et al. 2018). Finally, social norms and perceptions may affect relative evaluations that shape spending, as has been documented in analyses of data from the U.K., where low-income families dedicated a greater portion of their income to children in an effort to "catch up" with their affluent counterparts (Kornrich and Furstenberg 2013). where the age of consent was 19 years or older; (2) self-reported household 217 5 The cash transfer is a gift available through charitable organizations and is not taxable due to its source and unconditional nature. At study entry and upon consent to participate in the randomized controlled study, mothers received a letter detailing implications for taxes and receipt of other in Appendix Figure A1). Of these, 6,839 did not meet the inclusion criteria, and 227 341 declined to consent. A baseline interview was completed with the remaining 228 1,051 mothers. Of these 1,051 mothers, 1,003 agreed to receive cash gifts and 229 were randomized into the high-cash-gift or low-cash-gift groups. Randomization 230 into the high-cash-gift or low-cash-gift group occurred at the site level. Of the 231 1,003 mothers who were randomized, three were excluded because they notified 232 the interviewer within two days after completing the baseline interview that they 233 wanted to withdraw and stop receiving cash gifts. The result is a final sample of 234 1,000 mothers and infants recruited between May 2018 and June 2019. 235 Striking a balance between statistical power and project costs, 40% of the 236 recruited sample within each site was randomized to receive $333 monthly cash 237 gifts and 60% to receive $20 monthly gifts. With an enrolled sample of n=1,000 238 mother-infant dyads, and accounting for a predicted 20% attrition over longer-239 term follow-ups, the anticipated sample size of 800 dyads during subsequent 240 waves of data collection is estimated to provide 80% power to detect a .207 241 standard deviation impact at p<.05 in a two-tailed test on cognitive functioning 242 and family process outcomes. 6 Randomization successfully achieved baseline 243 equivalence across 30 characteristics for the full enrolled sample of 1,000 mother-244 infant dyads and within each site (see Appendix Table A1 for full sample; also 245 available in Noble et al. 2021). 246 The child age 1 survey (hereafter referred to as Wave 1 follow-up) began 247 in July 2019 and continued through June 2020 with an overall 94% completion 248 rate (see Appendix Table A2). Of the 37 characteristics measured at study entry 249 among the Wave 1 sample, four had small statistically significant differences by 250 cash gift group including race/ethnicity (whether identified as Black or American 251 Indian/Eskimo/Aleut), single parent (whether never married, single living with 252 partner, or biological father living in household), alcoholic drinking during 253 pregnancy, and reports of household receipt of benefits (see Appendix Table  254 A3). 7 Some of these statistical differences have very little substantive meaning by 255 way of magnitude of difference (e.g., household receipt of benefits is 95% versus 256 97%); nevertheless, as described below, all estimates are adjusted by these 257 baseline characteristics. As context for this study's analysis, the Wave 1 sample (n=931, as shown 272 in Appendix Table A3) is racially and ethnically diverse: per mothers' self-273 reports, 39% identify as Black, 42% Latina, <1% Asian/Pacific Islander, 1.5% 274 Native American, 11% white, non-Hispanic, and 7% multiple races/other. 275 Approximately one out of three infants were first-time births for the mother, and 276 one out of five mothers reported being married. Nearly 60% of mothers worked 277 for pay while pregnant, and 92% reported plans to return to work. Forty-one 278 percent reported that the biological father of the infant resided in the household. Wave 1 sample. It is made clear to mothers that the monthly cash gift will 307 continue regardless of their participation in the research. Indeed, 4MyBaby card 308 activity is observed for four mothers with deceased infants, one of four 309 incarcerated mothers, and all three low-cash-gift mothers who declined to 310 participate in the Wave 1 survey. As of June 2020 (the end of Wave 1 data 311 collection), we have data on over 62,800 transactions. 9 As shown in Appendix 312 Table A4, over the first 12 months, 3% (n=15) of the low-cash-gift mothers and 313 <1% (n=1) of the high-cash-gift mothers never used the card (from the time of 314 study enrollment or card activation). 10 In any given month, over 90% of the 315 mothers in the high-cash-gift group used the card. Two-thirds of high-cash-gift 316 mothers used the card every month. Very few transactions failed due to 317 insufficient funds or PIN problems. 11 Further, most of the cash gift tends to be 318 expended within days of the disbursement, and nearly all of it tends to be spent or 319 withdrawn from the card before the next monthly disbursement. The most 320 common transaction among high-cash-gift recipients is withdrawal of cash from 321 an ATM, averaging $1,435 in withdrawals, representing approximately one-third 322 of the $4,000 received annually. In contrast, $11 of the $220 received annually is 323 withdrawn as cash from an ATM among low-cash-gift recipients. The remainder 324 of the high cash gift is dispersed across a variety of point-of-sale venues, with the 325 largest amounts and most frequent transactions occurring at large chain stores and 326 various food-or grocery-related venues (see Appendix Table A5). One 327 implication of the high amount of ATM cash withdrawal among high-cash-gift 328 recipients is the limited ability to interpret expenditures from the debit card 329 transaction data alone, hence the importance of having additional information 330 from survey data. 331 332 V.
