New lecture on income inequality

[Harveyt47]

New data-centric lecture on changes in income inequality.

One area of interest is how income inequality has changed over time and space. i.e. has income inequality increased in country X over the past 100 years?

This paper looks at wage inequality in different areas of the US and does analysis of it over time.

Possible data source for US data is the Survey of Consumer Finances (SCF)

We should try to make this flow on from the wealth dynamics lecture. As well as making use of its code as much as we can.

Lecture could look something like this.

• overview: flowing on from wealth dynamics; premise of the questions we will aim to answer (has income inequality increased over time?); brief intro to the data we will use.
• Refresh of Gini coeff and Lorenz curves, a bit of code illustrating them.
• Description of the data. With some documentation on how we sourced it and cleaned it.
• Implementing Lorenz curve and Gini coefficient to the data. (graph Gini coefficients over time)
• Discussion and interpretation of the results.

CC: @najuzilu @shlff @sayaikegawa

[Harveyt47]

Also a good paper on this topic. They use US SCF data

[Harveyt47]

We could do this with Australian data but the only source I can think of which gives us what we need is HILDA and we won't be able to share that data with others.

Hence I think using US data is the best path.

[jstac]

Hi @Harveyt47, can't we share aggregated data? After all, lots of people show figures created from HILDA, which is essentially the same thing...

[Harveyt47]

Hi @jstac yes you are right we can share Aggregated data. See on page 6 here under 'Usage'.

What is the best way to aggregate it? I might be able to do it by area

[jstac]

Thanks @Harveyt47 . You showed me a Lorenz curve from HILDA previously. Can't we show that and the data points that define it?

[Harveyt47]

I used the household level data to make that curve, which I think is Unit Record Data so I would need to aggregate the data in some way to share the data points with others.

[jstac]

I see. I think we could just include the data points themselves (that is, the points that make up the curve), along with the plot of them, and explain where they come from (but that the data is confidential).

[Harveyt47]

Yes that sounds good. I can do that thanks @jstac

[Harveyt47]

So we could show something like this

[jstac]

Wow. That's a big spike. Just before the GFC. Fascinating. I wonder how closely this correlates with returns on assets (CC @shlff)?

Any earlier data available? Any estimates? Other sources of income data?

Anyway, we can definitely include that --- we can give as much info as possible to help people replicate ("To get access to the data, try this.." etc.)

How about data from US / Japan / China / European countries? The more the better. Then this page will be a great resource.

It would be great to combine modern charts with old historical stuff too...

[Harveyt47]

Hey @jstac HILDA only started in '01. I'll look for other sources of data. I know the ABS has some stuff on income data going back to 1994-95, see data here, but am unsure of the quality I can get it at (hopefully I can get more than percentiles).

Yeah I have some good documentation on how you can get the HILDA data and then how to use it.

I'm Looking into the US, I will use data from the SCF. This survey is a little tricky, there are weights that need to be applied because there is oversampling of high income household in the survey.

I will look into data on the other countries you mentioned ASAP.

[thomassargent30]

I recommend not having a stand alone lecture on income inequality. It should be bundled with a lecture on consumption inequality.

On Tue, May 12, 2020 at 9:28 PM Harvey Thompson notifications@github.com wrote:

Also a good paper https://urldefense.proofpoint.com/v2/url?u=https-3A__www.minneapolisfed.org_research_qr_qr3711.pdf&d=DwMCaQ&c=slrrB7dE8n7gBJbeO0g-IQ&r=hV4qLWg4zodvX-YQ-ETIbA&m=jD0GFzQCNCtF11Sma2zXs2bxJvG6TocCc_UQBBFKD7c&s=Ex86XJSJNApck9ZGLAVIThn2RxwoPP7QEutOlbYcHqo&e= on this topic. They use SCF data

