DSGE macro-models criticism, a round up. Part 6. Consumption
Looking at neoclassical macro-models through the lens of economic statistics. Today: consumption.
According to a Eurostat headline, “In 2014, CO2 emissions in the EU estimated to have decreased by 5% compared with 2013“. It is important to know if this decline was caused by technological progress or by lower consumption and/or investment. Alas, these data are too vague and fuzzy to answer such questions, as the article also states:
It should … be noted that imports and exports of energy products have an impact on CO2 emissions in the country where fossil fuels are burned: for example if coal is imported this leads to an increase in emissions, while if electricity is imported, it has no direct effect on emissions in the importing country
Can’t economic statisticians do better than this? Yes, they can. A Eurostat publication about this states:
“The EU-27 total of 7.8 tonnes of carbon dioxide emissions per inhabitant in 2011 was composed of three main elements:
- some 4.9 tonnes per inhabitant resulted from the consumption by households and governments of goods and services;
- a further 1.6 tonnes per inhabitant resulted from direct carbon dioxide emissions from private households in the EU-27 (for example, through the burning of fossil fuels for private vehicles or for heating);
- another 1.3 tonnes per inhabitant resulted from (production related to) fixed investments — also referred to as gross capital formation — in the EU-27 economy.
There was a slight reduction in CO2 emissions per inhabitant in the EU-27 between 2009 and 2011. Extended supply, use and input–output tables have been used to estimate carbon dioxide emissions induced by the final use of products within the EU-27 in 2011. Besides the carbon dioxide emitted by industries while processing products for final use, the estimates presented also take into account the carbon dioxide that is ‘embedded’ within the EU’s imports; these emissions arise from the worldwide production chains of goods that are imported into the EU-27. Carbon dioxide emissions that are embedded within products that are made in the EU but exported outside of the EU-27 are, in a similar vein, included in the accounts for non-member countries.
So, we do have empirical models which enable a thorough analysis of the (in this case: environmental) consequences of for instance consumption by households and of government consumption (which is not consumption by the government but consumption by households provided by the government, like a large part of education and health care and comparable activities). Which makes it all the more puzzling why neoclassical ‘macro’-models (so-called DSGE models) are in denial about the existence of government production (hence the parentheses around ‘macro’) and therewith ‘government consumption’ by households and stick the idea that only private purchases add ‘social utility'(whatever that is). When we investigate ‘consumption’ in these DSGE models the next five aspects require attention:
1) First, the central tenet of DSGE ‘macro’-models: social utility. Looking at this through the lens of statistics (actually: of metrics in a much more general sense) it does nots eem to be a valid scientific concept. It lacks a sound concept, a proper definition, a clear operationalization and an established way to measure it. The models assume some kind of relation between consumer purchases and utility – but this relation is not measured in any way. Romer, in his textbook Advanced macro economics, states that the concept of social utility is based upon the idea of permanent consumption, but at the same times states that permanent consumption stinks as an empirical concept (the end of chapter eight). To be clear: the social utility function is not based upon some kind of aggregation procedure of individual utility but seems to be an assumed emergent variable with (in some models) a quadratic shape or (in other models) a Cobb Douglas shape or even another shape. Or even, as shown by the work of Romer (same link, p. 386), a variable which can change its shape instantaneously, dependent on the point the researcher wants to make: “To see the issues involved, it is easiest … to assume that individuals have constant-relative-risk-aversion utility rather than quadratic utility“. Mind that it’s also not clear if this applies to individual consumers or to the aggregate of all consumers. Brrr.
2) As shown above, economic statisticians make a difference between household consumption purchased by households and household consumption provided by governments. DSGE models almost never make this distinction (look here) , surely not the models used by official institutes, like the ECB Eagle model. Models are of course simplify reality, just like maps. But this is not just a question of an abstract ‘small scale map’- but of a map which only shows private roads while government roads are left out of it. It will bring you nowhere. Still, it is used for policy purposes, as the incomplete and biased Eagle model was explicitly mentioned by Mario Draghi in his highly political recent speech about the perceived need for more labour market flexibility, less government and less product market rules. And do not underestimate the amount of government production and consumption. Look here for Eurostat data on EU government expenditure by function: among other things, government expenditure equal to 7,2% of EU 28 GDP was spent on health and 5% on education. And it’s also important as part of total consumption. Look here for a Eurostat estimate of total consumption (i.e. consumption items purchased by households plus goods and services provided by the government) of households: Actual Individual Consumption or AIC. According to the statisticians of Eurostat AIC is, compared with the household purchases concept of consumption used in DSGE models, a superior concept as
“Actual Individual Consumption consists of goods and services actually consumed by individuals, irrespective of whether these goods and services are purchased and paid for by households, by government, or by non-profit organisations. In international volume comparisons of consumption, AIC is often seen as the preferable measure, since it is not influenced by the fact that the organisation of certain important services consumed by households, like health and education services, differs a lot across countries.”
