Macronote #3A monetary VAR model for Sweden (or sub for a country of your choice), with forecasts for inflation, GDP, and policy rates
A small macro model
I have just finished an internal training session with Macrobond, but the instructor was, in my honest opinion, horrible. He tried to walk us through a Vector Auto Regression (VAR) analysis from scratch, but nothing worked and he seemed confused by every little detail. Small wonder if anyone listening managed to learn anything at all.
- Unfortunately, the instructor was me. To try to make amends, I thought I could send out a more polished version on what I built during the training session. And by publishing it externally, I hope clients will also get a whiff of what you can accomplish, quite effortlessly, with the Macrobond application.
And that was also my original idea for the internal training session; to show how easily you can construct, alter and publish more advanced statistical analysis in Macrobond than in any other application. We still lack testing for Granger-causality, but I’m sure if you keep those support tickets coming, it will soon find its way into the application.
So, what did I do? A recurring question for policy makers and practitioners alike is how monetary policy actually affects the economy. To do this, we have relied on a host of models, but one of the most wide-spread is the (simple) monetary VAR. This is also what I aimed to construct, for Sweden.
Normally, it consists of three variables: Inflation, the output gap (or unemployment rate) and the policy rate. Think Taylor rules (than you even get something of a “structural VAR”)! But, as usual, it is our imagination that sets the limits. To read more on this, I still think that Stock & Watson’s 1 2001 JEP-paper does an excellent job in explaining the use (and some shortcomings) of VAR for monetary analysis.
Now, since Sweden is the proverbial small open economy, I also used OECD’s forecast for Swedish world market growth as an exogenous variable in the VAR, but this is not necessary if you to perform the calculations for larger economies, such as the US or Euro Area. Anyhow, this is what Swedish world market growth looks like:
1 They should form a band. Especially after writing so much good stuff that I at times mistake them for a 70’s folk-rock duo…
Also, as endogenous variables I have used core inflation (CPIF ex Energy), GDP (vol) and the (real) repo rate. All sorts of alternatives could be used (again, think Taylor-rules). My calculations won’t be completely kosher, e.g., I chose to simply transform all the variables into % Y-o-Y growth rates, even though log diffs would of course be better (especially when we look at the Impulse-Response Function (IRF) later on). In Sweden, as the monetary policy regime (IT) began in 1993, I set that as the start date for the estimations.
The results of this little exercise are visible below (the statistical report is available in the Macrobond-documents).
First the inflation equation:
Then the GDP-equation:
And to tie the VAR-analysis down, the policy rate:
Under the (rather strong) assumption that the OECDs forecasts pan out, our model says the repo rate should rise around 100bp, but more than half of that this is explained by lower inflation. In the meantime, Swedish GDP-growth will hum along in a range of 3%-3½% y/y for the nearest future.
To spice things up, and achieve the “dynamic multipliers” normally sought after in a Monetary VAR, we can also perform Impulse-Response Functions and here the ordering of the variables is imperative. The ordering should be done by decreasing exogeneity (this is one of the reason’s to we would like to see the Granger causality test in MB!). The standard ordering of a monetary VAR puts inflation firmly at the top, with resource/output measure in second place and the monetary variable last, reacting to the other variables. I have used that very same ordering here.
Now, to my fellow Macrobonders, in defense of my less than compelling performance on the training session, I forgot to look closer at the lag structure of the exogenous variable, which made all the difference. Even the (Cholesky) IRFs look quite nice (the impulses are standard deviations and found in the report as the SDs for the equations):
Get these posts delivered to your inbox
Sign up now to get an email every Friday with the latest research posts from Macro n Cheese
DisclaimerWe don’t usually have views and opinions about economic and financial states of affairs, (not ones that we express publicly as a company, anyway). We do believe, however, that people can and do appreciate a variety of perspectives. What you’ve just read is the perspective of our resident chief economist. While we think he’s very smart, Macrobond Financial does not expressly endorse the views he presents here. And, as the old adage goes, you shouldn’t believe everything you read (not without finding the data, performing a few analyses and presenting it in a nice chart). We want to make it clear that we are not offering this information as investment advice. That being said, if you have the application you can easily check everything that’s mentioned here, and decide for yourself. If you don’t have the application, now you have a great reason to get it.