UNCERTAINTY & THE DYNAMICS OF R&D Nick Bloom (Stanford, CEP & NBER) January 2007.

UNCERTAINTY & THE DYNAMICS OF R&D

Nick Bloom (Stanford, CEP & NBER)

January 2007

UNCERTAINTY APPEARS TO VARY OVER TIME

Uncertainty appears to be counter-cyclical rising by 50% to 100% in recessions (Schwert, 1989)

Uncertainty also appears to jump 100% to 200% after major economic and political shocks (Bloom, 2006)

BUT WHAT IS THE REAL-OPTIONS IMPACT OF TIME VARYING UNCERTAINTY ON R&D?

Literature notes two effects of temporary rises in uncertainty:

– “Delay effect” generating a temporary slowdown in hiring, investment and productivity as firms pause activity

– “Caution effect” reducing the responsiveness of investment and hiring to any stimulus

But what about R&D – surely the extension is obvious?

In fact it is not - different adjustment costs of knowledge capital (versus physical capital & labor) generate different RO effects

Knowledge Capital Adjustment Costs

Simulation and Intuition

Implications

Physical capital usually modeled as cumulated investment

While knowledge capital usually modeled as cumulated R&D

KNOWLEDGE CAPITAL AND PHYSICAL CAPITAL HAVE SIMILAR LAWS OF MOTION….

ttt RGG G 1)1(

ttt IKK K 1)1(

Physical capital adjustment costs typically arise from directly changing the capital stock – for example resale loss on equipment

…BUT DIFFERENT ADJUSTMENT COSTS FOR CHANGING THEIR MOTION

)()( tKK KCIC t

Knowledge capital is intangible and not (easily) bought/sold. Instead it is adjusted by changing R&D, the flow rate

But the adjustment costs for changing R&D are similar to those for changing capital – for example resale loss on R&D equipment

)()( tGG GCRC t

So knowledge capital adjustment costs CG(ΔΔGt) are an order of difference apart from physical capital adjustment costs CK(ΔKt)

This is because adjustment costs for physical capital arise from changing the stock while for knowledge from changing the flow

This distinction plays a critical role in real-options effects

STOCK PHYSICAL CAPITAL AND, FLOW KNOWLEDGE CAPITAL ADJUSTMENT COSTS

Interestingly, Christiano, Eichenbaum and Evans (2005) assume flow adjustment costs for capital CK(ΔΔKt) – that is it is costly to change investment rates due to decision making costs etc....

Knowledge Capital Adjustment Costs

Simulation and Intuition

Implications

Set up a model – in summary1:

Firms uncertain over future “business conditions”

dYt = Yt (μ + σt dZt) dZt ~ N(0,1)

Uncertainty (σt ) varies over time, following an AR(1) process

σt = σt-1 + ρσ (σ* - σt-1) + σS St dSt ~ N(0,1)

BUILD MODEL OF R&D UNDER UNCERTAINTY (1)

1 Full details in the paper, program on www.stanford.edu/~bloom

There are adjustment costs for changing R&D

– Baseline assumes these are linear: C= λ |Rt – Rt-1|

– Also show results for quadratic costs: C= λ (Rt – Rt-1)2

Assume for tractability that labor and capital costless to adjust

Can show unique, continuous, unique analytical solution exists. But need numerical methods to solve for particular parameters

BUILD MODEL OF R&D UNDER UNCERTAINTY (2)

Figure 1: Higher uncertainty makes R&D more persistent over time and less responsive to current business conditions

Business Conditions, Log (yt)

Cur

rent

R&

D,

r t

Low uncertainty,σt=5%

Medium uncertainty,σt=20%

Lagged R&D, rt-1High uncertainty,σt=50%

The R&D “caution effect”

Figures 2a and 2b: The effect of uncertainty on R&D is negative if R&D is increasing, and positive if R&D is falling

σt=5%σt=50%

σt=20%

σt=5%

σt=20%

σt=50%

Business Conditions, Log (yt) Business Conditions, Log (yt)

Cur

rent

R&

D,

r t

Cur

rent

R&

D,

r tLagged R&D, rt-1

Lagged R&D, rt-1

An R&D“delay effect”

Figure 3: With only quadratic adjustment costs there are no real options effects of uncertainty on R&D

Business Conditions, Log (yt)

Cur

rent

R&

D,

r t

Lagged R&D, rt-1

σt=5% σt=50%

σt=20%

Knowledge Capital Adjustment Costs

Simulation & Intuition

Implications

Firms will be less responsive to external stimulus – like R&D tax credits – in periods of high uncertainty

Could be empirically investigated by estimating something like:

(rt is R&D, Δyt is change in business conditions, σt is uncertainty)

with the prediction is β4<0 and β5>0

“CAUTION EFFECT” IMPLICATIONS OF UNCERTAINTY FOR R&D

1543121 tttttttt ryryr

Impact of uncertainty on R&D depends on the change in R&D

“DELAY EFFECT” IMPLICATIONS OF UNCERTAINTY

Marginal impact of uncertainty on R&D

R&D decreasing*

R&D increasing

Knowledge stock decreasing + -

Knowledge stock increasing + -

*If R&D rates depreciate at rate δR, then condition is rt<(1- δR)rt-1

Impact of uncertainty on investment depends on change in capital

Marginal impact of uncertainty on investment

Investment decreasing*

Investment increasing

Capital stock decreasing* + +

Capital stock increasing - -

*After controlling for depreciation, Kt<(1- δK)Kt-1

“DELAY EFFECT” IMPLICATIONS OF UNCERTAINTY ON INVESTMENT (FOR COMPARISON TO R&D)

Higher uncertainty will tend to:

R&D– Reduce R&D when R&D is rising – i.e. coming out of a recession and start of a

boom– Reduce changes in R&D, inducing a more lagged response

Investment (or hiring)– Reduce investment when capital is rising – i.e. during a boom– Reduce level of response, flattening fluctuations

COMBINED “DELAY EFFECT” IMPLICATIONS

CONCLUSIONS

Uncertainty does appear to change strongly over time

This will induce important “delay” and “cautionary” real-options effects on investment and hiring

Will induce somewhat different “delay” and “cautionary” effects on R&D due to flow (rather than stock) adjustment costs

– slowing responsiveness– increasing persistence

Hope that future empirical research will test these predictions

BACK-UP

010

20

30

40

50

1960 1965 1970 1975 1980 1985 1990 1995 2000 2005Year

US stock market volatility 1962 to 2005

OPEC II

Monetary turning point

Black Monday*

Gulf War I

Asian Crisis

Russia & LTCM

9/11Enron

Gulf War II

Implied VolatilityActual Volatility

Afghanistan

JFK assassinated

Cuban missile

crisis

Cambodia,Kent State

OPEC I

Franklin National

An

nu

aliz

ed s

tan

dar

d d

evia

tio

n (

%)

Source: Bloom (2006)