Popular in Economics

Central Monetary Authorities

Macroeconomic modelling of CBDC: a critical review

European Central Bank Ulrich Bindseil

Central Monetary Authorities

Financing Private Credit

Federal Reserve Bank of New York Nina Boyarchenko

Economics

Failing Banks

Massachusetts Institute of Technology Emil Verner

Economic History

Golden Fetters

UC Berkeley Barry Eichengreen

Economic Inequalities

Visions of Inequality

City University of New York Branko Milanovic

Economics

School Finance Equalization Increases Intergenerational Mobility

Yale University Barbara Biasi

Economics

Olivier Blanchard on U.S. Pandemic-Era Inflation

Peterson Institute for International Economics Olivier Blanchard

Geopolitical Oil Price Risk and Economic Fluctuations

View more insights in Social Sciences - Economics - Central Monetary Authorities - Energy Economics

View more figures from Federal Reserve Bank of Dallas

View the page of Lutz Kilian

Read the Study

Next in Central Monetary Authorities

xtevent: Estimation and Visualization in the Linear Panel Event-Study Design

Simon Freyaldenhoven | Federal Reserve Bank of Philadelphia

Time-varying Persistence of House Price Growth

C.Y. Choi | Federal Reserve Bank of Dallas

What are the general equilibrium effects of time-varying geopolitical risk in oil markets?

The views presented here are those of the authors and do not represent the views of the Federal Reserve Bank of Dallas or the Federal Reserve System.

Image courtesy of interviewee. November 25, 2024

Log-in or Sign-up to Faculti
Currently viewing this subject insight as a guest. You have insight(s) remaining for this month. Login to view 8000+ figures on the platform.

Generating Workspace...

Export as PDF

This video discusses the complex relationship between oil prices, economic recessions, and uncertainties affecting them, particularly focusing on historical events from the 1980s onward. It begins by exploring the hypothesis that higher oil prices may cause recessions, established from previous trends, and then questions the linearity of this relationship based on the lack of economic expansion following a significant drop in oil prices in 1986.

Two main interpretations emerge: the influence of non-linear relationships between oil prices and the economy, particularly through oil price uncertainty, and the relatively minor role oil prices might play compared to broader economic factors like monetary policy changes initiated by Paul Volcker in 1979.

The discussion moves into the mechanics of how oil price uncertainty can lead to reduced investments and consumption, invoking models of irreversible investments and emphasizing that while oil-producing sectors may be notably affected, many other sectors exhibit less dependence on oil prices.

This video also elaborates on the implications of geopolitical events versus macroeconomic uncertainties, noting that large-scale crises (like financial crises) can create significant disturbances in oil price uncertainty without necessarily being linked to geopolitical risks. It emphasizes that caution is needed when interpreting these relationships due to the endogeneity of oil price uncertainty with respect to macroeconomic conditions.

Additionally, the discussion introduces a model that incorporates both oil price uncertainty and macroeconomic uncertainty, emphasizing how shocks to either can affect overall economic output significantly. The model aims to jointly determine the dynamics between oil price levels, their uncertainties, and broader macroeconomic responses.

This video concludes with remarks on practical implications for policymakers, suggesting the necessity of developing effective responses to shifts in geopolitical risk and the importance of understanding the potential economic impact of oil price fluctuations, particularly in light of recent global events.

Language:

Export as PDF

Export as PDF
Here are five key points from the video:

1. **Historical Convideo of Oil Prices and Recessions**: The video discusses the historical relationship between oil prices and economic recessions, particularly in reference to the 1970s oil price shocks and their perceived link to subsequent recessions. Notably, the 1986 drop in oil prices following the disintegration of OPEC, which did not lead to a significant economic expansion, suggests that the relationship between oil prices and the economy may not be as straightforward as once thought.

2. **Linear vs. Non-Linear Perspectives**: The author contrasts two interpretations of the relationship between oil prices and economic performance: one views the relationship as linear (wherein both increases and decreases in oil prices have predictable effects on the economy), while the other proposes a non-linear perspective that emphasizes oil price uncertainty as a significant factor influencing economic behavior, particularly investment.

3. **Role of Oil Price and Macroeconomic Uncertainty**: The video highlights that rising oil price uncertainty can inhibit investment decisions among firms due to concerns about future cash flows and economic conditions. This uncertainty can lead to reduced consumption and slower economic growth, framing oil price uncertainty as being more significant than previously acknowledged.

4. **Endogeneity of Oil Price Uncertainty**: The author argues that oil price uncertainty is influenced by broader macroeconomic conditions and thus should not be treated as an exogenous variable. This challenges existing empirical models that link oil price uncertainty to macroeconomic outcomes, suggesting they might underestimate the true impact of such uncertainty because they do not account for its interdependence with macroeconomic variables.

5. **Implications for Economic Models and Policy**: The video concludes that policymakers, such as central bankers, should carefully consider how geopolitical oil price uncertainty affects the macroeconomy. The findings imply that historical geopolitical oil price risks have not significantly driven macroeconomic volatility, with macroeconomic uncertainty being a larger contributing factor. This analysis necessitates reevaluating how empirical models are constructed and how responses to geopolitical risks are formulated by policymakers.

Export as PDF
Here are some key quotes extracted from the video:

1. "In the decade before that, in '73 and '79, we saw big oil price increases, which were followed by major recessions."

2. "If that's true, you would also expect that lower oil prices would cause expansions."

3. "What we learned from the '86 episode is that really in '79, oil prices weren't the major driver of a recession. It must have been something else."

4. "While the oil price went up, that made people worse off...the economy slows down."

5. "As oil prices came down, people were better off. They had more money to spend...that stimulus was offset by...all price uncertainty."

6. "This loose evidence that really I think has convinced macro [economists] that oil price uncertainty creates recessions."

7. "If oil price uncertainty goes up, you might say, 'Well, this used to make sense before price uncertainty went up. Now I'm not so sure, let me hold off.'"

8. "It becomes a little bit murky what's actually going on...which one is driving oil price uncertainty."

9. "There are surges which clearly have nothing to do with geopolitical events."

10. "It's not just the uncertainty, it's also that you expect the economy to slow down in the presence of downside risk."

11. "Downside risk is very subjective. It's very hard to get agreement on...volatility is very easy to measure."

12. "Most people say, well, volatility captures uncertainty...that is not the way most economic agents think about this."

13. "If you had a 40 percentage point increase in the probability of a major geopolitical oil disaster...you would still only expect output to be lower by at most 0.25%."

14. "More than half of the observed oil price uncertainty is macroeconomic uncertainty."

15. "The anticipation of a growth disaster...expected to lower GDP by 0.75 percentage points."

16. "Our model shows that the level shocks feed into oil price uncertainty."

These quotes highlight key insights regarding the relationship between oil prices, economic recessions, uncertainty, and investment behavior over different periods. They collectively clarify how the discussions integrate various economic theories and empirical observations to challenge previous assumptions about oil price impact on the economy.

X-TIMESTAMP-MAP=LOCAL:00:00:00.000,MPEGTS:90000

1
00:00:00.000 --> 00:00:02.525
Understand this perspective, I think you have to go back

2
00:00:02.525 --> 00:00:03.845
to the early 1980s.

3
00:00:04.545 --> 00:00:07.565
So in the decade before that, in 73

4
00:00:07.565 --> 00:00:11.365
and 79, we saw big oil price increases, which were followed

5
00:00:11.465 --> 00:00:12.565
by major recessions.

6
00:00:13.185 --> 00:00:15.325
And so a national conjecture was that, well,

7
00:00:15.325 --> 00:00:17.285
higher oil prices caused recessions.

8
00:00:18.175 --> 00:00:20.355
If that's true, you would also expect

9
00:00:20.465 --> 00:00:23.235
that lower oil prices would cause expansions.

10
00:00:24.115 --> 00:00:26.695
And that premise was put to a test in 1986

11
00:00:26.725 --> 00:00:30.135
because at that point the OPEC cartel dis integrated

12
00:00:30.515 --> 00:00:33.295
and the price of oil fell about as much

13
00:00:33.295 --> 00:00:36.255
as it had gone up in 79 and equally sharply.

14
00:00:36.795 --> 00:00:38.695
But there was no major expansion.

15
00:00:39.705 --> 00:00:41.245
So how do you make sense of that?

16
00:00:41.395 --> 00:00:45.085
Well, one interpretation of this evidence is to say, well,

17
00:00:45.085 --> 00:00:46.525
this relationship between oil prices

18
00:00:46.525 --> 00:00:48.325
and the economy is inherently linear.

