Danish data on the minimum wage: The hourly wage jumps up by 40pct at the discontinuity of minimum wage rules; employment falls by 33pct and total input of hours decreases by 45pct

Do Lower Minimum Wages for Young Workers Raise Their Employment? Evidence From a Danish Discontinuity. Claus Thustrup Kreiner,  Daniel Reck and  Peer Ebbesen Skov. Review of Economics and Statistics, March 04, 2019. https://doi.org/10.1162/rest_a_00825


Abstract : We estimate the impact of youth minimum wages on youth employment by exploiting a large discontinuity in Danish minimum wage rules at age 18, using monthly payroll records for the Danish population. The hourly wage jumps up by 40 percent at the discontinuity. Employment falls by 33 percent and total input of hours decreases by 45 percent, leaving the aggregate wage payment almost unchanged. We show theoretically how the discontinuity may be exploited to evaluate policy changes. The relevant elasticity for evaluating the effect on youth employment of changes in their minimum wage is in the range 0.6-1.1.

---
Minimumwages,setbylaworbycollectiveagreement,existin3/4ofOECDcountries(OECD,2015).IntheUnitedStates,minimumwageincreaseshavebeenhighonthepolicyagendainrecentyears,motivatedinpartbymanystudies ndingsmallemploymente ectsofminimumwagehikes.Somecities(e.g.LA,Seattle)andthestateofCaliforniahaverecentlylegislatedaminimumwagerateof$15,amuchhigherratethanthecurrentFederalminimumof$7.25perhour.Ashigherminimumwagesbecomecommon,policy-makersareconfrontedwithasecondquestion:shouldahighminimumwageapplytoeveryone?Inparticular,shoulditapplytoyoungerworkers?Youngworkersarelow-skilledandenterthelabormarketwithoutworkexperience,whichmakethempotentiallyvulnerabletohighminimumwages.ManyUSstatesandcities,includingCalifornia,Minnesota,SouthDakota,KansasCityandDesMoines,whichhaverecentlyincreasedtheirminimumwage,havedebated,andattimeslegislatedorplacedontheballot,anexceptionforyoungerworkers(Kreiner,ReckandSkov,2018).Similarly,manyEuropeancountrieswithhighminimumwageshavelowerminimumwagesforyoungerworkers(OECD,2015).Themainquestionweseektoansweris:Holdingtheadultminimumwage xedatagivenlevel,whatisthee ectofachangeintheminimumwageapplyingtoyoungworkersontheiremployment?ExistingUSevidenceandmostotherevidencecannotanswerthisquestionasitstudieschangesinaglobalminimumwageratherthanayouth-speci cminimumwage.Forexample,theelasticityofyouthemploymentwithrespecttotheminimumwageof0.075reportedbytheUSCongressionalBudgetO ceisbasedonchangesinaglobalminimumwage(CongressionalBudgetO ce,2014).OurempiricalevidenceexploitsalargediscontinuityinDanishminimumwagerulesoccurringwhenworkersreachage18.TheDanishcontextisidealforourpurpose.Denmarkhaslargechangesinminimumwagerateswhenworkersturn18(andnochangeatanyotherages)andahighadultminimumwagecomparabletothe$15levelinplaceinCaliforniaandunderconsiderationmoregenerallyintheUS.1Furthermore,wecanstudythee ectoftheagediscontinuityusinghigh-qualitymonthlydataonwages,employment,andhoursworkedfortheentireDanishworkforce.Ourmain ndingsarecontainedinFigure1,whichshowsthattheagediscontinuityinminimumwageshasalargeimpactonemploymentaroundage18.Weexplainthedetailsbehindtheconstructionofthedataset,measurementissues,andthesourceofidentifyingvariationbelow.Figure1aplotsaveragehourlywages,imputedbydividingreportedmonthlywagesbyreportedhoursworkedforeachindividual,asafunctionofage(measuredinmonths),fortwoyearsbeforeandaftertheir18thbirthday.TheaveragehourlywageratejumpsbyDKK46,orabout$7,correspondingtoa40percentchangeinthewagelevelatage18computedusingthemidpointmethod.Figure1bplotstheshareofindividualswhoareemployedbymonthlyage.Weobservea15percentage-pointdecreaseinemploymentatage18,whichcorrespondstoa33percentdecreaseinthenumberofemployedindividuals.Forcomparison,notethatthewageandemploymentratesdevelopsmoothlywhenindividualsturn17and19yearsold,andthatittakestwoyearsbeforetheemploymentrateisbackatthelevelitattainsjustbeforethejumpdownwardsatage18.Subsequentanalysesrevealthatthedropinemploymentwhenworkersturn18re