Re: st: RE: How to identify patterns of cyclicality using STATA?

[Natasha Xingyuan Che <ilovewordsworth@gmail.com>]

Thanks, Nick!  This is quite helpful.  I'm using a combination of your
suggestion and Kit's suggestion right now.  He is going to write up
his approach as a stata tip.  Thanks again!

-Natasha


On 11/9/11, Nick Cox <n.j.cox@durham.ac.uk> wrote:
> The other part of Natasha's question was to identify periods of recovery.
>
> Y at the previous peak is
>
> gen ypp = y if peak
> replace ypp = ypp[_n-1] if missing(ypp)
>
> So the code should be something like
>
> gen recovery = 0
> replace recovery = cond(recovery[_n-1] == 0 & peak, 1, cond(recovery[_n-1]
> == 1 & (y >= ypp), 0, recovery[_n-1])) if _n > 1
>
> which can all be done panelwise. However, this is not tested.
>
> Nick
> n.j.cox@durham.ac.uk
>
>
> -----Original Message-----
> From: owner-statalist@hsphsun2.harvard.edu
> [mailto:owner-statalist@hsphsun2.harvard.edu] On Behalf Of Nick Cox
> Sent: 09 November 2011 10:17
> To: 'statalist@hsphsun2.harvard.edu'
> Subject: RE: st: RE: How to identify patterns of cyclicality using STATA?
>
> Kit closed the thread, but I am going to open it again as Kit has not yet
> made his solution public and I have a corrected solution that may be of
> interest of others.
>
> We have an indicator -peak- for peaks and an indicator -trough- for troughs.
>
> Natasha pointed out that although a trough closes a phase, -sum(peak-trough)
> == 1- fails as a criterion for being in a phase because there can be one or
> more peaks between the peak starting the phase and the trough finishing it.
> She could have added, presumably, that there can be troughs before any peak
> that do not terminate a phase.
>
> Fair enough. Please consider
>
> gen phase = 0
> replace phase = cond(phase[_n-1] == 0 & peak, 1, cond(phase[_n-1] == 1 &
> trough, 0, phase[_n-1])) if _n > 1
>
> So the rules are
>
> 1. Initialise to 0.
>
> 2. Change your mind if we are not in phase and this observation is a peak
> (we've started)
> _or_ if we are in phase and this observation is a trough (we've stopped).
>
> 3. Phase is what it was previously otherwise.
>
> Note that the first observation needs care and that we can do this panelwise
> by prefixing
>
> by country:
>
> Then I think it is over to spell machinery.
>
> Nick
> n.j.cox@durham.ac.uk
>
> Kit Baum
>
> As it happens I was able to consult with Natasha face-to-face this
> afternoon, and we came up with a workable strategy to deal with this and
> other complications that arose in discussing her needs. I will write up a
> stylized version of that solution as a Stata Tip. Please consider the thread
> closed.
>
> Natasha Xingyuan Che
>
> That was very helpful.  Thanks.  But here is one additional
> complication in this specific example.  The cycles are not always in
> symmetry, i.e., there are more peaks than troughs.  For example, the
> two variables identifying peak and trough can look like this:
>
> peak   trough
> 0            0
> 1            0
> 0            0
> 0            0
> 1            0
> 0            0
> 0            1
> 0            0
> 0            0
>
> Now if I apply what you suggested: (gen period = sum(peak - trough)).
> The variable "period" would be equal to 1 for the spell between 1st
> and 2nd peak and the spell after the trough.  But the recession is
> actually the spell between the 2nd peak and the trough, for which the
> variable "period" would be equal to 2.
>
> I think the bottom line is I need to identify only those peaks that
> are followed by a trough in some future date.  Do you have any
> suggestions on how to do that?  Many thanks!
>
> Nick Cox
>
>> Check out
>>
>> SJ-7-2  dm0029  . . . . . . . . . . . . . . Speaking Stata: Identifying
>> spells
>>        . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .  N. J.
>> Cox
>>        Q2/07   SJ 7(2):249--265                                 (no
>> commands)
>>        shows how to handle spells with complete control over
>>        spell specification
>>
>> and -tsspell- (SSC)
>>
>> for some Stata technique in this area.
>>
>> Your first question appears to be answered, at least roughly, by
>>
>> bysort country (time) : gen period = sum(peak - trough)
>> tsspell , cond(period == 1)
>
> Natasha Xingyuan Che
>
>> I'm trying to identify business cycle peaks, troughs, and phases of
>> recession and recovery in a panel data set of countries's quarterly
>> output series.  I identified peaks and troughs, but am having trouble
>> marking the periods in between (i.e., recessions and recoveries).  I
>> would really appreciate it if any of you have suggestions about how to
>> do this.
>>
>> The specific definitions of the variables are the following --
>>
>> A peak occurs at Quarter t of a time series of Y_t if:
>> (Y_t - Y_t-1) > 0 and (Y_t - Y_t-2) > 0 and (Y_t - Y_t+1) > 0 and (Y_t
>> - Y_t+2) > 0
>>
>> A trough occurs at Quarter t if:
>> (Y_t - Y_t-1) < 0 and (Y_t - Y_t-2) < 0 and (Y_t - Y_t+1) < 0 and (Y_t
>> - Y_t+2) < 0
>>
>> A "recession" is a phase from peak to the next trough.  A "recovery"
>> is a phase from trough to the quarter when Y reaches the level of the
>> previous peak.
>>
>> I generated two 0/1 variables, peak and trough, as the following --
>>
>> tsset country time
>> gen peak=(lny>l.lny&lny>l2.lny&lny>f.lny&lny>f2.lny)
>> gen trough=(lny<l.lny&lny<l2.lny&lny<f.lny&lny<f2.lny)
>>
>> Now I want to generate two variables marking the quarters when the
>> economy is in recession and in recovery and calculate the respective
>> duration of the two.  But I'm not sure how to proceed.  Again, all
>> your inputs are very much appreciated.  Thank you so much for your
>> help!
>
> *
> *   For searches and help try:
> *   http://www.stata.com/help.cgi?search
> *   http://www.stata.com/support/statalist/faq
> *   http://www.ats.ucla.edu/stat/stata/
>
> *
> *   For searches and help try:
> *   http://www.stata.com/help.cgi?search
> *   http://www.stata.com/support/statalist/faq
> *   http://www.ats.ucla.edu/stat/stata/
>
*
*   For searches and help try:
*   http://www.stata.com/help.cgi?search
*   http://www.stata.com/support/statalist/faq
*   http://www.ats.ucla.edu/stat/stata/