19
00:00:48.865 --> 00:00:51.565
And so what we learned from the 86 episode is

20
00:00:51.875 --> 00:00:53.125
that really in 79,

21
00:00:53.265 --> 00:00:56.445
oil prices weren't the major driver of a recession.

22
00:00:56.545 --> 00:00:59.365
It must have been something else. What could that have been?

23
00:00:59.375 --> 00:01:02.285
While a natural conjecture would be maybe the shift in

24
00:01:02.485 --> 00:01:03.845
monetary policy under Paul Voker,

25
00:01:03.845 --> 00:01:05.565
which also took place in 1979.

26
00:01:06.665 --> 00:01:09.005
Now that leads you to conclude that oil prices are not

27
00:01:09.075 --> 00:01:11.565
that important, at least not relative to what people think,

28
00:01:11.985 --> 00:01:13.165
but there's another interpretation

29
00:01:13.625 --> 00:01:16.565
and that involves a non-linear relationship

30
00:01:16.605 --> 00:01:18.405
between oil prices and the economy.

31
00:01:18.835 --> 00:01:20.445
That involves oil price uncertainty.

32
00:01:20.795 --> 00:01:23.005
That interpretation In 79,

33
00:01:23.005 --> 00:01:24.845
what happened is while the oil price went up,

34
00:01:25.195 --> 00:01:26.525
that made people worse off.

35
00:01:26.545 --> 00:01:27.765
You have less money to spend

36
00:01:28.265 --> 00:01:30.965
and so your consumption goes down, the economy slows down.

37
00:01:30.985 --> 00:01:33.405
And that was made worse by an increase,

38
00:01:33.765 --> 00:01:35.925
a simultaneous increase in all price uncertainty,

39
00:01:36.375 --> 00:01:37.445
which reduced investment

40
00:01:37.545 --> 00:01:39.765
and made the CE recession that much worse.

41
00:01:40.935 --> 00:01:42.355
So what about 86?

42
00:01:42.865 --> 00:01:45.395
Well, clearly in 86, as the price of oil came down,

43
00:01:45.495 --> 00:01:46.595
people were better off.

44
00:01:46.595 --> 00:01:47.675
They had more money to spend.

45
00:01:47.855 --> 00:01:51.235
So there was a stimulus, but that stimulus was offset

46
00:01:51.695 --> 00:01:53.915
by the fact that as all prices came down,

47
00:01:54.015 --> 00:01:57.155
all present certainty went up, which then reduced investment

48
00:01:57.495 --> 00:01:59.475
and the reduction in investment offset

49
00:01:59.475 --> 00:02:00.515
the increase in consumption.

50
00:02:00.815 --> 00:02:02.995
So you ended up with not much of an expansion.

51
00:02:03.815 --> 00:02:06.075
And it's this loose evidence

52
00:02:06.075 --> 00:02:07.835
that really I think has convinced macro,

53
00:02:07.945 --> 00:02:10.235
many macro economists that there is something to this story

54
00:02:10.385 --> 00:02:12.595
that all person certainty creates recessions.

55
00:02:13.315 --> 00:02:15.655
Now economists of course also like to look at models.

56
00:02:15.655 --> 00:02:18.895
There has to be a story why, this would be recessionary.

57
00:02:19.195 --> 00:02:20.655
And they like to look at models

58
00:02:20.835 --> 00:02:23.535
of irreversible investment by firms.

59
00:02:24.635 --> 00:02:28.335
So the idea is that a firm decides whether

60
00:02:28.335 --> 00:02:29.455
to make an investment or not.

61
00:02:29.645 --> 00:02:31.495
They look at the net present value

62
00:02:31.495 --> 00:02:32.855
of the investment to decide that.

63
00:02:33.445 --> 00:02:34.935
That depends on the cash flow.

64
00:02:34.955 --> 00:02:37.375
And the cash flow depends on the prices

65
00:02:37.475 --> 00:02:38.655
of inputs and outputs.

66
00:02:39.115 --> 00:02:42.255
If those prices are sensitive to the price of oil,

67
00:02:42.645 --> 00:02:45.175
then if oil price uncertainty goes up, you might say, well,

68
00:02:45.205 --> 00:02:47.935
this used to make sense before price uncertainty went up.

69
00:02:48.035 --> 00:02:50.335
Now I'm not so sure, let me hold off.

70
00:02:50.685 --> 00:02:52.135
That means investment if

71
00:02:52.135 --> 00:02:53.855
that happens at the aggregate investment drops

72
00:02:54.075 --> 00:02:55.735
and you get a recession, right?

73
00:02:55.755 --> 00:02:58.655
So that's a very natural story. There are

74
00:02:59.205 --> 00:03:00.525
A number of problems with that story.

75
00:03:00.865 --> 00:03:03.685
One is it makes perfect sense if you're modeling oil

76
00:03:03.925 --> 00:03:06.645
producers or if you're modeling the petrochemical industry

77
00:03:06.895 --> 00:03:08.565
where the oil price is clearly very

78
00:03:08.565 --> 00:03:09.605
important for the cash flow.

79
00:03:10.305 --> 00:03:12.725
But most sectors of the economy aren't

80
00:03:12.725 --> 00:03:13.845
that exposed to oil, right?

81
00:03:13.845 --> 00:03:16.645
If you're talking about a computer manufacturer,

82
00:03:17.165 --> 00:03:19.565
a textile company, a service company,

83
00:03:20.015 --> 00:03:22.805
their cashflow isn't very dependent on the price of oil.

84
00:03:22.805 --> 00:03:24.885
So why should there be a big response in the aggregate?

85
00:03:24.885 --> 00:03:26.605
Because that's what most of the economy is,

86
00:03:27.735 --> 00:03:29.225
even when you have a sector

87
00:03:29.295 --> 00:03:31.185
that is exposed to oil price risk.

88
00:03:31.515 --> 00:03:33.065
Think of airlines, right?

89
00:03:33.325 --> 00:03:36.745
An airline company deciding whether to buy a new aircraft

90
00:03:36.765 --> 00:03:38.785
that's supposed to be in service for the next 20 years,

91
00:03:39.805 --> 00:03:42.095
that involves a very long,

92
00:03:42.095 --> 00:03:43.735
investment horizon, project horizon.

93
00:03:44.495 --> 00:03:47.515
Here's the problem. Suppose oil price uncertainty.

94
00:03:47.725 --> 00:03:51.025
Let's say oil price volatility goes up now, well,

95
00:03:51.025 --> 00:03:52.705
that means it's probably going to be high

96
00:03:52.705 --> 00:03:54.905
for the next couple of months, maybe quarters.

97
00:03:55.405 --> 00:03:58.225
But after that, basically our expectation of

98
00:03:58.225 --> 00:04:00.585
that volatility goes back to the unconditional variance,

99
00:04:00.675 --> 00:04:03.305
which is a constant, which means for the most part,

100
00:04:03.325 --> 00:04:06.025
not much is changing in that cash flow, right?

101
00:04:06.445 --> 00:04:08.465
And that means you wouldn't expect a big reaction

102
00:04:08.465 --> 00:04:09.705
either even in that case.

103
00:04:10.285 --> 00:04:13.745
And so that kind of says it's not clear whether

104
00:04:13.745 --> 00:04:15.705
that's the right way to think about this problem.

105
00:04:16.795 --> 00:04:19.215
In any case, these are very qualitative arguments.

106
00:04:19.215 --> 00:04:21.215
If you want to quantify this, you actually have

107
00:04:21.215 --> 00:04:22.775
to quantify oil press uncertainty.

108
00:04:22.775 --> 00:04:25.815
That's a non-trivial task. So we have to talk about that.

109
00:04:26.565 --> 00:04:29.065
But there is another issue, which is that

110
00:04:29.725 --> 00:04:32.505
oil press uncertainty is endogenous

111
00:04:32.505 --> 00:04:33.905
with respect to the macro economy.

112
00:04:34.285 --> 00:04:36.665
And so we really need to be careful not

113
00:04:36.665 --> 00:04:39.665
to confuse oil price uncertainty with macro uncertainty

114
00:04:39.665 --> 00:04:42.585
because an increase in macro uncertainty would affect oil

115
00:04:42.585 --> 00:04:43.785
prices in your cash flow.

