Present Bias: Lessons Learned And To Be Learned
American Economic Review: Papers & Proceedings 2015, 105(5): esent Bias: Lessons Learned and To Be Learned†By Ted O’Donoghue and Matthew Rabin*Present bias is an old idea. The notion thatpeople are susceptible to the over-pursuit ofimmediate gratification dates (at least) to theancient Greeks. In academic research, psychologists working with animals in the 1960s and1970s proposed “hyperbolic discounting”—afunctional form of discounting that generatespresent bias—as a natural way to representhow animals respond to time delays, and laterresearch in psychology extended this idea tohumans.1 Also in the 1960s, economists investigating general implications of time-inconsistentpreferences used as an example the now-popular β, δ functional form—which also generatespresent bias.2 But present bias really took holdin economics following David Laibson’s dissertation (Laibson 1994).The literature has blossomed in the past 20years. Research has led to a much better theoretical understanding of present bias, when andhow to apply it, and which ancillary assumptionsare appropriate in different contexts. Empiricalanalyses have demonstrated how present biascan improve our understanding of behaviorin various economic field contexts. While themodel is clearly not “correct”—no model is—for many contexts it is proving a useful, tractable, and (importantly) disciplined improvementin economic analysis. Nonetheless, there is stillmuch to learn.I. A Brief Overview of Present BiasLike exponential discounting, present bias isa model of discounting. One indication of thesuccess of present bias is that, much as for exponential discounting, most readers do not need areview of the structure of the model. Hence, herewe present only a brief summary.Suppose that intertemporal preferences fromthe perspective of period t can be represented by TU t τ t D(τ t) u τ , where u τ is instantaneous utility experienced in period τ and D (x) reflects the discounting associated with a delay x {0, 1, 2, . . . } . While more general variantsof present bias exist, the β, δ functional formtypically used assumes that1if x 0 D(x) { β δ x if x 0.With this functional form, β 1 correspondsto exponential discounting, while β (0, 1) reflects present bias. When β 1 , the model’spredictions may depend on an ancillary assumption about whether one is aware of how preferences change over time (sophisticated), unawareof how preferences change over time (naïve), orsomething in between (partially naïve).II. Some Lessons LearnedThis section summarizes some lessons learnedover the past 20 years. While these lessons havebeen learned well among those immersed in theliterature, we highlight these lessons for others.* O’Donoghue: Department of Economics, CornellUniversity, 482 Uris Hall, Ithaca, NY 14853 (e-mail: edo1@cornell.edu); Rabin: Department of Economics, HarvardUniversity, Littauer Center M-8, Cambridge, MA 02138(e-mail: matthewrabin@fas.harvard.edu). For helpful comments, we thank Dan Benjamin, Nava Ashraf, and other participants in our AEA Session.†Go to http://dx.doi.org/10.1257/aer.p20151085 to visitthe article page for additional materials and author disclosure statement(s).1See Ainslie (1992) for an overview.2See in particular Phelps and Pollak (1968) and Pollak(1968). These papers build on an earlier paper by Strotz(1956) that investigates general time-inconsistent preferences without special focus on the case of present bias.Lesson #1: Present Bias Operates onUtility.—All discounting models—exponential,present bias, or other—operate on the timing ofutility. Importantly, they do not operate on thetiming of purchases or on the timing of monetarypayments. Indeed, perhaps the most basic lessonfrom standard saving-consumption models isthat, in the absence of liquidity constraints and273
