Labour Reallocation and Productivity Dynamics: Financial Causes, Real - - PowerPoint PPT Presentation

▶
labour reallocation and productivity dynamics
SMART_READER_LITE
LIVE PREVIEW

Labour Reallocation and Productivity Dynamics: Financial Causes, Real - - PowerPoint PPT Presentation

Apr 03, 2023 261 likes •521 views

Labour Reallocation and Productivity Dynamics: Financial Causes, Real Consequences 1 Enisse Kharroubi Bank for International Settlements XI th Annual Seminar on Risk, Financial Stability and Banking Sao Paulo, Brazil August 10-12, 2016 1 The

slide-1
SLIDE 1

Labour Reallocation and Productivity Dynamics: Financial Causes, Real Consequences1

Enisse Kharroubi Bank for International Settlements XIth Annual Seminar on Risk, Financial Stability and Banking Sao Paulo, Brazil August 10-12, 2016

1The views expressed here are those of the authors and do not necessarily represent the views

  • f the BIS.

Borio, Kharroubi, Upper and Zampolli () Reallocation and Productivity August 10-12, 2016 1 / 25

slide-2
SLIDE 2

Introduction

Secular stagnation in advanced economies:

I (i) Prior to the crisis, growth was only modest despite signi…cant tailwinds. I (ii) After the crisis, the recovery has been abnormally weak, despite signi…cant

improvement in credit availability.

Increased credit dependence in emerging market economies

I (i) Growth has been more resilient, but has relied extensively on high

commodity prices and wide credit extension.

I (ii) weaker commodity prices and credit may jeopardize future growth. Borio, Kharroubi, Upper and Zampolli () Reallocation and Productivity August 10-12, 2016 2 / 25

slide-3
SLIDE 3

Introduction

What we do in this paper?

I We look at the e¤ects of credit booms on growth. We provide theory and

evidence that:

I Credit booms can have negative supply side e¤ects through labor reallocation I Labour reallocations have large implications for future productivity following a

…nancial crisis

Borio, Kharroubi, Upper and Zampolli () Reallocation and Productivity August 10-12, 2016 3 / 25

slide-4
SLIDE 4

Methodology and data

Methodology

Write total output (employment) as the sum of sectoral outputs (employments): y = ∑

s

ys and l = ∑

s

ls. Denote S number of sectors and x the unweighted (simple average) for xs y/l = 1 S ∑

s

ls l/S

  • (ys/ls) = ys/ls + cov
  • ls/l; ys/ls
  • Denote αs = ys/y sector s relative output size, and xg the growth rate of x

1+ (y/l)g =

  • 1 + (ls/l)g

1 + αs (ys/ls)g | {z }

com

+ cov (ls/l)g ;

  • 1 + (ys/ls)g

αs

  • |

{z }

alloc

Borio, Kharroubi, Upper and Zampolli () Reallocation and Productivity August 10-12, 2016 4 / 25

slide-5
SLIDE 5

Methodology and data

Data

We build a dataset with:

I 1 digit sector-level data: output, employment, prices (STAN, KLEMS, GGDC,

EUROSTAT)

I Macro data: real, …nancial and policy (STAN, KLEMS, BIS, EO, IMF)

Main requirement: sector-level data should provide a partition of the economy, i.e. allow to replicate the country-level data Dataset covers advanced economies, 21 OECD countries, over 1979-2009.

Borio, Kharroubi, Upper and Zampolli () Reallocation and Productivity August 10-12, 2016 5 / 25

slide-6
SLIDE 6

Methodology and data

Speci…cation

We estimate the following regression 2 4 (y/l)g com alloc 3 5

c,t

= αc + αt + βxc,t + βllg

c,t + θf g c,t + εc,t

Dependent variable:

I labour productivity growth (comp.) in country c in year t (non-OL periods)

Independent variables:

I Country and time e¤ects (focus on within-country idiosyncratic credit booms) I Control variables (government consumption, in‡ation, trade openness, etc..) I Employment growth (decreasing marginal returns to labour) I Credit growth (average for credit to GDP growth or credit to GDP deviation

from trend)

Borio, Kharroubi, Upper and Zampolli () Reallocation and Productivity August 10-12, 2016 6 / 25

slide-7
SLIDE 7

Productivity growth during credit booms

Bivariate correlations

We start estimating the baseline speci…cation without any control variable:

  • alloc

com

  • c,t

= αc + αt + θf g

c,t + εc,t

Growth in private credit to GDP correlates negatively with the allocation component of productivity growth, not with the common component.

