VoxEU Column Frontiers of economic research Global crisis

The invisible hand meets the invisible gorilla: The economics and psychology of scarce attention

Have economists been asleep at the wheel? This column reports from a conference on the psychology and economics of ‘scarce attention’. Among the ideas discussed is whether too much information can blind decision-making and whether this can explain why so many economists missed the warning signs of a crisis.

Since the financial crisis, many commentators have asked why so many economists failed to predict it – or even whether economics played a part in causing the crisis. A group of UK experts in 2009 attributed this failure to predict to a “psychology of denial” that had gripped the financial world as a whole.1

The economics profession has since continued to evaluate its own role in the financial catastrophe and subsequent economic crisis. The subject’s standard assumptions about how people take decisions and choose to behave have been a particular focus for scrutiny. Did the assumption of rational, self-interested choice, given the available information, in itself contribute to a dreadful misunderstanding on the part of regulators and policymakers about what could happen in the financial markets?

Behavioural economics offers several examples of alternative rules of thumb about behaviour that describe typical decisions more accurately than the standard assumptions. But to address the question systematically, economists will need to learn from psychology (specifically, the psychology of individual choice in situations where people are faced with a constant flow of information, as they are with many economic decisions).

These issues were addressed by participants in a recent workshop at the Toulouse School of Economics on the Psychology and Economics of Scarce Attention (see summary here). The key theme was whether cognitive scientists’ growing understanding of how, given people’s limited attention, ‘sense perceptions’ translate into behaviour offers any lessons for economists.

A vivid illustration of the potential parallels was offered by an example given by CNRS Research Director Kevin O’Regan of the Institut Paris Descartes de Neurosciences et Cognition, of an airplane crash (in a simulator) due to the pilot’s failure to see another aircraft in front of him as he landed. Why did the experienced pilot not see the airplane on the runway? Is this question in fact the same as asking why did most people not see the impending crash in the financial markets, when the information was available for those who looked carefully?

Reprinted from J.K. O'Regan, Why Red Doesn't Sound Like a Bell, Oxford University Press (2011) and courtesy of NASA.

‘Inattentional blindness’ is commonplace. The best-known example concerns what people about half the time fail to see when set the task of watching a video and counting passes between two teams of basketball players – a gorilla walking across the screen (Chabris and Simons 2010). The neuroscientists taking part in the workshop were not convinced the analogy between inattentional blindness and failure to predict was valid, but the economists thought the invisible gorilla could offer them a more fruitful set of assumptions than the invisible hand when it comes to how people form their preferences and make their decisions.

Given that attention is selective in dealing with the vast amount of sensory information that can be taken in from the outside world, both voluntary and involuntary factors could be involved in making that selection. The architecture of the brain holds the key to understanding the selection mechanisms. Many different areas of the brain deal with visual perception. Each has its own functional specialism, functioning concurrently and interactively. The system is complicated and more or less hierarchical, and all the senses come into play at the same time. There is a vigorous competition among neurons in the selective process.

The current view on scarce attention is therefore that it emerges from the architecture of the brain. For example in visual processing, information travels from lower to higher levels of the brain. The lower level neurons (in the primary visual cortex) are sensitive to simple features such as edges while the higher level ones are sensitive to objects and categories (such as faces). In addition, the lower level neurons are activated by visual features in specific locations while higher level neurons are not so specific, and the corresponding receptive field (the region of space in which the presence of a stimulus will alter the firing of that neuron) is larger. Both of these mean that each high-level neuron is associated with several low-level neurons – the latter are competing with each other in a ‘winner takes all’ contest. There is a loss of information, which is the neural basis of scarce attention.

There seems obvious scope for applying the lessons from psychology about how to direct attention effectively – for example, in finding appropriate visualisations of financial market data to help alert the authorities to future crises. The economists were also interested in broader lessons for the assumptions made about decision-taking in economic models. One area of economics where attention is a key factor is in advertising. Do we pay enough attention to ads to alter our purchasing choices - does advertising work?

The issue of the effectiveness of advertising has become more acute as a growing amount of both media and spending move online. Advertisers are considering where they can most effectively reach consumers, while consumers are faced with new types of advertising and, increasingly, ‘information overload’. Online ads generate plentiful data, and economic research is beginning to address some of these issues.