116
00:04:43.975 --> 00:04:46.345
Yeah, it would potentially have a very big effect on

117
00:04:46.455 --> 00:04:47.505
your net present value.

118
00:04:47.815 --> 00:04:50.305
Even though oil prices uncertainty by itself, if it,

119
00:04:50.305 --> 00:04:51.465
if it were exogenous, would not.

120
00:04:51.845 --> 00:04:55.065
And so when people try to link those all press uncertainty

121
00:04:55.085 --> 00:04:56.385
to cash flows,

122
00:04:56.405 --> 00:04:58.465
it becomes a little bit murky what's actually going on.

123
00:04:58.695 --> 00:05:01.465
Many people intuitively think of uncertainty

124
00:05:01.685 --> 00:05:04.545
as variability in the change in the price of oil.

125
00:05:04.545 --> 00:05:05.625
They're saying, well, the,

126
00:05:05.765 --> 00:05:07.625
the variability was pretty low in this period.

127
00:05:07.625 --> 00:05:08.745
It's higher in that period.

128
00:05:09.085 --> 00:05:10.625
So uncertainty has gone up

129
00:05:10.805 --> 00:05:13.385
or they run rolling window regressions to see

130
00:05:13.485 --> 00:05:15.585
how the variability has increased.

131
00:05:16.015 --> 00:05:17.625
That is not what uncertainty is about.

132
00:05:18.295 --> 00:05:21.505
Uncertainty intuitively is lack of predictability

133
00:05:21.925 --> 00:05:24.665
of the oil price because if you could perfectly predict the

134
00:05:24.665 --> 00:05:26.745
oil price, there would be no uncertainty, right?

135
00:05:27.025 --> 00:05:29.195
Decision making would be easy. That

136
00:05:29.195 --> 00:05:30.555
is not the world we live in.

137
00:05:31.015 --> 00:05:34.795
The question is how much is the oil price predictable given

138
00:05:34.825 --> 00:05:39.155
what we know at each point in time that can be measured?

139
00:05:39.325 --> 00:05:43.235
There is some nice work by Dorado et al in the a ER in 2015

140
00:05:43.545 --> 00:05:45.875
that spells out how to measure uncertainty in general.

141
00:05:46.295 --> 00:05:47.955
We apply that to the price of oil

142
00:05:48.175 --> 00:05:51.355
and we come up with an index of one quarter ahead.

143
00:05:51.615 --> 00:05:53.075
Oil price uncertainty. This is

144
00:05:53.075 --> 00:05:54.435
for the inflation adjusted price

145
00:05:54.575 --> 00:05:59.305
or global price of when you compare this index

146
00:05:59.405 --> 00:06:02.385
to the ofs, which is an applied volatility measure

147
00:06:02.385 --> 00:06:04.425
of oil price uncertainty used in financial markets,

148
00:06:04.895 --> 00:06:06.665
they're actually fairly highly correlated.

149
00:06:07.165 --> 00:06:09.545
The downside of the OS is it isn't available

150
00:06:09.725 --> 00:06:10.825
for a very long time period.

151
00:06:11.325 --> 00:06:13.145
And so if you're interested in the eighties

152
00:06:13.145 --> 00:06:15.945
and seventies, then you really need to go with our index.

153
00:06:16.595 --> 00:06:18.425
There are other indices people have worked with.

154
00:06:18.455 --> 00:06:21.385
They have worked, for example with text-based measures.

155
00:06:22.045 --> 00:06:24.585
So take a text-based measure of oil price uncertainty.

156
00:06:24.585 --> 00:06:26.385
You essentially go to major newspapers

157
00:06:26.765 --> 00:06:29.385
and you look up for combinations of the word oil price

158
00:06:29.385 --> 00:06:32.225
and uncertainty, and you count how many times that pops up.

159
00:06:32.735 --> 00:06:35.905
That is not a measure of lack of predictability.

160
00:06:36.285 --> 00:06:39.585
And indeed such an index would be very uncorrelated

161
00:06:39.735 --> 00:06:40.785
with what we're trying to do.

162
00:06:40.785 --> 00:06:42.745
And the same would be true for text-based measures

163
00:06:42.745 --> 00:06:44.105
of geopolitical risk, for example.

164
00:06:45.375 --> 00:06:49.275
So before we go into the details, we have to understand

165
00:06:49.345 --> 00:06:51.955
that all press uncertainty can be driven

166
00:06:53.035 --> 00:06:55.435
by geopolitical downside risk to all production,

167
00:06:55.695 --> 00:06:57.675
but it can also be driven by the macro economy

168
00:06:58.975 --> 00:07:02.025
that shows up very much when you look at the data, we find

169
00:07:02.025 --> 00:07:05.265
that ultra uncertainty surges may be associated

170
00:07:05.265 --> 00:07:06.465
with geopolitical events,

171
00:07:06.805 --> 00:07:08.985
but there are surges which clearly have nothing to do

172
00:07:08.985 --> 00:07:10.145
with geopolitical events.

173
00:07:10.845 --> 00:07:13.265
And when we do identify geopolitical events

174
00:07:13.295 --> 00:07:15.425
that shift the risk to oil production,

175
00:07:15.775 --> 00:07:17.185
sometimes they're associated

176
00:07:17.185 --> 00:07:19.625
with an increase in all press uncertainty, sometimes not.

177
00:07:20.395 --> 00:07:23.975
So the biggest spike in all press uncertainty since the

178
00:07:24.135 --> 00:07:26.735
1970s actually was in 2008, that was

179
00:07:26.735 --> 00:07:28.855
during the financial crisis that had nothing to do

180
00:07:28.855 --> 00:07:29.895
with geopolitics.

181
00:07:30.785 --> 00:07:33.095
There are events where geopolitics matters.

182
00:07:33.535 --> 00:07:35.055
A good example would be the invasion

183
00:07:35.055 --> 00:07:37.755
of Kuwait in 1990, but there are others.

184
00:07:38.015 --> 00:07:41.315
For example, the outbreak of the Iran Iraq war in 1980

185
00:07:41.585 --> 00:07:43.795
that had very little effect on our oppress circuit.

186
00:07:44.055 --> 00:07:45.475
So things are not mechanical

187
00:07:46.095 --> 00:07:49.395
and in many episodes it's not quite clear what's going on.

188
00:07:49.415 --> 00:07:52.475
For example, in 79 while you had the Iranian revolution,

189
00:07:52.655 --> 00:07:54.635
but you also had Paul Volcker, well,

190
00:07:54.635 --> 00:07:56.395
which one is driving all price uncertainty,

191
00:07:56.395 --> 00:07:58.635
which clearly went up in that episode, not so clear.

192
00:07:59.175 --> 00:08:00.635
Or you should take 2020,

193
00:08:01.255 --> 00:08:02.955
you had on the one hand the covid pandemic

194
00:08:02.975 --> 00:08:04.595
and the other hand you had the Saudi price war.

195
00:08:04.935 --> 00:08:07.475
Is that geopolitics? Is it macro? Is it some combination?

196
00:08:08.065 --> 00:08:09.995
That is something where you need modeling.

197
00:08:10.255 --> 00:08:12.035
The model is calibrated to us

198
00:08:12.135 --> 00:08:13.475
and global oil market data

199
00:08:14.235 --> 00:08:16.415
in the model we have endogenous oil prices

200
00:08:16.675 --> 00:08:18.975
and we have endogenous oil price uncertainty.

201
00:08:19.355 --> 00:08:21.975
Oil price uncertainty responds to level shocks

202
00:08:21.975 --> 00:08:25.335
and volatility shocks in, the, in the model,

203
00:08:25.795 --> 00:08:27.455
the model is highly non-linear.

204
00:08:27.555 --> 00:08:29.535
So there are eight state variables and six shocks.

205
00:08:29.535 --> 00:08:33.335
So you really require, fully non-linear solution methods

206
00:08:33.475 --> 00:08:34.895
as a beast to, to solve.

207
00:08:35.475 --> 00:08:37.255
Um, but it gives you a good fit in terms

208
00:08:37.255 --> 00:08:38.415
of first and second moments.

209
00:08:39.345 --> 00:08:43.725
The key feature of the model is that we model downside risk

210
00:08:44.625 --> 00:08:47.045
as opposed to a volatility.

211
00:08:47.315 --> 00:08:49.685
That is if oil price uncertainty

212
00:08:49.825 --> 00:08:52.205
or macro uncertainty in the model moves, it moves

213
00:08:52.205 --> 00:08:53.365
because of downside risk.