ectsadiscretechangeinjoblosswithoutanydiscretechangeinhiring(wedoobserveasmallanticipatoryslow-downinhiringasworkersapproachage18).Asimpleestimateoftheemploymentelasticity(theextensivemargin)withrespecttothewagechangeisobtainedbydividingtheestimatesofthepercentagechangesinemploymentandhourlywage.Thisgivesanelasticityaround-0.8.Whenlookingattotalhoursworked(theintensiveandextensivemargin),we ndanelasticityof-1.1,indicatingthatmostoftheresponseoccursalongtheextensivemargin.Recallthataunitelasticitywouldimplythattheaveragewagepaymentofallindividuals,includingbothemployedandnon-employedworkers,shouldstayunchangedwhenthewagerateisraised,becauseitse ectontheaveragewagepaymentisfullyo setbyadecreaseinemployment.Consistentwiththisreasoning,we ndnearlynoe ectonaverageearnings.Thisprovidesalternativeevidenceofatotalhoursworkedelasticityaround-1,notdependingonthemeasurementofhourlywages.Weuseeconomictheorytomotivateourempiricalspeci cationandtoshowthat,un-derreasonableassumptions,theestimatedemploymentelasticitymaybeusedtocalculatethee ectonyouthemploymentofachangeintheminimumwagespeci callyforyoungerworkers.First,weprovideasimplemodelinwhichtheelasticityweestimateusingtheagediscontinuityisexactlythesameastheelasticityneededforthedesiredcounterfactualpolicyanalysis.Inthemodel,workershaveexogenous,heterogeneousproductivitiesandarehirediftheirproductivityexceedstheminimumwage(correspondingtoahorizontaldemandforlabormeasuredine ectiveunits).Inthissimplesetting,cross-workere ectsarezero.Accordingtothisbasicmodel,wemaycomputetheconsequencesofincreasingtheminimumwageforyoungworkers(thoseunder18)uptothehigherlevelapplyingtoadultsbyusingourestimatedelasticity.Thiscalculationgivesa15percentagepointdropinyouthemployment,correspondingto33percentofinitialemployment.Amodelwithdownwardslopinglabordemandforlow-skilledworkwouldinsteadsug-gestthattherearecross-workere ects,implyingthatahigheryouthminimumwagemayincreaselow-skilledadultemployment.Suchcross-workere ectsposeapotentialthreattotheidenti cationstrategy.However,weshowthatonecanobtainalowerboundfortheyouthemploymentelasticitybyconsideringtheextremecaseofa xeddemandforlow-skilledwork(implyingthattheemploymente ectfromthediscontinuityanalysisisentirelydrivenbycross-workere ects).Thelowerboundmaybecomputedfromourestimatedelasticityandthewageshareofyoungerworkersinthelow-skilledlabormarket.Wethuscomputethewageshareoflow-skilledworkersunderage18,usingvariousde nitionsofthelow-skilledworkersthatareperfectlysubstitutableforworkersunderage18.Inthemostex-tremeofthesecalculations,inwhichonlyworkersaged18-19aredeemedtobe low-skilled substitutesforworkersunderage18,thelowerboundoftheyouthemploymentelasticitybecomes0.6.Increasingtheminimumwageforyoungworkersuptothelevelofadultwork-erswouldthendecreaseemploymentbyatleast11percentagepoints,or25percentofyouthemployment,whichisstillasubstantialemploymente ect.Wealsoembedoursimplemodelinanequilibriumsearchframeworkincorporatingdy-namicsforaging.Inaccordancewiththeempiricalevidence,themodelpredictsthatthedropinemploymentatage18re ectsadiscretechangeinjobloss,ratherthanadiscretechangeinhiring.Themodelalsopredictsspillovere ectsofanincreaseintheyouthminimumwageonadultemployment,butinthiscasethesignofthespillovere ectisambiguous.Inanycase,ourelasticityestimateisagainagoodapproximationofthee ectonyouthemploymentifyoungworkersconstitutealowshareoftotallow-skilledemployment."Additionalanalysisdemonstratesthatourinterpretationoftheempiricalresultsiscor-rectandstudiesheterogeneityinemploymente ectsacrossworkers.Mostimportantly,wedemonstratethatotherpoliciesthatchangewhenworkersturn18,suchastheeligibilityforDanishsocialwelfareprograms,arenotdrivingourresults.Wealsoshowthatthesizeoftheemploymentelasticityisonlyslightlylargerforworkersoflowerability,asproxiedbyschoolGPAin9thgradeortheincomeofparents.Finally,weprovidesuggestiveevidencethatjoblosseshavepersistente