Monetary policy, factor allocation and growth
1 like960 views
This document explores the impact of monetary policy on resource allocation and productivity growth post the Great Financial Crisis. It finds that the slope of the yield curve is closely related to how factor reallocation contributes to productivity, with flatter yield curves reducing such contributions. The analysis indicates a shift in monetary policy dynamics before and after the crisis, influencing the exit of unproductive firms and the growth of high productivity entrants.
Monetary policy, factor allocation and growth
- 1. Monetary policy, factor allocation and growth Ryan Banerjee, Enisse Kharroubi, Fabrizio Zampolli BIS-IMF-OECD productivity conference 10-11 January 2018, Paris Disclaimer: the views expressed are those of the presenters and not necessarily those of the BIS
- 2. Motivation 10 years after Great Financial Crisis (GFC) productivity growth remains weak Key crisis factors seem less relevant Credit conditions loose Banks have largely repaired their balance sheets Global growth is picking up What are the effects of monetary policy? Easier monetary policy essential to stabilise demand in the depths of the crisis But what are the effects on resource allocation? - Especially during extended period of exceptionally low interest rates 2
- 3. This paper How does monetary policy affect resource allocation? Could go in either direction Good: Easy monetary policy may help highly productivity firms expand by alleviating credit constraints Bad: Easy monetary policy can slow productivity enhancing reallocation if it reduces pressure on non-viable firms to exit New method to analyse contribution of factor reallocation to productivity Show the relationship between a productivity decomposition and panel regressions Use this insight to examine impact of monetary policy on contribution of reallocation to productivity growth 3
- 4. Main results Sectoral reallocation closely linked to the slope of the yield curve Rather than either short or long-term rates alone Shocks which flatten the yield curve Reduce the contribution of factor reallocation across sectors to aggregate productivity growth Evidence holds for both capital and labour reallocation but stronger for labour Not driven by properties of yield curve signalling the phase of the cycle Different way monetary policy conducted pre and post GFC Pre-crisis: yield curve flattening monetary tightening Post-crisis: yield curve flattening monetary easing May indicate state-dependency (credit market functioning, financial vs. real shocks, etc…) 4