274AEA PAPERS AND PROCEEDINGSuncertainty, choices between different streamsof monetary payments are driven entirely bymaximizing the present discounted value ofwealth, and preferences are irrelevant.The idea that discounting models, present-biased or not, operate on utility is not somuch a lesson learned over time as a fundamental lesson inherited from standard economics.However, we emphasize this lesson because it issometimes forgotten by those new to present bias.To highlight the importance of distinguishingutility flows from money flows, we note two natural intuitions for how present bias can lead people to delay money flows—the opposite of whatone would predict if one applied present biasdirectly to money flows. First, Laibson (1997)demonstrates how sophisticated present bias canlead people to choose to constrain their futureliquidity, and one way to do so is to defer moneyflows until the more distant future. Second,O’Donoghue and Rabin (1999b) demonstratehow naïve present bias applied to the effortrequired to optimize one’s retirement saving canlead people to delay the accrual of money flowsrather than move it forward.This lesson is also important when using quasi-linear preferences. In such models, if aperson purchases some good at a price p , theperson is assumed to experience a utility gainfrom consuming the good, and a utility cost of p interpreted as forgone consumption of othergoods. When applying present bias, one clearlymust be explicit about whether the utility fromthe purchased item is experienced immediatelyor over time. But when does that forgone consumption occur? In most applications, the mostnatural assumption is that, regardless of whenthe monetary payment is made, it is forgonefuture consumption.Finally, we note that, related to this lesson,economists are finally appreciating that the useof money alone is not appropriate for experiments that investigate time preferences. Whenchoosing between time-dated monetary payoffs,subjects’ time preferences are irrelevant unlessthey are liquidity constrained. For most experiments, it seems highly unlikely that subjectsare liquidity constrained in a relevant way.3 InMAY 2015r ecognition of this issue, recent experimentshave asked subjects to choose instead betweentime-dated utility flows, such as when to exertreal (unpleasant) effort.4Lesson #2: Present Bias is About Now.—Psychologists suggested a hyperbolic functionalform for discounting—i.e., D(x) 1/(1 kx) .Economists instead adopted the β, δ functionalform in large part for tractability, and this modelquickly became the workhorse for the literature.In the early years, we worried about whetherusing the β , δ functional form was restrictive.As time passed and our intuitions developed, webecame less worried. Indeed, many of us nowbelieve that, in fact, the β, δ functional form better captures the underlying psychology—thatthe vast majority of the action (relative to timeconsistency) is biased toward now.To be fair, there is little direct evidencethat compares different functional forms.Psychologists primarily compare the hyperbolicversus exponential functional forms, and do notconsider alternative functional forms. At thesame time, economists have primarily limitedourselves to studying the β, δ form.On a related note, researchers have pointedout, correctly, two confounds in time-preferenceexperiments: (i) payoffs received now might beviewed as certain while payoffs to be received inthe future might be viewed as uncertain, and (ii)payoffs to be received in the future might involvehigher transactions costs. Some have suggestedeliminating these confounds by attaching afront-end delay to all payoffs. If present bias isabout now, however, this approach also eliminates present bias as an influence on behavior.Indeed, when experiments with a front-enddelay find little evidence of deviations fromexponential discounting, such experiments provide support for the β , δ functional form relativeto the hyperbolic functional form.Lesson #3: Any Noticeable Short-TermDiscounting is Evidence of Present Bias.—Mostearly evidence on present bias emphasized timeinconsistency as the smoking gun for presentbias. Researchers are now more comfortable3Yet in many experiments that use monetary payoffs,many subjects seem not to behave in a wealth-maximizingway. It is an open question exactly what those subjects aredoing.4See, e.g., Augenblick, Niederle, and Sprenger (2013),who directly compare the use of money versus real effort,and find evidence of present bias only for the latter.