ectsonworkers.Twoyearsaftertheworkers'18thbirthdays,theemploymentrateisabout15percentagepointslowerforworkersloosingtheirjobatage18relativetoworkerswhokepttheirjob.Ourpapercontributestothesizableliteratureonminimumwagesandemployment,asreviewedinCardandKrueger(2015)andNeumarkandWascher(2008).Mostofthislit-eraturestudiesemploymente ectsofglobalminimumwagehikes,whileourfocusisonthee ectsofage-speci cminimumwages,whereevidenceislimited.NeumarkandWascher(2004)showthatcountrieswithhighminimumwagesalsotendtohavehighyouthunem-ployment,but,consistentwithourresults,thiscorrelationisweakerwhencountrieshavealowerminimumwageforyoungworkers.Onenewstudy,Kabátek(2015),analyzesanagediscontinuity,inthiscaseseveralsmallagediscontinuitiesinDutchminimumwages.Theobservedchangesinwagesandemploymentaroundworkers'birthdaysarethereforemuchsmallerandmoredi usethaninourcontext.Theimpliedemploymentelasticityisslightlysmallerthanours.Combiningonelargediscontinuitywiththoroughtheoreticalrea-soningandrichdataallowsustointerpretoure ectsinmoredetailandtoperformcrediblecounterfactualpolicyexercises.Ourresultsmaymakesomereadersconcernedabouttheimpactofglobalincreasesintheminimumwageonemployment,asubjectofintenseongoingdebate.SeveralDDstudies,mostfamouslyCardandKrueger(1994), ndlittletonoimpactofglobalminimumwagehikesonemployment.2Ourestimatesofthee ectofanincreaseinminimumwagesonemploymentaremuchlargerthanthosetypicallyestimatedforglobalminimumwagehikesusingDDdesigns.Therearethreefactorsthatcouldexplainthisdi erence.First,estimatesinexistingDDstudiesmightbeattenuatedbyshort-runfrictions(Baker,BenjaminandStanger,1999;Sorkin,2015;MeerandWest,2015;Aaronson,FrenchandSorkin,2017),whicharenotrelevantinoursetting.Second,ourstudyisbasedonahighminimumwagelevelcomparedtomostpreviousstudies.Minimumwagesmaynotbebindingatlowlevelsand,ifbinding,theymayincreaseemploymentduetolabormarketimperfections(Manning,2003).Third,ourresultsmightbedrivenbycross-agesubstitutionratherthanpurelyadisemploymente ectoftheminimumwage.The rsttwoofthesefactorssuggestthatourresultsaddressshortcomingsoftheexistingliteratureonglobalminimumwagehikes.However,thethirdisanimportantlimitationofourstudy'sabilitytospeaktothisdebate.Cross-agesubstitutionwouldimplythatweestimatehigheremploymentelasticityinoursettingthanwouldbeseenwithaglobalminimumwagechange.Theextenttowhichthisparticularfactordrivesourlargeestimatedeterminestheextenttowhichreadersshouldupdatetheirbeliefsabouttheemploymente ectsofglobalminimumwagehikes.Onthewhole,therefore,itisdi culttoimaginethatour ndingswillmakereaderslessconcernedaboutemploymente ectsofhighminimumwages,butwhetherandtowhatextenttheyshouldbemoreconcerneddependsonwhattheybelieveaboutthemechanismsbehindourresults.Ourworkalsocontributestothetheoreticalliteratureonthee ectsofminimumwages.MuchoftheliteratureattemptstorationalizeearlyDDstudies ndingsmallorevenpositiveemploymente ectsusingmodelswithmonopsonypowerorotherlabormarketimperfections(RebitzerandTaylor,1995;Manning,2003;Flinn,2006).Our ndingsoflarge,negativeemploymente ectsaroundage-basedminimumwagesalignbetterwithbindingminimumwagesinacompetitivelabormarketmodel.Theminimumwageliteratureoftenassumesthatworkers/jobsarehomogenouswithadownwardslopinglabordemandduetoade-creasingmarginalproductoflabor.Thisisincontrasttotheoptimalincometaxliteraturenormallyassumingheterogeneousproductivities(Mirrlees,1971).Ourexplanationsoftheempirical ndingsarebasedontheorywithheterogeneousproductivities,similartootherrecentminimumwageresearch(ClemensandWither,2016;ClemensandStrain,2017).Thefactthatsomeindividualslosetheirjobwhentheyturn18,whileotherskeeptheirjob,stronglysuggeststhatheterogeneousproductivityisanimportantaspectofthelow-skilledlabormarket.

