ciphergoth | The "mere token" effect

The "mere token" effect

Happened across a fascinating new cognitive bias today: the "mere token" effect. In summary:

If you offer people a choice between $300 in a week or $900 in a year, 62% of respondents choose the $300 in a week.
If you tell them that, immediately after they choose, you're going to give them $50, and then present them with the same choice, they make roughly the same decision.
If you offer them a choice between $50 now and $300 in a week, or $50 now and $900 in a year, suddenly 52% of respondents choose the $900 in a year option.

Obviously there's no rational reason to make a different choice in the second and third scenarios, the options in both scenarios are really the same, but the difference in the way they are presented makes a huge difference to people's responses.

It's worth noting that hyperbolic discounting with intertemporal bargaining, the model presented in "Breakdown of Will" that I've enthused about before, completely fails to predict this phenomenon.

Scope Insensitivity and the "Mere Token" Effect, Oleg Urminsky, 2006.

Update: I've edited the second point above: it used to read "If you give them $50, and then present them with the same choice, they make roughly the same decision". This gives the misleading impression that there's some difference in substance between the second and third points, which resulted in some discussions below. I hope that with the new phrasing, it's clear that there is no real difference at all between the second and third scenario; it is exactly the same choice phrased in two different ways. Updated: trying yet again with phrasing of the second choice.

Flat | Top-Level Comments Only

Wow. That's a new one on me, too.

But if they work for a bank, they'll ask you the prevailing interest rate and take out a calculator first to discount the forward amount :).

And then charge you $800 for the service :).

Could be rational. If you have no money, getting any money quick may be more important. Have just acquired money, one can then take a longer term view.

Probably a different effect: I remember reading a web article years ago that explained why some watch manufacturer made three models of a particular watch. I can't find the article or remember the details but it went along the lines of: if they made two versions, the cheap $10 watch and the expensive $30 watch, most people buy the cheap one. If they introduce a $100 watch and change nothing else, the exact same group of people will now choose the $30 watch.

Also there's the element of trust: anyone can *claim* to offer $900 in a year but will they pay up? Best call their bluff quickly.

However the $50 up front "proves" their intention to give you money so you will be more prepared to wait for them.

This is what i was going to say.

And in addition they are not so likely to change their already-made-up-mind from test1 to test2 when they are given the £50 as it would get them to question the intention of the £50 up-front (unless i misunderstood that and it is a different sample of people)

It is a different sample of people.

In all cases, the test is added to a customer satisfaction survey on a website. In all cases, you don't actually get any money until you've filled out all the options. In the second case, you get told you're going to get $50, you confirm that's what you're getting, and then you're presented with the choice; in the third you're presented with the choice with the $50 rolled in. So there is absolutely zero real difference between the two cases on which to hang a rational difference of decision.

Noodling more generally, it's funny how people say things like this in response to findings of cognitive bias. Usually, when people come up with what might be rational explanations for what appears to be irrational behaviour, new experiments test and falsify these rational explanations. This has happened a lot since Judgement Under Uncertainty was first published; the core findings are remarkably robust.

No, it can't be rational, because the second and third options are in practice exactly the same.

I sort of missed the second option and was going to say something like the above!

Ah, yes, I slipped on that distinction. (Though I dispute that in practice they are the same, because clearly there is a difference in outcome! In general there's a danger of ignoring the side-channel, which makes many more behaviours look irrational than is actually the case.)

So, alternative popsci explanation: Economics has the concept of a "sunk cost" - money you've already lost which should not affect your choice of whether to spend more money towards a certain goal or not. People allegedly have difficulty reasoning about this but maybe this experiment shows that on some level people understand the principle. In case (2) the $50 is a sunk cost for the experimenter, or perhaps a "sunk gain" for the subject. It can't be used to reason about the future behaviour or motivations of the experimenter, so has no effect. The promise of a future $50 indicates something about the experimenter's future behaviour however, and if made good on promotes trust in the distant reward of $900.

There's no difference in outcome, only a difference in presentation. In one instance people are encouraged to take the $50 into account when making the choice, and in the other they are not.

Unless by difference in outcome you mean that people make different choices, but I can't see how we can justify that difference by reference to itself!

I'm not sure I understand the scenarios as described.

(1) You ask them to choose $300 in a week or $900 in a year.
(2) You give them $50, then ask them to choose $300 in a week or $900 in a year.
(3) You ask them to choose $50 now and $300 in a week, or $50 now and $900 in a year.

So basically, giving them something now has no effect, but offering them the choice of something now does?

Yes, those are the three scenarios, and only in scenario 3 do people make the long-term choice. So whether or not giving them something now has an effect depends on whether you describe it as a part of both options, or a separate event that has nothing to do with the options.

It's perfectly rational if there's no trust - no expectation that the second or third payments in the future will materialise. By that logic, the experiment has established that $50 is a sum that people will 'punt' against the remote possibilty of a large sum in future; but if you offer them more money up-front, they'll take it far more seriously than an unlikely reward at some future date.