Borio, Kharroubi, Upper and Zampolli () Reallocation and Productivity August 10-12, 2016 7 / 25

slide-8
SLIDE 8

Productivity growth during credit booms

OLS estimates

(i.a) (ii.a) (iii.a) (i.b) (ii.b) (iii.b) Productivity Growth Common component Allocation component Productivity Growth Common component Allocation component Growth in private credit to GDP

  • 0.077**
  • 0.032
  • 0.045***

(0.0370) (0.0399) (0.0170)

Average private credit to GDP gap

  • 0.0729***
  • 0.0318
  • 0.0412*

(0.0131) (0.0549) (0.0228)

Employment growth

  • 0.372***
  • 0.514***

0.142**

  • 0.409***
  • 0.529***

0.120**

(0.080) (0.093) (0.058) (0.067) (0.094) (0.058)

Controls Yes Yes Yes Yes Yes Yes Country and time dummies Yes Yes Yes Yes Yes Yes Observations 108 108 108 108 108 108 R-squared 0.864 0.854 0.695 0.858 0.854 0.681

Productivity growth goes down during credit booms, mainly because of the allocation component. By contrast no statistically signi…cant e¤ect on the common component.

Borio, Kharroubi, Upper and Zampolli () Reallocation and Productivity August 10-12, 2016 8 / 25

slide-9
SLIDE 9

Productivity growth during credit booms

Decomposing the allocation component

During credit booms, allocation component goes down. Why?

1 Shifts in the distribution of sectoral productivity gains:

I Productivity slows-down in high employment growth sectors, and accelerates

in low employment growth sectors 2 Shifts in the distribution of sectoral employment shares:

I Employment picks-up in low-productivity gain sectors, and goes down in

high-productivity gain sectors

We test which part of the allocation component reacts negatively during credit booms.

Borio, Kharroubi, Upper and Zampolli () Reallocation and Productivity August 10-12, 2016 9 / 25

slide-10
SLIDE 10

Productivity growth during credit booms

Decomposing the allocation component

(i.a) (ii.a) (iii.a) (iv.a) (i.b) (ii.b) (iii.b) (iv.b) Allocation component due to shifts in the distribution of Allocation component due to shifts in the distribution of

Productivity Employment both Productivity Employment both

Private credit to GDP growth

  • 0.045***

0.002

  • 0.041***
  • 0.006

(0.017) (0.007) (0.016) (0.006)

Average private credit to GDP gap

  • 0.041*

0.006

  • 0.044**
  • 0.004

(0.023) (0.008) (0.021) (0.007)

Employment growth 0.142**

  • 0.049**

0.190*** 6.5e-05 0.120**

  • 0.049***

0.171***

  • 0.003

(0.058) (0.019) (0.054) (0.013) (0.058) (0.018) (0.054) (0.014)

Controls yes yes yes yes yes yes yes yes Country and time dummies yes yes yes yes yes yes yes yes Observations 108 108 108 108 108 108 108 108 R-squared 0.695 0.852 0.653 0.692 0.681 0.853 0.644 0.687

Main reason for productivity slows-down during credit booms:

I labour reallocation from high to low productivity gains sectors. Borio, Kharroubi, Upper and Zampolli () Reallocation and Productivity August 10-12, 2016 10 / 25

slide-11
SLIDE 11

Productivity growth during credit booms

Instrumenting credit booms

Do credit booms slow-down productivity or do productivity slow-downs cause credit booms? Latter is unlikely for two reasons:

I Productivity a¤ects credit through the cycle, but we focus on medium-term

developments (3-5 years)

I Credit may react to productivity in case of generalized slow-down, but

evidence in favour of labour reallocation

We test for reverse causality by instrumenting credit booms with …nancial reforms.