Suppressing thoughts often leads to a “rebound” effect; unpleasant thoughts were more prone to rebound in dreams than pleasant ones; may be support for an emotion‐processing theory of dream function

The effects of dream rebound: evidence for emotion‐processing theories of dreaming. Josie Malinowski, Michelle Carr, Christopher Edwards , Anya Ingarfill, , Alexandra Pinto. Journal of Sleep Research, March 12 2019. https://doi.org/10.1111/jsr.12827

Abstract: Suppressing thoughts often leads to a “rebound” effect, both in waking cognition (thoughts) and in sleep cognition (dreams). Rebound may be influenced by the valence of the suppressed thought, but there is currently no research on the effects of valence on dream rebound. Further, the effects of dream rebound on subsequent emotional response to a suppressed thought have not been studied before. The present experiment aimed to investigate whether emotional valence of a suppressed thought affects dream rebound, and whether dream rebound subsequently influences subjective emotional response to the suppressed thought. Participants (N = 77) were randomly assigned to a pleasant or unpleasant thought suppression condition, suppressed their target thought for 5 min pre‐sleep every evening, reported the extent to which they successfully suppressed the thought, and reported their dreams every morning for 7 days. It was found that unpleasant thoughts were more prone to dream rebound than pleasant thoughts. There was no effect of valence on the success or failure of suppression during wakefulness. Dream rebound and successful suppression were each found to have beneficial effects for subjective emotional response to both pleasant and unpleasant thoughts. The results may lend support for an emotion‐processing theory of dream function.

Foraging: age of peak productivity between 30 and 35 years of age, though high skill is maintained throughout much of adulthood

The Life History of Human Foraging: Cross-Cultural and Individual Variation. Jeremy Koster et al. bioXiv Mar 12 2019, https://doi.org/10.1101/574483

Abstract: Human adaptation depends upon the integration of slow life history, complex production skills, and extensive sociality. Refining and testing models of the evolution of human life history and cultural learning will benefit from increasingly accurate measurement of knowledge, skills, and rates of production with age. We pursue this goal by inferring individual hunters' of hunting skill gain and loss from approximately 23,000 hunting records generated by more than 1,800 individuals at 40 locations. The model provides an improved picture of ages of peak productivity as well as variation within and among ages. The data reveal an average age of peak productivity between 30 and 35 years of age, though high skill is maintained throughout much of adulthood. In addition, there is substantial variation both among individuals and sites. Within study sites, variation among individuals depends more upon heterogeneity in rates of decline than in rates of increase. This analysis sharpens questions about the co-evolution of human life history and cultural adaptation. It also demonstrates new statistical algorithms and models that expand the potential inferences drawn from detailed quantitative data collected in the field.

