Home > Uncategorized > DSGE macro models criticism 4. Capital.

DSGE macro models criticism 4. Capital.

Earlier posts in this series, which takes short looks at DSGE models using the lens of statistical concepts, were about money, market fundamentalism and unemployment.

There are at least three fundamental differences between ‘capital as statisticians measure it’ and ‘capital as a concept in neoclassical macro models’. Next to this, statisticians as well as neoclassical economists shy away from a crucial aspect of the ownership of capital: it’s forged in the fire of revolutions. Examples are the Protestant revolution (Cromwell, the Dutch Revolt, the  Glorious Revolution, expropriation of the massive wealth (land!) of many cloisters), the Enlightenment revolutions (the French revolution, the abolishment of slavery and the Civil War) or the decolonization revolutions. This last point won’t be elaborated but it is good to remember it when rating the statistics and the models – just read ‘Capital in the twenty first century’ to get an idea about the importance of slaves as a main asset on USA balance sheets before the civil war.

The three differences (to be elaborated below):

1) Statisticians treat ‘land’ and ‘unproduced assets’ as a separate category next to ‘financial assets’ and ‘fixed depreciable capital’ while the macro models lump these together (or take fixed private depreciable assets as the only meaningful kind of capital).

2) Statisticians use a hodge podge of methods to value different kinds of capital. This is a reflection of ‘the real world’ as different kinds of capital should be valued in different ways, for one thing  because not all investments  have a market price (see below). This does, however, lead to aggregation problems. Neoclassical models circumvent these problems by basically assuming that there is only one ‘consumption/investment’ good (Robinson Crusoe, or the representative consumer, chooses between consuming coconuts or investing in a stock of coconuts, mind that in this non-monetary example saving equals investment).

3) Neoclassical models seem to mix up the assets and the liabilities side of the balance sheet (for instance by assuming that all capital goods are owned by households and leased to companies). Economic statisticians rigourously use a quadruple accounting system which enables them, for one thing, to track sectoral changes in wealth and (albeit somewhat rudimentary) ownership.

Ad 1). To describe production and the distribution of income and wealth classical economists used the trichomoty ‘land, labour and capital’ why neoclassical economists use the dichomoty ‘labour, capital’. This enables neoclassical econmists to mix up the asset and the liability side of the balance sheet as no subdivision of the asset side has to be made while, an accounting convention, the sum of assets and the sum of liabilities are equal which means that you can, interchangeably, use the total value of assets or the total value of liabilities to estimate ‘capital’. But it disables them to track down  (changes of) the composition of the (ownership of) capital and to make analytical differences between financial capital (basically: owning some kind of liability), produced capital (planes, ‘intangible assets’ like copyright, computers) or unproduced assets (stocks of Lithium, land, aquifers, …). And to make an analytical difference between the distribution of income related to ownership of unproduced assets (think: Saoudi-Arabia) and related to the (legal and economic) ownership of the profits related to using the machinery to drill for oil. There is solid evidence that the change from the trichotomy to the dichotomy was meant to obfuscate out insight in exactly such processes. But, as stated, the statistics enable an analysis of such processes.

