Uncertainty Shocks & Business Cycles

The core of this section establishes a strong link between microeconomic uncertainty and business cycles. Using detailed Census microdata spanning 1972-2011 (Census of Manufactures and Annual Survey of Manufactures), the researchers construct novel empirical measures of uncertainty focusing on the dispersion of plant-level innovations to total factor productivity (TFP). A key finding is the countercyclical nature of this uncertainty; the dispersion of TFP shocks rises sharply during recessions. This is exemplified by Figure 1, which vividly illustrates a 76% increase in the variance of establishment-level TFP shocks during the 2007-2009 Great Recession, using a balanced panel of 15,752 establishments. The analysis goes beyond simple variance, also considering skewness and kurtosis, revealing that recessions at the micro level are characterized by a negative first-moment shock coupled with a positive second-moment shock. Furthermore, robustness checks using outlier-robust measures like the interquartile range (IQR) consistently confirm the countercyclical nature of the uncertainty, with the IQR of TFP shocks increasing by over 15% during recessionary periods. This finding aligns with similar research using alternative datasets and methodologies, demonstrating the robustness of the countercyclical microeconomic uncertainty phenomenon.

This subsection addresses a crucial caveat: the residual from the TFP regression might not perfectly capture unforecasted productivity shocks. To mitigate this concern, the study employs two strategies. First, it examines the cross-sectional spread of stock returns as an additional measure of uncertainty. Finding this spread to be countercyclical reinforces the initial findings. Second, the study investigates the relationship between establishment-level TFP shocks and parent firm stock returns. A significant positive correlation is detected, indicating that a substantial portion of the establishment-level TFP shocks represent new information to the market, strengthening the validity of using TFP shocks as a proxy for uncertainty. Further analysis attempts to remove any forecastable component of stock returns by controlling for quarterly factors such as firm size, market-to-book ratio, R&D intensity, and leverage (consistent with existing literature such as Bekaert, Hodrick, and Zhang (2012)). Even after this adjustment, the countercyclical relationship between uncertainty and recessions remains highly significant. The study also explores alternative indicators (e.g., plant-level output growth) which all confirm that dispersion significantly increases during recessions.

Expanding the analysis to the industry level, this subsection demonstrates a consistent pattern: higher uncertainty is associated with slower industry growth. Using the extensive Census dataset, the researchers examine the dispersion of productivity shocks within four-digit SIC industry-year cells. The significant size of the dataset (with a mean of 27.1 establishments per cell) allows for a robust investigation. Results reveal a strong correlation between within-industry dispersion of establishment TFP shocks and industry growth rates. Specifically, slower industry growth is linked to a sharp rise in within-industry dispersion of TFP shocks. Remarkably, this relationship holds true across various industries, suggesting a general pattern rather than an industry-specific anomaly. This finding is particularly significant in highlighting that the relationship between uncertainty and business cycles holds not just at the aggregate level, but also at the granular industry level. Though the direction of causality (uncertainty driving cycles versus endogenous amplification) is addressed, the paper suggests that both are likely involved.

This section details the structure of the Dynamic Stochastic General Equilibrium (DSGE) model used to analyze the effects of uncertainty shocks on business cycles. The model's key features include heterogeneous firms, adjustment costs for capital and labor, and a time-varying uncertainty process. This departure from standard models allows for a more nuanced understanding of firm-level responses to shocks. The productivity process in the model has both aggregate and idiosyncratic components, representing macroeconomic and microeconomic uncertainty respectively. Crucially, the model doesn't only allow for shocks to the mean (first-moment shocks) of productivity, but it also permits the variance (second-moment shocks) to vary over time, capturing periods of high and low uncertainty. This means productivity shocks can be small during normal times, but potentially larger when uncertainty is high. The inclusion of adjustment costs, a standard feature in the Real Business Cycle (RBC) literature, is crucial for capturing the lumpy nature of investment and hiring decisions. Firms' responses are thus not instantaneous but depend on the magnitude of uncertainty and the cost of adjustment.

This subsection presents a key finding: uncertainty shocks, simulated within the DSGE model, generate significant drops in Gross Domestic Product (GDP). Simulations show that these shocks can cause a decline in GDP of roughly 2.5%. This result highlights the quantitative importance of uncertainty shocks as a potential driver of business cycles, suggesting that uncertainty is not merely a secondary factor, but can be a primary force shaping macroeconomic outcomes. The model also reveals that simply introducing uncertainty shocks is insufficient to fully replicate observed consumption behavior over the business cycle. The simulations suggest the need to incorporate first-moment shocks to align model predictions with actual consumption patterns. Consequently, the research indicates that a comprehensive model of recessions should incorporate a combination of shocks that impact both the mean (negative first-moment) and variance (positive second-moment) of productivity.