I No evidence that …nancial reforms are related to productivity growth. Borio, Kharroubi, Upper and Zampolli () Reallocation and Productivity August 10-12, 2016 11 / 25

slide-12
SLIDE 12

Productivity growth during credit booms

Instrumenting credit booms

(i.a) (ii.a) (iii.a) (iv.a) (i.b) (ii.b) (iii.b) (iv.b)

Productivity Growth Common component Allocation component Employment- Allocation component Productivity Growth Common component Allocation component Employment- Allocation component

Private credit to GDP growth

  • 0.137***
  • 0.055
  • 0.082***
  • 0.080***

(0.046) (0.058) (0.027) (0.031)

Average private credit to GDP gap

  • 0.188***
  • 0.081
  • 0.107***
  • 0.102***

(0.069) (0.0750) (0.036) (0.040)

Employment growth

  • 0.681***
  • 0.934***
  • 0. 253***

0.351***

  • 0.873***
  • 1.012***

0.139 0.241**

(0.158) (0.175) (0.094) (0.096) (0.187) (0.178) (0.102) (0.0947)

Controls yes yes Yes yes yes yes yes yes Country and time dummies yes yes Yes yes yes yes yes yes J-stat 3.477 2.003 1.497 0.526 1.425 1.358 0.741 0.0599

  • p. value

(0.324) (0.572) (0.683) (0.913) (0.700) (0.715) (0.863) (0.996)

LM-test 18.26 18.26 18.26 18.26 14.97 14.97 14.97 14.97

  • p. value

(0.001) (0.001) (0.001) (0.001) (0.0048) (0.0048) (0.0048) (0.0048)

Observations 102 102 102 102 102 102 102 102 R-squared 0.491 0.503 0.140 0.133 0.346 0.466 0.048 0.138

Negative relationship between productivity growth and credit booms

I robust to instrumentation I driven by labour reallocation into low productivity gains sectors Borio, Kharroubi, Upper and Zampolli () Reallocation and Productivity August 10-12, 2016 12 / 25

slide-13
SLIDE 13

The model

The general framework

OLG model with entrepreneurs living 2 periods.

I Start with a bequest from the previous generation; choose, for life time, one

project type; raise funding and invest bequest and borrowing

I Reap output and pay claims back ; save and bequeath to the next generation;

raise funding and invest savings and borrowing

I Reap output and and pay claims back; consume. Borio, Kharroubi, Upper and Zampolli () Reallocation and Productivity August 10-12, 2016 13 / 25

slide-14
SLIDE 14

The model

The general framework

Entrepreneurs choose one of 2 available positive NPV projects: j 2 fi; hg innovative vs. housing projects

I Higher return project less sensitive to changes in credit supply F Trade-o¤ between total and pledgeable return: 1 + Ai > 1 + Ah but ρh > ρi I Return for newly-born entrepreneurs decreasing in sector size.

Denoting πj,t the pro…t rate on date-t investment in type-j projects, the equilibrium number nt of newly-born entrep. choosing innovative projects satis…es the break-even condition: πi,t (nt) [πi,t+1]

β 1+β = πh,t (nt) [πh,t+1] β 1+β

Borio, Kharroubi, Upper and Zampolli () Reallocation and Productivity August 10-12, 2016 14 / 25

slide-15
SLIDE 15

The model

Borrowing constraints and entrepreneurs’ pro…ts

Entrepreneurs borrow from the RoW (at 0 IR) and can default strategically.

I Recovery rate on loans issued at date t is 1 1/σt I Maximum borrowing per unit of net wealth for type-j projects is

dj = ρj σt/

  • 1 ρj σt
  • Pro…t rates on type-j projects satisfy

πj,t (nt) = 1 + Aj (nt) 1 ρjσt and πj,t+1 = 1 + Aj 1 ρjσt+1 LR-pro…ts on housing are more sensitive to credit supply shocks σ: 0 < ∂πi,t+1 ∂σt+1 < ∂πh,t+1 ∂σt+1 Equilibrium allocation nt satisfying: πi,t (nt) [πi,t+1]

β 1+β = πh,t (nt) [πh,t+1] β 1+β

depends on future credit supply shock σt+1, i.e. credit growth.