- 5. Related literature Factor reallocation boosts productivity growth in downturns eg Cabellero and Hammour (1994, 1996), Davis and Haltiwanger (1992), Mortensen and Pissarides (1994) Cleansing effect may be weaker if young productive firms exit eg Baden-Fuller (1989), Dunne et al (1989), Eslava et al (2010) Reallocation in US less intense during 2008-09 recession than in previous recessions Foster et al (2016) Productivity decompositions eg Baily et al (2001), Olley and Pakes (1996), Borio et al (2015) 5
- 6. Decomposing total factor productivity growth Assume aggregate TFP is an average of sectoral TFPs (𝐴𝐴𝑠𝑠𝑠𝑠) weighted by their shares in aggregate input (𝑥𝑥𝑠𝑠𝑠𝑠) (eg labour) 𝐴𝐴𝑡𝑡 𝑥𝑥 = ∑𝑠𝑠 ⁄𝑥𝑥𝑠𝑠𝑠𝑠 𝑥𝑥𝑡𝑡 𝐴𝐴𝑠𝑠𝑠𝑠 TFP growth can be decomposed into common and allocative components (where 𝑎𝑎𝑡𝑡 𝑥𝑥 = ⁄𝑥𝑥𝑠𝑠𝑠𝑠 𝑥𝑥𝑡𝑡 𝐴𝐴𝑠𝑠𝑠𝑠/𝐴𝐴𝑡𝑡 𝑥𝑥 ) 6 Average relative rate of labour accumulation across sectors Average size-weighted TFP growth across sectors Covariance of relative rate of labour accumulation and size-weighted TFP growth across sectors Common component Allocation component
- 7. From decompositions to regressions Could follow two stage approach eg Borio et al (2015) Compute common and allocation components Run cross-country, cross-time panel regression Alternative: use cross-sector variation in a single stage regression Note that the ̂𝛽𝛽 is proportional to allocation component 7
- 8. Regression specification Baseline regression: (𝑥𝑥𝑠𝑠/𝑥𝑥)𝑐𝑐𝑐𝑐+1 (𝑥𝑥𝑠𝑠/𝑥𝑥)𝑐𝑐𝑐𝑐 = 𝛼𝛼𝑠𝑠𝑠𝑠 + 𝛼𝛼𝑐𝑐𝑐𝑐 + 𝛼𝛼𝑠𝑠𝑠𝑠 + 𝛽𝛽1 + 𝛽𝛽2 𝑀𝑀𝑐𝑐𝑐𝑐 1 + ∆𝐴𝐴𝑠𝑠𝑠𝑠𝑠𝑠+1 𝐴𝐴𝑠𝑠𝑠𝑠𝑠𝑠 𝑎𝑎𝑠𝑠𝑠𝑠𝑠𝑠 𝑥𝑥 + 𝛽𝛽0 𝑍𝑍𝑠𝑠𝑠𝑠𝑠𝑠 + 𝜀𝜀𝑠𝑠𝑠𝑠𝑠𝑠 where 𝑀𝑀𝑐𝑐𝑐𝑐 is a vector of current and past macro variables including Interest rate surprises: 1 and 2 year lagged Sales growth: Current and 1 year lagged Credit growth: Current and 1 year lagged We are interested in the element of 𝛽𝛽2 which measures the sensitivity of factor allocation growth to the interaction between sectoral TFP growth and interest rate surprises 8
- 9. Data EU-KLEMS 10 countries 2000-2015. Austria, Belgium, Denmark, Finland, France, Germany, Italy, Netherlands, Sweden, United States Aggregate and sectoral value added, sales, employment, hours worked, capital stock and total factor productivity growth 13 sectors Additional datasets OECD: 1-year ahead forecast errors in policy rates and 10-y government bond yields as interest rate surprises BIS: aggregate data on credit to the non-financial corporate sector 9