---

THE LIFE HISTORY OF HUMAN FORAGING:CROSS-CULTURAL AND INDIVIDUAL VARIATIONJEREMY KOSTER1;2, RICHARD MCELREATH2;3, KIM HILL4, DOUGLAS YU5;6,GLENN SHEPARD JR.7, NATHALIE VAN VLIET8, MICHAEL GURVEN9,HILLARD KAPLAN10, BENJAMIN TRUMBLE4;11, REBECCA BLIEGE BIRD12,5DOUGLAS BIRD12, BRIAN CODDING13, LAUREN COAD8;14, LUIS PACHECO-COBOS15,BRUCE WINTERHALDER3, KAREN LUPO16, DAVE SCHMITT17, PAUL SILLITOE18,MARGARET FRANZEN19, MICHAEL ALVARD20, VIVEK VENKATARAMAN21,1Department of Anthropology, University of Cincinnati, Cincinnati OH 45221-03802Max Planck Institute for Evolutionary Anthropology3Department of Anthropology & Graduate Group in Ecology, University of California,Davis4School of Human Evolution and Social Change, Arizona State University5State Key Laboratory of Genetic Resources and Evolution, Kunming Institute of Zoology6School of Biological Sciences, University of East Anglia7Museu Paraense Emílio Goeldi8Centre for International Forestry Research9Department of Anthropology, University of California, Santa Barbara10Economic Science Institute, Chapman University11Center for Evolution and Medicine, Arizona State University12Department of Anthropology, Pennsylvania State University13Department of Anthropology, University of Utah14School of Life Sciences, University of Sussex15Facultad de Biologia, Xalapa Universidad Veracruzana16Department of Anthropology, Southern Methodist University17Department of Anthropology, Southern Methodist University18Anthropology Department, Durham University19Unaffiliated20Department of Anthropology, Texas A&M University21Institute for Advanced Study in Toulouse22Department of Anthropology, Dartmouth College23Departments of Biology and Anthropology, University of Richmond24Department of Anthropology, SIL International25Department of Conservation Biology, University of Göttingen, Germany and BiologyDepartment- FMIPA, Cenderawasih University, Papua Indonesia26Department of Geography, University of Helsinki27Institució Catalana de Recerca i Estudis Avançats (ICREA), Barcelona and Institut deCiència i Tecnologia Ambientals (ICTA), Universitat Autònoma de Barcelona28Institut de Ciència i Tecnologia Ambientals, Universitat Autònoma de Barcelona29Eco-anthropology & Ethnobiology Laboratory, UMR 7206 (CNRS-MNHN). Musée del’Homme - Muséum national d’Histoire naturelle, Paris30Metapopulation Research Centre (MRC), Department of Biosciences, University ofHelsinki31Faculty of Archaeology, Leiden University, Netherlands32School of Anthropology and Conservation, University of Kent33Department of Anthropology, Boise State University34Department of Food and Resource Economics, University of Copenhagen35School of Culture, History and Language, Australian National UniversityE-mail address:jeremy.koster@uc.edu.1.CC-BY-NC-ND 4.0 International licenseIt is made available under a (which was not peer-reviewed) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity.The copyright holder for this preprint. http://dx.doi.org/10.1101/574483doi: bioRxiv preprint first posted online Mar. 12, 2019;