Ad 2). How to value capital? No, I’m not going to bore you with the ‘Cambridge controversy’. It will be worse… I’m going to bore you with arcane statistical manuals. For some links to such manuals see the annexes of this publication. Statisticians basically use three methods to value capital. While one method is not used by them.Sometimes, they use market prices, like in the case of houses or bonds. To be able to do this, there has to be a lively secondary (i.e. second hand) market. Sometimes, the perpetual inventory method: ({original investment – depreciation + additional investment} divided by a proper deflator). Sometimes, they use replacement costs. The method which they do not use is discounting assumed future streams of monetary income by an assumed interest rate, as it turns out that, as interest rates can change by the day, this method does not yield stable results. Not all methods are fit for all kinds of fixed capital. ‘Replacement costs’  are not a good way to estimate the value of stocks of natural gas (instead, in the Centraal Bureau voor de Statistiek of the Netherlands uses a medium run average of gas prices us used, together with an estimate of the amount of ‘producable’ natural gas left). Dikes do not have a market value, for a number of reasons and ‘replacement costs’ are a better gauge to estimate their value than using the perpetual inventory method (think of the dikes as a gift of existing generations to future generations: provided that maintenance is optimal (which it has to be) depreciation is zero and the value can be equated to the costs of building them anew). According to neoclassical macro models, however (I’ve demystified the formula a little), “we … devine another variable to correspond to the price of capital (measured in terms of goods) a = b/c This is the marginal value (measured in goods) of an additional unit of installed capita” (2011 graduate college notes by Chris Sims). This marginal value is mainly based on the value of present and future ‘utility’ (and a little bit on the utilisation rate), as macro-utility (Yours? Mine? Ours? Voters? What about children? Illegal aliens? Demented people? Incarcerated people?) is not well defined or operationalized and as the formula excludes capital goods without a market price and non-produced assets or capital with a market priced owned by the government, this is not any kind of well conceptualized, defined, operationalized, let alone measured variable and totally at odds with the way capital is actually valued and it is not even in a theoretical way possible to compare it with the concept of capital used and the amounts measured by statisticians. Neoclassical economists nowadays also publish empirical DSGE models, like this recent IMF one which states that it’s special because it does not equate capital accumulation with savings. As it only uses a perpetual inventory method to estimate ‘capital’ it however excludes non-produced capital, while price changes (houses!) are excluded, too. I’m not sure if investments in houses are included at all, if so the article abstracts from the fact that owner occupied houses are not used in a market setting but in a partly monetary but clearly non-market household production setting (sorry guys, the real world is complex). Also, it uses a VAR approach (a multi dimensional running average which basically models the economy as a kind of spiders web: when a 3-D spiders web experiences a ‘shock’ everything starts to tremble and forward and backward linkages are very complicated – but it will return to equilibrium). Also (help): “The Kalman filter with state and measurement equations is used to fit the data to the model”. Despite this last sentence: using a VAR to estimate time series is throwing the DSGE model under the bus (as the time series do not contain independent estimates of utility) And the modeller accepts the definitions of the statistician, albeit in a twisted and implicit way.

Mind, however, that when it comes to capital statisticians are to an extent comparing apples with pears as they do use different valuation methods. Also, there is a strong case to be made that in a fiat money surrounding the value of existing houses (including land underlying houses) is decisively influenced by money creating mortgage lending by banks while prices of financial assets are influenced by central bank policies.

ad 3) There is a reason why balance sheets have two sides. One side shows the value of assets – the other side shows who ownership rights (whch are by definition future oriented) including rights to income related to using and/or owning these assets (profits, dividends, interests). This is, of course, interesting in its own right. Next to this, the statistics enable us to use a quadruple accounting concept. This enables us to analyse the consequences of, for instance, out of control mortgage lending related housing bubble and its consequences. A house price bubble will show up as an increase in the value of houses and (on the liabilities side) an increase in the value of household ownership of housing capital, an increase which however will be offset by an increase in gross mortgage debt – debts owed to the banks which on the asset side of their consolidated balance will see a related  increase in the value of mortgage debt owed to them by households. Now, imagine what happens when house prices decline again: on the asset side of the household balance the value of houses will decline, on the liability side net equity will decline (or even become negative…). On the balance sheets of the banks – nothing will happen. With means that interest income of the banks will, despite the decline of equity of households, not decline. A severe housing bust also leads to declining consumption, investments and decreasing incomes, this means that households have to pay a larger share of their income as interest while at the same time owning less capital…. Banks, however…  The point: the statistics enable such analyses. Neoclassical macro models don’t as they typically use a representative consumer who owns everything. Or maybe they use two consumers (which by necessity will lead you to using quadruple accounting methods).  But these sectors are still not congruent with the sectoral approach of the statistics. As they typically do not have a sector banks and do not use quadruple accounting methods. They are not up to the task set for them: explaining the real world in a way consistent with our measurements. which indicate that we have to use a multi-sector approach and different kinds of capital when we want to explain either production or the distribution of income. By the way – one kind of capital, the net international investment position, is very difficult to measure.

  1. No comments yet.
  1. No trackbacks yet.

Leave a Reply

Fill in your details below or click an icon to log in:

WordPress.com Logo

You are commenting using your WordPress.com account. Log Out / Change )

Twitter picture

You are commenting using your Twitter account. Log Out / Change )

Facebook photo

You are commenting using your Facebook account. Log Out / Change )

Google+ photo

You are commenting using your Google+ account. Log Out / Change )

Connecting to %s