— You are receiving this because you are subscribed to this thread. Reply to this email directly, view it on GitHub https://urldefense.proofpoint.com/v2/url?u=https-3A__github.com_QuantEcon_lecture-2Dpython_issues_212-23issuecomment-2D627722959&d=DwMCaQ&c=slrrB7dE8n7gBJbeO0g-IQ&r=hV4qLWg4zodvX-YQ-ETIbA&m=jD0GFzQCNCtF11Sma2zXs2bxJvG6TocCc_UQBBFKD7c&s=dHWiq36nwL2GTFLijaWGmyEse9-x4WDB3F8PgPV6qJc&e=, or unsubscribe https://urldefense.proofpoint.com/v2/url?u=https-3A__github.com_notifications_unsubscribe-2Dauth_AA63MTROHREHE6PNKZQXKQDRRIHXFANCNFSM4M6TEZAA&d=DwMCaQ&c=slrrB7dE8n7gBJbeO0g-IQ&r=hV4qLWg4zodvX-YQ-ETIbA&m=jD0GFzQCNCtF11Sma2zXs2bxJvG6TocCc_UQBBFKD7c&s=w9U4H81oCdmn3OMh45-pOTS4FRBHo82ARZs699YEHzY&e= .

[jstac]

Thanks for the comment @thomassargent30. Aside from general interest, the long game is to model the joint distribution of earnings and wealth, particularly at the tails. The need to match the joint distribution puts useful restrictions on the model. I'm thinking here of http://www.econ.nyu.edu/user/benhabib/earnings%20and%20wealth.pdf.

Is your point that consumption is more important for welfare? In any case, I think it's a good idea to include consumption if we can find good data.

[thomassargent30]

My reasons are several, including consumption being more important than welfare. There is an influential group of macroeconomists who have emphasized the greater equality in consumption than in income or in wealth. We want to attract them and their students to our lecture.

On Wed, May 13, 2020 at 4:42 PM John Stachurski notifications@github.com wrote:

Thanks for the comment @thomassargent30 https://urldefense.proofpoint.com/v2/url?u=https-3A__github.com_thomassargent30&d=DwMCaQ&c=slrrB7dE8n7gBJbeO0g-IQ&r=hV4qLWg4zodvX-YQ-ETIbA&m=sIkb96QFu3bbeTjdHCtAitb3l-8e0P_mD9RinRF-i84&s=TemJoUedljmDM3WSFKARUvOHb9vlJN495FLNr5xGMYg&e=. Aside from general interest, the long game is to model the joint distribution of earnings and wealth, particularly at the tails. The need to match the joint distribution puts useful restrictions on the model. I'm thinking here of http://www.econ.nyu.edu/user/benhabib/earnings%20and%20wealth.pdf.

Is your point that consumption is more important for welfare? In any case, I think it's a good idea to include consumption if we can find good data.

— You are receiving this because you were mentioned. Reply to this email directly, view it on GitHub https://urldefense.proofpoint.com/v2/url?u=https-3A__github.com_QuantEcon_lecture-2Dpython_issues_212-23issuecomment-2D628283697&d=DwMCaQ&c=slrrB7dE8n7gBJbeO0g-IQ&r=hV4qLWg4zodvX-YQ-ETIbA&m=sIkb96QFu3bbeTjdHCtAitb3l-8e0P_mD9RinRF-i84&s=xC_E7CeKaeT_5hatyFUgnDgtE0AdwZZNRsx8wmovI0E&e=, or unsubscribe https://urldefense.proofpoint.com/v2/url?u=https-3A__github.com_notifications_unsubscribe-2Dauth_AA63MTX23UNQOKH346MH4BDRRMO6ZANCNFSM4M6TEZAA&d=DwMCaQ&c=slrrB7dE8n7gBJbeO0g-IQ&r=hV4qLWg4zodvX-YQ-ETIbA&m=sIkb96QFu3bbeTjdHCtAitb3l-8e0P_mD9RinRF-i84&s=JkO9tzOJT3ChXGKJctnAKsy1ijmlkBPuDwViyYZj4Nk&e= .