214
00:08:54.065 --> 00:08:55.395
And so you might say, why is that?

215
00:08:55.395 --> 00:08:57.915
Most people say, well, water agility captures uncertainty.

216
00:08:57.915 --> 00:09:00.115
What else do you need? The problem is

217
00:09:00.265 --> 00:09:03.515
that is not the way most economic agents think about this.

218
00:09:04.215 --> 00:09:06.005
Think of a regular household,

219
00:09:06.085 --> 00:09:08.605
a consumer in the us you drive a car,

220
00:09:08.605 --> 00:09:10.165
you're worried about the price of gasoline.

221
00:09:10.785 --> 00:09:12.035
What keeps you up at night?

222
00:09:12.785 --> 00:09:16.395
Well, the possibility that oil production in the Middle East

223
00:09:16.395 --> 00:09:18.395
or oil exports from the Middle East might shut down.

224
00:09:19.015 --> 00:09:20.195
The price of oil goes up

225
00:09:20.415 --> 00:09:22.795
and you end up paying $5 a gallon at the pump.

226
00:09:23.135 --> 00:09:24.995
That's something that might keep you up at night.

227
00:09:26.005 --> 00:09:27.585
On the other hand, if I told you, well,

228
00:09:27.915 --> 00:09:30.545
there is also a chance that there will be a new technology,

229
00:09:30.545 --> 00:09:32.425
let's call it shale oil 2.0

230
00:09:32.935 --> 00:09:34.585
that makes oil production much easier

231
00:09:35.165 --> 00:09:36.625
and the price of oil is gonna drop

232
00:09:36.625 --> 00:09:38.985
and the price of gasoline is gonna go to 50 cents a gallon.

233
00:09:39.605 --> 00:09:42.065
Are you gonna stay up at night worrying about that? No.

234
00:09:42.125 --> 00:09:43.345
If it happens, that's fine.

235
00:09:43.365 --> 00:09:45.265
If it doesn't happen, that's fine too, right?

236
00:09:45.535 --> 00:09:47.625
It's not what makes you change your behavior.

237
00:09:48.285 --> 00:09:51.105
And so we argue it's the downside risk that really matters.

238
00:09:51.445 --> 00:09:53.905
Now, why haven't people worked with downside risk?

239
00:09:54.705 --> 00:09:56.225
Downside risk is very subjective.

240
00:09:56.535 --> 00:09:57.945
It's very hard to get agreement on

241
00:09:57.945 --> 00:09:59.305
what it is on how to measure it.

242
00:09:59.435 --> 00:10:01.145
Volatility is very easy to measure.

243
00:10:01.445 --> 00:10:03.985
And so I think economists have gravitated to what volatility

244
00:10:04.045 --> 00:10:05.465
for that reason, but we don't

245
00:10:05.465 --> 00:10:06.545
think that's the right way of doing it.

246
00:10:07.285 --> 00:10:09.565
So what is the downside risk?

247
00:10:09.585 --> 00:10:12.285
In our model, we have macroeconomic disasters.

248
00:10:12.825 --> 00:10:16.285
Uh, we model those after events such as the financial crisis

249
00:10:16.865 --> 00:10:18.365
or the COVID-19 recession.

250
00:10:18.575 --> 00:10:22.085
These are events that occur maybe once every 100 years,

251
00:10:22.105 --> 00:10:26.325
an expectation, and they involve, a, a disruption of a,

252
00:10:26.405 --> 00:10:27.725
a reduction in growth that lasts

253
00:10:27.725 --> 00:10:28.965
for about two or three years.

254
00:10:30.505 --> 00:10:32.365
The oil production disasters we model

255
00:10:32.365 --> 00:10:35.485
that create downside risk to oil production are modeled,

256
00:10:35.535 --> 00:10:39.765
after major global geopolitical production disruptions in

257
00:10:39.765 --> 00:10:40.925
73, 74.

258
00:10:41.385 --> 00:10:44.365
In 79 or in 1990 for example,

259
00:10:44.365 --> 00:10:48.085
they involve a 5% drop in global oil production.

260
00:10:48.585 --> 00:10:51.925
And that happens in expectation every 15 years or so, right?

261
00:10:51.925 --> 00:10:53.005
So they're a bit more frequent.

262
00:10:54.305 --> 00:10:56.875
What we concerned within the model is less the

263
00:10:56.875 --> 00:10:58.035
realization of a disaster.

264
00:10:58.185 --> 00:11:00.075
It's pretty clear what happens when you have a disaster.

265
00:11:00.495 --> 00:11:01.955
The interesting part from our point

266
00:11:01.955 --> 00:11:05.035
of view is the anticipation of disasters.

267
00:11:05.385 --> 00:11:06.435
What do people do

268
00:11:06.435 --> 00:11:08.395
because they're worried about such a disaster

269
00:11:09.075 --> 00:11:10.115
possibly being around the corner.

270
00:11:11.315 --> 00:11:14.525
Another point about the model is how we, model

271
00:11:14.665 --> 00:11:16.165
how uncertainty triggered

272
00:11:16.165 --> 00:11:18.125
by this downside risk affects the economy.

273
00:11:18.625 --> 00:11:20.445
We basically follow what the literature has done

274
00:11:20.445 --> 00:11:23.365
before We have a precautionary savings channel

275
00:11:23.665 --> 00:11:26.925
that's saying if there is say, increased macro risk,

276
00:11:26.985 --> 00:11:29.325
you worry about your future income, what do you do?

277
00:11:29.585 --> 00:11:31.125
You increase your precautionary savings.

278
00:11:31.275 --> 00:11:32.685
That means you reduce your consumption

279
00:11:32.905 --> 00:11:34.085
so the economy slows down.

280
00:11:35.195 --> 00:11:38.855
You also have a precautionary oil inventory demand channel

281
00:11:39.075 --> 00:11:41.735
that's saying if there's increased oil price uncertainty,

282
00:11:41.995 --> 00:11:44.735
you're gonna start building stocks in anticipation

283
00:11:44.735 --> 00:11:46.895
of something bad that might happen in the future.

284
00:11:47.355 --> 00:11:49.295
And of course, if there's increased macro risk,

285
00:11:49.295 --> 00:11:50.735
you think the economy is gonna slow down,

286
00:11:50.995 --> 00:11:52.775
you start selling off your stocks in,

287
00:11:52.775 --> 00:11:55.325
In that case, reducing your oil inventories.

288
00:11:56.345 --> 00:11:58.045
The third component that people like

289
00:11:58.045 --> 00:12:01.725
to talk about is the real options theory in Bernanki 1983.

290
00:12:02.545 --> 00:12:04.885
Um, it turns out, as we discussed in the paper,

291
00:12:04.995 --> 00:12:07.125
this is not tractable in general Libya.

292
00:12:07.235 --> 00:12:09.565
It's a partially be model that cannot be generalized.

293
00:12:10.365 --> 00:12:12.015
What we do instead is something

294
00:12:12.165 --> 00:12:14.735
that generates very similar effects without

295
00:12:14.735 --> 00:12:16.055
involving rail options.

296
00:12:16.675 --> 00:12:19.895
We impose limited substitutability between oil

297
00:12:19.955 --> 00:12:21.455
and capital in production,

298
00:12:21.955 --> 00:12:24.145
which means if anything ha is happening to oil,

299
00:12:24.145 --> 00:12:26.025
that spills over to what's happening with capital.

300
00:12:26.485 --> 00:12:29.305
And so if you had higher, oil price risk,

301
00:12:29.305 --> 00:12:31.105
that would lead to a decline in

302
00:12:31.105 --> 00:12:32.385
the expected return on investment.

303
00:12:33.005 --> 00:12:34.545
And that means you're gonna invest less.

304
00:12:34.575 --> 00:12:37.745
It's gonna reduce output, it's gonna cause a recession.

305
00:12:37.955 --> 00:12:40.345
Those are the key channels, in the model

306
00:12:40.445 --> 00:12:42.025
and I think also in the literature.

307
00:12:42.045 --> 00:12:43.705
So let me start with the oil production disaster.

308
00:12:44.205 --> 00:12:45.905
So we have an increase in the probability

309
00:12:45.905 --> 00:12:48.665
of an oil production disaster that is associated

310
00:12:48.665 --> 00:12:50.225
with higher oil present certainty.

