More on inflation, violence and identification

Chris has responded to my previous post, which he frames as a criticism of his research. I should state upfront that this does not strike me as entirely accurate, since I emphasized at various points that my concerns lay in the possible journalistic interpretation of his work. Some email correspondence between the two of us suggests that I am not alone in expressing such trepidations, but I digress. On then, to Chris’s response…

1) He begins by taking issue with my decision to focus on food prices, politely suggesting that I “may have missed” the fact that his non-discretionary index of living costs includes various other components (including rent, electricity, water, etc).

As it happens, I don’t think that I missed this at all. My reasons for focusing on food prices are quite simple. First, they provide relevant context to the real effects that I highlight in my post, i.e. agricultural shocks stemming from massive drought. This was done deliberately with the aim of illustrating the overriding message of my post: Attributing causation to any particular event is often very difficult, and we certainly have to bear real effects in the front of our minds when discussing the sources of inflation. (To reiterate, this is something that the Business Day article failed to do entirely.) Second, food prices provide an obvious segue to the other article that I discuss in my post, which concerns the role that monetary expansion had in driving up food prices and thus precipitating the Arab Spring. Such matters notwithstanding, however, I did happen to include the following passage in my original post:

To be clear, South Africans have also experienced sharp increases in the cost of amenities like electricity and water provision due to some boneheaded policy decisions and as a legacy of inefficient parastatal monopolies.

Chris may have missed that, though. (wink)

2) His second objection is that I am unfairly interpreting his research as a suggestion that food hikes are the only cause of violence. He quotes his references to “political grievances” as evidence that I haven’t read the article properly.

Again, however, this seems to be a misunderstanding of what I have written and the major point of my post. In the passage that he quotes, I'm not concerned with alternative causes of violence, but rather the underlying drivers of one particular cause, i.e. inflation. At the risk of repeating myself: To the extent that inflation does act as a trigger for social unrest and violence – and irrespective of whether that occurs alongside other factors such as political grievances or not – we need to understand what the underlying forces behind that inflation are. Any analysis that focuses only on the nominal effects of (quote unquote) “delinquent” monetary policy is simply misleading. Why? Well, because there may be very significant real price drivers occurring at the same time! This is something that the Business Day article completely failed to mention, and the same is true for The Telegraph article that I quoted in the second half of my post. I see nothing wrong with taking exception to such slipshod analysis.

3) Next issue: On my suggestion that one might baulk at the definitive description of this research as “proof” of the relationship between inflation and violence...  Well, I don’t have much to add here, since – again – this is a criticism of how the journalist chose to frame his article. “Proof” is simply too strong and simplistic a word to use given all this issues that I have raised. (Note: I see that this has happened elsewhere.)

4) The penultimate point that Chris makes in his reply extends beyond the article featured in Business Day.  I will summarize his argument as saying that the South African Reserve Bank (SARB) should abandon its focus on the headline CPI, because a) Non-discretionary inflation has been rising much faster, b) It cannot control which specific goods rise and fall in price, and c) It would better facilitate an environment of civil harmony by stabilizing the Rand against a basket of commodities.

Now, interestingly enough, subsequent to yesterday’s post I found this column that Chris has penned himself. (I’ll take it that we can safely assume away possibility of incorrect interpretation by a third party here.) He produces the below graph and proceeds to write: 

Seeing as Non-Discretionary goods price inflation has averaged well above the SARB’s price inflation target of 6% for most of the past seven years, low income groups’ standards of living are falling at a compounded rate relative to high income earners. [Emphasis mine.]

I don’t have the raw data to hand, but eye-balling the chart it doesn't seem at all obvious to me that non-discretionary goods have “averaged well above” the 6% inflation target. (Does it seem obvious to anyone else?) In fact, I’d hazard a guess that it averages a shade below the 6% mark. Certainly, the strongest statement that we can probably make about this series is that it fluctuates around that general level.