The study examines the implications of increased uncertainty for the effectiveness of government policy, specifically focusing on first-moment policies such as wage subsidies. A crucial finding is that higher uncertainty can significantly reduce the short-term effectiveness of such policies. The reason is that when uncertainty is high, firms adopt a more cautious stance, delaying investment and hiring decisions, even in the face of policy stimulus. This increased caution leads to a muted response to wage subsidies compared to a scenario without heightened uncertainty. The results suggest that the impact of first-moment policies can be substantially diminished (by more than two-thirds on impact, according to the model) when combined with second-moment shocks. This dampened response echoes findings in the literature on lumpy investment and further underscores the importance of considering both first- and second-moment shocks when designing macroeconomic policy interventions. The model therefore demonstrates that uncertainty can dramatically alter the effectiveness of policies designed to stimulate the economy during a downturn.

This section details the effects of uncertainty shocks on key macroeconomic variables as simulated by the DSGE model. An uncertainty shock leads to a significant and immediate drop in both investment and hiring. This is because the increased uncertainty causes firms to postpone investment and hiring decisions. This initial freeze in investment and hiring directly translates into a substantial fall in output – a drop of just over 2.5% within one quarter is observed in the simulations. However, the impact on consumption is more complex. Initially, consumption increases. This seemingly counterintuitive result is attributed to two factors: first, the increased misallocation of resources resulting from the uncertainty shock lowers the expected return on savings, making immediate consumption more attractive; second, the freeze on investment and hiring frees up resources that can be allocated to consumption. After this initial surge, however, consumption begins to decline as the capital stock falls below its ergodic distribution (the long-run average level), hours worked remain depressed, and misallocation of resources continues. The study's results show that a pure uncertainty shock, while leading to output and investment decline, yields an initial rise in consumption, which is not always consistent with real-world recessionary patterns. This counterintuitive consumption response motivates further investigation of modelling strategies.

The initial rise in consumption following a pure uncertainty shock is identified as an undesirable feature of the model. To address this, the authors explore alternative model specifications. They discuss extending the model to allow consumers to save in assets outside the domestic economy (like foreign assets), mirroring the approach in Fernandez-Villaverde et al. (2011). In an open economy, domestic uncertainty could lead to increased savings abroad, potentially explaining the initial consumption increase. The study also discusses the possibility of altering the utility function to reduce the consumption overshoot, but notes the significant increase in computational complexity involved in implementing these changes. Another option, modelling precautionary saving behaviour in response to uncertainty (as in Basu and Bundick, 2016), is also mentioned. However, adding nominal rigidities to account for such behaviour would significantly extend the scope of the model and falls outside the current study's objectives. The study ultimately demonstrates that including a negative first-moment shock (such as an aggregate productivity shock) in addition to the uncertainty shock eliminates the counterintuitive initial consumption increase, yielding a more realistic representation of recessionary dynamics.

This section assesses the robustness of the model's results to various parameter changes and compares the effects of uncertainty shocks in general equilibrium (GE) and partial equilibrium (PE) frameworks. Sensitivity analysis is performed by varying key parameters of the uncertainty process (macro- and micro-volatility, persistence, and frequency). The results are found to be robust, particularly with respect to changes in micro- and macro-volatility levels. A notable exception is the sensitivity to the persistence of the uncertainty shock. Lower persistence leads to shorter-lived impacts, highlighting the role of persistence in shaping the overall economic response. The analysis is further extended by comparing the model's output under different equilibrium frameworks. Comparing general equilibrium with partial equilibrium (with and without adjustment costs) reveals the relative contributions of different mechanisms, like real options effects and the Oi-Hartman-Abel effect, in shaping the economy's response to uncertainty shocks. This decomposition helps in understanding the intricate interplay between investment, labor, and consumption dynamics under uncertainty.

This subsection focuses on the impact of uncertainty on the effectiveness of stimulative policies, specifically wage subsidies. It's emphasized that the analysis is not about finding optimal policy, but rather understanding how uncertainty affects policy outcomes. Simulations demonstrate that the presence of uncertainty significantly reduces the effectiveness of wage subsidies. This occurs because increased uncertainty pushes firms further away from their hiring thresholds (the S in Ss policy rules). Consequently, many firms become less responsive to policy stimulus, resulting in a muted response to the wage subsidy. The effect of uncertainty on the policy response echoes findings from the lumpy investment literature and further underscores the limitations of policies that primarily target the mean of economic variables (first-moment policies) in the face of significant uncertainty. The study's results highlight the importance of understanding the interplay between uncertainty and policy design, and suggest that the effectiveness of traditional stimulative policies may be severely curtailed during periods of high uncertainty.