Based on Fed statements and central bank speeches dating back 25 years, the firm's economists, including Joseph Lupton, used a language model based on ChatGPT to detect the range of policy signals, effectively scoring them on a scale from likely to unlikely, calling it the Hawk-Dove Score.
Reading the markets
By designing the index based on a range of asset performances, the economists found that the AI tool can be useful in predicting potential changes in Fed policy — and sending out useful signals. For example, they found that when the model shows an increase in Fed speakers' dovishness, one-year Treasury yields rise.
"Early conclusions are encouraging," Lupton and his colleagues wrote in a note.
Το εργαλείο δίνει μια ώθηση στην προσπάθεια της Wall Street να έχει ένα ανταγωνιστικό πλεονέκτημα συναλλαγών. Νωρίτερα αυτό το μήνα, αρκετές έρευνες έδειξαν ότι το ChatGPT μπορεί να προσθέσει αξία σε εργασίες που σχετίζονται με την αγορά, όπως η αποκρυπτογράφηση της, ανάλογα με το αν οι δηλώσεις της Fed ήταν επιθετικές ή όχι.
Based on JPMorgan's model, while the reading for Fed statements has fluctuated in recent months, they remain near two-decade highs on interest rate policy, a clear sign of persistent hawkishness.
Fed Hawk-Dove Score
The Fed is expected to raise its key interest rate next week by another 25 basis points to 5,25 percent, economists polled by Bloomberg forecast.
Investors have focused on the Fed's policy path after an aggressive inflation-fighting tightening campaign sent stocks and bonds tumbling last year. While asset prices have since recovered, analysts have continued to bet on rate cuts later this year, defying Fed Chairman Jerome Powell's warnings that rate cuts in 2023 is not στο μυαλο του.
According to JPMorgan's model, a 10-point increase in the Fed Hawk-Dove Score now translates to about a 10% increase in the probability of a 25-basis-point hike at the central bank's next policy meeting, or vice versa.
The Hawk-Dove Score, also available for the European Central Bank and the Bank of England, is expected to be rolled out to more than 30 central banks around the world in the coming months.