Visualisation is the trendy way to represent data these days, but sight is not our only sense. Via Dave Giles I see that some researchers are exploring sonification as a way to represent information. They claim that

…with complex data series one can often hear patterns or persistent pitches that would be difficult to show visually. Musical pitches are periodic components of sound and repetition over time can be readily discerned by the listener.

Giles has previously covered the topic and it’s well worth having a look at the series he examines in that post. I know I’ve spent far too much time staring blankly at a volatile time series plot before attempting various transformations just so that my eyes can make sense of it. If there is a better way that uses other senses to quickly discern patterns in the data then I’m all for it!

For a more populist example, here’s a cellist playing a time series of temperature readings: A song of our warming planet

Experience may not improve judgement

We’ve all met hardened cynics in our professional lives. Those people who think the worst of those they meet at every turn because they’ve been burned so many times. They give nobody the benefit of the doubt and look down on new staff for their hopelessly naivety and gullibility. The question posed by a group of researchers in the latest EJ is whether judges are similarly afflicted by experience.

They take a panel of UK Competition Commission decisions from 1970–2003 and evaluate the effect of the chairman’s experience on the probability of an adverse finding. Using a panel of that size allows them to control for various effects such as the chairman’s age.

Using a unique data set of companies investigated under UK competition law, we find very strong experience effects for chairmen of investigation panels, estimated from the increase in experience of individual chairman. Probit and IV probit regressions indicate that replacing an inexperienced chairman with one of average experience increases the probability of a ‘guilty’ outcome by approximately 30% and, after chairing around 30 cases, a chairman is predicted to find almost every case guilty.

Housing fuelled consumption boom?

In the EJ:

There is strong evidence that house prices and consumption are synchronised. There is, however, disagreement over the causes of this link. This study examines if there is a wealth effect of house prices on consumption. Using a household-level panel data set with information about house ownership, income, wealth and demographics for a large sample of the Danish population in the period 1987–96, we model the dependence of the growth rate of total household expenditure with unanticipated innovations to house prices. Controlling for factors related to competing explanations, we find little evidence of a housing wealth effect.

Housing wealth and consumption – an upper bound, not an estimate

Via Economist’s View, I saw this paper by Case, Quigley, and Shiller (REPEC).  Let me start with the positives, which are many:

  1. They are excellent writers,
  2. They get to the point – and have a clear idea of how important this sort of issue is for trying to understand cyclical phenomenon,
  3. They have pulled together a large data source consistently, which is a lot of work.

So as you can tell from that, I have a lot of respect for them, their work, and what they are doing.  However, I have a giant misgiving about the way they’ve framed the result they have found.  Fundamentally they HAVE NOT estimated the causal impact of housing wealth on retail sales/consumption.  In the introduction they do not make this claim hunting down an “association” … but by the conclusion this is what they are starting to claim they have done:

The importance of housing market wealth and financial wealth in affecting consumption is an empirical matter … we do find strong evidence that variations in housing market wealth have important effects upon consumption.

These descriptions are veering on causal, which is very inappropriate in a situation where you have an obvious, and likely significant, case of omitted variable bias!

Let us think about this.  Demand for housing is similar to demand for other durable goods – when confidence is high, unemployment is low, income expectations are elevated, and financial conditions are good, demand for both will rise, pushing up prices.  As a result there are some “third variables” that will drive up demand for both.  They cover this off at the end by stating:

Underlying our analysis is an assumption that it is useful to think of causality as running from wealth components to consumption, and not that, for example, the two are determined by some third variable, such as general confidence in the economy. We believe even more strongly that these new results demonstrate that it is useful to think of consumption as determined in accordance with the models we have presented. In consulting this evidence, recall that our measure of housing wealth excludes wealth changes due to changes in the size or quality of homes, changes that are likely to be correlated with consumption changes merely because housing services are a component of consumption. We have alluded elsewhere to others’ evidence using data on individuals that the reaction of consumption to stock market increases is stronger for stockholders than for non-stockholders (Mankiw and Zeldes, 1991), and that the reaction of consumption to housing price increases is stronger for homeowners than for renters. This lends additional credibility to our structural models when compared to a model that postulates that general confidence determines both consumption and asset prices.

To think about this point let’s think about housing.  Housing is a durable consumer good.  As the price of housing goes up relative to other goods and services, then given other goods and services constitute a “normal good”, spending on other goods and services should fall!  Of course, it also constitutes a transfer of wealth from homeowners to renters – and as a result, we have to ask about these separate markets in order to figure out what is going on.

