This document summarizes discussions from a conference on financial factors and productivity. It discusses three papers that found credit frictions have significant negative impacts on firm productivity. More bank forbearance during crises leads to less "cleansing" of unproductive firms but also less growth after crises. While bank recapitalization aims to strengthen banks, it can also act like forbearance. The document discusses reconciling these results and their implications for policies that preserve credit supply.

1. Discussion on
“Financial frictions
and within firm
performance”
session
BIS-IMF-OECD conference
“Weak productivity: the role of
financial factors and policies”
Paris, 10/11 January 2018
Filippos Petroulakis
DG Economics
Supply Side, Surveillance and Labour
The views presented here are mine
alone and may not necessarily
reflect those of the European
Central Bank, or the Eurosystem

2. Rubric
www.ecb.europa.eu ©
Putting it all together….
• Romain’s and Nicola’s papers have very similar focus and results.
– Slight deviation in mechanism: Romain focuses more on firm-level balance sheet
weaknesses, Nicola’s paper on credit supply.
– Both find credit frictions have important long lasting productivity consequences.
– Policy implications? Policies that preserve credit supply appear obvious.
• Vahid’s paper:
– More forbearance led to less cleansing: fewer losses during crisis, lower growth after
crisis.
– Bank recapitalisation acted as forbearance.
• Can we reconcile these results?
– A priori all possible in theory.
– Important difference 1: US vs Italy vs most of Europe + JP + KR.
• Banks much less important in US than Europe. Deep and diverse financial system able to
marshal private funds to support weak but salvageable banks.
• Unclear whether this would have been possible in vulnerable EA countries.
– Important difference 2: aggregate vs within-firm growth perspective.
• Is it possible that what is good for one is not good for the other?
2

3. Rubric
www.ecb.europa.eu ©
Cleansing Effect of Banking Crises
• Central idea of the creative destruction hypothesis.
– Reallocation may be stronger in recessions because OC is lower.
– Unproductive firms affected more by the shock and exit/downsize.
– Intuitively appealing but unclear empirical reality, esp. regarding labour flows.
– Caballero and Hammour (1994), Darby-Haltiwanger-Plant (1986), Hall (2005), Shimer
(2005, 2012), Elsby et al. (2009, 2013), Foster et al. (2016), Bartelsman et al. (2016).
• This paper: focus on the cleansing effect of banking crises
– Important and timely paper in its own right.
– Provides a novel approach to the cleansing literature.
• Can potentially tell us why there is less cleansing than expected, by going beyond aggregates.
• Main result: more forbearance leads to less destruction/cleansing during
crisis, but less job creation after.
– Consistent with the idea of trading-off temporary losses for future gains.
– Also consistent with cleansing hypothesis.
– Also finds that bank recapitalisation acts like forbearance.
– While the forbearance story makes sense, the recapitalisation result is more complex.
3

4. Rubric
www.ecb.europa.eu ©
Cleansing Effect of Banking Crises
• Spain vs Italy (and the importance of recaps).
– Spain engaged in swift recap & restructuring, growth and NPL reduction came earlier.
– Did growth bring NPLs down, or did lower NPLs promote growth?
• Find that TARP acted like forbearance.
– Are the forbearance and recap stories consistent?
• Forbearance means weak banks stay alive, but TARP should have strengthened them – was TARP
recap insufficient?.
– Cannot account for general equilibrium consequences of TARP.
• E.g. preventing firesales.
– Cannot distinguish between credit support and forbearance.
• Berger et al (2017a,b) & Li (2013) find TARP led to more credit supply and reduced systemic risk.
– Half of TARP went to national banks/IBs – can these be apportioned to states?
• What role for policy? Are these results generalizable to other countries?
– Authors find that bank restructuring (i.e. failures) resulted in higher growth after the crisis.
• Private investors/other banks were able to step-in and prevent credit collapse?
• Would be better to show this holds after removing systemic banks.
– US financial system deep and diversified – banks not as important as elsewhere.
• Cannot necessarily replicate this elsewhere.
4

5. Rubric
www.ecb.europa.eu ©
Cleansing Effect of Banking Crises – Other issues
• Can the strategy provide sufficient variation?
– Well-documented consolidation of the US banking system pre-crisis.
– Interstate bank branching – national banks may dominate local markets.
• Remaining variation may only be relevant for a small share of firms.
• Show market-share of local banks by MSA.
– Similarly for the IV – distance from DC may not be relevant for large banks.
• Would be good to show more checks about differential shocks.
– Control for pre-crisis construction employment/exposure to shale oil shock.
– Control for policy e.g. differential exposure to stimulus.
• Drop state capitals and DC itself – rising state employment?
– Alter timing conventions: 2007 and 2010 were growth years.
• Other comments:
– Unclear from paper whether local Feds influence regulatory decisions.
• If so, maybe distance from local Fed more important?
– Paper argues that fewer bank failures than 80s imply less cleansing – but numbers of
banks less than half pre-GFC than pre-S&L.
5

