Sahm (2019) makes the case for lump-sum payments to people in a recession to boost consumer expenditure. These payments should be large enough to be salient and maintain spending over (precautionary) saving e.g. on durables. Crucially, these payments must also be timely to provide a quick and early response to a recession.
The challenge for a stimulus response is knowing when to enact it. The National Bureau of Economic Research (NBER) is responsible for the official identification of recessions in the US. However, this takes time and we’d ideally have something that tells us whether we’re in a recession in as close to real time as possible.
This is where the Sahm Rule comes in as a trigger for the payments. By the rule, payments would begin when (Sahm, 2019; Page 76):
the three-month average national unemployment rate rises by at least 0.50 percentage points relative to its low in the previous 12 months
Sahm (2019) shows that this is effective as a US recession indicator, signalling a recession within months of its start. This is far quicker than the NBER and, applied to historical data, has not resulted in a false positive since 1970.
The whole paper is worth a look. It’s very readable and, as a policy paper, covers a range of other considerations including the role of policymakers’ commitment, practicalities of implementation, and policy evaluation.
The rest of this post reproduces the Sahm Rule recession indicator in Python.
Download the data
If you’re running this as a notebook, you’ll need to store your own API key to
I’m not going to worry about the accompanying metadata etc and will focus
instead on downloading the series values. The FRED API call of interest is
function below makes that API call using the
Requests library, checks that the
request was successful (status code 200) and returns a
object of the data.
We can then download the data. Strictly, to simulate the situation a policymaker would be in, I should be using the real-time unemployment figures available at the time. For that I’d need to make calls to the ALFRED API, rather than the FRED API. I’m not going to bother with that here. Maybe next time…
Update 21/05/2020: Re-run to extend the series to April 2020, when the impacts of COVID-19 become very apparent.
2019-09 3.5 2019-10 3.6 2019-11 3.5 2019-12 3.5 2020-01 3.6 2020-02 3.5 2020-03 4.4 2020-04 14.7 Freq: M, Name: UNRATE, dtype: float64
Calculate the Sahm Rule recession indicator
Having downloaded the series, calculate the Sahm Rule recession indicator:
2019-09 0.0 2019-10 -0.0 2019-11 -0.1 2019-12 0.0 2020-01 0.0 2020-02 0.0 2020-03 0.3 2020-04 4.0 Freq: M, Name: UNRATE, dtype: float64
We can then plot the series:
See here for this post as a Jupyter notebook.
Sahm, C. ‘Sahm Rule recession indicator’ [SAHMCURRENT], retrieved from FRED,
Federal Reserve Bank of St. Louis
US Bureau of Labor Statistics ‘Unemployment rate’ [UNRATE], retrieved from
FRED, Federal Reserve of St. Louis
Sahm, C. (2019) ‘Direct stimulus payments to individuals’ in Boushey, H., Nunn,
R., Shambaugh, J. (eds) Recession ready: Fiscal policies to stabilize the
American economy, The Hamilton Project and the Washington Center for