[Harveyt47]

Hi @thomassargent30 , I think it's a really good idea to add consumption inequality. I think it will make for interesting discussion. We should be careful with the data for consumption as it can be harder to measure accurately.

@jstac regarding finding the data, I have already found consumption data for Italy see page 9 here. I also have Australian household expenditure on groceries, food and meals eaten outside the home and more, see page 92 here. There is also US expenditure data from the Consumer Expenditure Surveys.

[jstac]

That's great @Harveyt47 , well done. I suppose we can treat the consumption data in the same way, looking at the Gini coefficient. I'm excited to see this come together.

[Harveyt47]

That sounds like a good idea. @jstac I was wondering if you knew who put together the SCF_data.txt file here

[jstac]

Maybe Natasha or Juna?

This is a very nice summary of the data in income inequality / Ginis: https://ourworldindata.org/income-inequality

[Harveyt47]

@jstac that link has a cool data source. The data source has microdata collected from 50 countries. See here.

[najuzilu]

@jstac I like the way they tell the story. I'm a +1 on replicating some of their work in python.

[Harveyt47]

Hi @jstac

@najuzilu , @sayaikegawa and I had a meeting about the inequality lecture and decided to look for more aggregated data instead of microdata as it is easier to find for more countries and as the type of data we want to look at (gini coefficient over time, top 10% income share), using countries statistical institutes will provide enough of the data to present those figure.

We will, however, also explore LIS, which has microdata for a number of countries, however we are uncertain about users ability to access it.

We want to focus some of the discussion on how changes in industrial structure (skill-biased technological change and globalisation) coincided with changes in gini coefficients top income earn income shares.

[jstac]

Thanks @Harveyt47 , sounds like a good plan. It will be interesting to learn something about those correlations you mention at the end, if they exist. (Although, at the first stage, it might be best to focus purely on description. Let's see what has happened before we try to figure out why it has happened.)

Let me know if you think I can help with access.

[shlff]

Thanks @jstac , @thomassargent30 , @Harveyt47 and @najuzilu for your excellent comments and work. I'd like to briefly introduce and share a small part of my thesis work (under @jstac 's supervision) on joint income-wealth distribution and dynamic inequality. Please see repo, where I did two things:

1. Use Kuhn, Schularick and Steins (2020)'s methodology to construct a dataset SCF+ from the survey data Survey of Consumer Finances (SCF) mentioned by @Harveyt47 above.

2. Estimate a joint income-wealth distribution and dynamic income-wealth Gini coefficients for the U.S. households from 1950 to 2016 across three wealth-level samples (whole, bottom 95% and bottom 90%) and plot Ginis.

You can find the dataset SCF+ and code for generating income-wealth Ginis here, where

• A README file is included, and
• I list all variables with labels in this dataset.

Hope my work could be of a little help to our work on these lectures. G'Day!

[jstac]

Thanks @shlff , I think it is ideal for the US case.

How hard would it be to convert this to Python, do you think?

[shlff]

Thanks @shlff , I think it is ideal for the US case.

How hard would it be to convert this to Python, do you think?

Thanks @jstac . I think it would not be too hard to convert it into Python. The Stata do file consists of two parts:

• The first part is to clean data and construct aggregate income and wealth variables for each observation;
• The second part with many loops is to merge data, construct joint income-wealth distributions and generate Gini coefficients across time.

I converted the first part into Python already, and plan to convert the second part tomorrow. If possible, then hopefully I could report my progress to you by tomorrow night.

[ericvd-ucb]

Hey @Harveyt47 did this noteboook ever get made? Looking for an empirical hostorical data Gini inequality notebook!

[jstac]

Hi @ericvd-ucb , thanks for prompting us on this. I'll be glad to hear about how your project pans out.

Somewhat connected: As part of this activity we ended up adding some functionality to QuantEcon related to inequality. See https://quanteconpy.readthedocs.io/en/latest/tools/inequality.html.

@shlff and @natashawatkins, do you have anything that might be useful to @ericvd-ucb ?