0
$\begingroup$

I obtained the historical CPI data available in monthly and quarterly format from a data repository. It seems that the Q1, Q2, etc. data are the average of the monthly data.

I thought that if I were to get quarterly data from monthly data, I can just take the March, June, Sept, and December closing figures. If I were to get the quarterly data for stock price or stock index, should I can just take the 4 month closing price or would I have to take the average of 3 months in each quarter?

$\endgroup$
  • 1
    $\begingroup$ What do you want to do using stock prices? $\endgroup$ – london Dec 23 '17 at 16:21
  • $\begingroup$ I want to investigate the relationship between stock index and macroeconomic variables using econometric method $\endgroup$ – Josephine PM Dec 23 '17 at 16:53
  • $\begingroup$ Because you need to deflate the stock prices, take the averages of the monthly data to match with the CPI magnitudes $\endgroup$ – london Dec 23 '17 at 19:38
0
$\begingroup$

The bottom line is that we never want to discard information we can use, unless it we have good reason to believe it's "noise".

You can imagine an extreme case, with four quarterly closing values: 10, 10, 10, 10. Is that an accurate snapshot of the market in those times, if that's all the data you have? Yes, surely. But what if the monthly data looked like: 16, 14, 10, 4, 6, 10, 16, 14, 10, 4, 6, 10.

While the quarterly closings give us some information, there's clearly important information in the monthly data that we lose. Ideally we would do our analysis at the monthly level here, but maybe that's not possible due to, say, the frequency of the other variables in your analysis. If we take the quarterly mean, we get: 13.3, 6.7, 13.3, 6.7. That conveys a lot more of the information we have about this data, even if we're still discarding some of it by resampling to quarterly.

| improve this answer | |
$\endgroup$
  • $\begingroup$ Thanks! meaning that if monthly data are available, but due to some reason i still need the data in quarterly form, then it is better to take the quarterly mean than taking the quarterly closing value. Am I right? $\endgroup$ – Josephine PM Dec 26 '17 at 4:31
  • $\begingroup$ @JosephinePM There are actually more ways to resample than just mean, but generally speaking, yes. $\endgroup$ – Jeff Dec 26 '17 at 7:01
0
$\begingroup$

Given that this is a macroeconomic study, the cost to write a script to average your data is simple and effectively costless. If you obtain your variables based off of monthly formats, then it would be wise to also obtain your other variables also in the same format.

Each time you simplify, you may potentially lose small but significant outliers in your data that are averaged out as part of your calculations. But if you have more granular data at hand, you can revisit your figures and revise your analysis as needed.

In other words, if you can do your analysis month-by-month I would recommend that you do so, it is relatively simple to scale up and analyze it by quarter as needed; it is however difficult to 'scale' down, don't mix and match two variables measured by different time segments.

| improve this answer | |
$\endgroup$

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.