Heuristics and Biases Part I: Understanding Brexit from a Behavioural Economics Perspective - The Attribute Substitution Heuristic

Image: Shutterstock

by HK Lim (hklim [at] thetroublewitheconomics.com)

This is the first article in a series that examines how the Brexit referendum outcome can be understood through the lens of behavioural economics. We begin by first examining how limitations introduced by traditional economic thinking's reliance on the foundation of rational choice theory as a basis for economic decision making might have given rise to voter biases and the employment of heuristics and a subsequent deviation from "traditional" economic rationality. The usual argument by proponents of the rational choice school and its adherents is that individual economic rationality should prevail whenever the financial stakes are high enough, as would be expected in the case of the Brexit referendum. The outcome of the referendum clearly paints a very different picture, one of behavioural biases writ large.

The shock of Brexit and the day Britain woke to a changed reality

Friday 24th June 2016, the day Britain and the rest of Europe (and really the whole world) woke to a decidedly changed reality. In the Brexit referendum on Britain's membership in the European Union (EU), the British electorate had with a 51.9% majority (17,410,742 votes) voted to leave the EU versus a 48.1% vote (16,141,242 votes) to remain, stunning political analysts, pollsters, politicians both in the UK and across Europe, EU bureaucrats, as well as a deeply divided British electorate.

Source: Bloomberg

On the back of this result, Britain being the second largest economy within the EU as well as the fifth largest economy in the world, the Pound saw the biggest single-day fall against the greenback, the  leader of the Conservatives, David Cameron announced his resignation as Prime Minister, and Scottish First Minister Nicola Sturgeon immediately announced Scotland's desire to remain a part of the EU and reignited the Scottish devolution debate amidst calls for a new Scottish referendum in light of the United Kingdom's changed circumstances.

Source: Bloomberg

Framing the Brexit referendum question: "Should the United Kingdom remain a member of the European Union or leave the European Union?"

One interesting feature of this referendum was how broadly the referendum question was phrased, "Should the United Kingdom remain a member of the European Union or leave the European Union?" which arguably left open the option to interpret this question in many different ways. As a consequence, the run-up to the Brexit referendum was marked by vigorous campaigning in both the remain and leave camps that ranged the gamut from appeals calling on voters to carefully consider the aggregate economic costs and benefits of Britain's membership in the EU (and also constant reminders emphasising the potential costs to Britain of leaving the EU) to arguments that primarily channeled voter emotions of fear and/or the discomfort with future economic uncertainty.

While a myriad of economic justifications (of various degrees of complexity and plausibility) were bandied about by both the remain and leave campaigns in the run-up to the referendum, aggregate regional voting patterns across Britain painted a starkly vivid picture that highlighted the influence of deep socioeconomic divisions on vote choice that also revealed identifiable geographic variations in voter preferences and psychology. At the heart of these divisions were differing perspectives on the inequality associated with wage variations between regions as well as varying perceptions of the actual benefits of EU membership and the economic effects of the associated European immigration into Britain. These geographical variations also suggested the possibility that behavioral factors could have played a significant role in influencing the referendum's somewhat unexpected outcome through voter perception and behaviour. Well, the outcome was unexpected at least to the remain campaigners, the pollsters and the prediction and betting markets as well as even those who had campaigned to leave for this latter group appeared to be caught especially flat-footed by the referendum result with no clear plans on how to proceed with enacting Britain's departure from the EU.

We begin by examining how the limitations introduced by traditional economic thinking's reliance on the foundation of rational choice theory as a basis for economic decision making might have given rise to voter biases and the employment of heuristics and a subsequent deviation from "traditional" economic rationality. The usual argument by proponents of the rational choice school and its adherents is that individual economic rationality should prevail whenever the financial stakes are high enough, as would be expected in the case of the Brexit referendum. The outcome of the referendum clearly paints a very different picture, one of behavioural biases writ large.

Regional variations in voter preferences and divisions in aggregate geographic voter psychology

Before diving into a consideration of how behavioral considerations could have influenced the referendum result, a quick discussion of the aggregate regional voter patterns is in order. By regional totals, Scotland (62% remain vs. 38% leave), Northern Ireland (55.8% remain vs. 44.2% leave) and London (59.9% remain vs. 40.1% leave) opted to remain in the EU, while the remaining regions in England and Wales opted to leave.

