Description
ascol converts daily data of asset prices or returns to weekly, monthly, quarterly, or yearly frequencies. When converting asset prices to a lower frequency, ascol selects the last price in the given period. For converting asset returns, ascol offers two possibilities – either to sum the daily returns or find products of the daily returns. The first choice is used with daily log returns while the second is used with daily simple returns (Detailed discussion is given below). ascol requires that the existing data has a time variable that tracks daily dates. If the data is already tsset or xtset, ascol will automatically pick the time and panel variables from the previous tsset or xtset declarations. In case the data is not already set for time or panel dimensions, then the time variable has to be set by using the option timevar(varname).
How to Install it
ascol can be installed from SSC by typing the following line of code in the Stata command window
ssc install ascol
Options
ascol has the following options.
1. returns
This option can be entered as returns(simple) or returns(log). If we wish to convert daily returns to a lower frequency we shall use this option. This option can be used with two variations: simple returns and log returns. For detailed discussion, examples, and comparisons of simple and log returns, please visit this page . See the following details that explain when to use which of the two sub-options:
1.1 returns(simple)
If daily returns have already been calculated with the following formula;
simple ri = ( Price[i] - Price[i-1] ) / Price[i-1] ... (Eq. 1)
Then the appropriate method to convert the returns to n-period cumulative returns would be;
Cumulative n-period simple returns = (1+simple_r1) * (1+simple_r2) *(1+simple_r3) * (1+simple_rn) - 1 ... (Eq. 2)
By invoking option returns(simple), ascol applies Eq. 2 to find n-period cumulative returns.
1.2 returns(log)
If daily returns have already been calculated with the following formula;
log_ri = ln(Price[i] / Price[i-1]) ... (Eq. 3)
Then the appropriate method to convert the returns to n-periods cumulative returns would be to just sum the daily returns. By invoking option returns(log), ascol sums the daily returns to find n-periods cumulative returns. Therefore, users must exercise care in selecting the appropriate option in converting daily returns to n-period cumulative returns
2. prices
If the data in memory are asset prices, we shall use the option prices. Please note that option return and prices cannot be combined together. To collapse prices to the desired frequency, the program finds the last traded prices of the period.
3. Frequency Options
ascol has the following options for data conversion:
toweek converts from daily to weekly frequency
tomonth converts from daily to monthly frequency
toquarter converts from daily to quarterly frequency
toyear converts from daily to yearly frequency
4. timevar(varname) and panelvar(varname)
ascol needs a variable that tracks daily dates. If the data is already tsset, ascol will automatically pick the time variable. Therefore, there will be no need to use the option timevar(). Similarly, if the data is already xtset, ascol will pick both the time and panel variables from the previous xtset declarations. Again, there will be no need to use the options timevar() or panelvar(). However, if the data has duplicates or has other reasons that do not allow the tsset or xtset declarations, then we shall have to inform ascol about the time and/or panel variables of the data set through options
timevar(varname) and panelvar(varname).
5. keep(all) or keep(vars)
When we convert data from daily to a lower-frequency such as weekly, monthly, etc., we end up with repeated values of the converted variable. We often just need one value of the variable per cross-sectional unit and time-period. Therefore, the repeated observations are not needed and should be dropped. This is what the Stata’s collapse command does. The default in ascol is to collapse the data to a lower frequency and delete all other variables except the newly created one. However, there might be circumstances when we want to retain all the observations without collapsing the data set. Towards this end, we can use the option keep(all)
or keep(vars)
. keep(all) will keep the data set as it was before running the command, while keep(vars)
will collapse the data to a lower frequency and keep all the variables of the data set. Here is the summary:
keep(all) conversion happens without collapsing the data and without deleting other variables
keep(vars) conversion happens without deleting other variables; data collapses to a lower frequency
6. generate(newvar)
This is an optional option to specify the name of the new variable. If left blank, ascol will automatically name the new variable as varname_frequency.