Borio, Kharroubi, Upper and Zampolli () Reallocation and Productivity August 10-12, 2016 15 / 25

slide-16
SLIDE 16

The model

Main results

Equilibrium properties:

I The size of the innovative sector decreases with credit growth, dnt/dσt+1 < 0 I Short-run pro…ts are lower on housing than on innovative projects,

πh,t (nt) < πi,t (nt)

Intuitions:

I LR-pro…t for housing more sensitive to σ ) LR-pro…t for housing increase

relative to innovative projects ) SR-pro…t for housing must decrease relative to innovative projects ) housing sector needs to expand and innovative sector needs to shrink.

I LR-pro…t larger for housing ) SR pro…t must be lower for housing. Borio, Kharroubi, Upper and Zampolli () Reallocation and Productivity August 10-12, 2016 16 / 25

slide-17
SLIDE 17

The model

Credit booms, reallocation and growth

Productivity growth writes as gt+1 = Bt+1 Bt = 1 1 + β [ntπi,t (nt) + (1 nt) πh,t (nt)] Credit growth hurts productivity growth by reallocating entrepreneurs from the innovative sector to the housing sector. ∂gt+1 ∂σt+1

  • [πi,t (nt) πh,t (nt)]

∂nt ∂σt+1 | {z }

  • +
  • ntπ0

i,t (nt) + (1 nt) π0 h,t (nt)

  • ∂nt

∂σt+1 | {z }

+/

Borio, Kharroubi, Upper and Zampolli () Reallocation and Productivity August 10-12, 2016 17 / 25

slide-18
SLIDE 18

The model

Allocation and the dynamics of productivity

Extend the previous framework to allow for time dependence:

I Innovative projects bene…t positive knowledge externalities: SR returns for new

entrepreneurs: Ai (nt1) with A0

i (nt1) > 0

Equilibrium allocation is such that entrepreneurs should break-even: 1 + Ah (nt) 1 ρhσt =

  • 1 + Ai (nt1)

1 ρi σt πi,t+1 πh,t+1

  • β

1+β

Productivity growth writes as gt+1 = Bt+1 Bt = 1 1 + β 2 4nt + (1 nt) πi,t+1 πh,t+1

  • β

1+β

3 5 πi,t (nt1, σt)

Borio, Kharroubi, Upper and Zampolli () Reallocation and Productivity August 10-12, 2016 18 / 25

slide-19
SLIDE 19

The model

Allocation and the dynamics of productivity

A larger innovative sector size raises future productivity growth:

I Pro…ts on innovative projects depend positively on the size of the innovative

sector

I The size of the innovative sector exhibits positive time persistence

dπi,t (nt1, σt) dnt1 > 0 and dnt dnt1 > 0 ) dgt+1 dnt1 > 0 How does the credit shock a¤ect these relationships?

I With higher σt, the size of the innovative sector has a larger e¤ect on growth I With higher σt, the size of the innovative sector exhibits less time persistence

d2πi,t (nt1, σt) dσtdnt1 > 0 and d2nt dσtdnt1 < 0 ) d2gt+1 dσtdnt1 ? 0

Borio, Kharroubi, Upper and Zampolli () Reallocation and Productivity August 10-12, 2016 19 / 25

slide-20
SLIDE 20

The model

Allocation and the dynamics of productivity

When innovative projects are su¢ciently credit-insensitive, i.e. ρi is su¢ciently low, then

I A larger innovative sector raises future productivity growth I Easy (tight) credit conditions dampen (amplify) the positive e¤ect of a larger

innovative sector on future productivity growth

dgt+1 dnt1 > 0 and d2gt+1 dσtdnt1 < 0

Borio, Kharroubi, Upper and Zampolli () Reallocation and Productivity August 10-12, 2016 20 / 25

slide-21
SLIDE 21

Labor reallocation’s implications for productivity

Empirical methodology

1 We de…ne recessions as periods starting with turning points in GDP to working-age population. 2 Decompose productivity growth over the 3-year period prior to the recession: yc,t/lc,t yc,t3/lc,t3 = allocct + comct 3 Build a dummy variable -Fc,t- for recessions associated with a …nancial crisis 4 Estimate for each horizon h = f1; 2; . . . ; 8g the speci…cation: yc,t+h/lc,t+h yc,t/lc,t = αc,h + βhxc,t +

