RDP 2021-11: Smells Like Animal Spirits: The Effect of Corporate Sentiment on Investment Read me file

This ‘read me’ file contains information on the files used to generate the results presented in RDP 2021-11.

The analysis was done using the statistical programming language Stata version 16.1.

The code is distributed as is, without warranty, and is solely for the purpose of replicating the results. Any alternative use of the code is not supported.

Data

The file ‘rdp-2021-11-graph-data.xlsx’ contains the publically available data used to directly plot the figures.

Morningstar

Most of the data underpinning the analysis presented in the paper comes from the Morningstar database.

The analysis should be able to be replicated to some extent with other datasets with publicly listed company information.

Refinitiv Eikon

Company-level estimates for the end-of-year share prices and number of shares outstanding are drawn from Refinitiv Eikon.

Estimates of equity analyst forecasts for 1-year ahead earnings per share also come from Refinitiv Eikon.

Due to third-party provider provisions, researchers will only be able to replicate the analysis if they have access to datasets with information on Australian publicly listed companies.

Connect 4

Annual financial reports for Australian listed companies were obtained under a subscription to Connect4. The counts of positive, negative and uncertain words for each company and year were produced following the procedure outlined in the paper. These counts are included in the supplementary information.

ABS

Data in Figure 2 are constructed using publicly available data from the Australian Bureau of Statistics (ABS) national accounts release, as well as aggregated information from the BLADE firm-level data environment. Only users with registered access to the ABS DataLab will be able to reproduce the firm-level estimates.

Code

The Stata files do not include data to run the code, as users must have access to the underlying unit record data. Once company-level data are acquired the replication file will allow the code to run with appropriate adjustments to company-level variables.

To produce the necessary output, do files should be run in the following order (with appropriate adjustments to the names of variables for the underlying company-level data if using alternative sources of information):

To set-up most of the underlying data files you need to run the DO file:

Do file: Corporate Sentiment – Setup.do

Inputs: morningstar_csv (or morningstar_raw.dta), Eikon.xlsx; D21 244102 Listed Companies – Locations.xlsx; shprice_1997_2020.dta; sharesoutstanding.dta; connect4_*.csv files (for various years between 2003 and 2020)

Outputs: morningstar_semiannual.dta; sentiment.dta; ms_listing.dta; shprice_eikon.dta

Most of the analysis presented in the paper (including Figures and Tables) are obtained by running:

Do file: Corporate Sentiment – Analysis.do

Inputs: morningstar_semiannual.dta; sentiment.dta; shprice_eikon.dta; eps1.dta; ms_listing.dta; turnover.dta (all these DTA files are combined into ms_sentiment_analysis.dta)

Outputs: regression tables, summary tables and graphs

However, for the analysis of the link between sentiment and business surveys, run the file:

Do file: Corporate Sentiment – Business Surveys.do

Inputs: GDP.xlsx, Business Surveys.xlsx, ms_sentiment_analysis.dta

Outputs: regression table

30 November 2021