Добавил:
Upload Опубликованный материал нарушает ваши авторские права? Сообщите нам.
Вуз: Предмет: Файл:
Eviews5 / EViews5 / Docs / EViews 5 Users Guide.pdf
Скачиваний:
152
Добавлен:
23.03.2015
Размер:
11.51 Mб
Скачать

Chapter 28. Working with Panel Data

EViews provides you with specialized tools for working with stacked data that have a panel structure. You may have, for example, data for various individuals or countries that are stacked one on top of another.

The first step in working with stacked panel data is to describe the panel structure of your data: we term this step structuring the workfile. Once your workfile is structured as a panel workfile, you may take advantage of the EViews tools for working with panel data, and for estimating equation specifications using the panel structure.

The following discussion assumes that you have an understanding of the basics of panel data. “Panel Data” beginning on page 210 provides background on the characteristics of panel structured data.

We first review briefly the process of applying a panel structure to a workfile. The remainder of the discussion in this chapter focuses on the basics working with data in a panel workfile. Chapter 29, “Panel Estimation”, on page 901 outlines the features of equation estimation in a panel workfile.

Structuring a Panel Workfile

The first step in panel data analysis is to define the panel structure of your data. By defining a panel structure for your data, you perform the dual tasks of identifying the cross-sec- tion associated with each observation in your stacked data, and of defining the way that lags and leads operate in your workfile.

While the procedures for structuring a panel workfile outlined below are described in greater detail elsewhere, an abbreviated review may prove useful (for additional detail, see “Describing a Balanced Panel Workfile” on page 53, “Dated Panels” on page 223, and “Undated Panels” on page 228).

There are two basic ways to create a panel structured workfile. First, you may create a new workfile that has a simple balanced panel structure. Simply select

File/New/Workfile... from the main EViews menu to open the Workfile Create dialog. Next, select Balanced Panel from the Workfile structure type combo box, and fill out the dialog as desired.

874—Chapter 28. Working with Panel Data

Here, we create a balanced quarterly panel (ranging from 1970Q1 to 2020Q4) with 200 cross-sections.

When you click on OK, EViews will create an appropriately structured workfile with 40,800 observations (51 years, 4 quarters, 200 cross-sections). You may then enter or import the data into the workfile.

More commonly, you will use the second method of structuring a panel workfile, in which you first read stacked data into an unstructured workfile, and then apply a structure to

the workfile. While there are a number of issues involved with this operation, let us consider a simple, illustrative example of the basic method.

Suppose that we have data for the job training example considered by Wooldridge (2002), using data from Holzer, et al. (1993).

These data form a balanced panel of 3 annual observations on 157 firms. The data are first read into a 471 observation, unstructured EViews workfile. The values of the series YEAR and FCODE may be used to identify the date and cross-section, respectively, for each observation.

To apply a panel structure to this workfile, simply double click on the “Range:” line at the top of the workfile window, or select Proc/Structure/Resize Current Page... to open the

Workfile structure dialog. Select Dated Panel as our Workfile structure type.

Panel Workfile Display—875

Next, enter YEAR as the Date series and FCODE as the Crosssection ID series. Since our data form a simple balanced dated panel, we need not concern ourselves with the remaining settings, so we may simply click on

OK.

EViews will analyze the data in the specified Date series and

Cross-section ID series to determine the appropriate structure

for the workfile. The data in the workfile will be sorted by cross-section ID series, and then by date, and the panel structure will be applied to the workfile.

Panel Workfile Display

The two most prominent visual changes in a panel structured workfile are the change in the range and sample information display at the top of the workfile window, and the change in the labels used to identify individual observations.

Range and Sample

The first visual change in a panel structured workfile is in the Range and Sample descriptions at the top of workfile window.

For a dated panel workfile, EViews will list both the earliest and latest observed dates, the number of cross-sections, and the total number of unique observations.

876—Chapter 28. Working with Panel Data

Here we see the top portion of an annual workfile with observations from 1935 to 1954 for 10 cross-sections. Note that workfile sample is described using the earliest and latest observed annual frequency dates (“1935 1954”).

In contrast, an undated panel workfile will display an observation range of 1 to the total number of observations. The panel dimension statement

will indicate the largest number of observations in a cross-section and the number of crosssections. Here, we have 92 cross-sections containing up to 30 observations, for a total of 506 observations. Note that the workfile sample is described using the raw observation numbers (“1 506”) since there is no notion of a date pair in undated panels.

You may, at any time, click on the Range display line or select Proc/WF Structure & Range... to bring up the Workfile Structure dialog so that you may modify or remove your panel structure.

Observation Labels

The left-hand side of every workfile contains observation labels that identify each observation. In a simple unstructured workfile, these labels are simply the integers from 1 to the total number of observations in the workfile. For dated, non-panel workfiles, these labels are representations of the unique dates associated with each observation. For example, in an annual workfile ranging from 1935 to 1950, the observation labels are of the form “1935”, “1936”, etc.

In contrast, the observation labels in a panel workfile must reflect the fact that observations possess both cross-section and within-cross-section identifiers. Accordingly, EViews will form observation identifiers using both the cross-section and the cell ID values.

Here, for example, we see the labels for a panel annual workfile structured using ID as the crosssection identifier, and DATE as the cell ID (we have widened the observation column so that the full observation identifier is visible). The labels have been formed by concatenating the ID value, a hyphen, and a two-digit representation of the date.

Note that EViews generally uses the display formats for the cross-section and cell ID series when forming the observation label. For example, since ID is displayed using 7 fixed char-

Соседние файлы в папке Docs