VOL. 105 NO. 5PRESENT BIAS LESSONSwith a simpler argument based on calibration.Exponential discounting does not permit anynoticeable discounting over short horizonsbecause such discounting would compound topredict counterfactually severe discounting overlonger horizons. Present bias, in contrast, is allabout noticeable short-term discounting.5To illustrate, suppose the only thing weobserved about Johnny is that he cares 1 percent more about his utility today than tomorrow.If Johnny were an exponential discounter, thisobservation would imply a yearly discount factor of (0.99) 365 0.026 . This is clearly counterfactual: nobody cares 40 times more aboutnow than one year from now. Hence, Johnnycould not be an exponential discounter.What about smaller short-term discounting?Under exponential discounting, caring just 0 .1percent more about today than tomorrow impliesa yearly discount factor of (0.999) 365 0.694 ,or caring 1 percent more about today thannext week implies a yearly discount factor of (0.99) 52 0.593 . At first glance, these numbers might appear reasonable, because estimatessuggest that discounting one year from now by0.69 might be realistic. However, exponentialdiscounting implies that the same 0.69 appliesbetween any two years—e.g., people wouldneed to care only 0.69 as much about eight yearsfrom now as seven years from now. This is againcounterfactual. Observing more than infinitesimal short-term impatience is sufficient to rejectexponential discounting, and quite consonantwith present bias.Lesson #4: Naïvete Makes Sense, andDoesn’t Always Lead to “Crazy” Behavior.—Early work on present bias in economics focusedon the assumption of complete sophistication, inpart based on a belief that naïvete would leadto “crazy” behavior. In our own work, we highlighted how complete or partial naïvete are validassumptions in the sense that they can be modeled in a rigorous fashion (see O’Donoghue andRabin 1999a, 2001, and subsequent research).65This argument is analogous to calibration arguments forhow any noticeable risk aversion over modest stakes is evidence against expected utility.6Prior to our work, research on naïvete was scarce.Strotz (1956) introduced the distinction between sophistication and naïvete, but his formal analysis considered only sophistication. Pollak (1968) formally analyzed naïvete, but275Moreover, it is not the case that naïvete in general predicts “crazy” behavior, at least not inthe sense of predicting unrealistic behavior. Insome situations, the predictions of naïvete areidentical to the predictions of sophistication. Inother situations, naïvete predicts very damagingbehavior, but of the sort that, unfortunately, weobserve way too often in the world. For instance,while it is inconsistent with exponential discounting or sophisticated present bias for a person to predict hundreds of times that she’ll starta diet, quit smoking, or write a referee reporttomorrow when she won’t, these seem to be thetypes of behaviors that we observe. More andmore research is suggesting that models thatincorporate naïvete (at least to some degree)seem to better explain behavior.Lesson #5: There is a Natural Intuitionfor How to Identify the Parameters of PresentBias.—Economists often estimate parametervalues in structural models. Over time, researchers have developed a good intuition for how toestimate the parameters of present bias (i.e., β and δ ). Specifically, one needs data on multipletypes of choices, some which involve trade-offsbetween immediate utility and future utility,and others which involve trade-offs betweenfuture utility and further-future utility—i.e.,some decisions which are heavily influenced bypresent bias ( β ), and other decisions which areprimarily influenced by longer-term discounting ( δ ). Two applications illustrate this intuitionnicely. Angeletos et al. (2001) study present biasin the context of savings-consumption decisions,and describe how identification can come froma combination of a household’s credit-card borrowing to finance current consumption (heavilyinfluenced by present bias) and a household’ssavings accumulation for retirement (primarily influenced by longer-term discounting).DellaVigna and Paserman (2005) study presentbias in the context of job search, and describehow identification can come from a combinationof a person’s effort to search for a job (heavily influenced by present bias) and a person’s reservation wage applied to job offers (primarilyinfluenced by longer-term discounting).only as a methodological approach to prove a result aboutsophistication. Akerlof (1991) was the first to seriously consider implications of naïvete, although his analysis was notframed in terms of present bias.