Source: Bloomberg

The referendum results also indicated how polarised Britain was on the referendum issue, with regions that were expected demographically to vote remain voting overwhelmingly so, and regions that were expected to vote to leave also voting overwhelmingly to leave. To paraphrase the headline from an article in the  Guardian, "if you have money, you vote to stay, if you don't, you vote to leave." As it turns out, a significant predictor for a regional vote to remain was the proportion of the region's population with a university degree, consistent with the view that such individuals would have the greatest labour mobility (both within the UK and also within the EU) as well as also enjoy access to the highest paying jobs and were hence most likely to have benefited (and continue to benefit) from Britain's (continued) membership in the EU. Thus, voter patterns in the Brexit referendum also revealed deep divisions along socioeconomic lines. Similarly, younger voters overwhelmingly voted to remain in the EU while older voters overwhelmingly voted to leave. (Further referendum voter demographic pattern analysis is described here.)

Brexit as a failure of economic rationality? Hardly. 

One of the key questions raised post-referendum was whether there had been a failure of the majority of those who voted leave (or at the very least a significant proportion of this group) in "properly recognising" the supposed adverse economic impact of leaving the EU, based on the implicit assumption that supposed "experts" could be relied upon (to say nothing of their credibility) to aid the voter in understanding the ramifications of this decision. For now putting aside the subtler issues regarding the distinction between immediate short term adjustment costs contrasted against the longer-term uncertainty in both Britain and Europe's economic outlook given the dynamic effects of Britain's (and Europe's) response to the situation as well as changes in the EU membership as a consequence, the vagaries of the global economic cycles.

These considerations raise the associated question of whether there was a failure of economic rationality in segments of the British electorate who voted to leave. What does this all have to do with behavioural economics? Well, quite a lot actually. The very nature of the supposed economic arguments showcased in the referendum debate (particularly as articulated on the side of the remain campaigners) was hugely complex, often confusing and it was generally difficult for the typical voter to be expected to make sense of and evaluate all or even the majority of the arguments for or against Brexit, much less arrive at a well reasoned rational conclusion whether to vote remain or leave. This difficulty was further compounded by the contentious issue of European immigration and its perceived adverse impact on British jobs and wages, as well as growing perceptions of a broad increase in wage inequality between regions (and also between job categories), the reduction or stagnation of wages outside of a few major cities (for example London and Manchester), plus the failure of arguments based on trickle-down economics to convince the average voter of the benefits of the current economic system to say nothing of the actual distribution of such benefits. Taken together, we have all the ingredients for behavioural economics' main actors to play starring roles.

A quick caveat is in order, this article does not aim to examine or evaluate the economic arguments for or against Britain's membership in the EU, and the behavioral perspective discussed is not intended to detract from the argument that some of the leave voters may have in effect been casting a "protest vote" against the political establishment, the current economic system and its major players, and/or the supranational institution that is the EU. Rather, this behavioural approach is aimed at offering an additional perspective in understanding voter behaviour in the referendum via the insights of behavioural economics that are beginning to gain (an albeit) gradual acceptance within the ranks of mainstream economic thinking.

The attribute substitution heuristic: the role of perceptions of widening wage inequality between regions and the issue of immigration

What is the attribute substitution heuristic?  

In behavioral economics, a heuristic refers to a shortcut or a rule of thumb that assists, facilitates or simplifies complex decision making. We often unconsciously employ what behavioral economists term the attribute substitution heuristic when we are grappling with a complex or difficult question by instead answering a related simpler question in place of the original one. For example, in a home buying decision, instead of determining what price range might constitute a comfortably affordable home from the perspective of one's ability to afford the associated mortgage (based on one's expectations of current and future income and also taking into account related employment uncertainty dependent on future economic business cycles), buyers instead (often in consultation with mortgage advisors) typically work backwards by calculating the largest mortgage (and hence property value) they can afford given their current income coupled with often also an overly optimistic expectation of the projected growth in their future income as well as well as their future job prospects (this example is elaborated in the revised and updated edition of Dan Ariely's book "Predictiably Irrational").