311
00:12:50.955 --> 00:12:52.845
That means if agents observe

312
00:12:52.845 --> 00:12:54.605
that higher oil present certainty, they're going

313
00:12:54.605 --> 00:12:56.165
to increase their oil storage demand

314
00:12:56.165 --> 00:12:58.125
because you wanna build inventories,

315
00:12:58.125 --> 00:12:59.365
as a precautionary measure

316
00:12:59.875 --> 00:13:02.445
that results in a persistent increase in oil inventories.

317
00:13:02.505 --> 00:13:04.725
And of course in an increase in the price of oil.

318
00:13:05.665 --> 00:13:07.445
Now the increase in the price

319
00:13:07.525 --> 00:13:11.325
of oil in turn is gonna generate a modest drop in investment

320
00:13:11.585 --> 00:13:13.645
and therefore a modest drop in output.

321
00:13:13.945 --> 00:13:15.885
And there's little response in consumption

322
00:13:16.065 --> 00:13:17.525
or in output uncertainty.

323
00:13:18.805 --> 00:13:21.825
In contrast, if you now play the game of having a,

324
00:13:22.045 --> 00:13:24.465
an increase in the probability of a map for disaster,

325
00:13:24.465 --> 00:13:26.665
essentially a growth disaster, well,

326
00:13:26.665 --> 00:13:29.625
that will give you a sharp rise in output uncertainty

327
00:13:30.005 --> 00:13:32.545
and an immediate sharp drop in investment and output

328
00:13:32.695 --> 00:13:35.385
because, it's not just the uncertainty, it's also

329
00:13:35.385 --> 00:13:37.585
that you expect, the economy

330
00:13:37.685 --> 00:13:39.785
to slow down in the presence of downside risk.

331
00:13:39.785 --> 00:13:41.985
That's kind of the difference between volatility shocks and,

332
00:13:41.985 --> 00:13:43.225
and disaster shocks.

333
00:13:43.645 --> 00:13:46.545
So investment is going to tank, output is gonna tank.

334
00:13:46.655 --> 00:13:48.705
There's gonna be little response in consumption.

335
00:13:49.165 --> 00:13:51.345
Uh, like in the oil disaster probability case.

336
00:13:51.685 --> 00:13:54.625
The key difference is that these effects tend

337
00:13:54.625 --> 00:13:56.825
to be about three times as big, right?

338
00:13:56.925 --> 00:13:59.185
So macro disasters are more potent.

339
00:13:59.255 --> 00:14:03.465
That makes a lot of sense because oil disasters are hampered

340
00:14:03.485 --> 00:14:05.585
by the low share of oil in output,

341
00:14:05.965 --> 00:14:08.425
but the macro disaster basically affects everything, right?

342
00:14:08.425 --> 00:14:10.025
So, so intuitively that makes sense.

343
00:14:10.765 --> 00:14:12.225
Now, as actual

344
00:14:12.365 --> 00:14:15.785
and expected demand for oil drops in response to this,

345
00:14:15.795 --> 00:14:17.505
macro disaster probability shock,

346
00:14:17.715 --> 00:14:20.745
there is a sharp drop in the price of oil,

347
00:14:21.125 --> 00:14:23.825
and that means inventories persistently decline.

348
00:14:24.775 --> 00:14:25.955
So at the same time,

349
00:14:25.955 --> 00:14:27.915
you get increases in oil price uncertainty

350
00:14:27.935 --> 00:14:29.035
and output uncertainty.

351
00:14:29.415 --> 00:14:33.035
And so what's the key difference apart from the magnitudes

352
00:14:33.035 --> 00:14:34.795
here, is that in the case

353
00:14:34.795 --> 00:14:36.715
of an oil production probability increase,

354
00:14:37.095 --> 00:14:39.755
you build inventories here you reduce inventories.

355
00:14:39.935 --> 00:14:41.475
So the mechanics of how this is

356
00:14:41.475 --> 00:14:43.075
translated are a little bit different.

357
00:14:43.095 --> 00:14:46.595
If you go to the early literature, I mentioned 19 19 83,

358
00:14:46.785 --> 00:14:49.635
that literature focused on partial equilibrium models

359
00:14:49.825 --> 00:14:51.930
with exogenous, that is geopolitically

360
00:14:51.930 --> 00:14:53.485
driven oil price uncertainty.

361
00:14:53.485 --> 00:14:54.685
And there are two problems with that.

362
00:14:55.105 --> 00:14:57.845
One, as I mentioned, oil price uncertainty is endogenous.

363
00:14:58.345 --> 00:15:01.165
Second, it's well documented that the implications

364
00:15:01.165 --> 00:15:03.965
of such partial equilibrium models need not hold up

365
00:15:04.025 --> 00:15:05.045
in general equilibrium.

366
00:15:05.065 --> 00:15:06.725
So it's important to look at this from a

367
00:15:06.725 --> 00:15:07.845
general equilibrium point of view.

368
00:15:08.225 --> 00:15:11.485
And there are actually very few papers that have done that,

369
00:15:11.755 --> 00:15:13.525
that have worked with this in general equilibrium.

370
00:15:13.835 --> 00:15:17.765
They typically focus on stochastic volatility shocks either

371
00:15:18.325 --> 00:15:19.565
directly to exogenous oil prices

372
00:15:19.985 --> 00:15:21.685
or to exogenous oil production.

373
00:15:22.665 --> 00:15:25.005
Now, what we demonstrate in the paper is that when you do

374
00:15:25.005 --> 00:15:26.685
that, we can build that into our model

375
00:15:26.685 --> 00:15:28.325
as an alternative specification.

376
00:15:28.945 --> 00:15:30.965
You don't get much of a response of the economy

377
00:15:31.225 --> 00:15:34.005
and the intuition is that with a volatility shock,

378
00:15:34.395 --> 00:15:37.285
uncertainty goes up because it could be better

379
00:15:37.305 --> 00:15:39.285
or worse than you expected as your baseline.

380
00:15:39.875 --> 00:15:42.525
Well, why would there be a big response intuitively, right?

381
00:15:42.715 --> 00:15:44.845
However, if I tell you there's a disaster

382
00:15:45.235 --> 00:15:47.285
with increased probability around the corner,

383
00:15:47.625 --> 00:15:49.725
that's not just changing uncertainty,

384
00:15:49.945 --> 00:15:52.045
that's also changing the expectation of the future.

385
00:15:52.065 --> 00:15:54.125
And that gives you a much bigger effect,

386
00:15:54.175 --> 00:15:55.925
which intuitively explains why

387
00:15:56.565 --> 00:15:59.005
downside risk explains why these things are really potent.

388
00:15:59.005 --> 00:16:01.085
Something you can't generate with volatility shock.

389
00:16:01.305 --> 00:16:05.205
So our model is the first general equian model

390
00:16:05.665 --> 00:16:08.605
of the transmission of upside risk to oil prices

391
00:16:08.985 --> 00:16:10.965
and downside risk to the macro economy,

392
00:16:11.185 --> 00:16:13.285
how this affects the oil price of the global economy.

393
00:16:13.905 --> 00:16:17.005
We are also the first model in which oil price uncertainty,

394
00:16:17.255 --> 00:16:19.445
macro uncertainty, the economy

395
00:16:19.745 --> 00:16:21.965
and the price of oil are all jointly

396
00:16:21.965 --> 00:16:23.285
and simultaneously determined.

397
00:16:23.675 --> 00:16:26.285
Usually people make some of these things exogenous,

398
00:16:26.285 --> 00:16:28.005
feed them into the model from the outside,

399
00:16:28.265 --> 00:16:30.605
we let the model determine how important these things are.

400
00:16:30.705 --> 00:16:32.485
The first is policymakers.

401
00:16:32.485 --> 00:16:34.165
Central bankers, for example, tend

402
00:16:34.165 --> 00:16:36.685
to be concerned about geopolitical oil person certainty.

403
00:16:37.025 --> 00:16:38.445
And that that makes a lot of sense, right?

404
00:16:38.445 --> 00:16:42.625
We understand that what our analysis ads is that it shows

405
00:16:42.625 --> 00:16:46.185
that it takes a very large increase in the probability

406
00:16:46.245 --> 00:16:47.585
of an oil production disaster

407
00:16:48.085 --> 00:16:50.185
to have a sizable macroeconomic impact.

408
00:16:50.925 --> 00:16:52.145
So to give you some numbers,

409
00:16:52.805 --> 00:16:56.345
if you had a 40 percentage point increase in the probability

410
00:16:56.485 --> 00:17:00.265
of a major geopolitical oil disaster, so you go from 10%

411
00:17:00.265 --> 00:17:03.085
to 50%, that's pretty big, right?