- 10. 10 Summary statistics – productivity growth Common component accounts for most of average labour and capital productivity growth, but allocation component accounts for most of average TFP growth Allocation component volatility is about 1/3 of (total factor) productivity growth volatility. Mean Std. dev. Labour productivity growth 1.03% 2.24% Allocation component 0.06% 0.70% Common component 0.97% 2.31% Capital Productivity growth -0.55% 1.99% Allocation component 0.02% 0.60% Common component -0.56% 2.01% TFP growth (weighted by employment shares) 0.13% 1.68% Allocation component 0.11% 0.56% Common component 0.03% 1.70% TFP growth (weighted by hours shares) 0.20% 1.71% Allocation component 0.13% 0.54% Common component 0.07% 1.70%
- 11. 11 Interest rate shocks Interest rate shocks: 𝐹𝐹𝐸𝐸{𝑆𝑆𝑆𝑆,𝐿𝐿𝐿𝐿} = 𝑖𝑖{𝑆𝑆𝑆𝑆,𝐿𝐿𝐿𝐿}𝑡𝑡 − 𝐸𝐸𝑡𝑡−1(𝑖𝑖{𝑆𝑆𝑆𝑆,𝐿𝐿𝐿𝐿}𝑡𝑡) (positive accommodation) Post 2010 most action from long-term interest rates
- 12. 12 Dependent variable: Relative employment growth Variable (1) (2) log(𝑒𝑒𝑒𝑒𝑒𝑒 𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠)𝑐𝑐𝑐𝑐𝑐𝑐−1 -0.0955*** -0.0951*** (0.0186) (0.0185) 𝑅𝑅𝑅𝑅𝑅𝑅𝑅𝑅 𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠 𝑔𝑔𝑔𝑔𝑔𝑔𝑔𝑔𝑔𝑔𝑔𝑐𝑐𝑐𝑐𝑐𝑐 0.127*** 0.127*** (0.0382) (0.0381) 𝑅𝑅𝑅𝑅𝑅𝑅𝑅𝑅 𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠 𝑔𝑔𝑔𝑔𝑔𝑔𝑔𝑔𝑔𝑔𝑔𝑐𝑐𝑐𝑐𝑐𝑐−1 0.0656*** 0.0679*** (0.0209) (0.0210) 𝑇𝑇𝑇𝑇𝑇𝑇𝑔𝑔𝑔𝑔𝑔𝑔𝑔𝑔𝑔𝑔𝑔(𝑆𝑆𝑆𝑆𝑆𝑆𝑆𝑆 𝑊𝑊𝑊𝑊𝑊𝑊𝑊𝑊𝑊𝑊𝑊𝑊𝑊𝑊𝑊)𝑐𝑐𝑐𝑐𝑡𝑡 0.161 0.189* (0.111) (0.112) 𝐹𝐹𝐹𝐹𝑆𝑆𝑆𝑆,𝑐𝑐𝑡𝑡−1 ∗ 𝑇𝑇𝑇𝑇𝑇𝑇𝑔𝑔𝑔𝑔𝑔𝑔𝑔𝑔𝑔𝑔𝑔(𝑆𝑆𝑆𝑆)𝑐𝑐𝑐𝑐𝑐𝑐 10.35*** (3.321) 𝐹𝐹𝐹𝐹𝑆𝑆𝑆𝑆,𝑐𝑐𝑡𝑡−2 ∗ 𝑇𝑇𝑇𝑇𝑇𝑇𝑔𝑔𝑔𝑔𝑔𝑔𝑔𝑔𝑔𝑔𝑔(𝑆𝑆𝑆𝑆)𝑐𝑐𝑐𝑐𝑐𝑐 3.032 (3.207) 𝐹𝐹𝐹𝐹𝐿𝐿𝐿𝐿,𝑐𝑐𝑡𝑡−1 ∗ 𝑇𝑇𝑇𝑇𝑇𝑇𝑔𝑔𝑔𝑔𝑔𝑔𝑔𝑔𝑔𝑔𝑔(𝑆𝑆𝑆𝑆)𝑐𝑐𝑐𝑐𝑡𝑡 -8.829** (3.762) 𝐹𝐹𝐹𝐹𝐿𝐿𝐿𝐿,𝑐𝑐𝑡𝑡−2 ∗ 𝑇𝑇𝑇𝑇𝑇𝑇𝑔𝑔𝑔𝑔𝑔𝑔𝑔𝑔𝑔𝑔𝑔(𝑆𝑆𝑆𝑆)𝑐𝑐𝑐𝑐𝑡𝑡 1.854 (3.365) 𝐹𝐹𝐹𝐹𝑆𝑆𝑆𝑆𝑆𝑆 𝑆𝑆𝑆𝑆 𝑐𝑐𝑡𝑡−1 ∗ 𝑇𝑇𝑇𝑇𝑇𝑇𝑔𝑔𝑔𝑔𝑔𝑔𝑔𝑔𝑔𝑔𝑔(𝑆𝑆𝑆𝑆)𝑐𝑐𝑐𝑐𝑡𝑡 -9.158*** (2.822) 𝐹𝐹𝐹𝐹𝑆𝑆𝑆𝑆𝑆𝑆 𝑆𝑆𝑆𝑆 𝑐𝑐𝑡𝑡−2 ∗ 𝑇𝑇𝑇𝑇𝑇𝑇𝑔𝑔𝑔𝑔𝑔𝑔𝑔𝑔𝑔𝑔𝑔(𝑆𝑆𝑆𝑆)𝑐𝑐𝑐𝑐𝑡𝑡 -2.507 (2.587) 𝐹𝐹𝐹𝐹𝐹𝐹𝐹𝐹𝐹𝐹 𝑒𝑒𝑒𝑒𝑒𝑒𝑒𝑒𝑒𝑒𝑒𝑒𝑒𝑒 ( 𝛼𝛼𝑠𝑠𝑠𝑠, 𝛼𝛼𝑐𝑐𝑐𝑐, 𝛼𝛼𝑠𝑠𝑠𝑠) Yes Yes Obs 1,651 1,651 R-squared 0.562 0.562 Short-rates: Monetary easing t-1 Higher growth contribution from reallocation Yield curve flattening reduces contribution from reallocation F-test cannot reject hypothesis: sum of short and long interest rate coefficients is zero Robust to inclusion of credit gap and sales growth (current + 1 lag) Baseline results – employment reallocation
