Hello
Assalamu Alaikum
Though my question is not linked to the DSGE estimation and code, yet it is linked to the DSGE.
Denton Method is used for the data quarterization. Is it a credible enough to be used for the frequency conversion? What about the indicators, like can I use the Gross Fixed Capital Formation as an indicator variable for financial series (such as bank deposits, advances)?
What exactly are you trying to do? I don’t think there is a general answer. If your data is available at a lower frequency than the one desired, there are essentially two ways out. You use indicators to disaggregate using e.g. the Denton method. Whether that is a good option generally depends on the available indicators and how good they approximate the true temporal dynamics. The alternative is to work with mixed frequency data, i.e. employ the model dynamics to infer the dynamics.
Thank you sir.
I can try the first option of finding an appropriate indicator.
But the second option; to “employ the model dynamics to infer the dynamics” is a puzzle. Is there any simple way to adopt this approach?
Is it preferable over the “indicators based disaggregation”?
However the answer is quite a clear way out for beginners like me".
There is a rather long literature on mixed-frequency estimation. See Data with different frequency - #2 by jpfeifer
I tried denton method with indicator. For data like the GDP, it divided with 4 and applied some extrapolation function (with something like addition of quadratic difference). But for the series like the interest rate, to my surprise, it did the same. Here series cannot be divided by 4. Instead it should try to keep on the past’s (and/or subsequent observations’) average. And from here I tried to find some alternate. Like using the Eviews with “qudratic match average” performs a better alternate.
I am trying to find some procedure of the denton method. If someone could guide me for extrapolation of the interest rate or similar series.
Why do you even need that for interest rates? They should be available at much higher frequencies.
Historical data is available at low frequency. Number of series like the Rates of return on advances are available at low frequency. Then, Non-financial data is also available at low frequency. These data series cannot be divided by 4.