Accessible Version
When productivity booms, do interest rates rise? Accessible Data
Figure 1. Impulse responses of selected variables to a TFP news shock
Figure 1 presents a multi-panel chart showing impulse response functions over a 40-quarter (10-year) horizon. The figure displays the dynamic responses of key interest rate variables to a positive total factor productivity (TFP) news shock.
Key findings depicted in the figure:
- The 10-year Treasury yield declines following the TFP news shock, contrary to standard macroeconomic theory predictions
- The expected real short rate 5-to-10 years ahead (a measure of r*) also decreases in response to the shock
- The responses demonstrate that anticipated increases in productivity growth are associated with lower, not higher, long-term interest rates
Figure 2. Impulse responses of core variables to an identified technology shock
Figure 2 displays a five-panel chart showing impulse response functions over a 40-quarter horizon for the core variables in the VAR model. Each panel shows the percentage or level response to a one-standard-deviation TFP news shock, with 68% confidence bands.
The five panels show responses for:
- TFP (Total Factor Productivity): Shows a gradual, persistent increase over time. The largest effects on the level of TFP materialize several years after the initial shock, reaching maximum impact around quarters 20-40. This trajectory is consistent with the gradual diffusion of productivity-enhancing technologies.
- Real GDP per capita: Displays a positive response that builds gradually over time, reflecting the eventual realization of higher productivity.
- Real Business Fixed Investment per capita: Shows an immediate, sharp positive response that occurs well before the bulk of the TFP increase materializes. This front-loading is consistent with firms anticipating higher returns on capital and responding to news about future productivity improvements.
- Real Consumption per capita: Exhibits a positive response that increases over time as households anticipate benefiting from higher incomes.
- S&P 500 (deflated by CPI): Demonstrates an immediate, substantial positive response, rising sharply upon impact and remaining elevated. This immediate reaction, occurring before the main TFP gains materialize, confirms that the productivity gains were anticipated by economic agents well in advance.
The responses of investment and the stock market rising immediately, long before the bulk of the TFP increase occurs, are consistent with the interpretation that these shocks contain advance information about future productivity improvements.
Figure 3. Impulse responses of auxiliary variables to an identified technology shock
Figure 3 presents a five-panel chart showing impulse response functions over a 40-quarter horizon for auxiliary interest rate and term premium variables. Each panel shows the response to the same TFP news shock identified in Figure 2, with 68% confidence bands. The five panels show:
- Nominal 10-year Treasury yield: Falls in response to the positive TFP news shock. The decline is statistically significant and persistent, contradicting the standard theoretical prediction that anticipated productivity growth should raise interest rates.
- Real 10-year Treasury yield: Exhibits a similar declining pattern to the nominal yield, demonstrating that the fall in nominal yields is not driven by declining inflation expectations but reflects movements in real rates.
- Real federal funds rate: Shows a modest positive response that rises temporarily before returning toward baseline, indicating short-term monetary policy tightening in response to the positive shock.
- Decomposition of real 10-year yield:
- Expectations component: Rises modestly, mirroring the path of the real federal funds rate shown in panel 3
- Term premium component: Exhibits a large, persistent, and statistically significant decline that more than offsets the rise in the expectations component
The net result is that the decline in the term premium dominates, causing the overall real 10-year yield to fall despite the modest increase in expected short rates.
Figure 4. Estimated TFP news shocks
Figure 4 displays a time series chart covering the period from 2022Q4 to 2026Q2. The chart shows an eight-quarter (two-year) moving average of the model's estimated TFP news shocks.
Key features:
- Horizontal axis: Time, spanning from 2022Q4 through 2026Q2
- Vertical axis: Magnitude of TFP news shocks (in standardized units)
- Zero line: A horizontal line at zero helps identify positive versus negative shock periods
Pattern over time:
- 2022Q4-2023 early: The estimated shocks are initially negative, reflecting the 2022 equity market downturn and pessimism about future productivity
- 2023-2026Q2: The shock realizations turn positive on average and remain elevated, showing a sustained improvement in productivity expectations
This pattern is consistent with the view that rapid developments in artificial intelligence technology, particularly following the release of ChatGPT in November 2022, have progressively brightened prospects for future productivity growth over the 2023-2026 period.
Figure 5. VAR forecast error for real 10-year Treasury yields
Figure 5 presents a time series chart showing a historical decomposition of the real 10-year Treasury yield from 2022Q4 to 2026Q2. The chart illustrates how well TFP news shocks explain the model’s forecast error of the real 10-year Treasury yield.
The chart contains two key lines:
1. Black line (Total VAR forecast error): Represents the difference between the actual real 10-year Treasury yield and what the VAR model predicted in the absence of shocks. This line shows:
- The model expected long-term rates to revert toward their low pre-pandemic baseline
- The forecast error widens steadily over time, reaching 200 basis points by early 2026
- This large positive error indicates that actual yields remained much higher than the model's baseline projection, even though the actual yield increased by only 90 basis points between 2022Q3 and 2026Q2
2. Blue line (Contribution of TFP news shocks): Shows how much of the forecast error can be attributed to TFP news shocks. This line shows:
- 2022Q4-early 2023: Initially positive contribution, as negative TFP news shocks during the 2022 downturn pushed yields higher than the baseline
- 2023-2025: Steadily declining contribution as positive TFP news shocks (driven by AI optimism) exerted downward pressure on rates
- 2026Q2: The net contribution approaches zero or becomes slightly negative
Key interpretation: By 2026Q2, the cumulative contribution of TFP news shocks since the launch of ChatGPT is near zero, suggesting that news about future productivity growth has had virtually no net impact on current 10-year Treasury yields. The large remaining forecast error must be explained by other macroeconomic factors, not AI-related productivity expectations.
Figure 6. Impulse response of the expected real short rate 5-to-10 years ahead
Figure 6 displays a single-panel impulse response function over a 40-quarter (10-year) horizon, with 68% confidence bands. The figure shows the dynamic response of the expected real short rate 5-to-10 years ahead—a commonly used market-based measure of r* (the long-run equilibrium real interest rate).
Key features:
- Response pattern: Following a positive TFP news shock, the expected real short rate 5-to-10 years ahead declines significantly
- The decline is statistically significant (the confidence band does not include zero)
- The decline is persistent, lasting throughout most of the 40-quarter horizon
- The magnitude of the decline contradicts standard macroeconomic theory
Theoretical vs. Empirical findings:
- Standard theory prediction: r* should rise in response to an anticipated, persistent increase in TFP growth, as higher expected future productivity should increase the natural rate of interest
- Empirical finding: The expected real short rate 5-to-10 years ahead actually falls, not rises
Implications: This finding calls into question modeling approaches that estimate r* by enforcing a tight relationship between r* and the medium-run growth rate of the economy, such as the Laubach-Williams (2003) framework. The results suggest that the relationship between productivity growth expectations and the equilibrium real interest rate is more complex than standard models assume.