- 13. 13 Dependent variable: Relative capital growth Variable (1) (2) log(𝑐𝑐𝑐𝑐𝑐𝑐𝑐𝑐𝑐𝑐𝑐𝑐𝑐𝑐 𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠)𝑐𝑐𝑐𝑐𝑐𝑐−1 -0.0611*** -0.0614*** (0.0180) (0.0184) 𝑅𝑅𝑅𝑅𝑅𝑅𝑅𝑅 𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠 𝑔𝑔𝑔𝑔𝑔𝑔𝑔𝑔𝑔𝑔𝑔𝑐𝑐𝑐𝑐𝑐𝑐 0.0272 0.0259 (0.0501) (0.0504) 𝑅𝑅𝑅𝑅𝑅𝑅𝑅𝑅 𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠 𝑔𝑔𝑔𝑔𝑔𝑔𝑔𝑔𝑔𝑔𝑔𝑐𝑐𝑐𝑐𝑐𝑐−1 0.0161 0.0172 (0.0366) (0.0365) 𝑇𝑇𝑇𝑇𝑇𝑇𝑔𝑔𝑔𝑔𝑔𝑔𝑔𝑔𝑔𝑔𝑔(𝑆𝑆𝑆𝑆𝑆𝑆𝑆𝑆 𝑊𝑊𝑊𝑊𝑊𝑊𝑊𝑊𝑊𝑊𝑊𝑊𝑊𝑊𝑊)𝑐𝑐𝑐𝑐𝑡𝑡 0.0493 0.105 (0.279) (1.117) 𝐹𝐹𝐹𝐹𝑆𝑆𝑆𝑆,𝑐𝑐𝑡𝑡−1 ∗ 𝑇𝑇𝑇𝑇𝑇𝑇𝑔𝑔𝑔𝑔𝑔𝑔𝑔𝑔𝑔𝑔𝑔(𝑆𝑆𝑆𝑆)𝑐𝑐𝑐𝑐𝑐𝑐 8.071* (4.179) 𝐹𝐹𝐹𝐹𝑆𝑆𝑆𝑆,𝑐𝑐𝑡𝑡−2 ∗ 𝑇𝑇𝑇𝑇𝑇𝑇𝑔𝑔𝑔𝑔𝑔𝑔𝑔𝑔𝑔𝑔𝑔(𝑆𝑆𝑆𝑆)𝑐𝑐𝑐𝑐𝑐𝑐 3.640* (2.158) 𝐹𝐹𝐹𝐹𝐿𝐿𝐿𝐿,𝑐𝑐𝑡𝑡−1 ∗ 𝑇𝑇𝑇𝑇𝑇𝑇𝑔𝑔𝑔𝑔𝑔𝑔𝑔𝑔𝑔𝑔𝑔(𝑆𝑆𝑆𝑆)𝑐𝑐𝑐𝑐𝑡𝑡 -8.181* (4.260) 𝐹𝐹𝐹𝐹𝐿𝐿𝐿𝐿,𝑐𝑐𝑡𝑡−2 ∗ 𝑇𝑇𝑇𝑇𝑇𝑇𝑔𝑔𝑔𝑔𝑔𝑔𝑔𝑔𝑔𝑔𝑔(𝑆𝑆𝑆𝑆)𝑐𝑐𝑐𝑐𝑡𝑡 -4.039 (4.512) 𝐹𝐹𝐹𝐹𝑆𝑆𝑆𝑆𝑆𝑆 𝑆𝑆𝑆𝑆 𝑐𝑐𝑡𝑡−1 ∗ 𝑇𝑇𝑇𝑇𝑇𝑇𝑔𝑔𝑔𝑔𝑔𝑔𝑔𝑔𝑔𝑔𝑔(𝑆𝑆𝑆𝑆)𝑐𝑐𝑐𝑐𝑡𝑡 -7.269 (4.605) 𝐹𝐹𝐹𝐹𝑆𝑆𝑆𝑆𝑆𝑆 𝑆𝑆𝑆𝑆 𝑐𝑐𝑡𝑡−2 ∗ 𝑇𝑇𝑇𝑇𝑇𝑇𝑔𝑔𝑔𝑔𝑔𝑔𝑔𝑔𝑔𝑔𝑔(𝑆𝑆𝑆𝑆)𝑐𝑐𝑐𝑐𝑡𝑡 -3.743* (2.033) 𝐹𝐹𝐹𝐹𝐹𝐹𝐹𝐹𝐹𝐹 𝑒𝑒𝑒𝑒𝑒𝑒𝑒𝑒𝑒𝑒𝑒𝑒𝑒𝑒 ( 𝛼𝛼𝑠𝑠𝑠𝑠, 𝛼𝛼𝑐𝑐𝑐𝑐, 𝛼𝛼𝑠𝑠𝑠𝑠) Yes Yes Obs 1,651 1,651 R-squared 0.559 0.562 Yield curve flattening reduces contribution from reallocation F-test cannot reject hypothesis: sum of short and long interest rate coefficients is zero Weaker significance than labour regressions (measurement issues) Demand (real sales growth) less influence on capital allocation Baseline results – capital reallocation
- 14. Just the signalling properties of the yield curve? 14 Flatter yield curve forecasts slower growth (eg Estrella and Hardouvelis (1991) and productivity tends to be procyclical BUT: fluctuations in productivity driven by common component Signalling fluctuations in yield curve slope affect average productivity across all sectors – not reallocation between sectors Correlation between forecast errors in the yield curve slope and TFP growth in the following two years is close to zero
- 15. Conclusions Slope of the yield curve closely linked to contribution of reallocation to productivity growth Surprise flattening of the yield curve reduces contribution of reallocation As monetary policy operations have changed with short- rates hitting the ZLB May suggest some state dependency Two possible interpretations Pre-crisis, tighter credit conditions reduce growth of high productivity entrants but post-crisis, loose conditions slow exit of unproductive firms. Higher productivity growth sectors hold more short term debt. 15
- 16. EXTRA SLIDES 16