It’s not just that neoclassical models are in denial about ‘government consumption’. They also do not investigate why governments produce such a lot of education and Healthcare (among other things). Look here for a paper by June Sekera about the lack of a theory of ‘government choice’ or ‘government action” which takes account of the fact that non-market systems of production and distribution, like the family, do not just exist because ‘perfect markets’ are not possible for this kind of production (and no, dear neoclassicals, the family is not a ‘hidden market’: market transactions are characterized by the fact that prices are negotiated before transactions take place (or not), which is not what happens in a family. The hunter who does not catch a prey does not lose his right to gathered food because of this). Romer, in the fourth (!) edition of his neoclassical textbook (see the link above) does not even start to discuss such important conceptual issues and restricts consumption to household purchases. According to June Sekera (in a mail): “It’s astonishing that Romer has no role for government as producer. From what I can see, in his model government only purchases the output of firms. His slighting of the crucial role of government as producer is all the more amazing considering that his wife was head of Obama’s Council of Economic Advisors. Maybe they don’t pillow-talk about economics.” It is valid to argue that the government is, sometimes or often, an inefficient or ineffective or even repressive producer (though the yardstick to measure this can’t be the market, as this institution knows its own inefficiencies). But it is bonkers to ignore government production.
3) A comparable argument can be made for ‘Non Profit Institutions Serving Households’ like churches, unions, sports clubs, the AAA, Greenpeace and whatever, which are measured by economists. Though in a quantitative sense not as important as the government these organizations are crucial to civil society as well as households, ignoring them (like Draghi implicitly does in his speech mentioned above, by using a neoclassical model) might all too easily lead to an eroding of the very basis of our society.
3) The lack of a sound concept of consumption in neoclassical macro of course comes at a cost. The Romer textbook mentioned above (fourth edition…) is unusually garbled when it comes to empirics – probably because the concept of ‘utility’ itself is elusive and garbled. It takes too much space to mention all his mistakes in his chapter about consumption, but this excerpt gives an idea:
“Now consider the differences between blacks and whites. The relative variances of permanent and transitory income are similar in the two groups … But blacks’ average incomes are lower than whites’; as a result, the estimate of a [the constant of an estimate of consumption in a simple regression of income and consumption, M.K.] for blacks is lower than the estimate for whites … To see the intuition for this result, consider a member of each group whose income equals the average income among whites. Since there are many more blacks with permanent incomes below this level than there are with permanent incomes above it, the black’s permanent income is much more likely to be less than his or her current income than more. As a result, blacks with this current income have on average lower permanent income; thus on average they consume less than their income. For the white, in contrast, his or her permanent income is about as likely to be more than current income as it is to be less; as a result, whites with this current income on average have the same permanent income, and thus on average they consume their income.”
Wow. Men are, on average, taller than women and will often buy larger trousers. But even when the relative variance of both groups is equal this does not mean that when I take a sample of one from te group of women and the woman in question (who sometimes wears high heels and sometimes walks barefoot) turns out to be 1,90 meter meter and I take another sample of one and happen to draw the same women that she’s suddenly 1,80 or 1,60 meter because are on average smaller than men… (technically I should have introduced shoes with and without high heels and measure the woman while she is wearing shoes in this example, but that makes it too complicated)! Romer acts as if this is the case however, and assumes that this causes the woman in question buy small trousers. See also this Lars Syll post which makes the same points albeit in a more abstract way. The entire chapter by Romer on consumption is stuffed with comparable garbage, continuously mixing up unique individuals with averages with aggregates. Look here for some information about Cole and Ohannian, two economists who make comparable mistakes about unemployment during the Great Depression.
4) According to Romer, the neoclassical consumption function (the ‘Euler equation’) supposes a rational equilibrium between consumption when you’re eighteen and consumption when you’re eighty with the additional assumption that the eighteen year old knows his or her income for the rest of his or her life. Alas, real consumption seems to be pretty much a function of present income, according to Romer hismelf – at least for individual households. Now, it is of course possible that total consumption is a kind of emergent variable which can be described with a social utility based Euler equation (if only we could measure social utility, of course…). But the DSGE models I’ve seen do not show this. There is no thorough discussion at all of the long run development of consumption and saving, pensions, life expectancy, family formation and demographic developments and the like – the sector households is not even properly defined. Mind that data about this are available! Crucial assumptions are not checked in any serious way.
5) The lack of attention paid to culture and history (this applies to an extent to economic statistics, too). Household size has been declining for at least a century. Household technology has changed dramatically (light bulbs, television, the washing machine, houses,…). The number of people over 80 years of age is rising dramatically. This list can of course be extended: households and household technology are changing all the time. Any discussion of intertemporal consumption should take this in consideration, surely as many of these changes require the purchase of consumer durables (houses!). Durables (cars, smartphones) are by the way a weak spot of DSGE models as well as of national account statistics, surely when it comes to how these are used and how these change our behaviour.
Rational intertemporal optimization of consumption requires that we cut CO2 emissions – a lot. We need input-output models to help us do this. Alas, the assumption of the neoclassical macro models that rational intertemporal optimization is what we do is wrong. Ironically, this is shown by the very fact that macro-economists hardly use these input-outputs anymore – and do use neoclassical models. We can do better than that.
Earlier posts in this series, which consists of concise posts looking at DSGE models using the lens of statistical concepts, were about money, market fundamentalism, unemployment, capital and the intertemporal budget constraint for the government.