412
00:17:03.715 --> 00:17:06.805
Then you would still only expect output to be lower

413
00:17:07.065 --> 00:17:09.645
by at most 0.25%.

414
00:17:10.145 --> 00:17:12.245
And that number could be smaller depending on

415
00:17:12.245 --> 00:17:15.045
how a model the persistence of the oil disaster.

416
00:17:15.195 --> 00:17:16.845
This is kind of, the, the,

417
00:17:16.965 --> 00:17:19.525
a more optimistic case that makes it bigger.

418
00:17:20.745 --> 00:17:25.245
So that's saying many geopolitical events say when the

419
00:17:25.245 --> 00:17:29.085
houthis, attacked shipping, that's not likely

420
00:17:29.265 --> 00:17:31.725
to be a big deal, simply

421
00:17:31.725 --> 00:17:32.765
because the probability

422
00:17:32.785 --> 00:17:34.965
of a major old production disruption are quite small.

423
00:17:35.145 --> 00:17:36.245
It takes much more than that.

424
00:17:37.425 --> 00:17:40.565
The second point is not unrelated, which is to say

425
00:17:41.765 --> 00:17:45.445
historically geopolitical oil press risk has not been a

426
00:17:45.445 --> 00:17:48.045
major driver of macroeconomic volatility.

427
00:17:48.695 --> 00:17:50.085
And, and that's intuitively

428
00:17:50.115 --> 00:17:53.445
because historically these risks have been fairly small.

429
00:17:53.825 --> 00:17:55.645
Now that doesn't mean they couldn't be important in the

430
00:17:55.645 --> 00:17:58.325
future, but just historically they have not been a big

431
00:17:58.325 --> 00:17:59.525
deal in the data.

432
00:17:59.795 --> 00:18:03.205
More than half of the observed all press uncertainty is

433
00:18:03.205 --> 00:18:04.445
macroeconomic uncertainty.

434
00:18:05.235 --> 00:18:06.415
Now, that's important

435
00:18:06.415 --> 00:18:10.055
because it suggests that previous empirical studies that try

436
00:18:10.055 --> 00:18:12.615
to relate all price uncertainty to macroeconomic outcomes

437
00:18:13.615 --> 00:18:16.855
actually captured effects of macroeconomic uncertainty,

438
00:18:16.985 --> 00:18:18.815
which made these results look much bigger

439
00:18:18.815 --> 00:18:20.055
than they actually should have been.

440
00:18:20.115 --> 00:18:22.885
That's point number two. Point number three is

441
00:18:23.115 --> 00:18:25.925
that the anticipation of a growth disaster

442
00:18:26.585 --> 00:18:28.605
is actually quite quantitatively important.

443
00:18:28.825 --> 00:18:30.525
Now, that's not surprising when it comes

444
00:18:30.525 --> 00:18:32.965
to the macro economy because there's been early work

445
00:18:32.965 --> 00:18:36.245
by Guo in 2012 that showed this.

446
00:18:36.505 --> 00:18:38.845
Uh, he focused on the impact on the financial sector.

447
00:18:39.235 --> 00:18:41.565
What we found out that's also true for the macro economy

448
00:18:41.745 --> 00:18:44.325
and for oil markets, and that's something that's really new.

449
00:18:44.325 --> 00:18:46.125
People have not fully appreciated that.

450
00:18:46.505 --> 00:18:49.525
So give you a sense, a 40 percentage point increase in the

451
00:18:49.525 --> 00:18:52.405
probability of a growth disaster, well, that's expected

452
00:18:52.405 --> 00:18:55.205
to lower GDP by 0.75% points.

453
00:18:55.205 --> 00:18:57.445
That's three times as big as the number

454
00:18:57.625 --> 00:18:58.845
for the corresponding oil disaster.

455
00:18:58.945 --> 00:19:00.205
Let focus on two examples.

456
00:19:00.785 --> 00:19:05.005
One example has to do with the implications of our analysis

457
00:19:05.505 --> 00:19:08.045
for empirical work on the effects

458
00:19:08.045 --> 00:19:09.805
of world price uncertainty shocks on the economy.

459
00:19:10.075 --> 00:19:11.765
There's a large literature on that.

460
00:19:12.435 --> 00:19:13.805
What kind of models do people use?

461
00:19:14.185 --> 00:19:18.565
One model are recursive linear vector auto regressive models

462
00:19:18.995 --> 00:19:21.125
with the uncertainty measure ordered first.

463
00:19:21.385 --> 00:19:23.485
That's very much in the tradition of work done

464
00:19:23.485 --> 00:19:24.965
by Nick Bloom in other contexts.

465
00:19:25.985 --> 00:19:28.405
The second are nonlinear models

466
00:19:28.535 --> 00:19:30.685
where you have a large error term

467
00:19:30.835 --> 00:19:32.805
that generates conditional hetero spasticity.

468
00:19:33.225 --> 00:19:35.925
And this conditional variance feeds into the conditional

469
00:19:35.925 --> 00:19:37.925
mean specification providing feedback.

470
00:19:38.965 --> 00:19:42.145
And the third is stochastic volatility VR models

471
00:19:42.145 --> 00:19:43.585
that are similar in spirit,

472
00:19:43.925 --> 00:19:46.545
but you model the traditional ity a bit differently.

473
00:19:47.815 --> 00:19:50.115
Now, all three models have one thing in common.

474
00:19:50.385 --> 00:19:52.675
They treat oil price uncertainty as exogenous.

475
00:19:52.695 --> 00:19:54.795
And our model showed, well, clearly it's not,

476
00:19:54.925 --> 00:19:56.035
which tells you immediately

477
00:19:56.035 --> 00:19:57.995
that all these estimates must be inconsistent.

478
00:19:59.135 --> 00:20:00.275
The second point is

479
00:20:00.275 --> 00:20:03.395
that when you look at stochastic volatility VR models,

480
00:20:04.105 --> 00:20:08.395
they treat the oil price uncertainty shock as independent

481
00:20:08.395 --> 00:20:09.515
of the level shocks.

482
00:20:10.165 --> 00:20:11.385
But our model shows

483
00:20:11.575 --> 00:20:13.825
that the level shocks feed into oil price uncertainty.

484
00:20:14.205 --> 00:20:15.265
So that's not a good assumption.

485
00:20:16.135 --> 00:20:17.875
Now, the guardian mean bar models,

486
00:20:18.715 --> 00:20:19.795
they go the opposite route.

487
00:20:19.945 --> 00:20:22.475
They're saying the level shock to the price of oil

488
00:20:22.575 --> 00:20:24.595
and the volatility shock to the price of oil.

489
00:20:24.595 --> 00:20:26.555
They're one in the same shock. They're perfectly correlated.

490
00:20:27.095 --> 00:20:28.385
Well, that's not right either.

491
00:20:28.445 --> 00:20:29.745
The truth is somewhere in between.

492
00:20:30.245 --> 00:20:32.105
So both of these models have some additional

493
00:20:32.105 --> 00:20:33.465
problems beyond assuming.

494
00:20:33.525 --> 00:20:34.905
So Gen 80, the question is

495
00:20:35.405 --> 00:20:37.945
how good are these models or, or differently?

496
00:20:38.165 --> 00:20:40.545
How bad are they? How biased are they relative to

497
00:20:40.545 --> 00:20:41.625
what the true response is?

498
00:20:42.005 --> 00:20:43.825
And what can we learn about empirical Morgan?

499
00:20:43.825 --> 00:20:46.705
And maybe is there a way of trying to, get

500
00:20:46.705 --> 00:20:48.925
Better estimates of the impact of all present uncertainty?

501
00:20:48.945 --> 00:20:52.605
That's one example. The other goes back to the theory.

502
00:20:53.345 --> 00:20:55.245
An obvious question to ask is

503
00:20:55.665 --> 00:20:57.845
how central banks should respond

504
00:20:58.265 --> 00:21:00.085
to shifts in geopolitical risk,

505
00:21:00.995 --> 00:21:03.575
and that will require making the model richer.

506
00:21:03.635 --> 00:21:05.135
We will have to introduce sticky prices.

507
00:21:05.315 --> 00:21:07.575
We have to introduce policy reaction functions.

508
00:21:07.995 --> 00:21:09.415
But ultimately the question is,

509
00:21:09.515 --> 00:21:11.855
if you're a policymaker, what should you do?