- 17. Dependent variable: Relative employment growth Variable (1) (2) (3) (4) Log(emp share)cst-1 -0.0955*** -0.0924*** -0.0955*** -0.0951*** (0.0187) (0.0185) (0.0186) (0.0185) Real sales growthcst 0.124*** 0.127*** 0.127*** 0.127*** (0.0385) (0.0392) (0.0382) (0.0381) Real sales growthcst-1 0.0657*** 0.0653*** 0.0656*** 0.0679*** (0.0214) (0.0211) (0.0209) (0.021) Size-weighted TFP growthcst 0.137 0.163 0.161 0.189* (0.111) (0.117) (0.111) (0.112) FESR,ct-1×Size-weighted TFP growthcst 7.478** 10.35*** (2.938) (3.321) FESR,ct-2×Size-weighted TFP growthcst 2.233 3.032 (2.478) (3.207) FELR,ct-1×Size-weighted TFP growthcst -3.581 -8.829** (2.816) (3.762) FELR,ct-2×Size-weighted TFP growthcst 2.271 1.854 (2.899) (3.365) FESLOPE,ct-1×Size-weighted TFP growthcst -9.158*** (2.822) FESLOPE,ct-2×Size-weighted TFP growthcst -2.507 (2.587) Fixed effects (𝛼𝛼𝑠𝑠𝑠𝑠 , 𝛼𝛼𝑐𝑐𝑐𝑐 , 𝛼𝛼𝑠𝑠𝑠𝑠) Yes Yes Yes Yes Observations 1,651 1,651 1,651 1,651 R-squared 0.559 0.556 0.562 0.562 17 Short-rates: Monetary easing t-1 Higher growth contribution from reallocation Yield curve flattening reduces contribution from reallocation F-test cannot reject hypothesis: sum of short and long interest rate coefficients is zero Baseline results – employment reallocation
- 18. Dependent variable: Relative capital growth Variable (1) (2) (3) (4) Log(emp share)cst-1 -0.0611*** -0.0617*** -0.0613*** -0.0614*** (0.0180) (0.0185) (0.0185) (0.0184) Real sales growthcst 0.0272 0.0255 0.0259 0.0259 (0.0501) (0.0502) (0.0505) (0.0504) Real sales growthcst-1 0.0161 0.0175 0.0179 0.0172 (0.0366) (0.0354) (0.0356) (0.0365) Size-weighted TFP growthcst 0.0493 0.0401 0.0728 0.105 (0.279) (0.117) (1.264) (1.177) FESR,ct-1×Size-weighted TFP growthcst 8.071* 8.321* 6.974 (4.179) (4.344) (5.276) FESR,ct-2×Size-weighted TFP growthcst 3.640* 3.529 3.632* (2.158) (2.220) (2.188) FELR,ct-1×Size-weighted TFP growthcst -8.181* -7.996** -7.728* (4.260) (4.030) (4.262) FELR,ct-2×Size-weighted TFP growthcst -4.039 -4.784 -5.022 (4.512) (4.483) (4.616) FESLOPE,ct-1×Size-weighted TFP growthcst -7.269 (4.605) FESLOPE,ct-2×Size-weighted TFP growthcst -3.743* (2.033) Fixed effects (𝛼𝛼𝑠𝑠𝑠𝑠 , 𝛼𝛼𝑐𝑐𝑐𝑐 , 𝛼𝛼𝑠𝑠𝑠𝑠) Yes Yes Yes Yes Additional controls interacted with size-weighted TFP growth None Credit gap Credit gap Sales growth Observations 1,651 1,651 1,651 1,651 R-squared 0.559 0.556 0.562 0.562 18 Yield curve flattening reduces contribution from reallocation F-test cannot reject hypothesis: sum of short and long interest rate coefficients is zero Weaker significance than labour regressions (measurement issues) Demand (real sales growth) less influence on capital allocation Baseline results – capital reallocation