  • (1 Fc,t) θ0

h + Fc,tθ1 h

  • .allocc,t

+

  • (1 Fc,t) µ0

h + Fc,tµ0 h

  • .comc,t + εc,t,h

Borio, Kharroubi, Upper and Zampolli () Reallocation and Productivity August 10-12, 2016 21 / 25

slide-22
SLIDE 22

Labor reallocation’s implications for productivity

LP estimates

Dependent variable: Aggregate Productivity Growth

(1) (2) (3) (4) (5) (6) (7) (8) Allocation component × FC 0.885** 2.297*** 2.393*** 2.998*** 3.971*** 4.562*** 5.026*** 5.678***

(0.377) (0.612) (0.819) (1.071) (1.176) (1.250) (1.348) (1.510)

Allocation component × NFC 0.252** 0.449*** 0.360 0.303 0.431 0.606* 0.618* 0.732*

(0.103) (0.151) (0.243) (0.306) (0.300) (0.303) (0.318) (0.371)

Common component × FC 0.310 0.735** 0.886* 1.237** 1.692*** 1.974*** 2.237*** 2.453***

(0.224) (0.331) (0.441) (0.542) (0.510) (0.528) (0.569) (0.640)

Common component × NFC 0.0938 0.359** 0.362* 0.474* 0.632** 0.814*** 1.062*** 1.315***

(0.109) (0.169) (0.209) (0.265) (0.282) (0.291) (0.314) (0.362)

Controls yes yes yes yes yes yes yes yes Country dummies yes yes yes yes yes yes yes yes Observations 81 81 81 81 81 81 81 81 R-squared 0.589 0.756 0.742 0.733 0.742 0.731 0.741 0.749 H0: Alloc × FC = Alloc × NFC 0.118 0.006 0.022 0.018 0.005 0.003 0.002 0.002 H0: Alloc × FC = Com × FC 0.129 0.001 0.013 0.026 0.017 0.013 0.011 0.010

In FC-recessions, the allocation component has

I a positive and signi…cant e¤ect on the subsequent productivity path I a larger e¤ect than in non FC-recessions (1st row in F-tests) I a larger e¤ect than the common component (2nd row in F-tests) Borio, Kharroubi, Upper and Zampolli () Reallocation and Productivity August 10-12, 2016 22 / 25

slide-23
SLIDE 23

Simulating productivity

Financial crisis vs. labor allocation

Using estimation results, we simulate the path of productivity under two set

  • f assumptions

I the occurrence or the absence of a …nancial crisis I a high or a low contribution of labour reallocation to productivity growth

(75th vs. 25th percentile)

We end up with four cases:

1

Strong labour reallocation contribution and no …nancial crisis

2

Weak labour reallocation contribution and no …nancial crisis

3

Strong labour reallocation contribution and occurrence of a …nancial crisis

4

Weak labour reallocation contribution and occurrence of a …nancial crisis

Borio, Kharroubi, Upper and Zampolli () Reallocation and Productivity August 10-12, 2016 23 / 25

slide-24
SLIDE 24

Simulating the path of productivity

Financial crisis vs. labor allocation

Misallocation and the occurrence of a …nancial crisis lead to productivity stagnation.

Borio, Kharroubi, Upper and Zampolli () Reallocation and Productivity August 10-12, 2016 24 / 25

slide-25
SLIDE 25

Conclusions

Takeaways:

I Credit booms come hand-in-hand with labour reallocations hurting aggregate

productivity growth

I Labour reallocations have large implications for subsequent productivity,

especially following …nancial crisis

Possible mechanisms

I Credit booms bene…t disproportionately sectors/activities with "good"

collateral (consistent with Cecchetti and Kharroubi 2015)

I Credit booms reduce the time persistence of labour allocation (to be tested)

Next steps:

I Investigate these stylized facts at the micro-level I What role for policy? Borio, Kharroubi, Upper and Zampolli () Reallocation and Productivity August 10-12, 2016 25 / 25