510
00:21:11.955 --> 00:21:14.575
What's the best way of responding to fluctuations?

511
00:21:15.095 --> 00:21:16.215
Fluctuations of geo palatal risk?

512
00:21:16.215 --> 00:21:17.495
And obviously that's something

513
00:21:17.495 --> 00:21:18.935
that's very much in the headlines right now.

Exploring Oil Price Dynamics and Economic Impact

00:00:00 - Understand this perspective, I think you have to go back to the early 1980s. So in the decade before that, in 73 and 79, we saw big oil price increases, which were followed by major recessions. And so a national conjecture was that, well, higher oil prices caused recessions. If that's true, you would also expect that lower oil prices would cause expansions. And that premise was put to a test in 1986 because at that point the OPEC cartel dis integrated and the price of oil fell about as much

00:00:36 - But there was no major expansion. So how do you make sense of that? Well, one interpretation of this evidence is to say, well, this relationship between oil prices and the economy is inherently linear. And so what we learned from the 86 episode is that really in 79, oil prices weren't the major driver of a recession. It must have been something else. What could that have been? While a natural conjecture would be maybe the shift in monetary policy under Paul Voker,

00:01:06 - Now that leads you to conclude that oil prices are not that important, at least not relative to what people think, but there's another interpretation and that involves a non-linear relationship between oil prices and the economy. That involves oil price uncertainty. That interpretation In 79, what happened is while the oil price went up, that made people worse off. You have less money to spend and so your consumption goes down, the economy slows down. And that was made worse by an increase, a simultaneous increase in all price uncertainty,

Oil Price Dynamics and Economic Implications

00:01:37 - and made the CE recession that much worse. So what about 86? Well, clearly in 86, as the price of oil came down, people were better off. They had more money to spend. So there was a stimulus, but that stimulus was offset by the fact that as all prices came down, all present certainty went up, which then reduced investment and the reduction in investment offset the increase in consumption. So you ended up with not much of an expansion. And it's this loose evidence that really I think has convinced macro,

00:02:10 - that all person certainty creates recessions. Now economists of course also like to look at models. There has to be a story why, this would be recessionary. And they like to look at models of irreversible investment by firms. So the idea is that a firm decides whether to make an investment or not. They look at the net present value of the investment to decide that. That depends on the cash flow. And the cash flow depends on the prices of inputs and outputs.

00:02:42 - then if oil price uncertainty goes up, you might say, well, this used to make sense before price uncertainty went up. Now I'm not so sure, let me hold off. That means investment if that happens at the aggregate investment drops and you get a recession, right? So that's a very natural story. There are A number of problems with that story. One is it makes perfect sense if you're modeling oil producers or if you're modeling the petrochemical industry where the oil price is clearly very important for the cash flow.

Impacts of Oil Price Uncertainty on Diverse Industries

00:03:12 - that exposed to oil, right? If you're talking about a computer manufacturer, a textile company, a service company, their cashflow isn't very dependent on the price of oil. So why should there be a big response in the aggregate? Because that's what most of the economy is, even when you have a sector that is exposed to oil price risk. Think of airlines, right? An airline company deciding whether to buy a new aircraft that's supposed to be in service for the next 20 years, that involves a very long,

00:03:44 - Here's the problem. Suppose oil price uncertainty. Let's say oil price volatility goes up now, well, that means it's probably going to be high for the next couple of months, maybe quarters. But after that, basically our expectation of that volatility goes back to the unconditional variance, which is a constant, which means for the most part, not much is changing in that cash flow, right? And that means you wouldn't expect a big reaction either even in that case. And so that kind of says it's not clear whether

00:04:16 - In any case, these are very qualitative arguments. If you want to quantify this, you actually have to quantify oil press uncertainty. That's a non-trivial task. So we have to talk about that. But there is another issue, which is that oil press uncertainty is endogenous with respect to the macro economy. And so we really need to be careful not to confuse oil price uncertainty with macro uncertainty because an increase in macro uncertainty would affect oil prices in your cash flow.

Understanding Oil Price Uncertainty: Predictability and Measurement

00:04:46 - your net present value. Even though oil prices uncertainty by itself, if it, if it were exogenous, would not. And so when people try to link those all press uncertainty to cash flows, it becomes a little bit murky what's actually going on. Many people intuitively think of uncertainty as variability in the change in the price of oil. They're saying, well, the, the variability was pretty low in this period. It's higher in that period. So uncertainty has gone up or they run rolling window regressions to see how the variability has increased.

00:05:18 - Uncertainty intuitively is lack of predictability of the oil price because if you could perfectly predict the oil price, there would be no uncertainty, right? Decision making would be easy. That is not the world we live in. The question is how much is the oil price predictable given what we know at each point in time that can be measured? There is some nice work by Dorado et al in the a in 2015 that spells out how to measure uncertainty in general.

00:05:48 - and we come up with an index of one quarter ahead. Oil price uncertainty. This is for the inflation adjusted price or global price of when you compare this index to the ofs, which is an applied volatility measure of oil price uncertainty used in financial markets, they're actually fairly highly correlated. The downside of the OS is it isn't available for a very long time period. And so if you're interested in the eighties and seventies, then you really need to go with our index.

Analyzing Oil Price Uncertainty: Geopolitical vs. Macroeconomic Factors

00:06:18 - They have worked, for example with text-based measures. So take a text-based measure of oil price uncertainty. You essentially go to major newspapers and you look up for combinations of the word oil price and uncertainty, and you count how many times that pops up. That is not a measure of lack of predictability. And indeed such an index would be very uncorrelated with what we're trying to do. And the same would be true for text-based measures of geopolitical risk, for example. So before we go into the details, we have to understand

00:06:53 - by geopolitical downside risk to all production, but it can also be driven by the macro economy that shows up very much when you look at the data, we find that ultra uncertainty surges may be associated with geopolitical events, but there are surges which clearly have nothing to do with geopolitical events. And when we do identify geopolitical events that shift the risk to oil production, sometimes they're associated with an increase in all press uncertainty, sometimes not.

00:07:24 - 1970s actually was in 2008, that was during the financial crisis that had nothing to do with geopolitics. There are events where geopolitics matters. A good example would be the invasion of Kuwait in 1990, but there are others. For example, the outbreak of the Iran Iraq war in 1980 that had very little effect on our oppress circuit. So things are not mechanical and in many episodes it's not quite clear what's going on. For example, in 79 while you had the Iranian revolution,

Modeling Oil Price Uncertainty: Geopolitical vs. Macroeconomic Factors

00:07:54 - which one is driving all price uncertainty, which clearly went up in that episode, not so clear. Or you should take 2020, you had on the one hand the covid pandemic and the other hand you had the Saudi price war. Is that geopolitics? Is it macro? Is it some combination? That is something where you need modeling. The model is calibrated to us and global oil market data in the model we have endogenous oil prices and we have endogenous oil price uncertainty. Oil price uncertainty responds to level shocks and volatility shocks in, the, in the model,

00:08:27 - So there are eight state variables and six shocks. So you really require, fully non-linear solution methods as a beast to, to solve. but it gives you a good fit in terms of first and second moments. The key feature of the model is that we model downside risk as opposed to a volatility. That is if oil price uncertainty or macro uncertainty in the model moves, it moves because of downside risk. And so you might say, why is that? Most people say, well, water agility captures uncertainty.

00:09:00 - that is not the way most economic agents think about this. Think of a regular household, a consumer in the us you drive a car, you're worried about the price of gasoline. What keeps you up at night? Well, the possibility that oil production in the Middle East or oil exports from the Middle East might shut down. The price of oil goes up and you end up paying $5 a gallon at the pump. That's something that might keep you up at night. On the other hand, if I told you, well,

Understanding Downside Risk in Oil Production

00:09:30 - let's call it shale oil 2.0 that makes oil production much easier and the price of oil is gonna drop and the price of gasoline is gonna go to 50 cents a gallon. Are you gonna stay up at night worrying about that? No. If it happens, that's fine. If it doesn't happen, that's fine too, right? It's not what makes you change your behavior. And so we argue it's the downside risk that really matters. Now, why haven't people worked with downside risk? Downside risk is very subjective. It's very hard to get agreement on

00:09:59 - Volatility is very easy to measure. And so I think economists have gravitated to what volatility for that reason, but we don't think that's the right way of doing it. So what is the downside risk? In our model, we have macroeconomic disasters. we model those after events such as the financial crisis or the COVID-19 recession. These are events that occur maybe once every 100 years, an expectation, and they involve, a, a disruption of a, a reduction in growth that lasts for about two or three years.