2KOSTER ET AL.THOMAS KRAFT9, KIRK ENDICOTT22, STEPHEN BECKERMAN12, STUART A. MARKS23,THOMAS HEADLAND24, MARGARETHA PANGAU-ADAM25, ANDERS SIREN26, KAREN10KRAMER13, RUSSELL GREAVES13, VICTORIA REYES-GARCÍA27, MAXIMILIEN GUÈZE28,ROMAIN DUDA29, ÁLVARO FERNÁNDEZ-LLAMAZARES30, SANDRINE GALLOIS31,LUCENTEZZA NAPITUPULU28, ROY ELLEN32, JOHN ZIKER33, MARTIN R. NIELSEN34,ELSPETH READY2, CHRISTOPHER HEALEY35, AND CODY ROSS2March 11, 201915Abstract. Human adaptation depends upon the integration of slow life history, complexproduction skills, and extensive sociality. Refining and testing models of the evolution ofhuman life history and cultural learning will benefit from increasingly accurate measurementof knowledge, skills, and rates of production with age. We pursue this goal by inferringindividual hunters’ of hunting skill gain and loss from approximately 23,000 hunting records20generated by more than 1,800 individuals at 40 locations. The model provides an improvedpicture of ages of peak productivity as well as variation within and among ages. The datareveal an average age of peak productivity between 30 and 35 years of age, though high skillis maintained throughout much of adulthood. In addition, there is substantial variationboth among individuals and sites. Within study sites, variation among individuals depends25more upon heterogeneity in rates of decline than in rates of increase. This analysis sharpensquestions about the co-evolution of human life history and cultural adaptation. It alsodemonstrates new statistical algorithms and models that expand the potential inferencesdrawn from detailed quantitative data collected in the field.Keywords:Human evolution, foraging skill, hunting, life history, Bayesian data analysis30.CC-BY-NC-ND 4.0 International licenseIt is made available under a (which was not peer-reviewed) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity.The copyright holder for this preprint. http://dx.doi.org/10.1101/574483doi: bioRxiv preprint first posted online Mar. 12, 2019;

LIFE HISTORY OF HUMAN FORAGING IN 40 SOCIETIES31. IntroductionAs a slow-developing primate, humans exhibit puzzling life history traits. Primates ingeneral, and especially the apes, have slow life histories, with late age of first reproductionand singleton births. But even compared to other hominoids, humans have longer child-hoods, shorter inter-birth intervals, and extended post-reproductive lifespans (Jones 2011).35That is, human children are slower to develop and more dependent, but we nonetheless havemore of them, more quickly. These traits are plausibly unique to the genusHomo, but thetiming and adaptive origins of the human life history strategy remains unsettled (Schwartz2012).One way for humans to ease the costs of expensive childhoods is through alloparental40investments from highly productive adults (Kramer 2010). There are at least two majorquestions lurking within, however. The first is: Which individuals provide allocare? Anyanswer to this question will have implications for how selection operates on other aspectsof life history. The second: Is childhood itself more than just a period required for growinglarge and physically adept? Is it also required for individuals to learn complex, culturally-45evolved skills (Gurven et al. 2006)? What role does childhood play in the cultural evolutionof complex, productive skills in the first place (Henrich and McElreath 2003)?Any satisfactory model of human life history must address the integration of growth,reproduction, cognitive development, skill development, sociality, and cultural evolution.This is not easy. As a result, existing models make progress by omitting some features. The50most advanced attempt we know is the optimal control model of González-Forero et al.(2017). While this model omits cultural dynamics for acquired skills, it does successfullyintegrate growth, cognitive and skill development, and reproduction in overlapping gen-erations. By solving for the optimal life history, the model suggests natural selection fordelayed growth, early investment in cognition, and delayed reproduction. The brain gets55big first, and only then the body, because this allows a longer window of learning and ulti-mately higher adult productivity. These results are similar to theembodied capital hypothesis(Kaplan et al. 2000), in which highly productive foraging and food sharing by adult mensupports alloparental investments in offspring. From this point of view, human life historytraits stem from the highly complex human foraging niche, which selects for delayed mat-60uration by requiring an extended period of learning before adults are able to achieve highproductivity. In contrast, Hawkes et al. (1998) emphasize provisioning of grandchildren bypost-reproductive women, which selects for longer lifespans. This perspective sees child-hood as a consequence of prolonged lifespan, not a trait that needs to be explained as havingits own direct function (Charnov 1993). A spectrum of models exists, in which adult forag-65ing is variably influenced by size, skill, and culturally-transmitted knowledge, and differentamounts of time are needed for individuals to acquire and perfect adult skills.To develop and test models, anthropologists have used observational studies of subsis-tence hunting, with a focus on variation across the lifespan. For example, Walker et al.(2002) and Gurven et al. (2006) report data from the southern Neotropics that subsistence70hunters achieve high proficiency only after reaching advanced ages, roughly 35 to 45 yearsold. Because hunters achieve adult size and strength much earlier in life, these results areconsistent with the embodied capital hypothesis and its emphasis on the gradual masteryof cognitively complex hunting strategies. But comparative data from other contexts havebeen scarce. Among the few other empirical studies, some find slow skill development (e.g.,75Ohtsuka 1989) while others do not (Bird and Bliege Bird 2005)..CC-BY-NC-ND 4.0 International licenseIt is made available under a (which was not peer-reviewed) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity.The copyright holder for this preprint. http://dx.doi.org/10.1101/574483doi: bioRxiv preprint first posted online Mar. 12, 2019;