Notice that in answering the more difficult question of "what size home and mortgage quantum is comfortably affordable?", the home buyer has instead substituted the question, "what is the maximum mortgage I can afford given my current and expected future income?" This easier latter question is easily answered using a standard mortgage calculator or spreadsheet to back out the maximum mortgage within the buyer's budget, but is an altogether different question from the original. So in effect, the potential home buyer has conveniently (well depending on the perspective) substituted a simpler but wrong question in place of the original question asked. In all likelihood, the potential home buyer will end up with a mortgage at the upper end of his or her affordability range (hence reducing the buffer for uncertain economic events), or even exceed the range of (comfortable) financial affordability altogether. Additionally, the potential home owner is also likely to exhibit an overconfidence bias in overly optimistic expectations of future income and employment prospects, and simultaneously underestimate the role of chance in his or her future, thereby compounding the original bias in selecting too large a mortgage. Clearly, this isn't a good outcome.

The attribute substitution heuristic and the Brexit referendum

This attribute substitution heuristic is also directly relateable to the Brexit referendum, where instead of evaluating the more complex and difficult question of the net economic benefit to Britain of remaining within the EU in deciding to vote remain or to leave (notwithstanding questioning the inherent lumpiness and typically unequal distribution of such economic benefits across individuals both by geographic area as well as by social economic strata), a voter may instead substitute a more readily answerable, simpler question. Instead of assessing whether the decision to remain in or leave the EU would be economically beneficial in the future to you and your family, voters (the swing voters as and particularly those those who voted to leave) would instead simply ask themselves questions similar to the following:

1. "Do I feel that Britain's EU membership has widened income/wealth inequality?"
2. "How do I feel about the European immigration resulting from Britain's membership in the EU?"
3. "Has European immigration on balance benefited me and my family?"
4. "Has European immigration contributed to a reduction or stagnation of wages in Britain?"
5. "Has European immigration resulted in a widening of income/wealth inequality in Britain?"
6. "Is immigration a positive for Britain?" (this last question was actually posed as part of a poll at the end of a Telegraph article in the run up to the referendum as can be seen here)

and their variations. All these questions can also be similarly posed with respect to a voter's perception and expectations of Britain's future prospects and need not be focused exclusively on past experiences.

Median wage variations across regions in the United Kingdom

The perception of a reduction or stagnation of wages in regions outside of the larger British metropolises (and perceptions of a widening income and/or wealth inequality across regions) is one possible reason why the attribute substitution heuristic might have played a major role in the framing of the referendum vote for those who voted to leave. This appears to be particularly true in the English regions outside of London (which overwhelmingly voted to leave) coinciding with the  regions with lower median incomes. Note that Scotland's interests in the outcome of the Brexit referendum are unique in that Britain's membership in the EU is perceived to be a significant factor in favour of Scotland remaining part of the United Kingdom (as per the 2014 Scottish referendum) and thus outlier demographic variations for Scotland should be considered separately from the rest of the UK.

Figure 1: Median full-time wages by region 

Source: ONS, Annual Survey of Hours and Earnings: 2015 Provisional Results

From the Annual Survey of Hours and Earnings: 2015 Provisional Results by the Office of National Statistics (ONS):

"In April 2015, London topped the regional list for median earnings for full-time employees, at £660 per week. Employees here earned £108 more per week than the next highest, the South East (£552), and £132 more than the median for the whole of the UK (£528). The high pay in London is largely due to a high proportion of its labour force being employed in high-paying industries and occupations, and also because many employees are entitled to allowances for working in the capital. The regional pattern has remained fairly consistent since the series began in 1997, with London and the South East consistently topping the list.

At the local authority level, earnings vary significantly. In April 2015 full-time employees working in the City of London had the highest median gross weekly earnings (£921) and those working in North East Derbyshire had the lowest (£389)."

As is clear from the infographic in Figure 1 as well as the accompanying text from the ONS reproduced above, by region, London topped the list for median earnings for full-time employees in the 2015 ONS survey and exceeded the median for the whole of the UK by £122 per week (23.1% higher than the UK median), and this regional pattern of median wage variation has remained fairly consistent since the series began first in 1997, with London and the South East consistently topping the list. This however, does not take into account part-time median incomes that would be even more relevant in areas with a significantly underemployed workforce. It is thus not surprising that the London region voted overwhelmingly for Britain to stay in the EU.

Does raising the economic stakes produce conventional economic rationality in voter decisions?