Intent-to-treat estimates 333 The intent-to-treat (ITT) estimates are derived by the following model implicated 334 by the random assignment design: 335 where Z is an indicator of whether a mother is in the high-cash-gift group, and 337 thus is the causal estimate of receipt of the high-cash gift. Y is the outcome of 338 interest for mother or the household and ε is the error term. The model includes 339 the following baseline covariates (X) with the goal of improving precision in the 340 estimate: mother's characteristics (mother's age, maximum education level 341 attained, race and ethnicity, marital status, general health, an indicator of maternal 342 depressive symptoms, and cigarette and alcohol consumption during pregnancy), 343 household characteristics (number of children born to mother, number of adults in 344 the household, father living with the mother, household income, and household 345 net worth), baby's birth characteristics (weight at birth and gestational age), a site-346 based fixed effect, and an indicator for the switch from in-person to phone 347 interviews. Robust standard errors are produced via Huber-White adjustments for 348 heteroskedasticity. Given the implementation success of the debit card 349 mechanism, the ITT estimate captures the effects of a net positive income shock 350 of $313, essentially equivalent to a treatment-on-the-treated interpretation. 351 All main statistical analyses are pre-registered. We address the possibility 352 of false positives by estimating the statistical significance of conceptually similar 353 outcomes; that is, we capture a common or similar domain or "family," by 354 generating a familywise error rate of outcomes using step-down resampling 355 methods (Westfall and Young 1993; results shown for outcomes in Tables 4 and  356 5, and in Appendix Table A10). The regression-adjusted model also includes an 357 indicator for whether the Wave 1 follow-up survey was conducted in person or by 358 phone (detailed further below), as well as the child's age in months at the time of 359 the Wave 1 survey. Child age at the Wave 1 follow-up serves as a within-360 treatment group proxy for the length of time receiving the cash gift, in case the 361 length of time of cash gift receipt might affect behavior and as a control for child-362 specific and related outcomes that might differ by child age (e.g., transitioning 363 from breastmilk or formula to milk). While over 60% of the Wave 1 interviews 364 were completed within the intended two-week window before or after the child's 365 first birthday (equivalent to child age 11 to 13 months), just under 40% occurred 366 when the child was older than 14 months. The average child age at the Wave 1 367 data collection time was 13.1 months (sd=2.1) for the low-cash-gift group and 368 12.6 months (sd=1.5) for the high-cash-gift group. Wave 1 survey in person. Thus, the pandemic generated two potentially 383 12 There is baseline equivalence across groups with respect to reasons for nonresponse related to the study focal child being deceased, study focal child no longer being in mother's custodial care, and mother being incarcerated. There were more refusals to participate in the Wave 1 follow-up among low-cash-gift mothers (n=6) than among high-cash-gift mothers (n=0; small difference is statistically significant at p<0.05), and more low-cash-gift mothers (n=38) were not found or were not able to be contacted for the Wave 1 follow-up than high-cash-gift mothers (n=12; small difference is statistically significant at p<0.05).