We all agree that no single measure of price changes is perfect. Indeed, it is precisely for this reason that we have constructed so many different indices in the hopes of getting a better sense of how “inflation” is playing out in the economy. Central banks like the SARB choose to follow a preferred metric – like the CPI – for a number of reasons, most of them very sensible. As it happens, the sheer volatility of commodities is a key reason why some CBs prefer “core” to “headline” inflation measures. Trying to conduct monetary policy in response to a simple basket of commodity prices would not only be incredibly difficult due to the inherent volatility (and the fact that the CB is more or less powerless to  stop these short-run swings), but potentially counterproductive because of the amplifying effect that it could have on consumption cycles. (For more discussion, see Matt Rognlie’s excellent posts on this subject: here, here, and here.)

5) As for his final point, that peer review is not superior to insights that bring in paying clients… Well, clearly that is not what I meant by “cracking” the problems of identifying a causal link between price increases and the uprisings in the Arab world. (Mind you, if he did accurately predict these events in advance of them happening then I certainly am impressed.) So, while I regard the profit mechanism as essential as the next economist, that has nothing to do with my concerns about getting to grips with some very obvious identification problems. That said, allow me to make a broader concluding remark: Just as no-one should suggest that peer-review is infallible, we should never confuse profitability with validity. Even psychics have been doing a roaring trade for centuries. It doesn't make them right.

Did monetary expansion cause the Arab Spring?

South Africa's top economic and financial daily, Business Day, ran an article yesterday referring to research conducted in part by my old school friend and occasional commentator on this blog, Chris Becker.

Proof that high inflation leads to more public violence 

NEW research appears to show a direct link between inflation and social violence. In the months before the Marikana massacre, in which more than 30 miners died, there was a spike in nondiscretionary inflation — the inflation the poor experience — from 3% to more than 10%. The same is true of the xenophobic attacks in 2008. Just before these attacks, nondiscretionary inflation surged to 20%. The recent violence in Sasolburg was also preceded by an acceleration in inflation.

One might blanch at the definitive description (i.e. "proof") given to an in-house research document that, as far as I can tell, is unscrutinised by outside review. Certainly, I can immediately think of a host of problems that would need to be accounted for before we even begin to talk about proper causation.

That said, I don't doubt that food shortages and price hikes can, and do, trigger civil unrest and social upheaval. The idea is eminently plausible and there have been many attempts to quantify this relationship (more on this below). I commend Chris for trying to establish a more systematic understanding of the issue in the South African context.

However, I find it striking that this particular article makes no mention whatsoever of the real factors that have been driving high food prices in recent years. You know, massive crop failures due to historic droughts in the former Soviet Union, North America, and elsewhere... That kind of thing. In fact, here's a timely case study on South Africa that pinpoints these exact issues, which is itself part of a broader research programme linking food riots and political instability to agricultural supply-side shocks (in particular, those related to climate).

In contrast, the singular premise of the above BD article seems to be that food hikes -- and subsequent violence -- are entirely the fault of "delinquent" monetary policy.[*] I'm certainly not suggesting that loose monetary policy can't lead to inflation. Rather, the failure to acknowledge these severe real shocks makes any kind of simple analysis very misleading. (I should say that I am going strictly on the article here; Chris and his co-authors may well try to account for real factors in their actual research. At least, I sincerely hope so.) 

However, my faith in journalistic competence is somewhat shaken by the inevitable reference to -- you guessed it -- Shadowstats, the preferred purveyor of hyperinflation statistics for conspiracy theorists freedom lovers everywhere!™ Furthermore, statements like "The conclusion[...] is inescapable: inflation leads to violence" are more or less misleading in the same sense as the suggestion that increasing the temperature of your bath water will lead to you being boiled alive. There may may a kernel of truth therein, but it is clearly important to recognise that this is a matter of degree.

The passage that really caught my eye, however, was the following.

Becker conducted similar research internationally and found that countries experiencing the highest levels of social upheaval, such as Syria, Tunisia, Egypt and Algeria, embarked on huge monetary expansion in the months before the outbreak of violence. This monetary expansion translated into sharp increases in inflation just before the outbreak of violence. In Egypt and Tunisia, the violence culminated in the overthrow of the previous governments.

Woah. Let's just back up there a bit. We are now treading very dangerous territory as far as correctly identifying causation goes. It strikes me as as borderline irresponsible to intimate that the proximate cause of the "Arab Spring" was loose monetary policy. There are a myriad, interwoven factors at play and it would take a highly skilled statistician, armed with reams of data, to tease out the underlying drivers from concurrent symptoms. The fact is I've yet to see a paper on this subject make it through the peer-review process to journal publication... and I'm pretty certain that this is precisely due to the difficulties in attributing causation. With respect to my friend, I'm not convinced that he has managed to crack the problem that has stymied so many others.