It is thus not implausible that local variations in economic prospects as perceived via variations in the median wage by region compared to more affluent regions can produce very different regional  perspectives on wage associated inequality as it relates to the Brexit referendum choice. There is an argument advanced by mainstream economic thinking (subscribers to the notion of "homo economicus" from rational choice theory as opposed to those espousing more nuanced behavioural perspectives) that if the financial stakes are raised (here in the form of the risk to wages and the subsequent possibility of further economic losses if Britain leaves the EU, further exacerbated by loss aversion), economic rationality with regards to the referendum decision will prevail. Tacit in this is of course the assumption that a plausibly convincing economic case can be made for Britain to remain in the EU coupled with the possibility of persuading portions of the electorate that do not directly benefit economically from EU membership in their sacrifice for the greater good. Thus the notion of rationality itself is very subjective and hinged on its definition. The very notion of the more narrowly defined economic rationality as a factor in justifying a remain vote from the portion of the electorate that does not directly see the redistributive benefits of the current economic arrangement with the EU is thus even less persuasive.

The arguments for remain based off the net economic benefits of EU membership are thus likely to be most compelling in regions like London (or in local authority areas like the City of London), where a higher proportion of the labour force is employed in high-paying industries anchored by multinational corporations and often involved in occupations like finance and banking. These areas are the ones most likely to directly enjoy (and perceive) the economic benefits of Britain's EU membership and the associated advantages offered by London's access to the single European market either through their jobs or via spillover opportunities from companies operating out of London serving the EU market. This would include for example employees in the financial sector (for example, foreign exchange traders for the Euro based in London) or sandwich shops servicing those individuals working for London-based multinational corporations servicing the EU market. 

In other regions (or local authority areas), it is less likely that voters will immediately perceive such direct economic benefits of Britain's membership in the EU and hence can be expected to reframe the referendum issue on more tangible terms related to their perception of the effect of Britain's EU membership on the median wage in their region compared to the more affluent regions/areas, and use this as a basis for assessing their feelings regarding the state of wage inequality across Britain.

Are we all expert economists and statisticians? The availability heuristic.

Of course, it is also not reasonable to expect every individual doing such an assessment to embark on a thorough evaluation of the totality on wage distribution data, and it is quite likely that individual voters will base their perspectives on their own circles and the experiences within their own communities, hence again producing a secondary bias via the "availability" heuristic. In the case of the availability heuristic, individuals typically draw generalised statistical conclusions of the frequency of occurrence for an event based off the ease with which they can recall particular examples of this event from memory or personal experience. So if one were to see depressed wages in one's immediate community, this would likely prompt the individual to over-generalise how widespread the phenomena of depressed wages might be hence biasing their subsequent evaluation of the issue (and thereby biasing the answering of the attribute substitution's replacement question), and vice versa. This is just one way the attribute substitution heuristic (combined with the availability heuristic) could have influenced and exacerbated differential viewpoints in the interpretation of the referendum vote decision.

The attribute substitution heuristic, the immigration issue and the experts

The topic of immigration was a hot topic in the run up to the referendum and duly exploited by some of the leave campaigners to sway the electorate's emotions in the vote (the "affect" heuristic also likely played a role here, more on this later). Notwithstanding the reduced faith in "experts" in reporting and forecasting net migration into the UK, immigration has been a polarising issue for the British electorate. The earlier discussion on regional wage variations segues naturally into the topic of how immigration also factors into the attribute substitution heuristic as evidenced by the direct mention of immigration in the sample of simpler substituted questions proposed above under the attribute substitution heuristic (Q2-6), for example to quote the earlier Telegraph example, "Is immigration a positive for Britain?"

A BOE report from 2015 found some evidence of the impact of immigration on wages (see "The impact of immigration on occupational wage: evidence from Britain" from the BOE Staff Working Paper 574), where it was estimated that in some low-skilled jobs a 10% increase in migrants resulted in a fall in wages for native workers of between 0.5% and 1.9%. While that might not seem like a lot, for someone working a retail job earning £17,000 a year, a cut of 1.9% translates potentially into £323 in lost wages. So it is entirely possible that regions characterised by lower median wages are also more likely to be affected by wage declines due to immigration, and voters examining the referendum choice from the perspective of immigration's effect on their wages are likely to be less quickly convinced of the economic benefits of EU membership and the accompanying free migration across EU member states (particularly into Britain).