276MAY 2015AEA PAPERS AND PROCEEDINGSResearchers have been less successful at generating well identified parameter estimates fornaïvete. In principle, one needs data on choicesthat reflect (perhaps indirectly) people’s predictions for their own future behavior—e.g.,purchases now intended to be consumed in thefuture, or decisions now that impact the futureprices that will be faced. In practice, clean datathat permit precise identification of the degree ofnaïvete has proven hard to find. More often, wesee either evidence that indicates at least somesophistication (e.g., commitments) or evidencethat indicates at least some naïvete (e.g., commitments that don’t work as intended or clearlyinefficient procrastination).Lesson #6: Welfare Analysis is Doable.—When preferences are time-inconsistent, welfareanalysis becomes tricky because there are multiple preferences that one could use. While therehave been growing pains—some economistsinitially suggested that, given this issue, weshouldn’t do welfare analysis at all—economistsseem to have accepted that welfare analysis isdoable. In particular, the most important lessonon this dimension is that one should be rigorousand precise in exactly the way that economistsusually are—be very clear about what assumptions one is making about how to assess welfare,and assess whether one’s welfare conclusionsare robust to other assumptions about how toassess welfare.While there is no agreed-upon welfare criterion, we have argued for the use of “long-runutility,” wherein we use the intertemporal utilityfunction U t evaluated from a prior (or long-run)perspective, which is equivalent to using β 1 .The early literature suggested instead using aPareto criterion in which intertemporal utility from all perspectives is taken into account.It turns out that these two approaches frequentlyyield the same conclusions. Based on suchresults, we conjecture that long-run utility willin the end be seen as best single criterion.77There also exist other exotic welfare criteria which (webelieve) are less in the spirit of traditional economics. Buteven for these, we conjecture that the ancillary assumptionsneeded to fit economists’ intuitions about welfare will makethose models line up with the long-run utility criterion.III. Lessons To Be LearnedDespite the progress of the past 20 years,there is still much to be learned. We next discusssome important open questions.Question #1: How Can We Improve thePredictions of Present Bias?—Typically ourmodels explain only some of the variation in thedata, and it is natural to seek ways to improve ourmodels. A popular approach among researchers is to enrich the model of present bias. Twopotentially important ways to enrich the modelhave been discussed. First, one might incorporate heterogeneity in present bias, and therebyexplain some of the variation in behavior acrossindividuals. Indeed, more and more research isfinding correlations between measures of present bias—e.g., from a survey—and field behaviors. Second, one might incorporate ways inwhich the magnitude of present bias depends oncontext, and thereby explain some of the variation in behavior across contexts. Quantitativeestimates of discounting do tend to vary acrosscontexts, and some models (e.g., dual-processmodels) explicitly incorporate context-specificdiscounting.8While such enrichments are surely useful, weworry that researchers are perhaps excessivelyfocused on the details of present bias, and notfocused enough on other details. Present biasmakes no predictions about behavior independent of (i) utility functions (what people likeand don’t like), (ii) the timing of decisions, and(iii) constraints and transactions costs. In seeking to improve our models, we must not forget to be careful in accounting for these otherfactors that standard economic theory deemsrelevant.We worry, for example, about attempts toexplain heterogeneity in behavior primarily dueto heterogeneity in present bias. Heterogeneityin cigarette consumption, for instance, is farmore likely due to heterogeneity in tastes forcigarettes, or in prices and extent of peer pressure toward cigarettes experienced in one’syouth. We also think there are reasons to resistoverexcitement about variation of present8One might also search for improvements to the functional form of present bias. We conjecture, however, thatsuch improvements are unlikely to be important.