confounding complexities for the Wave 1 data: a data collection change from in-384 person to phone interviews, and a differential experience of being interviewed 385 before the pandemic (in person) vs. after the onset of the pandemic (by phone). 386 Based on our review of announcements and mobile phone activity, the 387 four metropolitan areas in this study over the March to June 2020 period 388 experienced similar timing of local guidance and stay-at-home ordinances, thus 389 limiting variation in timing across sites that could be captured through more Nevertheless, we investigated how sensitive the Wave 1 ITT estimates 414 derived from the main specification findings described below may be to these 415 issues by re-estimating the ITT estimates adjusting the main specification with 416 weighting. We applied inverse probability weights to adjust the baseline 417 characteristics of the low-cash-gift group to the baseline characteristics of the 418 high-cash-gift for the n=931 Wave 1 sample, the n=605 pre-pandemic Wave 1 419 sample, and the n=325 Wave 1 pandemic sample (Appendix Tables A6b-A6d). 420 We applied non-response weight to adjust the n=606 pre-pandemic sample and 421 the n=931 Wave 1 sample to reflect the baseline characteristics of the n=1,000 422 full study sample (Appendix Tables A6e-A6f). The findings on key selected pre-423 registered outcomes do not differ in economic or statistical significance with the 424 exception of the within pre-pandemic sample analysis (Appendix Table A6d) 425 signifying the importance of relying on the full Wave 1 (n=931) sample. 426

VI.
Impacts on household income and family investment 428 Tables 1 through 5 present the ITT findings on pre-registered outcomes  429 and pre-registered summary indices of economic well-being as well as an 430 expanded set of outcomes adjusted for considerations of multiple testing bias. 431 These tables include ITT estimates converted to effect size (ES), calculated as the 432 ratio of the estimated difference between the high-and low-cash-gift groups 433 divided by the standard deviation of the low-cash-gift group. A.
Household income 441 Table 1 shows that household income, using a scaled measure of inflation- There were no statistically detectable differences in receipt of government 470 assistance and benefits between high-and low-cash-gift families. Thus the cash 471 gift did not (mechanically) crowd out government assistance as might be expected 472 given the agreements in place to minimize risk of ineligibility due to the cash gift. 473 At least two-thirds of the high-cash-gift families reported some receipt of social 474 benefits at Wave 1 (as shown in Appendix A9, 64% reported receiving SNAP, 475 72% reported receiving WIC, and 66% reported receiving Medicaid). One 476 exception is that high-cash-gift families were less likely to report receiving 477 housing assistance (by 7.7 percentage points, or 33%) compared with low-cash-478 gift families, an effect that is statistically significant only among families in the 479 New York City site. 480

B.
Maternal time in work and in child-enriching activities 481 Table 2  (ES=0.13). High-cash-gift mothers were less likely to report "rarely or never" 496 engaging in reading or telling stories, and were 6.5 percentage points (or 8%) 497 more likely than low-cash-gift mothers to read books or tell stories a few times 498 per week or more. Translating the categorical responses to a continuous measure 499 of minutes spent on each activity, 14 the impacts are equivalent to high-cash gift 500 mothers increased time spent reading and telling stories to infants by 501 approximately 2.7 minutes more per week, or a 10% increase. 502

C.