4KOSTER ET AL.More and better estimates of age-related foraging skill are necessary inputs into all evo-lutionary models of human life history. Associations between brain development, culturalknowledge, physical skill, and foraging performance at each age constrain the models wespecify: quantitative and representative estimates of these variables are needed to param-80eterize optimal life history models like González-Forero et al. (2017). Variation acrossindividuals informs models of food sharing and other investments, both within and be-tween generations. Variation across sites and contexts informs models of tradeoffs and howindividuals cope with them.In principle, skill and production in other subsistence economies is equally relevant to85understanding human life history. Garden production and animal husbandry depend uponthe same cognitive and developmental foundations as hunting and gathering. We focus onsubsistence hunting for two reasons. First, the data are easier to model than are gardeningand herding—hunting returns are easier to identify with specific individuals and labor al-locations. Second, hunting is practiced, to some extent, everywhere. It is both a primitive90economy and a modern one that has endured the emergence of other subsistence strate-gies. The breadth of hunting in diverse ecological settings provides a compelling range ofevidence.Studies of hunting returns are nevertheless inferentially challenging. A typical outcomevariable, such as kilograms of harvested meat, may be a mixture of zeros and skewed positive95values that violate assumptions of conventional regression models (McElreath and Koster2014). The available foraging data often exhibit imbalanced sampling of individuals and agegroups. Predictor variables may be missing or measured with uncertainty. These problemsare surmountable in any individual study, but comparative inferences are challenging whenstudies rely on heterogeneous statistical solutions.100In this paper, we address the inferential and comparative challenges within a novel sta-tistical framework. We assemble the largest yet data base of individual human huntingrecords, comprising over 21,000 trips from 40 different study sites. These data elucidatethe extent to which the ontogeny and decline of hunting skill are attributable to individual-level or site-level factors, and the comparative analysis help to mitigate over-generalization105from individual studies. The results of this study consequently inform subsequent theorizingabout the evolution of life history traits in humans.Our statistical approach accepts the imperfections of the sample and conservatively poolsinformation, both among individuals within sites and among sites within the total sample.The goal is not to substantiate any particular theoretical model of human evolution, nor to110pretend that the data are sufficient for all inferential objectives. Rather, the goal is to showwhat can be inferred from a statistical approach that uses all available data and treats missingdata and measurement error conservatively. One of the most important aims is to highlightthe limits of existing data and approaches so that future empirical and inferential projectscan make further progress.115Our analysis supports the general conclusion that skill peaks between 30 and 35 years ofage, well after the age of reproductive maturity. Peak skill is typically not much higher thanskill during early adulthood, however. Declines with age are typically slow—an average56 year old has the same proportion of maximum skill as an average 18 year old. There isconsiderable variation both among sites and individual hunters within study sites. Variation120among individuals is described more by heterogeneity in the rate of decline than the rateof gain. Partly owing to heterogeneous data collection methods across sites and anticipated.CC-BY-NC-ND 4.0 International licenseIt is made available under a (which was not peer-reviewed) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity.The copyright holder for this preprint. http://dx.doi.org/10.1101/574483doi: bioRxiv preprint first posted online Mar. 12, 2019;

