These are just a few potential issues that HANK and RANK models might face, and researchers continue to work on improving these models to better understand the dynamics of modern economies.

Identification or sequence of combination among

variables

its close to ardl and parch method

very similar to lm serial Lagrange multiplayers

Bzuayehu Alemu

thank you for a systematic characterization of HANK. I understand the basic characters of HANK models.

Is there no serious problem with HANK?

This is an example of introductions to HANK economics:

• There is significant recent interest in how inequality between consumers may affect the monetary transmission mechanism and how inequality is affected by monetary policy. Traditional New Keynesian (NK) models rely on a setting with a representative agent (RA) and thus by definition do not allow this topic to be analyzed.
• The RA setup is analytically very convenient, however, and the hope, perhaps, has been that its main conclusions are robust to extensions to multiple agents. With new analytical and computational tools available, a burgeoning literature is now beginning to explore the conjecture that most of the RA model’s conclusions are indeed robust to consumer heterogeneity.
• Kaplan et al.(2016) coin the term HANK models to represent HA extensions to the NK model, thus making RANK an appropriate term of NK models using the RA assumption, with no pun intended.

From

Tobias Broer, Niels-Jakob H. Hansen, Per Krusell, and Erik Öberg (2016) THE NEW KEYNESIAN TRANSMISSION MECHANISM: A HETEROGENOUS-AGENT PERSPECTIVE. National Bureau of Economics Research, Working Paper 22418.

After having read and analysed many articles with New Keynesian (and New Classical) models, I find that most of these models are rather stupid. Now I have studied the NBER article “The new Keynesian mechanism: A heterogenous-agent perspective” and think, it belongs to the most stupid ones. I wonder, in which economy (bubble?) NK economists live to invent or accept their unrealistic assumptions. New Keynesians are, of course, no Keynesians, I fear, only a few of them have ever read Keynes’ GT.

My arguments are in the same order as the article and the numbers of equations. So let us start with production. There is one representative firm which “produces the final good” with a continuum of intermediate goods. The production function is of CES type (1), but with integral instead of sum. The continuum means that we have infinite intermediate goods – even more: one cannot count them or identify one. To make the matter tractable and to change to a finite number of goods and firms, let us change from ∫ to ∑, i,e. Y = (∑Yia)1/a with i = 1, 2, …, n the number of intermediate goods and firms producing them and a = (e-1)/e (e as the elasticity of substitution, assumed 6). Strangely enough this production takes place without labour and capital. Obviously, the contribution to production is the same per “unit” of inputs, which means we can use the same measurement (e.g. kg). If one uses this measurement for the final good, too, it is easy to show that the output/input-relation is falling with higher n.

The intermediate goods are produced by a continuum of firms indexed j el.(0,1) with a simple production function: Yj = Nj, where Nj is labour input (working time), which itself is the Dixit-Stiglitz aggregate (instead of simply the sum) of labour inputs from (a continuum of) households (equ.(2)) with different “types of labour” (but the same productivity per hour). This is a rather rare bulk of nonsense assumptions. If we change from ∫ to ∑ in (2), we can show that Dixit-Stiglitz aggregation leads to a reduction of labour input: The working hours used in production is much lower than those supplied by the households, and the gap increases with the number of households. By the way, one should never assume a continuum of goods, persons, firms etc. or (what is the same) or take real numbers as indices.

For the price setting of the intermediate firms, the authors introduce a strange assumption, namely that they can reset (i.e. change) their price against the previous period (quarter) only with a given probability. I have never heard of such a rule. Obviously, the authors assume some higher power, who plays at dice and tells the firms who is allowed to change its price. The price equation (3) is (not only) therefore fake. The “accounting equation for the evolution of real wages” (4) remains a mystery, at least I do not understand it. For the interest rate (5), a so-called Taylor rule is applied. For a study showing the influence of interest rate policy (“shocks”), I would prefer to take the interest rate as exogenous.