VOL. 105 NO. 5PRESENT BIAS LESSONSbias across contexts. Such variation is often confounded with variation in the utility function and constraints across contexts—indeed,we conjecture that the latter is likely far morepronounced.Question #2: How Important is TemporalAggregation?—Data come in different frequencies—e.g., consumption data might comeat a monthly or quarterly frequency. Moreimportantly, data often come at a frequencythat arguably reflects the net effects of a seriesof underlying decisions. In such cases, empirical analyses typically develop a model at thesame frequency as the data—e.g., if the datacome at the monthly frequency, then a periodin the model is assumed to be one month.Such analyses ignore the underlying temporalaggregation.Under exponential discounting, we suspectthis issue doesn’t matter much. Under presentbias, in contrast, it could be quite important.For instance, suppose data come at the monthlyfrequency, and reflect the net behavior of 30daily decisions each impacted by a small present bias. If one uses this data to estimate amodel in which a period is a month, estimatedimpatience will be very large. Moreover, thatestimated impatience would not predict welldecisions on simple trade-offs between utilitynow versus utility one month from now. Theright way to approach such data would be toexplicitly model how a series of underlyingdecisions aggregate into predictions at the frequency of the data.Question #3: How to Assess the Impact ofPresent Bias Against Other Phenomena?—Present bias is being incorporated into moreand more analyses. However, the success ofpresent bias has perhaps been to the detrimentof other potential improvements to economicmodels of intertemporal choice. Indeed, economists are sometimes prone to misattributebehaviors to present bias that more likely aredue to other shortcomings of the classical economics model.Four intertemporal phenomena seem particularly relevant. First, there is the old ideaof habit formation wherein one’s utility fromconsumption depends on one’s own past consumption. Second, there is projection biaswherein one’s decisions are distorted by mis-277predictions of future tastes (Loewenstein,O’Donoghue, and Rabin 2003). Third, thereis anticipatory utility wherein one experiencesutility now from anticipating future consumption (Loewenstein 1987). Fourth, there isint
Present Bias: Lessons Learned and To Be Learned . front-end delay to all payoffs. If present bias is about now, however, this approach also elimi- . Early work on present bias in economics focused on the assumption of complete sophisti
(4 Hours) Biasing of BJTs: Load lines (AC and DC); Operating Points; Fixed Bias and Self Bias, DC Bias with Voltage Feedback; Bias Stabilization; Examples. (4 Hours) Biasing of FETs and MOSFETs: Fixed Bias Configuration and Self Bias Configuration, Voltage Divider Bias and Design (4 Hours) MODULE - II (12 Hours) .
CHAPTER 11 Conservatism Bias 119 CHAPTER 12 Ambiguity Aversion Bias 129 CHAPTER 13 Endowment Bias 139 CHAPTER 14 Self-Control Bias 150 CHAPTER 15 Optimism Bias 163 Contents vii 00_POMPIAN_i_xviii 2/7/06 1:58 PM Page vii. CHAPTER 16 Mental Accounting Bias 171 CHAPTER 17 Confirmation Bias 187
DC Biasing BJT circuits There is numerous bias configuration of BJT circuits. Some of the common configuration of BJT circuit includes 1. Fixed-bias circuit 2. Emitter-bias circuit 3. Voltage divider bias circuit 4. Collector-feedback bias circuit 5. Emitter-follower bias circuit 6. Common base circuit Fixed Bias Configuration
ES-5: PREVENTING SOCIAL BIAS Controlling Social Bias involves understanding, identifying, and actively countering bias. It is important to reflect on the nature of bias and how it comes about before attempting to control social bias. Bias is a part of human nature because we all naturally prefer familiar things and familiar ways of thinking.
A sensor bias current will source from Sensor to Sensor- if a resistor is tied across R BIAS and R BIAS-. Connect a 10 kΩ resistor across Sensor and Sensor- when using an AD590 temperature sensor. See STEP 4 Sensor - Pins 13 & 14 on page 8. 15 16 R BIAS R BIAS-SENSOR BIAS CURRENT (SW1:7, 8, 9, 10)
4.4.1. Averaging total cloud amount and frequencies of clear sky and precipitation. 12. 4.4.2. Averaging methods for cloud types. 13. 4.4.3. Bias adjustments for cloud type analyses. 14 (partial-undercast bias, abstention bias, clear-sky bias, sky-obscured bias, night-detection bias) 5.
Diode Reverse Bias Reverse bias is the condition that prevents the current to flow through the diode. It happens when connect the voltage supply in a reverse bias, as shown in Figure (12). The p-region to the negative bias of the supply and the n-region to the positive bias of the supply. The reverse bias happens because the positive side of the voltage
hubungan antara asupan asam folat dengan kadar Hb dengan nilai p 0,64. Kata Kunci : asupan fe, asupan folat, kadar hb, tb paru . Abstract . Tuberculosis pulmonary can lead to various metabolic disorders and system disturbances in the body, one of which is synthetic disorder of Hemoglobin levels. Some nutrients which can influence the synthetic of Hemoglobin levels are iron (Fe) and folic .