6KOSTER ET AL.Table 1. Study sites and their numerical and text codes. See the help fileof thecchuntspackage for related citations.Number Code CountryGroupDataset incchuntspackage1CRE CanadaCreeWinterhalder2 MYA BelizeMayaPacheco3 MYN NicaraguaMayangnaKoster4 QUIEcuadorQuichuaSiren5 ECH ColombiaEmbera Chami Ross6 WAO EcuadorWaoraniFranzen7 BARVenezuelaBariBeckerman8 INUCanadaInuitReady9 MTS PeruMatsigenkaYu_et_al10 PIRPeruPiroAlvard11 CLBColombiaVan_Vliet_et_al_South_America_sites12 PME VenezuelaPumeKramer_Greaves13 TS1BoliviaTsimaneFernandez_Llamazares14 TS2BoliviaTsimaneReyes-Garcia15 TS3BoliviaTsimaneTrumble_Gurven16 ACH ParaguayAcheHill_Kintigh17 GB1GabonCoad18 GB2GabonVan_Vliet_et_al_Gabon19 GB3GabonVan_Vliet_et_al_Ovan20 CN1DR CongoVan_Vliet_et_al_Phalanga21 GB4GabonVan_Vliet_et_al_Djoutou22 BK1CameroonBakaGallois23 BK2CameroonBakaDuda24 CN2CongoVan_Vliet_et_al_Ingolo25 CN3CongoVan_Vliet_et_al_Ngombe26 BFACentral African Republic Bofi and AkaLupo_Schmitt27 CN4DR CongoVan_Vliet_et_al_Baego28 BISZambiaValley BisaMarks29 HEH TanzaniaNielsen30 DLG RussiaDolganZiker31 BTKMalaysiaBatekVenkataraman_et_al32 PN1IndonesiaPunanGueze33 PN2IndonesiaPunanNapitupulu34 AGT PhilippinesAgtaHeadland35 MRT AustraliaMartuBird_Bird_Codding36 NUA IndonesiaNuauluEllen37 NIM IndonesiaNimboranPangau_Adam38 NEN Papua New GuineaNenHealey_Nen_PNG39 MAR Papua New GuineaMaringHealey40 WOL Papua New GuineaWolaSillitoe.CC-BY-NC-ND 4.0 International licenseIt is made available under a (which was not peer-reviewed) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity.The copyright holder for this preprint. http://dx.doi.org/10.1101/574483doi: bioRxiv preprint first posted online Mar. 12, 2019;

LIFE HISTORY OF HUMAN FORAGING IN 40 SOCIETIES7either the same or a different functional relationship with age.) Most sites contribute pri-marily cross-sectional data, while a few others exhibit impressive time series. The statisticalframework is designed to make use of all these data.1503. The life history foraging modelSince skill cannot be directly observed, what is required is a model with latent age-varyingskill. This unobservable skill feeds into a production function for observable hunting re-turns. In this section, we define a framework that satisfies this requirement. We explainit one piece at a time, with a focus on the scientific justification. The presentation in the155supplemental contains more mathematical detail, and the model code itself is available toresolve any remaining ambiguities about the approach. Our framework was developed andreviewed in the initial grant proposal (NSF #1534548) prior to seeing the assembled sam-ple. Therefore, whatever the model’s flaws, they do not include being designed specially forthese observations or chosen to produce a desired result.160One advantage of the latent skill approach is that it allows us to use different observationsfrom different contexts—both solo and group hunting, for example—to infer a commonunderlying dimension of skill. But modeling even the simplest foraging data benefits fromthis approach, as hunting returns often are highly zero-augmented. Separate productionfunctions for zeros and non-zeros are needed to describe such data. In principle, more165than one dimension of latent skill could be modeled. We restrict ourselves to only one inthe current analysis. With more detailed data, describing additional dimensions should bepossible.We implemented the model both as a forward simulation and as a statistical model. Theforward simulation generates data with known parameter values, which are used to confirm170that the estimated statistical model can recover the parameters. The code is available as partof thecchuntsR package.3.1.Latent skill model.One of the simplest life history models is the von Bertalanffy(1934) asymptotic growth model. We use this model to represent the increasing compo-nents of hunting skill as a function of age. These increasing components include knowledge,175strength, cognitive function, and many other aspects that contribute to hunting success andincrease but decelerate with age. For convenience, label the composite of these componentsknowledge. Assume that the rate of change in knowledge with respect to agexis given bydK=dx=k(1