Another way in which the attribute substitution heuristic could have worked via a voter's views on immigration is related to the notion of "a just noticeable difference" (JND) associated with the Weber-Fechner Law. The JND of some variable is proportional to the magnitude of the variable itself. For example, while it is less likely for you to notice if one of your car headlights is blown (especially when driving in a well lit city), going from one working headlight to none is definitely a JND. In the same way, rather than focusing on the percentage distribution of immigrants across Britain, it is typically more useful to consider the rate of growth of immigration within each region. A high value for this growth rate is more likely to constitute a JND and give the perception that immigration into that region is high. Based on the Labour Force Survey's data on foreign born persons living in the UK, it can be seen in Table 1 below that immigration across regions between 1995-2014 has in general exceeded 100% for most regions, which lends credence to the individual perception that immigration into the UK has been relatively high over the past two decades.

Table 1: Number of foreign born persons by region

Source: Oxford's Migration Observatory

According to Oxford's Migration Observatory's publication "Migrants in the UK: An Overview":

"The UK population was 13.1% foreign-born and 8.5% non-British citizens in 2014

The share of foreign-born people in the UK’s total population increased by over 50% between 1993 and 2014, i.e. from 7 to nearly 13.1%. During the same period, the share of foreign citizens rose from 3.6 to 8.5%, while that of recent migrants increased from 1.4 to 2.7%. There was a significant percentage increase in the share of foreign-born people in the UK’s total population during the 2004-2008 period."

So it is not altogether unimaginable that voters from regions with the greatest percentage increases in immigration may be more sensitive to the immigration issue raised by the Brexit referendum seeing as they are experiencing the largest growth (which are definitely more than the JND) in migration into the UK.

A further interesting insight on this issue regards the origin and citizenship of the foreign born residents (Note the reference here is to residents) in the UK. Again, according to the Oxford Migration Observatory's publication "Migrants in the UK: An Overview":

"India is the most common country of birth among the foreign-born, but Poland tops list of foreign citizens in the UK
 
India, Poland, and Pakistan  are the top three countries of birth for the foreign-born (Table 3) accounting respectively for 9.2, 9.1 and 6.0% of the total, followed by Ireland and  Germany. India and Poland remain the top two countries of citizenship of foreign citizens, with Poles being the biggest group, accounting for about 15% of the total." 

Table 2: Top 10 sender countries by country of birth and nationality, UK 2014

Source: Oxford's Migration Observatory

And as evidenced in Table 2 above, at least based on nationality, European nationals form the bulk of the share of the top 10 sender countries of foreign born persons in the UK as of 2014, and it is again not unforeseeable that voters in regions with the highest rates of immigration may associate Britain's membership in the EU with downward pressures on regional wages (and rising housing costs).

Taken together, it then becomes clearer how immigration pressures and depressed wages in regions with the highest rate of immigration are likely to contribute to the attribute substitution of the Brexit referendum question of whether the net economic benefits of Britain's membership in the EU warrant a vote to remain with the simpler question, "Is immigration a positive for Britain?" (Further analysis on the immigration question can be found in Oxford's Migration Observatory's "Migrants in the UK: An Overview" report here.)

Concluding thoughts on the attribute substitution heuristic and the Brexit referendum

Thus, as the aforementioned discussion illustrates, the attribute substitution heuristic may have played a legitimate role in reframing the Brexit referendum debate, and prompted voters in regions with lower median incomes to consider the referendum question from the perspective of whether EU membership may have contributed to widening wage inequality/differentials between regions (and also possibly between occupations) due to educational attainment related differential opportunities and the related impact of the effect of increased immigration from Europe into Britain on working class wages.

Despite the best efforts of remain campaigners to frame the Brexit referendum debate in terms of the net economic benefit of Britain's membership in the EU, the complexity of the economic arguments are likely to have prompted leave voters to substitute any such consideration of the referendum question with a simpler question that was more easily swayed by emotions. This was to be expected particularly if the public's faith in purported "experts" on the relevant economic issues had already been shaken. And of course, as highlighted earlier, it is entirely possible that a proportion of the electorate that voted to leave simply wanted to send a "protest" vote to the establishment to express their displeasure at being marginalised by the current economic arrangement. This economically disenfranchised group was more than willing to take a chance with another roll of the die, given that their existing economic situation was already pretty dire to begin with. And despite the risk that  things could possibly get worse, they would rather gamble on the chance that it could get better. I'll have more to say about this in a follow-up piece.

In Part II, the discussion will turn to focus on how cognitive biases could have influenced key players in the Brexit referendum itself. Stay tuned.