Unemployment rate vs deficit spending

[Orion1]

I attempted the following plot on Wolfram Alpha:

plot United States unemployment rate,(United States federal deficit/United States GNP)

The resulting graph is as attachment.

According to my understanding of this graph, the derivative of the United States unemployment rate is a direct result of deficit spending by the United States Federal Government.

The greater the amount of deficit spending the exponentially higher the unemployment rate.

Are there any economists here that can improve upon my calculation?
[/Color]
Reference:
http://www.wolframalpha.com/input/?...ed+States+federal+deficit/United+States+GNP)"

 

[Gokul43201]

What is it that you are trying to calculate? If anything, you can extract a correlation between unemployment and deficits. Without a proposal for a mechanism, you have provided no basis for a causation.

Moreover, a naive inspection of your plot shows a lag in the deficit relative to unemployment. And it's hard to not notice that rising unemployment causes governments to go into spending overdrive in an attempt to stave off job losses, stimulate economic activity, or provide more unemployment benefits. So there's clearly an argument for the causation to be primarily in the opposite direction to your claim.

 

[Ivan Seeking]

Moreover, a naive inspection of your plot shows a lag in the deficit relative to unemployment. And it's hard to not notice that rising unemployment causes governments to go into spending overdrive in an attempt to stave off job losses, stimulate economic activity, or provide more unemployment benefits. So there's clearly an argument for the causation to be primarily in the opposite direction to your claim.

What's more, if we assume causation, then the graph suggests that the government succeeds in reducing unemployment through spending, but it takes a few years for the full effect to be seen.

 

[imiyakawa]

We must understand OPs position, from 1990 onwards there seems to be identical correlation, it doesn't seem to be a lagging indicator, which doesn't help with the hypothesis that unemployment is causing deficit spending.

I suggest everyone thinks about the actual events that are happening in line with the changes in the gradients.

Our limits in infering given this data are:
-One cannot conclude that unemployment is causing deficit spending.
-One cannot conclude that deficit spending is causing unemployment.
-One CAN conclude that global events are causing both unemployment and deficit spending.

(1) beginning of 2000, Stock market crashes -- Fiscal Stimulus AS WELL AS inevitable unemployment. Both have a root cause, (A) isn't causing (B), (B) isn't causing (A), (C) is causing (A) and (B).

(2) 1990s, general domestic economic prosperity within USA (call this (C)). This caused (A) reducing unemployment, (B) reducing deficit. Again, (C) caused (A) and (B).

(3) 2010, SubPrime mortgage crisis, 383 mortgage-related financial institutions bust! (ml-implode.com), (A) is not causing (B), (B) is not causing (A), (C - collapsing financial system), is causing but (A - unemployment) and (B - a responsive deficit).

 

[Ivan Seeking]

We must understand OPs position, from 1990 onwards there seems to be identical correlation, it doesn't seem to be a lagging indicator, which doesn't help with the hypothesis that unemployment is causing deficit spending.

Really? I clearly see the spending [the deficit] lagging the unemployment rate before and after 1990. Where on the graph [approximate date] do you see the opposite?

 

[imiyakawa]

Really? I clearly see the spending [the deficit] lagging the unemployment rate before and after 1990. Where on the graph [approximate date] do you see the opposite?

Just the small patch between 1993.5 to 2002.5 :p

There looks like identical correlation.

ALthough you of course have grounds to argue against that patch. 2001 you could argue that the gradient of the deficit increased markedly not only in response to the stock market crash but in response to quickly growing unemployment.

 

[russ_watters]

Moreover, a naive inspection of your plot shows a lag in the deficit relative to unemployment. And it's hard to not notice that rising unemployment causes governments to go into spending overdrive in an attempt to stave off job losses, stimulate economic activity, or provide more unemployment benefits. So there's clearly an argument for the causation to be primarily in the opposite direction to your claim.

Also, high unemployment lags behind a drop in GDP and a drop in GDP has a direct impact on tax revenue.