Child-specific expenditures 503 Table 3 presents impacts on child-related expenditures. High-cash-gift 504 mothers spent $65.02 more in the 30 days prior to the survey interview on child-505 specific goods relative to expenditures by low-cash-gift mothers, an increase that 506 was both economically (20%) and statistically (p<0.05) significant. The increased 507 spending in the month prior to the survey interview included books ($7.38) and 508 toys ($16.80), each of which was statistically significant at the 5% level, and 509 clothes ($27.25) and diapers ($8.67), marginally statistically significant at the 510 10% level. (Note that the item on child-specific electronic goods or devices did 511 not statistically differ between the high-and low-cash-gift groups.) Expenditures 512 on these items for children in this age group are not readily available in nationally 513 representative samples, as we describe below. However, these expenditure 514 amounts are comparable to the $370 per month estimate for these same goods, The high-cash-gift families not only increased total dollars allocated to 521 these items but also increased the incidence of child-specific expenditures. High-522 cash-gift families were statistically more likely to purchase books by 11 523 percentage points (20% increase), toys by 3.5 percentage points (25% increase), 524 and children's clothing by 5.5 percentage points (6%). For the Wave 1 in-person 525 pre-pandemic sample, interviewers observed high-cash-gift families' homes as 526 more likely to have children's books (25%), compared with homes of low-cash-527 gift families (21%; ES=0.14; findings on this and related child-specific 528 expenditures for the pre-pandemic sample are shown in Appendix Table A8). 529 Table 3 further shows that high-cash-gift families were slightly more likely to 530 purchase child-specific durable goods as measured via an index (marginally 531 statistically significant); in particular, they were 7.6 percentage points more likely 532 than low-cash gift families to have purchased a high chair since the birth of the 533 infant. 534

D.
General core expenditures and household economic hardship 535 Table 4  Economists and social science scholars posit that increased income among 563 families residing in poverty will improve material well-being and reduce 564 hardships such as hunger and homelessness. We find no statistically detectable 565 changes across a variety of pre-registered measures of material well-being or 566 hardship. Maternal reports related to housing, 16 health, or bill payment 17 generally 567 track national estimates, with no statistically detectable differences between the 568 high-and low-cash-gift families (see Appendix Table A8 for item-level 569 breakdowns of the indices presented in Table 4). 570 Impacts of the cash gift on food hardship are mixed. Paradoxically, 571 maternal reports of food insecurity are higher in the high-cash-gift families than in 572 the low-cash-gift families (0.14 ES), largely driven by higher reports of not being 573 able to afford balanced meals. Fifteen percent of high-cash-gift families would be 574 considered food insecure. As a point of comparison, 14.8% of households with 575 children and 35.3% of households with income below the poverty line in 2020 576 16 Approximately one-fifth of mothers reported missing rent or a mortgage payment (four times higher than national estimates). Nearly 7% of mothers reported being evicted (close to average eviction filing rates in 2016; Desmond 2016), and 8% reported being homeless or in a group shelter since the baby's birth. 17 For example, utilities comprise up to 20% of net income of households in the lowest income decile, with one-third of U.S. households reporting difficulties meeting energy needs by either forgoing other necessities to meet energy bills, receiving disconnection notices, or keeping a home at an unhealthy or unsafe temperature. Six million households nationally received heating assistance alone, with an average benefit of $371 per household annually, covering about half of each family's heating bills. A third or more of mothers reported missing a utility payment (slightly higher than estimates for the general population), with mothers reporting spending on average of $228 in utilities and $160 in cable payments a month. Less than 10% of mothers reported having utilities cut off in the prior 12 months, whereas closer to 40% reported having cable cut off.
were food insecure (USDA Economic Research Service 2021). Hardships related 577 to housing quality, access to consumer durables (working heat or air conditioning, 578 clothes washer or dryer), and neighborhood safety also did not statistically differ 579 (at p<0.05) between the high-and low-cash-gift families. Further, impacts on 580 these outcomes are not confounded by increased residential moves among high-581 cash-gift households. 18 The high-and low-cash-gift families had no statistical 582 differences in reported ownership of a working car or access to a smartphone or 583 equivalent device or a tablet or desktop/laptop with an internet connection.

E.
Financial stress and subjective well-being 600 18 In our study, 40% of mothers reported at least one residential move since the time of baby's birth, and 6% report moving three or more times (by comparison, 10% of families in the Moving to Opportunity study moved three or more times over the 10-to 15-year follow-up; Ludwig et al. 2013). Nearly two-thirds of mothers reported that their housing condition is good or excellent.