00:10:32 - that create downside risk to oil production are modeled, after major global geopolitical production disruptions in 73, 74. In 79 or in 1990 for example, they involve a 5% drop in global oil production. And that happens in expectation every 15 years or so, right? So they're a bit more frequent. What we concerned within the model is less the realization of a disaster. It's pretty clear what happens when you have a disaster. The interesting part from our point

Impact of Oil Price Uncertainty on Economic Behavior

00:11:05 - What do people do because they're worried about such a disaster possibly being around the corner. Another point about the model is how we, model how uncertainty triggered by this downside risk affects the economy. We basically follow what the literature has done before We have a precautionary savings channel that's saying if there is say, increased macro risk, you worry about your future income, what do you do? You increase your precautionary savings. That means you reduce your consumption

00:11:35 - You also have a precautionary oil inventory demand channel that's saying if there's increased oil price uncertainty, you're gonna start building stocks in anticipation of something bad that might happen in the future. And of course, if there's increased macro risk, you think the economy is gonna slow down, you start selling off your stocks in, In that case, reducing your oil inventories. The third component that people like to talk about is the real options theory in Bernanki 1983. it turns out, as we discussed in the paper,

00:12:07 - It's a partially be model that cannot be generalized. What we do instead is something that generates very similar effects without involving rail options. We impose limited substitutability between oil and capital in production, which means if anything ha is happening to oil, that spills over to what's happening with capital. And so if you had higher, oil price risk, that would lead to a decline in the expected return on investment. And that means you're gonna invest less. It's gonna reduce output, it's gonna cause a recession.

Impact of Oil and Macro Disasters on Investment and Output

00:12:40 - and I think also in the literature. So let me start with the oil production disaster. So we have an increase in the probability of an oil production disaster that is associated with higher oil present certainty. That means if agents observe that higher oil present certainty, they're going to increase their oil storage demand because you wanna build inventories, as a precautionary measure that results in a persistent increase in oil inventories. And of course in an increase in the price of oil. Now the increase in the price of oil in turn is gonna generate a modest drop in investment

00:13:13 - And there's little response in consumption or in output uncertainty. In contrast, if you now play the game of having a, an increase in the probability of a map for disaster, essentially a growth disaster, well, that will give you a sharp rise in output uncertainty and an immediate sharp drop in investment and output because, it's not just the uncertainty, it's also that you expect, the economy to slow down in the presence of downside risk. That's kind of the difference between volatility shocks and,

00:13:43 - So investment is going to tank, output is gonna tank. There's gonna be little response in consumption. like in the oil disaster probability case. The key difference is that these effects tend to be about three times as big, right? So macro disasters are more potent. That makes a lot of sense because oil disasters are hampered by the low share of oil in output, but the macro disaster basically affects everything, right? So, so intuitively that makes sense. Now, as actual

Oil Price Uncertainty and Economic Impacts

00:14:15 - macro disaster probability shock, there is a sharp drop in the price of oil, and that means inventories persistently decline. So at the same time, you get increases in oil price uncertainty and output uncertainty. And so what's the key difference apart from the magnitudes here, is that in the case of an oil production probability increase, you build inventories here you reduce inventories. So the mechanics of how this is translated are a little bit different.

00:14:46 - that literature focused on partial equilibrium models with exogenous, that is geopolitically driven oil price uncertainty. And there are two problems with that. One, as I mentioned, oil price uncertainty is endogenous. Second, it's well documented that the implications of such partial equilibrium models need not hold up in general equilibrium. So it's important to look at this from a general equilibrium point of view. And there are actually very few papers that have done that, that have worked with this in general equilibrium.

00:15:18 - directly to exogenous oil prices or to exogenous oil production. Now, what we demonstrate in the paper is that when you do that, we can build that into our model as an alternative specification. You don't get much of a response of the economy and the intuition is that with a volatility shock, uncertainty goes up because it could be better or worse than you expected as your baseline. Well, why would there be a big response intuitively, right? However, if I tell you there's a disaster

Modeling Downside Risk and Oil Price Uncertainty

00:15:47 - that's not just changing uncertainty, that's also changing the expectation of the future. And that gives you a much bigger effect, which intuitively explains why downside risk explains why these things are really potent. Something you can't generate with volatility shock. So our model is the first general equian model of the transmission of upside risk to oil prices and downside risk to the macro economy, how this affects the oil price of the global economy. We are also the first model in which oil price uncertainty,

00:16:19 - and the price of oil are all jointly and simultaneously determined. Usually people make some of these things exogenous, feed them into the model from the outside, we let the model determine how important these things are. The first is policymakers. Central bankers, for example, tend to be concerned about geopolitical oil person certainty. And that that makes a lot of sense, right? We understand that what our analysis ads is that it shows that it takes a very large increase in the probability of an oil production disaster

00:16:50 - So to give you some numbers, if you had a 40 percentage point increase in the probability of a major geopolitical oil disaster, so you go from 10% to 50%, that's pretty big, right? Then you would still only expect output to be lower by at most 0.25%. And that number could be smaller depending on how a model the persistence of the oil disaster. This is kind of, the, the, a more optimistic case that makes it bigger.

Assessing Geopolitical Oil Risks and Macroeconomic Uncertainty

00:17:25 - houthis, attacked shipping, that's not likely to be a big deal, simply because the probability of a major old production disruption are quite small. It takes much more than that. The second point is not unrelated, which is to say historically geopolitical oil press risk has not been a major driver of macroeconomic volatility. And, and that's intuitively because historically these risks have been fairly small.

00:17:55 - future, but just historically they have not been a big deal in the data. More than half of the observed all press uncertainty is macroeconomic uncertainty. Now, that's important because it suggests that previous empirical studies that try to relate all price uncertainty to macroeconomic outcomes actually captured effects of macroeconomic uncertainty, which made these results look much bigger than they actually should have been. That's point number two. Point number three is that the anticipation of a growth disaster

00:18:28 - Now, that's not surprising when it comes to the macro economy because there's been early work by Guo in 2012 that showed this. he focused on the impact on the financial sector. What we found out that's also true for the macro economy and for oil markets, and that's something that's really new. People have not fully appreciated that. So give you a sense, a 40 percentage point increase in the probability of a growth disaster, well, that's expected to lower GDP by 0.75% points. That's three times as big as the number

Reevaluating Economic Models of Oil Price Uncertainty

00:18:58 - Let focus on two examples. One example has to do with the implications of our analysis for empirical work on the effects of world price uncertainty shocks on the economy. There's a large literature on that. What kind of models do people use? One model are recursive linear vector auto regressive models with the uncertainty measure ordered first. That's very much in the tradition of work done by Nick Bloom in other contexts. The second are nonlinear models

00:19:30 - that generates conditional hetero spasticity. And this conditional variance feeds into the conditional mean specification providing feedback. And the third is stochastic volatility VR models that are similar in spirit, but you model the traditional ity a bit differently. Now, all three models have one thing in common. They treat oil price uncertainty as exogenous. And our model showed, well, clearly it's not, which tells you immediately that all these estimates must be inconsistent.

00:20:00 - that when you look at stochastic volatility VR models, they treat the oil price uncertainty shock as independent of the level shocks. But our model shows that the level shocks feed into oil price uncertainty. So that's not a good assumption. Now, the guardian mean bar models, they go the opposite route. They're saying the level shock to the price of oil and the volatility shock to the price of oil. They're one in the same shock. They're perfectly correlated. Well, that's not right either. The truth is somewhere in between.

Evaluating Models and Central Bank Response to Geopolitical Risks

00:20:32 - problems beyond assuming. So Gen 80, the question is how good are these models or, or differently? How bad are they? How biased are they relative to what the true response is? And what can we learn about empirical Morgan? And maybe is there a way of trying to, get Better estimates of the impact of all present uncertainty? That's one example. The other goes back to the theory. An obvious question to ask is how central banks should respond to shifts in geopolitical risk,

00:21:17 - We will have to introduce sticky prices. We have to introduce policy reaction functions. But ultimately the question is, if you're a policymaker, what should you do? What's the best way of responding to fluctuations? Fluctuations of geo palatal risk? And obviously that's something that's very much in the headlines right now.

Faculti AI can make mistakes

Research

Policy

Analysis