Example Data Set
Let us generate a dummy data set for our example. Copy the following and run from Stata do editor
clear
set obs 1000
gen date=date("1/1/2012" , "DMY")+_n
format %td date
tsset date
gen pr=10
replace pr=pr[_n-1]+uniform() if _n>1
gen simpleRi=(pr/l.pr)-1
gen logRi = ln(pr/l.pr)
save stocks, replace
Example 1: From Daily to weekly – simple returns
Suppose we have already generated daily simple returns using Equation 1, we shall convert them to weekly returns with:
use stocks, clear
ascol simpleRi, toweek returns(simple)
ascol is the program name, simpleRi is the stock return variable in our data set, toweek is the program option that tells Stata to convert daily data to weekly frequency, and the returns(simple) option tells Stata that our simpleRi variable has simple stock returns and therefore ascol will apply Equation 2 above to find cumulative weekly returns. Please note that we did not use the option timevar(varname) and panelvar(varname) as our data is already tsset.
Example 2: From Daily to weekly – log returns
Suppose we have already generated log returns using Equation 2, we shall convert them to weekly returns with:
use stocks, clear
ascol logRi, toweek returns(log)
ascol is the program name, logRi is the stock return variable in our data set, toweek is the program option that tells Stata to convert daily data to weekly frequency, and the returns(log) option tells Stata that our logRi variable has log stock returns. Therefore ascol will just sum the returns within each week to find cumulative weekly returns. Please note that we did not use the option timevar(varname) and panelvar(varname) as our data is already tsset.
Example 3: From Daily to monthly – prices
use stocks, clear
ascol pr, tomonth price
pr is the variable name that has stock prices data, tomonth
option specifies conversion from daily to a monthly frequency, and the price
specifies that the conversion is needed for stock prices data.
Converting Data to Other Frequencies
From daily to quarterly, option toquarter
or toq
is to be used.
ascol pr, toq price
From daily to yearly, option toyear
or toy
is to be used.
ascol pr, toy price
Example 4: Conversion without collapse – keep all observations and variables
We shall use the option keep(all)
to retain all variables and observations in the data set. After conversion, you can see that there are duplicate values in the newly created variable week_simpleRi.
use stocks, clear
ascol simpleRi , toweek returns(simple) keep(all)
Example 5: Collapsing by time variables only – keep variables
We shall use the option `keep(vars)` to retain all variables while collapsing the data to a lower frequency. After conversion, you can see that there are no duplicate values of the newly created variable. week_simpleRi
.
use stocks, clear
ascol simpleRi , toweek returns(simple) keep(vars)
Questions on ascol:
My ascol command returns the error “Invalid subscript” | Answer is here on the Statalist |
Great work sir……….
Actually, I used it several times and I double checked the monthly prices, but I found wrong prices.
Saleh
ascol keeps the last price in a given period. In the case of monthly prices, ascol would keep the last price of that month. When you say that you get wrong prices, what exactly is not correct. What command did you use and in what way the output had an error? Please reply with relevant details.
Dear Dr Shah,
I am trying to compute firm-specific weekly returns for my panel data. I came across your website after reading your post on Statalist on changing returns frequency. Do you think the following code makes sense for calculating weekly returns on my panel data?
bys firm_id (date): gen dreturns = ln(pr[_n]/pr[_n-1])
ascol dreturns, toweek returns(log)
thank you!
i am using ascol but the keep option do not work
ascol closing, keep(all) toweek prices
Please post a data example and the error you are getting from using the ascol code.
I keep getting the following error:
asrol may not be combined with by
I am running the following code:
clear
set obs 1000
gen date=date(“1/1/2012” , “DMY”)+_n
format %td date
tsset date
gen pr=10
replace pr=pr[_n-1]+uniform() if _n>1
gen simpleRi=(pr/l.pr)-1
gen logRi = ln(pr/l.pr)
save stocks,replace
ascol simpleRi, toweek returns(simple)
Please re install asrol with the following code
ssc install asrol, replace
If the above does not solve the problem, then run the following code and send the asrol.smcl file that is created by the code. After your run the following code, the asrol.smcl file will be available in the current folder