“Each household can reset (set) its (nominal) wage each period”, but a part (chosen with a certain probability th) of them must supply their labour at the same wage as in the quarter before – a very special economy(?)! Every household chooses its consumption and labour supply by maximizing a utility function for an infinite future. Other than in usual functions, the time preference rate ß is multiplied by this probability. This means, that utility is lower, when more households can (re)set their wage, and the utility is zero, when wages are flexible (th = 0). A critique of models with such utility functions can be found in my short article “Nonsense Economics”.

The statement that Cjt equals Ct above (9) contradicts the market clearing Ct = Yt, because Ct is total consumption (total production is consumed), whereas Cjt is only the consumption of household i. The nominal wage equation (10) is a curiosity. As a part of households change their wage whereas the others must stay with the wage of the preceding quarter, one has to calculate an average. Normal people would simply take the weighted average of both periods, some statisticians might prefer a geometric mean. Not so economists infected by Dixit-Stiglitz aggregation. They somehow mix arithmetic and geometric mean calculation to get to a more complicated (more impressive) formula, which delivers wrong results (underestimation for substitution elastities>1).

To transform the RANK model to a HANK model, the authors add “a representative capitalist” to the “unit mass of workers”. That is really a curiosity in an economy without capital. The capitalist does therefore neither save nor invest and has no active function. Nevertheless, he cashes the profits and uses this sum for consumption. The consumption of the workers is reduced by this amount.

I think it does not pay to go into the solutions of the numerical model, because of the highly unrealistic assumptions and mistakes, the results cannot be accepted.

Anton Rainer Bzuayehu Alemu

you have a very different opinion on HANK. I hope you will give comments on the other's post.

Dear Anton,

thank you for your long comment. Is it a citation from one of your papers?

You refer to numbered equations but they do not appear in the comment.

Dear Yoshinori,

I am surprised, that you seem to take New Keynesian and New Classical models serious, although you are critical ageinst GE models, especially what concerns "market clearing" via prices.

My comment was not only on HANK, but also on RANK. The numbers of the equations of the NBER article are in my comment, but I have not rewritten the euations there.

I do not see me in contradiction to Bzuayehu and Adnan, it is true that solving NK models may difficult. In most cases, it is even not possible. In usual NK models one can e.g. cann calculate the growth rates of consumption and labour supply )via Lagrange multipliers, but not the absolute values of these variables for the first period. My article on "Nonsense Economics" is on ResearchGate.

That the NBER model cannot be sensibly solved can be seen from the first two euations.

Yi in (1) and Nj in (2) are not (cannot be) functions of i and j, respectively. To make the equations tractable, one can assume they are the same (constant) for all i and j or one can leave the continuum, which means a change from integral to sum:

(1) Y = (sum(Yia)1/a and (2) N = (sum(Yja)1/a , where a=(e-1)/e with e as elasticity of substitutions, which may be different in (1) and (2), but is assumed 6 in both in the article. These equations lead to rather strange results.

Supplement to my last answer

(unfortunately I typed on "Add" before completing)

In the annexed tables, you can find some simulations to the Dixit-Stiglitz aggregation and concerning the calculation of averages. The simple example of producing bread ("Brot") shows that these models are rather a caricature than serious economics.

Dear Anton,

Maistream has a "good" tactics to change their topic every 10 to 20 years. IN this way, all young economists were obliged to follow what is happening there. Only a few convinced mavericks can resist this temptation. In this way, majority of young hopeful economists are attracted to the gravity field of mainstream economics.

RANK, HANK, and TANK are new brands (not so glorious as DSGE or RBC but attractive enough to draw attention of many naive young economists). That is why I am interested in those "new" trends.

Anton Rainer Bzuayehu Alemu Adnan Ali Chaudhary John B. Davis Édgar Hernando Sánchez Cuevas Diethard Pallaschke Victor Cruz-e-Silva

On a related note, have you seen Sargent's paper comparing Old Keynesian and HANK: http://www.tomsargent.com/research/HAOK_HANK.pdf?

Although I hold Sargent a "marginalist economist" (more interested in mathematical methods than in economics), I have read this paper. I regret this waste of time. But we should accept his proposal on the last page to remove the K from RANK, because New Keynesian models are, in effect, anti-Keynesian.