You seem to be saying the same thing as

deiberateblank and

hughe above; see my replies to them. There is no practical difference between the second and third options whatsoever; they are the same thing presented in two different ways.

They are only the same after the fact, in the same way that the probability of a fair coin landing heads is 0.5 before you throw it, but 0 or 1 afterwards.

No, they are entirely the same. I've edited the post to try and make this clearer.

Hmm. The original wording was clear, if wrong. The new wording is less clear, so I went to the paper. So the distinction is whether the token is presented as a separate event that will just happen anyway, or explicitly as a result of the choice, even though it'll just happen anyway in all available choices.

It looks like you've chosen the percentages for 1 and 3 from study 1a, which only considered cases 1 and 3. Case 2 was introduced in study 1c, which used a $100 'token' and a choice between $300 and $1000, resulting in percentages choosing the $1000 of 63%, 67% and 73% for the three cases.

Yes, I'm slightly mixing the results; they don't seem to have been as consistent as they should have been in their experiments, and in particular the psychological difference between $900 and $1000 skews things. I don't think it alters the thrust of the assertion though; people treat options 1 and 2 the same, but 3 differently.

This might depend on the precise wording used, but I wonder if it's a question of people not starting to listen properly until part way through the sentence?

Suppose option 2 says "You get $50 now, and also you can choose $300 in a week or $900 in a year", and option 3 says "You get $50 now and $300 in a week, or $50 now and $900 in a year". Someone who wasn't listening properly for the first four words of each would naturally treat option 2 exactly like option 1, but would treat option 3 very differently and might easily consider that money right now trumps money in a week in addition to the extra $600 a year later.

Except the paper states that it was an online survey, and participants were asked to type back the reward values to confirm comprehension.

(Though admittedly, seeing "Flurgius sjnc sdun isudnc iusdcu sdm $100 sdkjn kjnsd sdjn sdjkcn skjdc sdjn kjn jnsdc $1000 sdkjcn jkjsnc", followed by "TYPE IN THE AMOUNT OF THE FIRST REWARD: $|_______|", admits of correctly answering the question without fully taking in the implications. I'm sure they were more careful than that though.)

Very interesting and depending on my personal circumstances at the time of asking I suspect I would respond in the following fashion

Scenario One. Take the $300, I can't really explain but the idea of having to wait a year means I'm less likely to trust the other person to pay the $900

Scenario Two & Three. Slightly more inclined to go for the $900 but still worried about the year waiting for it, can't explain why though.

Fascinating!

Interesting. It puzzles me what kind of "explanation" for this phenomenon would be helpful. As I understand the assertion is that the representation of the offer in scenario 2 & 3 is different while the substance of the offer is identical. Therfore the argument is that the subjects are making an irrational choice (although it isn't clear in which case) as they are responding differently to offers which from the point of view of substance (i.e. cash in this case) are the same. The typical explanation for such phenomena is that the subjects have misinterpreted the offers, which are slightly odd, usually because they mistake them for types of offers that are more familar to them but which are actually substantively different (often having money involved skews the results, I haven't checked the study, but they often repeat the offer as a game with points instead of money and don't find the same effect - again I don't have a reference).
So to my question, what is a useful explanation of this phenomenon? There can be no explanation by logic as in order to apply logic you have to convert option 2 & 3 into identical propositions which leaves you dry. I have discussed above an explanation by analogy. My view is that usually human beings base decision making of this type on heuristics not logic - they try to think of outcomes of similar choices that have been offered to them in the past. You can play around with defining your offer in way that looks like one that was a good bet in the past versus one that has looked like a bad bet in the past. You can play around with the concreteness and the abstractness of the offer (money versus points). The common human strategy while definitely irrational seems to be a somewhat successful option evolutionarily so far which is probably the most you can say. I'd guess that the strictly logical approach to such questions might have flaws too. There'd probably be too many pieces of information to take into account to fully assess the value of $300 now versus $900 later and in order to avoid stasis you just have to break the loop - the precise phrasing of the offer leads to the average human breaking the loop one way or the other. I bet if it was £300 now versus £1m at the end of the year - logical behaviour would reassert. Anyway I've gone on way long - anyone else want a go?

Argh! No fair!

Yet another cognitive bias to watch out for. I like to think I'm relatively good at this sort of thing. My personal preference for deferred gratification options sometimes errs on the too-strong/time-discount-too-low side, which is unusual. (I've been known to save treats long past their edibility dates, and not particularly regretted it.) But nevertheless - I strongly suspect I'm somewhat prone to this when making snap judgements under pressure. And so I'll be trying to bear this one in mind, along with all the others I'm aware of ... adding to the risk of cognitive overload, making me more prone to errors of judgement. Whereas people seeking to exploit this in offers have the time to consider (and can pick just one).

Actually, though, those who do seek to exploit that sort of flaw (whether legit sales staff or scam artists) tend not to restrict themselves to a single strategy, and spotting any such tactics tends to be a red flag for me, and once I'm actively suspicious I'm very hard to convince.