6. Rubric
www.ecb.europa.eu ©
Credit Supply and Productivity Growth
• Important contribution using a fantastic dataset.
– Universe of lender-borrower relationships allows for clean removal of demand shocks, and
hence isolates impact of supply shocks.
– Authors exploit dataset to recover a number of interesting results.
• Big role for credit frictions – credit crunch of 2007-09 could account
for 12.5-30% of TFP fall.
– Effect persistent so could be even more important.
• Innovative estimation of the production function.
– Accounting for credit frictions in determining inputs and hence TFP.
• Concave relationship between credit and productivity growth.
– Productivity losses from negative shocks larger than gains from positive shocks.
– Crucial point: stability of credit matters, positive shocks cannot undo negative ones.
6

7. Rubric
www.ecb.europa.eu ©
Credit Supply and Productivity Growth
• Theoretically ambiguity about the effect of credit on productivity.
– Credit booms lead to contractions by shifting resources to non-tradables (Benigno et al.
’15) & unconstrained low productivity firms (Gopinath et al. ’17).
– Presumably not the case for R&D intensive (tradable?) constrained firms.
– Can consider differential analysis by sector? Who drives the result?
• Using cutting-edge approach to estimate the effect of credit shocks
from existing relationships.
– Elasticity of total credit granted wrt credit supply from existing relationships is ≈1.
– Strong evidence of sticky bank-firm relationships; v. important result on its own.
– Also find no assortative matching between firms and banks -> preventable.
• Importance of steady credit supply.
– Implicit support for prompt policy action.
– But why? Competition dynamics (foreign or domestic)? Continuous learning?
• Run this model at a sectoral level.
7

8. Rubric
www.ecb.europa.eu ©
Credit Supply and Productivity Growth – Other
issues
• Argue that productivity dispersion may be due to credit constraints
– Gopinath et al show that credit in Spain pre-crisis flew to unconstrained but not
necessarily productive firms.
• Data cover 1997-2013; would be useful to conduct time-varying
analysis.
– Concavity of φ could be changing.
• Dig deeper: the Italian slowdown of the past 2 decades is a big
policy topic – can this framework contribute to the debate?
– Credit frictions cannot be relevant for the duration of the slowdown.
– The various exercises (patents, exports) are useful and interesting, but cannot tell us
why Italy went from circa 90% of US GDP/capita to 60% in 20 years.
• Effect on output stronger than on inputs -> TFP affected by credit
– Why? ICT/innovation important – what about labour?
• Young firms affected more?
8

9. Rubric
www.ecb.europa.eu ©
Financial Frictions and the Great Productivity
Slowdown
• Very similar focus with Manaresi & Pierri.
– Interest is on within-firm growth as well.
– Cross-country perspective, “event-study” approach.
• Very similar results.
– “TFP hysteresis” – enduring TFP gap for affected firms.
– Firms that entered crisis with balance sheet vulnerabilities experienced much deeper
reductions in TFP.
– Effect magnified if accompanied by a credit shock.
• More emphasis on firm balance-sheet channel.
– Manaresi & Pierri focus instead on credit crunch, controlling for firm FEs.
– Pre-crisis trends not different between the two groups of firms.
• Similar finding to Manaresi & Pierri; difference in pre-crisis fundamentals not important.
– Corollary: unconstrained firms with low idiosyncratic productivity may survive because
constrained firms with high idiosyncratic productivity cannot borrow.
• Underlines the interaction between allocative efficiency and within-firm growth.
9

10. Rubric
www.ecb.europa.eu ©
Financial Frictions and the Great Productivity
Slowdown
• Firm-level balance sheet weakness.
– Provides evidence for the balance-sheet recession view.
• Roll-over risk and debt-overhang -> constrained borrowers deleverage.
– This is an important contribution of the paper: excessive indebtedness can lead to long-
lasting consequences.
• Indebtedness did not matter for the dot-com recession.
– Is the corollary that financial sector leverage is the root cause of the problem?
– What is the appropriate policy response?
• Would emergency credit supply be sufficient?
– Eggertson-Krugman: need fiscal policy to take up spending shortfall.
• Major channel identified: intangible investment.
– Intangible investment unpledgeable; hence more sensitive to financial frictions.
– But physical capital also affected. Significant investment shortfall in AE (WEO April ’15).
– Fast depreciating or illiquid physical capital may also be unpledgeable.
– Insolvency framework amplifier?
10

11. Rubric
www.ecb.europa.eu ©
Financial Frictions and the Great Productivity
Slowdown – other issues
• Bank-level shocks: identified through credit ratings -> clean but
severely limits sample.
– Solution: Andrews & Petroulakis use 1st PC from PCA on 7 BS and IS variables – Storz
et al (‘17) show high correlation with credit rating.
• Would be useful to shore-up identification by showing that pre-
crisis rollover-risk unrelated to ΔCDS.
– Would not change main message but would illustrate that can disentangle between firm
and bank vulnerabilities.
• Flesh-out policy implications.
– Briefly touches upon macro-prudential issues but discussion on appropriate policy
response also useful.
– Can we see difference of policies across countries in the data?
– Insolvency regime? Low barriers to restructuring could alleviate indebtedness issues.
11