Alle innlegg av Erik Ø. Sørensen

The effect of fast and slow decisions on risk taking

Etter en lang og treg prosess har endelig samarbeidsprosjektet med Magnus Johanneson, Caroline Bonn m.fl. – som starta under fra sabbatsåret i Stockholm – kommet på trykk i Journal of Risk and Uncertainty. Open access.

Abstract: We experimentally compare fast and slow decisions in a series of experiments on financial risk taking in three countries involving over 1700 subjects. To manipulate fast and slow decisions, subjects were randomly allocated to responding within 7 seconds (time pressure) or waiting for at least 7 or 20 seconds (time delay) before responding. To control for different effects of time pressure and time delay on measurement noise, we estimate separate parameters for noise and risk preferences within a random utility framework. We find that time pressure increases risk aversion for gains and risk taking for losses compared to time delay, implying that time pressure increases the reflection effect of Prospect Theory. The results for gains are weaker and less robust than the results for losses. We find no significant difference between time pressure and time delay for loss aversion (tested in only one of the experiments). Time delay also leads to less measurement noise than time pressure and unconstrained decisions, and appears to be an effective way of decreasing noise in experiments.

Testing av AD8307 RF effektmeter

RF-power meter, sensor board.

Jeg har ambisjoner om å bygge meg et 50-Ohm terminert RF effektmeter til testing av filtre og andre småkretser. Mange har bygd slike før meg – feks. har Roger Hayward, KA7EXM, skrevet en fin artikkel i QEX, men jeg tenkte det ville være en fin øvelse å legge ut kort og lage en kontroller selv. Selve detektorkretsen er uansett veldig enkel og kommer ganske rett fra Analog Devices dokumentasjon. Meteret skal kunne måle -70 til 7 dBm og AD8307 skal levere 25mV/dBm. Hayward foreslo litt kompensasjon på inngangen som jeg også har implementert.

Jeg la ut kretsen og bestilte kretskort fra Osh Park den 19/12, og i dag fikk jeg tre kort i posten – to uker med jula i mellom, og alt kosta bare 15 usd inkludert frakt. Ikke ille! Kretskortene er lilla og ser flotte ut. Litt lodding og kortet var ferdig til test i benk. Til endelig bruk ser jeg for meg å montere detektoren i en støpt aluminiumsboks, og at jeg bygger dette inn i en større boks med et med en AD-konverter og display grensesnitt. Jeg ser at mine opprinnelige tilkoblingsplaner (BNC for RF og DB-9 for strømforsyning og signal) ikke vil passe på den måten jeg tenkte, men en tilpasning av dette får vente.

For testing kobla jeg til 9V fra labforsyninga mi. Det er en 78L05 som regulerer 5V lokalt på kortet. I alt trakk kortet bare 10mA. Jeg føder signal fra min HP8657a generator gjennom en ebay-levert meter-lang RG316 coax med BNC kobling til kortet. Jeg leser detektorspenninga på mitt Uni-T 61E multimeter. Det er litt lavfrekvent støy i målingene på lavt nivå (kanskje bare på den åpne benken min), litt løpende gjennomsnitt i meteret burde kunne løse det.

Her er utspenninga for noen forskjellige frekvenser:

20170103-test

Bare ved de høyeste frekvensene er det mulig å skjelne punktene fra hverandre, det ser ut som om det er litt mer demping i kretsen da. Noe er nok pga dårlig testoppsett, men kanskje kan jeg justere litt i frekvenskompensasjonen eller kretskortutlegget dersom jeg gjør endringer for å tilpasse tilkoblingene bedre (tror SMA gjennom boksen ville være bedre og ta mindre plass, og bedre kontrollert impedanse på RF-delen ville kanskje hjulpet noe). Men alt i alt synes jeg tallene taler om suksess: God linearitet i 1-500MHz, med behov for litt nivåjustering av de høyeste frekvensene. Og kurven passer fint i henhold til ADs datasheet med 25mV/dBm:

## 
## Call:
## lm(formula = V ~ dBm + frequency)
## 
## Residuals:
##       Min        1Q    Median        3Q       Max 
## -0.032368 -0.006030 -0.001346  0.004840  0.029123 
## 
## Coefficients:
##                Estimate Std. Error t value Pr(>|t|)    
## (Intercept)   2.082e+00  4.597e-03 452.899  < 2e-16 ***
## dBm           2.528e-02  6.161e-05 410.256  < 2e-16 ***
## frequency5    7.500e-03  5.952e-03   1.260 0.212987    
## frequency25   1.519e-02  5.952e-03   2.552 0.013528 *  
## frequency100  5.878e-03  5.952e-03   0.987 0.327736    
## frequency250 -2.348e-02  5.952e-03  -3.944 0.000229 ***
## frequency400 -4.192e-02  5.952e-03  -7.043 3.21e-09 ***
## frequency500 -8.236e-02  5.952e-03 -13.836  < 2e-16 ***
## ---
## Signif. codes:  0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
## 
## Residual standard error: 0.01263 on 55 degrees of freedom
## Multiple R-squared:  0.9997, Adjusted R-squared:  0.9996 
## F-statistic: 2.41e+04 on 7 and 55 DF,  p-value: < 2.2e-16

Mine røffe benkmålinger gir 25.28mV/dBm i gjennomsnitt. Det lavere nivået ved 500MHz tilsvarer omtrent 3.2dB, mens ved 250 MHz ligger målingene mindre enn 1dB lavere enn ved 1MHz nivået.

Når en bestiller kort fra Osh Park får en alltid 3 stk. Dersom noen har lyst på et av overskuddskortene til sin egen lab må de bare gi beskjed!

What Explains the Gender Gap in College Track Dropout? Experimental and Administrative Evidence

Sammen med kollegaer har jeg et notat i årets utgave av American Economic Review (Papers and Proceedings), om «What Explains the Gender Gap in College Track Dropout? Experimental and Administrative Evidence«. Med Ingvild Almås, Alexander W. Cappelen, Kjell G. Salvanes og Bertil Tungodden.

Sammendrag: We exploit a unique data set, combining rich experimental data with high-quality administrative data, to study dropout from the college track in Norway, and why boys are more likely to drop out. The paper provides three main findings. First, we show that family background and personal characteristics contribute to explain dropout. Second, we show that the gender difference in dropout rates appears both when the adolescents select into the college track and after they have started. Third, we show that different processes guide the choices of the boys and the girls of whether to drop out from the college track.

Fairness and family background

With Ingvild AlmåsAlexander W. Cappelen,  Kjell G. Salvanes and Bertil Tungodden.

Abstract: Fairness preferences fundamentally affect individual behavior and play an important role in shaping social and political institutions. However, people differ both with respect to what they view as fair and with respect to how much weight they attach to fairness considerations. In this paper, we study the role of family background in explaining these heterogeneities in fairness preferences. In particular, we examine how socioeconomic background relates to fairness views and to how people make trade-offs between fairness and self-interest. To study this we conducted an economic experiment with a representative sample of 14-15 year-olds and matched the experimental data to administrative data on parental income and education. The participants made two distributive choices in the experiment. The first choice was to distribute money between themselves and another participant in a situation where there was no difference in merit. The second choice was to distribute money between two other participants with unequal merits. Our main finding is that there is a systematic difference in fairness view between children from low socioceconomic status (SES) families and the rest of the participants; more than 50 percent of the participants from low SES families are egalitarians, whereas only about 20 percent in the rest of the sample hold this fairness view. In contrast, we find no significant difference in the weight attached to fairness between children from different socioeconomic groups.

Available as NHH discussion paper 2015/15. 

Fairness in bankruptcy situations: an experimental study

Nytt arbeidsnotat ute i dag, med Alexander W. Cappelen, Roland I. Luttens og Bertil Tungodden (Discussion Paper 17/2015, Department of Economics, NHH Norwegian School of Economcis). Har tatt litt tid å få dette paperet ferdig, deilig å ha en ferdig versjon!

Abstract: The pari passu principle is the most prominent principle in the law of insolvency. We report from a lab experiment designed to study whether people find this principle a fair solution to the bankruptcy problem. The experimental design generates situations where participants work and accumulate claims in firms, some of which subsequently go bankrupt. Third-party arbitrators are randomly assigned to determine how the liquidation value of the bankrupt firms should be distributed between claimants. Our main finding is that there is a striking support for the pari passu principle of awarding claimants proportionally to their pre-insolvency claims. We estimate a random utility model that allows for the arbitrators to differ in what they consider a fair solution to the bankruptcy problem and find that about 85 percent of the participants endorse the proportional rule. We also find that a non-negligible fraction of the arbitrators follow the constrained equal losses rule, while there is almost no support in our experiment for the constrained equal awards rule or other fairness rules suggested in the normative literature. Finally, we show that the estimated random utility model nicely captures the observed arbitrator behavior, in terms of both the overall distribution of awards and the relationship between awards and claims.

Linker: Til paperappendiks med noen flere resultater og Instrukser/screenshots fra eksperiment.