Search earnings calls

How to search earnings calls

Searching earnings calls involves two steps:

1. Crafting a query. A query can be simple or complex; for example, it can simply be a collection of terms such as supply chain, supplier, or it can be thousands of keywords chained and modified by our query language. To craft a good query, it is useful to understand that our search engine operates at the sentence level: For each sentence that satisfy your search options, the search engine checks whether the query can be satisfied.

We created our Keyword Tool to help you create a list of relevant keywords and recommend you use it when crafting your query.

1. Specifying additional search options. Once you have your query, you can specify additional search options to target specific earnings calls, or to focus on specific parts within earnings calls:

   • Start Date and End Date: Specify the date range for your search. The tool will only search earnings call transcripts within this period.

   • Section of Earnings Calls: Select a section (presentation or Q&A) of the earnings calls to narrow down your search.

   • Affiliation of Speakers: Choose the affiliation of speakers (corporate executives or external participants) to filter results based on who is speaking.

   • Include Adjacent Sentences: Toggle this option to include one sentence before and after by the same speaker, if available.

   • Search Name: You can name your search for easier reference later.

Example of a search for keywords related to 'ChatGPT'

We currently do not offer to specify specific industries or firms in the search form. However, once the search is done you can focus on any particular firm or industry using the output file as well as in the Snippet Tool.

Output files

Each search generates a set of output files designed to facilitate further downstream analyses:

• Firm level exposure, risk, and sentiment (CSV file). This is the key output file that can be used for downstream analysis. All variables are defined precisely here, but broadly it contains the following:

  • Measures of topic exposure, risk, and sentiment. These are integer counts for the number of sentences in that earnings call that satisfy the query.

  • Additional variables including the company name, industry, country of headquarter, ISIN, and a few other identifiers.

• Firm level unconditional risk and sentiment (CSV file). This file has been requested by multiple clients. It contains, for each earnings call:

  • The number of sentences with at least one risk synonym.

  • The number of sentences with at least one positive sentiment word.

  • The number of sentences with at least one negative sentiment word.

• Crosswalk to GVkey (CSV file). This file provides a link table to the GVKey identifier. Its purpose is to facilitate merging our output with data from Capital IQ. Please also see here.
• Search metadata (TXT file). This file contains all search parameters, including a unique search identifier, for reference.

A few additional details that are useful to keep in mind:

• All CSV files are at the earnings call level (one row, one earnings call) for earnings calls during the time period specified.

• There is an identifier for each earnings call from the NL Analytics ecosystem. This identifier (earningscallID) is unique, stable, and not reused. It can be used not only to link the CSV files, but also to link across searches.

• The variable nr_sentences currently does not adjust when you restrict the search to a particular section or speaker, or when you allow the search to use sentence triples. It is always just the total number of sentences of that earnings call. We understand this might not be intuitive and are working on a solution.

• The CSV files always contain all earnings calls within the date range specified in the search, even if no sentences satisfy the query.

Query language

Note that our search engine is case insensitive by default.

OR

Finds sentences containing at least one of the specified keywords. Equivalent to using a comma between phrases.

Example: The query china OR lockdown finds sentences containing either keyword.

AND

Ensures sentences contain all specified keywords.

Example: The query china AND lockdown finds sentences containing both keywords.

AND NOT

Finds sentences with the first keyword but not the second.

Example: The query silicon AND NOT valley finds sentences with "silicon" but not "valley".

Quotation marks ["..."]:

Finds exact matches for the enclosed keyword.

Example: The query motor AND NOT "Motor" finds sentences with "motor", "mOTOR", and other case variations except for "Motor".

Parentheses [(...)]:

Groups keywords, useful with AND or AND NOT operators.

Example: To find all sentences containing the keywords "investment," "investments," and "investing" but not "investor" you would type

(investment OR investments OR investing) AND NOT investor

Wildcards [*] or [?]:

• * allows zero or more arbitrary characters until the next word boundary.

Example: To find all sentences containing any word that starts with "invest" (which includes, for example, "investment," "investing," and "investments"), you would type

invest*

• ? allows zero or one arbitrary character until the next word boundary.

Example: To find all sentences containing any word that starts with 'investment' and allows an additional character (for example 'investments'), you would type

investment?

Capitalization

By default, the search is case insensitive.

Example: The query supply chain finds all sentences containing the phrase "supply chain" regardless of capitalization.

To respect capitalization, please enclose the phrase with quotation marks.

The Search earnings calls form allows you to search through transcripts of earnings calls. The key component of a search is a query, which can be as simple one or more keywords. However, you can craft more tailored queries using specific rules described below. Additionally, our search options enable you to target specific parts of the speech.