 

[russ_watters]

What's more, if we assume causation, then the graph suggests that the government succeeds in reducing unemployment through spending, but it takes a few years for the full effect to be seen.

Right: you can't assume causation, you have to demonstrate it with logic. The was the problem with the OP!

 

[Ivan Seeking]

Right: you can't assume causation, you have to demonstrate it with logic. The was the problem with the OP!

Indeed.

 

[Matterwave]

Perhaps it's best if you also included a regression on the two variables. Instead of graphing them over time, plot the data points for each year on a graph with deficit spending on the x-axis and unemployment on the y-axis. This should give you some better indication if there is correlation (on an instantaneous basis). You can then expand your analysis by plotting say, the unemployment rate at time t and the deficit spending at time t+3 or something to see if there is supposedly lagged correlation.

This is all correlation...it's very hard to come up with causation arguments without first correcting for every omitted variable...omitted variable bias is incredibly hard to get rid of in something as complicated as this haha.

 

[hamster143]

Are there any economists here that can improve upon my calculation?

Not an economist but I can surely improve on that calculation:

plot United States unemployment rate, (United States federal receipts/GNP), (United States federal outlays/GNP)

2001 and 2008 recessions are clearly accompanied by sharp drops in government revenues. Government spending did not increase substantially during the 2001 recession, and it only started to rise with considerable delay during the 2008 recession.

 

[imiyakawa]

Perhaps it's best if you also included a regression on the two variables. Instead of graphing them over time, plot the data points for each year on a graph with deficit spending on the x-axis and unemployment on the y-axis. This should give you some better indication if there is correlation (on an instantaneous basis). You can then expand your analysis by plotting say, the unemployment rate at time t and the deficit spending at time t+3 or something to see if there is supposedly lagged correlation.

This is all correlation...it's very hard to come up with causation arguments without first correcting for every omitted variable...omitted variable bias is incredibly hard to get rid of in something as complicated as this haha.

THIS. .. THIS. THIS. THIS!

Yes!

Although such a regression will be basically nonsense. As matterwave said, omitted variable bias. This means there's an X2 that's causing both X1 and Y1, and this X2 isn't in your model. This makes it look like you have very neat correlation, but you actually don't! (E.G., when comparing student test scores to class size, an example of X2, X3, X4 would be income, geographic area, and whether there are computers in the classroom). An example of that X2 in what we're talking about are real world events, such as recessions, etc. Recessions are causing correlated directional changes in both GDP and unemployment, but it's not in you rmodel, and so it makes unemployment look like it's having a huge causal effect when it's not!

I don't know how you're going to account for this, frankly.
----------------------------------------------------
In addition,
There is some reason to expect this kind of negative correlation after all!
Inflation and employment are negatively correlated. It's not unwise to suspect that the government would risk a deficit during times of non-inflation. This will make it seem like a deficit is causing unemployment. Another omitted variable!

 

[Orion1]

unemployment rate vs government spending...

plot United States unemployment rate, (United States federal receipts/GNP), (United States federal outlays/GNP)

Given that causation has not been established and there is an omitted variable bias, it still appears that the government spending rate drives the unemployment rate.

[/Color]
Reference:
http://www.wolframalpha.com/input/?...s/GNP),+(United+States+federal+outlays/GNP)+"

 

[WhoWee]

You might want to consider using GDP.

 

[hamster143]

Given that causation has not been established and there is an omitted variable bias, it still appears that the government spending rate drives the unemployment rate.

Here's your causation.

Federal spending comes in two major categories, discretionary and mandatory.

Discretionary spending (e.g. deciding to spend extra $5 billion on the space shuttle program, or to save $5 billion by shutting down Tevatron) does not vary much (except when the government decides to start a new war). And it is a small part of the total bill. In the last pre-recession budget, non-defense discretionary spending accounted for less than 20% of all spending.