Clicks on ads online account for 98% of Google’s revenues. Hal Varian, Chief Economist at Google, said that the position of an ad on the screen determines how frequently people click on it. The choice of placement is partly cultural – depending for example on whether people read left to right in the relevant language – and partly evolves as a result of Google learning from users’ and advertisers’ choices over time. The positioning of each ad on the online page is therefore determined by the value to each advertiser of the position they end up with; and the difference between their bids is a measure of the value of attention. If people were equally likely to click on any ad on the page, advertisers would not be willing to bid more to be placed at the top of the page.

The conventional wisdom about the increasing commercial pressure on both the newspaper and the broadcasting industries is that their revenues are being undermined by both consumers’ unwillingness to pay for anything online and by far lower revenues for online than for offline advertising. Varian argued that it is wrong to claim offline ads are vastly more expensive, however, because the relevant comparison is the price of attention, not the price of the ad itself.

Offline readers generate 88% of newspapers’ ad revenue in the US, and spend an average 24 minutes a day reading the paper. Online readers generate 12% of ad revenue and spend just 1.2 minutes a day reading. Offline advertising is therefore cheaper per minute of reading as, compared to online reading, it gets 20 times the attention for seven times the cost. The reason seems to be that online reading is done at work, offline at home. That makes the real challenge for newspapers getting people to spend longer reading them online, Varian concluded.

David Reiley (2011), Principal Research Scientist at Yahoo! Research, described the results of an experiment looking at over a million customers who purchased from a large US retailer and were also users of Yahoo! The customers were randomly assigned to treatment and control groups. Looking at data for each customer on ad exposure and weekly purchases at this retailer, both online and in stores, showed statistically and economically significant impacts of the advertising on sales. The effect persisted for weeks after the end of an advertising campaign, and the total effect on revenues exceeded the retailer's expenditure on advertising. An advertising campaign costing $25,000 brought additional sales over two weeks of $83,000 (plus or minus $70,000 – the confidence interval is wide because retail sales data show great variability). After five weeks, a $33,000 ad campaign is estimated to have increased sales by $250,000 plus or minus $190,000. Most of the effect of online ads on sales occurred through physical sales in the stores, however.

Other evidence presented at the workshop suggested that personalising online ads is often less effective than sending generic messages. The tailoring of ads to the individual according to their browsing behaviour – known in the industry as ‘dynamic retargeting’ – is the focus of enormous hope among advertisers for greater effectiveness thanks to the relevance of the ads displayed to the individual. But evidence from an online experiment presented by Anja Lambrecht of London Business School, using randomised dynamic retargeting by a travel firm, suggested that a generic ad was more likely to induce a consumer to purchase than a specific one, increasing the probability of conversion by 60%. The only people for whom the specific ad was more effective were those who had already visited a review website to search within a specific product category. Lambrecht suggested that this distinction may indicate that many consumers do not start out with well-defined preferences, and use the search process to refine their preferences. But if they already know enough about what they want to buy to visit a review site, they can be presumed to have already reasonably fixed preferences. The lesson for the advertising industry is that greater effectiveness will require greater insight than they have at present into consumers’ decision-making, needing to know how firmly their preferences are already fixed, and consequently what level of detail a targeted ad needs to provide to be effective (Lambrecht and Tucker 2011).

References

Chabris, CF and DJ Simons (2010), The Invisible Gorilla, and Other Ways Our Intuitions Deceive Us, Crown

Lambrecht, Anja and Catherine Tucker (2011), “When does retargeting work? Timing information specificity”, Working Paper Series.

IDEI-TSE (2011),The Economics and Psychology of Scarce Attention: Executive Summary.

Reiley, David (2011), “Does retail advertising work?”, 1 June.

Stewart, Heather (2009), “This is how we let the credit crunch happen, Ma'am …”, The Observer, 26 July.


1 In November 2008, Queen Elizabeth II asked economists to explain why the credit crunch had happened. Six months later they provided a three-page document signed by some of the country’s top economists blaming “a failure of the collective imagination of many bright people”. See Heather Stewart’s (2009) column in the Observer.

1,260 Reads