'Frinstance, I can't see how I'm ever going to get my fogged double glazing panels replaced, since it requires interacting with double glazing sales people, and I've yet to meet ones who don't scream "Run away! Do not buy!" to me. It's life-limitingly foolish to invite known vampires in to your home ... but what if vampires are the only ones who can fix your windows?

Another hit-and-run-point: it'd make for a clearer comparison if you were consistent in the percentages reported - so the base is 62% of people choosing the $300 now option, and the changed 'mere token' scenario has 48% of people choosing the $300 now option.

That's only 14% of people changing their minds - it looks like it could well be significant, but it's clearly not as universal/strong a phenomenon as some others. Including the base effect of dramatic discount rates of which the 'mere token' effect is a modulation.

Um. I'm probably being really stupid, but I read the 'revised' second and third options as ...

Second: I get $50 now, and then get to chose between $300 and $500.

Third: I get to chose between $50, $300 and $500.

... which seem completely different, and I can't seem to read em any other way. Like I said, probably me being a dumb.

oops, I meant $900's for the $500's of course. I really am a dumb today.

First option:

Would you like
$300 in a week, or
$900 in a year?
[OK]

Second option:

You will receive $50 as soon as you press OK. In addition, would you like
$300 in a week, or
$900 in a year?
[OK]

Third option:

Would you like
$50 now and $300 in a week, or
$50 now and $900 in a year?
[OK]

That's definitely the clearest explanation of it!

No idea why the difference in percentage, though...

That's definitely the clearest explanation of it!

Yes it is, but there's problem. It's not what the study tested. The percentages from the study

ciphergoth quotes don't directly relate to the second and third options. It may be the case that if the experimenters had done the research that

ciphergoth suggests, they'd have got the percentages he states, but the study itself doesn't state that.

For a start, the $50 was not offered 'now' or 'as soon as you've pressed OK'. Additionally, it wasn't always $50. In option three it was offered in three days (page 12 of the paper). In option 2 $100 (not $50) was offered as a separate reward, but still in three days (page 15 of the paper). This was not an instant reward. Whether or not that changes the results is a matter for conjecture, but it still means that the situation wasn't as

ciphergoth presents it. Personally, I think a reward offered in three days is a conceptually different thing from one offered instantly.

That said, if you only look at 1c and don't mix in the results from the earlier survey, I think

ciphergoth is right that options 2 and 3 are logically identical, but for some reason aren't seen as such by the participants. And I think there results do still need explaining. So in a sense, I'm quibbling, but I'm a stickler for accuracy in reporting, especially when what's being reported is a counter-intuitive result that appears to show people behaving irrationally.

Here, as best I can present them, are the options from 1c of the paper (pages 15 & 16), where at least the quantities of money are consistent, so the results should be directly comparable. Please check my working with the original paper - if

ciphergoth can slip up in reporting the results, so can I:

Option 1:

Choose between:

* $300 in a week
* $1000 in a year

37% choose $300; 63% choose $1000.

Option 2:

Choose between:

* $100 in three days and $300 in a week
* $100 in three days and $1000 in a year

27% choose $300; 73% choose $1000.

Option 3:

First, agree to receive $100 in three days. Then choose between:

* $300 in a week
* $1000 in a year

33% choose $300; 67% choose $1000

Note that the study includes z- and p-values for the results, but my statistics is too rusty to be certain I'm interpreting them correctly. If anyone else would like to check the original paper and report, I'd be grateful.

And, just to reiterate, I agree that options 2 and 3 are, on the face of it, logically identical but presented differently, and that the fact that people do choose differently between the two options is unexpected and worthy of investigation.

I'm not surprised, and I'd even say I expected something like that to be uncovered after learning of the very common bias where people under-value the future, either compared to the present or incrementally for different intervals. One way to think about it is that in casual situations people seem to apply an excessively high depreciation, like *0.25 per year, when calculating the NPV of the offer.

My guess is that, in more serious and controlled situations the same people would apply a more reasonable depreciation, like *0.8 a year or better. It seems that the "small amount of money now" fixes this for the casual situation of the experiment. It's not clear to me if the monery really suppresses a bias (skewed valuation) or if it addresses a straightforward issue of trust that otherwise confounds the experiment. I would certainly doubt whether I'd see the 1-week or 1-year payments, and fifty-one weeks less of doubt is valuable.

Another possible confounding factor is the "closed world" formulation of the experiment. In a closed world, "broke now" is really bad but "money in my pocket" may be comfortable enough to allow me to wait for the better pay-off in a year. I might be broke again next week, but who cares.

Oh, believe me, no matter what the circumstances, the default state of almost anyone is 'broke now'. Insert smiley.

Flat | Top-Level Comments Only

The "mere token" effect

no subject

no subject

no subject

no subject

no subject

no subject

no subject

no subject

no subject

no subject

no subject

no subject

no subject

no subject

no subject

no subject

no subject

no subject

no subject

no subject

no subject

no subject

no subject

no subject

no subject

no subject

no subject

no subject

no subject

no subject

no subject

no subject

no subject

no subject