I'll leave you with a final thought on this question of monetary expansion and the Arab Spring. A quick Google search on the topic throws up an article by Andrew Lilico that appeared in The Telegraph: How the Fed triggered the Arab Spring uprisings in two easy graphs (4 May 2011).  After demurely suggesting that most analysts are simply too afraid or short-sighted to "join the dots between the Federal Reserve’s second phase of quantitative easing and these revolutions [in the Middle East and North Africa]", Lilico bravely plunges forth to do exactly that. True to his word, he also produces two graphs, the most important of which appears below.

Now, I don't know about you, but that graph seems to show a rise in food prices that precedes the Fed's sharp increase in asset purchases... by several months. I am glad to report that this discrepancy wasn't lost on readers at the time. One commentator sardonically observes: "In my experience causes occur before effects."

It is a matter of some debate among economists how inflation manifests itself in the economy during times of monetary expansion. (E.g. Some of you may recall the rather heated discussions on Cantillon Effects that occurred in the blogosphere only recently.)  Well, it is a relief to know that the issue has been resolved thanks to the careful work of Mr Lilico. It turns out that expansionary U.S. monetary policy is so potent that it can positively impact the price of global commodities with a negative lag of several months!

Note (13/02/13): Follow-up here.

___

[*] To be clear, South Africans have also experienced sharp increases in the cost of amenities like electricity and water provision due to some boneheaded policy decisions and as a legacy of inefficient parastatal monopolies. I've covered these issues numerous times before on this blog and elsewhere.

Are charts of oil priced in gold really that impressive?

Okay, one more gold-related post before I go home...

Following on from my last post, I've just clicked through to Chris's website and seen a post titled, Petrol Price in Gold Terms, in which he argues that the recent rise in South African petrol prices are "not owed to higher petrol prices, but a much weaker Rand, caused by the Reserve Bank". [I assume by petrol prices he obviously means oil prices.] He continues, "In hard currency terms, the price of petrol is unchanged since 2002."

Chris isn't alone in making this argument, which is a favourite among gold fans. That said, I've never found it as persuasive or profound as others seem to do. If you read my previous post then you may already have guessed why, but here is the key passage:

Looking at [gold] prices first, we can see that these have been undeniably impacted by the rising costs of producing an ounce of gold. This can be put down to a number of things, but chief on that list would be rising energy costs (since mines are incredibly energy consumptive)...

Energy is a fundamental input in mining activity. Gold mines, which are deeper and more complicated to run than virtually all other mining operations, are clearly no exception. In that light, why wouldn't we expect the price of gold to track what is happening in the oil market? It would be roughly analogous to me saying that cupcakes have stayed at a constant price... in terms of flour.

Now, if you're about to argue that what I've said would hold for coal but not oil... Fuggedaboutit. The movements of coal and oil movements track each other very closely. So much so that oil prices are widely used as a proxy for coal prices when forecasting and hedging in the electricity industry (since they are also more liquid). This is true even in countries where oil plays an insignificant role in power generation. And, of course, mining companies still consume vast quantities of oil during their day-to-day operations regardless of which energy source fuels their electricity needs.

Anyway, to illustrate here is the price of gold per per barrel of oil since 1971, followed by the same for several commodities. These series were picked more or less at random and are taken from the World Bank's Data Centre. (Click to enlarge.)

Not much to choose between them, if you ask me. The point here is that virtually all commodities exhibit some kind of long-term mean relationship with oil. Indeed the increasing linkage between energy and non-energy goods was a driving factor in the commodities boom of recent years. Now, of course, scale matters here and it might be misleading just to eyeball separate charts. As one last treat then, here is a single chart containing the above commodities plus a few extra, normalized in terms of their respective units. (I pick 2002 = 1 for no better reason than this is the year that Chris used in his initial post.) Again, I think the message is pretty clear.

THOUGHT FOR THE DAY: All this talk about gold, but when are we going to have a serious discussion about the Fertilizer Standard, or the Soy Bean Standard?