Ana Mª Abad Enrique M. Quilis

Transcription

1 SOFTWARE TO PERFORM TEMPORAL DISAGGREGATION OF ECONOMIC TIME SERIES Ana Mª Abad Enrique M. Quilis Instituto Nacional de Estadística

2 CONTENTS Basic Matlab library Additional interface (in Visual Basic) to use the library in an Excel environment

3

4

5

6

7

8

9 MATLAB LIBRARY Univariate procedures WITHOUT INDICATOR WITH INDICATOR Boot-Feibes-Lisman Stram-Wei Denton (additive & proportional) Fernández Chow-Lin Litterman Santos-Cardoso Guerrero

10 MATLAB LIBRARY Multivariate procedures TRANSVERSAL van der Ploeg TRANSVERSAL AND TEMPORAL Rossi Denton Di Fonzo

11 APPLICATIONS Boot-Feibes-Lisman Chow-Lin Santos-Cardoso Di Fonzo

12 BOOT-FEIBES-LISMAN close all; clear all; clc; % DATA: Non-farm, non-government employees. % UNIT: Persons (in thousands) % SOURCE: Short-term Labor Survey (Encuesta de Coyuntura Laboral) % SAMPLE: Quarterly, Y = load('ecl.prn'); % % Inputs for td library % Type of aggregation: average ta=2; % Minimizing the volatility of d-differenced series d=1; % Frequency conversion: quarterly to monthly s=3; % Name of ASCII file for output file_sal='ecl.sal'; % Calling the function: output is loaded in a structure called res res=bfl(y,ta,d,s); % Calling printing function tduni_print(res,file_sal); edit ecl.sal; % Calling graph function tduni_plot(res);

13 BOOT-FEIBES-LISMAN

14 BOOT-FEIBES-LISMAN **************************************************** TEMPORAL DISAGGREGATION METHOD: Boot-Feibes-Lisman **************************************************** Number of low-frequency observations : 36 Frequency conversion : 3 Number of high-frequency observations : 108 Degree of differencing : 1 Type of disaggregation: average (index). High frequency series (columnwise): Elapsed time:

15 close all; clear all; clc; CHOW-LIN % ANNUAL DATA: Gross Added Value. Construction. % UNIT: 1995 Constant prices. Euros (thousand of millions). % SOURCE: National Accounts. % SAMPLE: Annual, Y = load('cst.anu'); % QUARTERLY DATA: Synthetic indicator. Construction. % Basic source: Construction Short-term Survey % UNIT: Euros (thousand of millions). % SOURCE: Quarterly National Accounts. % SAMPLE: Annual, (last data is a forecast) x = load('cst.ind'); % % Inputs for td library % Type of aggregation: sum ta=1; % Frequency conversion: annual to quarterly s=4; % Method of estimation: 1 (MLE), 0 (WLS) type=1; % Name of ASCII file for output file_sal='cst.sal'; % Calling the function: output is loaded in a structure called res res=chowlin(y,x,ta,s,type); % Calling printing function output=1; % Include series td_print(res,file_sal,output); edit cst.sal; % Calling graph function td_plot(res);

16 CHOW-LIN: output res: a structure res.meth ='Chow-Lin'; res.ta = type of disaggregation res.type = method of estimation res.n = nobs. of low frequency data res.n = nobs. of high-frequency data res.pred = number of extrapolations res.s = frequency conversion between low and high freq. res.p = number of regressors (including intercept) res.y = low frequency data res.x = high frequency indicators res.y = high frequency estimate res.y_dt = high frequency estimate: standard deviation res.y_lo = high frequency estimate: sd - sigma res.y_up = high frequency estimate: sd + sigma res.u = high frequency residuals res.u = low frequency residuals res.beta = estimated model parameters res.beta_sd = estimated model parameters: standard deviation res.beta_t = estimated model parameters: t ratios res.rho = innovational parameter res.aic = Information criterion: AIC res.bic = Information criterion: BIC res.val = Objective function used by the estimation method res.r = grid of innovational parameters used by the estimation method

17 CHOW-LIN

18 CHOW-LIN

19 CHOW-LIN

20 CHOW-LIN

21 CHOW-LIN

22 CHOW-LIN

23 CHOW-LIN

24 CHOW-LIN

25 CHOW-LIN

26 CHOW-LIN

27 CHOW-LIN **************************************************** TEMPORAL DISAGGREGATION METHOD: Chow-Lin **************************************************** Number of low-frequency observations : 25 Frequency conversion : 4 Number of high-frequency observations: 101 Number of extrapolations : 1 Number of indicators (+ constant) : 2 Type of disaggregation: sum (flow). Estimation method: Maximum likelihood. ** High frequency model ** Beta parameters (columnwise): * Estimate * Std. deviation * t-ratios Innovational parameter:

28 CHOW-LIN AIC: BIC: Low-frequency correlation (Y,X) - levels : yoy rates : High-frequency correlation (y,x) - levels : yoy rates : High-frequency volatility of yoy rates - estimate : indicator : ratio : High-frequency correlation (y,x*beta) - levels : yoy rates :

29 CHOW-LIN High frequency series (columnwise): * Estimate * Std. deviation * 1 sigma lower limit * 1 sigma upper limit * Residuals Elapsed time:

30 clc; clear all; close all; SANTOS-CARDOSO % ANNUAL DATA: Gross Added Value. Construction. % UNIT: 1995 Constant prices. Euros (thousand of millions). % SOURCE: National Accounts. % SAMPLE: Annual, Y = load('cst.anu'); % QUARTERLY DATA: Synthetic indicator. Construction. % Basic source: Construction Short-term Survey % UNIT: Euros (thousand of millions). % SOURCE: Quarterly National Accounts. % SAMPLE: Annual, (last data is a forecast) x = load('cst.ind'); % % Inputs for td library % Type of aggregation: sum ta=1; % Frequency conversion: annual to quarterly s=4; % Method of estimation type=0; % Name of ASCII file for output file_sal='cst.sal'; % Innovational parameter is set a priori rho = 0.63; % Calling the function: output is loaded in a structure called res res=ssc_fix(y,x,ta,s,type,rho); % Calling printing function output=1; % Do not include series td_print(res,file_sal,output); edit cst.sal; % Calling graph function td_plot(res);

31 SANTOS-CARDOSO

32 SANTOS-CARDOSO

33 SANTOS-CARDOSO **************************************************** TEMPORAL DISAGGREGATION METHOD: Santos Silva-Cardoso **************************************************** Number of low-frequency observations : 25 Frequency conversion : 4 Number of high-frequency observations: 101 Number of extrapolations : 1 Number of indicators (+ constant) : 2 Type of disaggregation: sum (flow). Estimation method: Weighted least squares. ** High frequency model ** Beta parameters (columnwise): * Estimate * Std. deviation * t-ratios Dynamic parameter:

34 SANTOS-CARDOSO Long-run beta parameters (columnwise): Truncation remainder: expected y(0): * Estimate * Std. deviation * t-ratios AIC: BIC: Low-frequency correlation (Y,X) - levels : yoy rates : High-frequency correlation (y,x) - levels : yoy rates : High-frequency volatility of yoy rates - estimate : indicator : ratio : High-frequency correlation (y,x*beta) - levels : yoy rates :

35 SANTOS-CARDOSO

36 SANTOS-CARDOSO

37 SANTOS-CARDOSO

38 DI FONZO clc; clear all; close all; % ANNUAL DATA: Gross Added Value. Y1=Agriculture. % Y2=Fisheries. % UNIT: 1995 Constant prices. Euros (thousand of millions). % SOURCE: National Accounts. % SAMPLE: Annual, % QUARTERLY DATA: Synthetic indicators. % Basic source: Ministry of Agriculture. Statistics dept. % UNIT: Euros (thousand of millions). % SOURCE: Quarterly National Accounts. % SAMPLE: Quarterly, % QUARTERLY CONSTRAINT: Gross Added Value. % z=agriculture + Fisheries. % UNIT: 1995 Constant prices. Euros (thousand of % millions). % SOURCE: Quarterly National Accounts. % SAMPLE: Quarterly,

39 % % Inputs for td library % Type of aggregation ta=1; % Frequency conversion s=4; % Model for innovations type=1; % Name of ASCII file for output file_sal='difonzo.sal'; DI FONZO % Number of high frequency indicators linked to each low frequency % aggregate f=[1 1]; % Calling the function: output is loaded in a structure called res res=difonzo(y,x,z,ta,s,type,f); % Calling printing function mtd_print(res,file_sal); edit difonzo.sal; % Calling graph function mtd_plot(res,z);

40 DI FONZO

41 DI FONZO

42 DI FONZO

43 DI FONZO

44 DI FONZO

45 DI FONZO

46 DI FONZO

47 DI FONZO ******************************************************* TEMPORAL DISAGGREGATION METHOD: Multivariate di Fonzo ******************************************************* --- Number of low-frequency observations : 8 Frequency conversion : 4 Number of high-frequency observations : 40 Number of extrapolations : Type of disaggregation: sum (flow). --- Model for the innovations: random walk. --- High frequency series (columnwise): * Point estimate High frequency series (columnwise): * Std. desviation Elapsed time:

48 ADDITIONAL DEVELOPMENTS Further integration of the Matlab library with Excel via add-ins Extended diagnostics Enhancement of dynamic models: the Autoregressive Dynamic Linear model ADL(1,1) (see Proietti, 2004)