In this section, we turn to ITT estimates of mothers' perspectives on 601 financial strain and subjective well-being. Table 5 shows that the high cash gift 602 marginally increased an index of economic stress (ES=0.10), particularly driven 603 by the subcomponents of the scale related to subjective assessments of household 604 spending totaling more than household income, an adjusted impact of 8 605 percentage points. There were no statistically detectable differences between the 606 high-and low-cash-gift groups in reports of worry over expenses, incidence of 607 setting aside emergency funds, or having the ability to cover a month of expenses 608 without income. 19 Although some mothers reported paying remittances, receiving 609 the high cash gift did not statistically affect the amount of money given to support 610 others not living in the household, as shown in Table 4  Several measures in the survey aimed to capture mothers' overall 618 subjective welfare (as distinct from psychological or mental health). Along this 619 19 Some insights may be gleaned on these subjective assessments of financial strain by comparing reports at the Wave 1 follow-up with reports at study entry on these same items. Approximately one-third of mothers reported financial worry at study entry and 35% reported spending more in total than their income. Thus, reports of financial worry and spending more in total than their income increased for both groups by the time of the Wave 1 follow-up, though slightly more for mothers in the high-cash-gift group. We also note that although the survey question indicated to count the cash gift as income with respect to the item on spending more, we cannot rule out mothers' mental earmarking of the cash gift differently from their perception of counting income in their reporting of this item. Mothers in both groups were less likely to report not being able to cover expenses without income at the Wave 1 follow-up relative to study entry (from 48% to 38% unadjusted in the low-cash-gift group and from 45% to 54% in the high-cash-gift group). The difference, however, between the high-and low-cash-gift group at study entry, and at the Wave 1 follow-up, is statistically similar (with a difference of ES 0.11). dimension, there were no statistically detectable differences in high-and low-620 cash-gift mothers' happiness or on an 8-item index designed to capture maternal 621 perceptions of agency and hope (e.g., meeting and pursuing goals, problem 622 solving despite discouragement; details shown in Appendix Table A9). Cities (15 in the high-cash-gift group; 15 in the low-cash-gift group), were 668 recruited to participate in 1.5-to 2-hour interviews every 9 to 12 months over the 669 follow-up period. The intention of these interviews is not to draw conclusions that 670 are generalizable to all BFY participants, but rather to gain insights into the 671 mothers' perspectives and experiences related to time and money allocations that 672 are difficult to capture in conventional quantitative data collection strategies. 673 First, mothers explicitly talk about mental earmarking of money for the 674 baby. For example, Bianca, 20 a 24-year-old Black mother of one from New 675 Orleans in the high-cash-gift group, said, "I feel like that's my baby's money for 676 the month. I make sure she gets her Pampers and wipes, everything she needs out 677 of it." Similarly, Jade, a 26-year-old Black mother of two from New Orleans in 678 the low-cash-gift group, said, "It's for the baby, so it's for her. So I just spend it 679 only on her…. I mean it's just $20, just, you know…. Whatever; it helps." The 680 importance to mothers of having money mentally designated for their children 681 was not taken lightly. Raven, a 31-year-old Black and American Indian mother in 682 New Orleans in the high-cash-gift group, encapsulates this phenomenon. Raven 683 says that without the BFY money, "I would not be able to give him more of what 684 he wants…like, that money is strictly for him. Whatever he picks up, whatever he 685 grabs, whatever he lacks is what that goes towards." 686 Second, mental earmarking of money for the child did not always translate 687 to spending on child-specific goods. For example, Isabella, a 30 year old, Latina 688 mother of three, and her husband both saw the BFY dollars as "children's 689 money." However, they differed in what they believed was the wisest way of 690 using the money to contribute to their baby's well-being. As Isabelle says "I can't 691 pause my kids, you know what I'm saying? Like if I need something in the house, 692 that money will be spent for the fact that I need it in the house. It could be stuff to 693 clean my house-it's going to be spent." In Isabella's mind, taking care of 694 immediate needs-like buying cleaning supplies in order to raise her son in a 695 clean house-trumped consideration of the long-term, amorphous goals like 696 saving that her husband preferred. 697 Third, mothers indicated how financial strain was top of mind. Many 698 mothers found themselves in the maternity ward amid difficult financial 699 20 In line with common practice, we use pseudonyms to protect mothers' identities.