Big fluctuations you see in the chart come from mandatory spending, they reflect recession-time increases in safety-net spending on Medicaid, unemployment insurance, welfare, and food stamps. These all increase automatically as a consequence of rising unemployment and poverty.

Likewise, fluctuations in federal receipts are causally linked to unemployment (fewer people working mean fewer taxes collected).

 

[mheslep]

The OP is attempting to show, I believe, how spending or the deficit drives unemployment, not the other way around.

 

[hamster143]

The OP is attempting to show, I believe, how spending or the deficit drives unemployment, not the other way around.

And we're trying to explain to him that he's wrong.

 

[mheslep]

And we're trying to explain to him that he's wrong.

? Not in this thread. The only arrows thrown are ones casting doubt on the causality, not that he's wrong.

 

[hamster143]

? Not in this thread. The only arrows thrown are ones casting doubt on the causality, not that he's wrong.

He is wrong because he posits causality in one direction, when in actuality the causality goes in the opposite direction.

 

[mheslep]

He is wrong because he posits causality in one direction, when in actuality the causality goes in the opposite direction.

That does not mean his thesis is wrong. Causality might go both ways.

 

[russ_watters]

Given that causation has not been established and there is an omitted variable bias, it still appears that the government spending rate drives the unemployment rate.

No, it really doesn't. The recent data patterns appear pretty much exactly superimposed to me!

IIRC, people are eligible for unemployment compensation a week after they become unemployed. So when someone becomes unemployed, there is an immediate drop in the taxes they pay and a week later a rise in the benefits they receive. I don't see how there could be any question that the major year-to-year fluctuations in deficit are caused by unemployment fluctuations and not the other way around.

 

[russ_watters]

He is wrong because he posits causality in one direction, when in actuality the causality goes in the opposite direction.

Yes, the long-term fluctuation in deficit could result in long-term fluctuation in unemployment... but the OP didn't attempt that argument!

 

[John Creighto]

I attempted the following plot on Wolfram Alpha:

plot United States unemployment rate,(United States federal deficit/United States GNP)

The resulting graph is as attachment.

According to my understanding of this graph, the derivative of the United States unemployment rate is a direct result of deficit spending by the United States Federal Government.

The greater the amount of deficit spending the exponentially higher the unemployment rate.

Are there any economists here that can improve upon my calculation?
[/Color]
Reference:
http://www.wolframalpha.com/input/?...ed+States+federal+deficit/United+States+GNP)"

Aside from what the causation actually is it is a striking relationship in the latter part of the graph. I do not know how these documents work so I don’t know what to infer about the methodology. Is the computation more complex then shown or is it exactly as shown? Is there any way to see the source code?

What I wonder is if the tighter relationship in the latter part of the graph is not a result of how the graph is drawn, then what quantities are held constant in the latter period of the graph that weren’t held constant in the earlier period of the graph? For instance how did inflation and the interest rate change over these time periods, how did government spending change, and how did tax revenue change? Additionally, how did the demographics of government tax revenue change thought these periods. Reducing taxes on the supper rich would have likely made the deficit more sensitive to unemployment as the national unemployment rate is not likely as correlated with the unemployment rate of the wealthiest Americans.

 

[Orion1]

You might want to consider using GDP.

plot United States unemployment rate, (United States federal receipts/GDP), (United States federal outlays/GDP)

Done. At the end of the graph, the recession begins in late 2007.
[/Color]
Reference:
http://www.wolframalpha.com/input/?...ts/GDP),+(United+States+federal+outlays/GDP)"

 

[John Creighto]

plot United States unemployment rate, (United States federal receipts/GDP), (United States federal outlays/GDP)

Done. At the end of the graph, the recession begins in late 2007.
[/Color]
Reference:
http://www.wolframalpha.com/input/?...ts/GDP),+(United+States+federal+outlays/GDP)"

That's interesting that the correlation is stronger when deviding by GNP then GDP. I suppose this makes sence.