As a result, the point that homeowners and renters behave differently is VERY useful, and justifies the study.  However, it in no way supports ignoring omitted variables and just deciding that the model is causal – in fact the way they have dismissed OVB is far too casual, given that there was no effort to deal with it (FE estimators deal with unobserved heterogeneity that is constant through time – this is not the case with our OV’s).

The evidence here appears to point at the fact that changes in house prices are a good proxy for changes in access to credit – hardly surprising given that housing is an asset and a durable consumer good.  When trying to understand the tendency of movements in retail spending, and the set of risks going forward for such spending, using house prices as a proxy for a set of “real structural” variables is useful.  However, this evidence is far from suggesting a causal relationship – and even further from suggesting that there is anything policy relevant here (as we need to understand the structure of the relationship in order to understand how changing policy settings will change outcomes – a change in policy settings can change the fundamental relationship between variables, think Lucas Critique!).

When the authors began to discuss this as causal, they should have stated that this provides an “upper bound” on the impact of housing wealth on consumption – and that more detailed analysis would be required.  They could even have gone further and stated that “given the size of the link, it is more likely that there is a tendency for higher house prices to drive up consumption” – that would have been mildly contentious, but reasonable.  As it is, their comments that they are estimating the size of a causal link are misleading.

Is advertising evil?

Vox says the data supports Matt’s priors:

There is an old debate in economic theory… about whether advertising increases or decreases the prices of consumer goods. Some have argued that advertising provides information to consumers, such as information on prices or the existence of products. This information increases the degree of competition in a market, and thereby lowers consumer prices. On the other hand, there is the view that advertising changes the preferences of consumers, for example by shifting demand curves outwards, increasing the monopoly power of brands or decreasing elasticities of substitution. All these effects should lead to an increase of market prices.

…advertising increased consumer prices in some industries such as alcohol, tobacco and transportation, in which the persuasive effect dominates. But it also decreased consumer prices in other industries such as food. …those industries which exhibit the informative price include more information in their advertisements, consistent with the interpretation of informational and persuasive forces of advertising.

The aggregate effect is informative, which means that, on average, advertising decreases consumer prices.

Also, a perspective from inside advertising.

The issue of assumptions

In a recent post, James recently raised an essential, and fascinating, point on assumptions.  To borrow his own words:

Assumptions are often made for tractability, rather than realism, yet still influence our conclusions. It isn’t possible to control for the unrealistic assumptions; if it were we wouldn’t have made them. That means our conclusions will be biased by assumptions we’ve made only for convenience and we need to bear that in mind when considering the policy implications of our models.

This reminded me of an essay by Maki which can be found in “New Directions in Economic Methodology“.  The essay was title “Reorienting the assumptions issue“.

In this essay, Maki does a number of interesting things – but in terms of this specific issue his key insight was to differentiate between “core” and “peripheral” assumptions.  This is an insight we have borrowed many times when talking here on the blog.

A core assumption “constitutes the theory” while a peripheral assumption does not.  What does this mean?  Well it means that the “core” makes up the central set of necessary assumptions required to achieve a certain result!

It is these core assumptions that need to be “realistic” in some sense of the word – while peripheral assumptions are merely there to increase tractability, or make a result easier to interpret.  Fundamentally, our core result should only rely on our core assumptions.

Core assumptions in econometrics

For those who are statistically inclined, this also fits neatly inside our view of econometric theory.  Fundamentally, the assumptions we make about error terms in a regression in econometrics are akin to checks regarding whether we have all the appropriate “core” assumptions for the econometric model and the question we are asking with that model.

For example, we may be trying to explain some independent variable (y) with a set of (presumed to be) exogenous dependent variables (x).  However, it turns out one of these x’s turns out to be correlated the error term – so that this variable is “endogenous“!  If we wanted to estimate the impact of this dependent variable on our independent variable through typical OLS, our estimate will be biased (this is the justification for instrumental variables).  In this case, the model is likely underspecified in some way – such as having missing “core” variables (omitted variable bias), or a missing “causal relationship” (reverse causation between the independent and dependent variables).

A conclusion

As the above example hopefully showed, what dictates a “core” assumption in a model depends strong on what question we are trying to answer with the model!  An analysis of what constitutes “core” assumptions for our analysis, and then a discussion regarding whether these core assumptions correspond to a realistic set of assumptions, should be an essential component of all model building exercises in economics.