circumstances. The cash gift was a wholly unexpected relief, and among many 700 high-cash-gift mothers, it immediately went toward expenditures such as buying a 701 car seat, paying an electricity bill, purchasing cleaning supplies, or purchasing 702 related items to provide a safe home environment for the baby. 21  continue that." Two other moms described using BFY money to invest in their 726 side businesses. Camille, introduced above, sold hair bows, tutus, and birthday 727 banners through her Instagram account. She used one month of the BFY money to 728 buy the craft supplies she needed to continue making the items she sold. 729 Alexandra, a 24-year-old Black mother of one in the Twin Cities, also saw the 730 $333 in BFY money as enabling her to invest in a better future for herself and her 731 family. With she and her husband both working, they were stressed and still 732 barely getting by, so Alexandra began to stay home with their child and pursued a 733 career as a writer; one month, she used the BFY money to invest in a writing 734 course to support this effort. Mothers also described how the BFY money enabled 735 them to handle their financial situations differently than they otherwise would, 736 either not having to ask kin for assistance, or working fewer hours while still 737 being able to cover expenses, though neither of these behaviors were captured as 738 significant differences between gift groups in the quantitative survey data. 739 Finally, mothers conveyed how they viewed the cash gift as money 740 independent from a government system, differently from other types of assistance. 741 Over 90% of mothers reported receipt of some type of government assistance at 742 study entry. Experiences with public benefit systems offer a context for mothers' 743 perspectives of the cash gift. For example, high-cash-gift mothers expressed fears 744 that the money would stop prior to the pre-established 40 months (now extended 745 to 52 months). Tonya, introduced above, had been receiving the BFY money for 746 more than a year, yet she still described herself as "very wary…that it will be 747 there." Such reactions align with mothers' expressions of simultaneous surprise 748 and gratitude for the cash gift, using words like "blessing" or "thank you, God" 749 when discussing first learning about the cash gift. This portrays the money's 750 arrival as akin to a miracle-and therefore a rare occurrence that cannot be 751 explained or necessarily trusted to continue. Also, mothers described frustrating 752 experiences with public benefit programs, with applications seemingly wrongfully 753 denied or benefits simply not showing up as expected from one month to the next. 754 Life had taught many that depending on anyone or anything other than themselves 755 was a risk. To the extent that mothers felt unsure about the money's regular food was slightly higher among high-cash-gift versus low-cash-gift families, none 773 of these individual spending categories statistically differed for the high-versus 774 22 Other research indicates that families consider multiple factors when moving, including being open to neighborhoods that are unsafe in order to access enough space within an apartment and have access to necessary public transit, as well as to find landlords who are willing to accept them as tenants (Rosenblatt and DeLuca 2012). As in other domains of life, therefore, families' residential choices may be more constrained than their finances alone would suggest. We could anticipate that such factors would limit the impact of the additional BFY dollars in families' abilities to make particular purchases, especially in arenas in which market and non-market forces may constrain supply or quality (e.g., housing, childcare).       Estimates on diapers, books, toys, and cigarettes/tobacco come from the diary survey of 3,000 households per quarter. Information on all other expenditure categories come from the interview survey drawing from 6,000 households per quarter. Estimates adjust age of mother, number of children in household, race/ethnicity of mother, marital status, total household size, and state/year/quarter fixed effects. Standard errors in parentheses, clustered at household level. 2 Estimates adjust for covariates as described in Tables 1-5. Coefficients derived from OLS estimates of monthly expenditure in each row on monthly average total household income other than gift; and gift ($333 or $20. The p-value difference is from Wald test of equality of coefficient on income and on gift.