Literature Review and Simulations – The Supply Shocks – Part II

Since we have received lots of questions recently, we proceed by Q&As.

Are your models only/mainly about the demand-side of the economy?

We use different types of models, each model captures the supply-side of the economy, somehow. For instance, some models include the (broad) nominal exchange rate or a measure of core import prices which already reflects supply-disruptions, at least to some extent. In general, all Phillips curve models include variables that capture supply-side shocks, and the micro-Phillips curve is essentially about the supply-side, as explained here.

What is the elasticity of core inflation to import prices?

We estimate that a 10% movement in the broad FX, results ceteris paribus in a 2.5%-3% movement in (relative) core import prices and about 3 tenths on core inflation after 2-3 quarters. The result applies to the US, the Euro area, and Japan. This means that we can have a visible impact on core inflation but we would need a pretty sizable FX shock. (Please also note that an exchange rate is a relative price: an appreciation of a currency is a depreciation of another one. Therefore, it is pretty unlikely to have a contemporaneous inflation acceleration across regions driven by such a shock)

What if we get another round of supply disruptions?

The current shock is expected to put (if any) only a few tenths of upper pressure on core PCE (same results for core HICP and core CPI in Japan). In the last couple of weeks some shipping cost indexes (i.e. Drewry world container index) have started to increase again. We assess the potential upside risks using our models, augmented with either (i) the Harper-Petersen shipping index (chart), or (ii) the NY Fed global supply chain pressure index (chart), or (iii) the Drewry world container index (chart). Here is a summary of our findings:

  • In a properly specified model, the three indexes are not (or only weakly) significant in the pre-Covid sample and do not always have the right sign (for instance, augmenting our “main” model with the NY Fed supply chain index and stopping the estimation in 2019:Q4 results in a negative (!) and not-significant coefficient, while extending the sample until 2023:Q4 results in a 0.28* coefficient). The identification, for all indexes, comes from the Covid-generated spike (so you have to be very careful). Figure 1 below makes this point by plotting the NY Fed global supply chain pressure index (yellow line) and the YoY of CPI core goods (excluding used cars and trucks). The point is clear: even looking at the portion of the US CPI most influenced by supply disruptions, it is hard to see a strong correlation in the pre-Covid sample.

Figure 1. NY Fed Global Supply Chain Pressure Index and CPI Core Good Ex Used Autos and Trucks

  • The inclusion of the indexes in our Phillips curve models (assuming a random walk path for the indexes going forward) results in a forecast in line with the current one when using the Drewry index (chart), in a forecast a bit below 2% when using the Harper index (chart), and a bit above 2% when using the NY Fed global supply chain index (chart). (Note: all charts refer to core PCE – the charts for US core CPI, core HICP, and Japan core CPI are available upon request)
  • The results using the Harper and the Drewry index should be taken with caution because the series are relatively short (available from 2001 and 2011, respectively).
  • Only some lags of the indexes are significant in the regressions, although they are often significant jointly.
  • In terms of elasticities for core PCE price inflation, we estimate that:
    • A 1,000US$ increase in the Harper index results in 25bps increase on impact (1 quarter) and up to 7 tenths after 3-4 quarters.
    • A 1,000US$ increase in the Drewry index results in 10bps on impact and up to 20bps after 3-4 quarters.
    • A 1 standard deviation increase in the NY Fed global supply chain index (that is, a +1 of the index) results in 28bps increase on impact and about 50-60bps after 3-4 quarters.

(The estimates for core HICP are more muted, consistent with the results in Banbura et al. (2023) reviewed in part I. We estimate that the indexes have limited effect on core HICP on impact and can raise it by 10-30bps after 3-4 quarters for the same size of the shocks listed above)

  • Averaging across the three indexes and multiplying the estimated elasticities by the 2020-2022 delta, we get that around 2½pp of the acceleration of core PCE price inflation between 2020 and 2022 can be attributed to supply disruptions. These estimates are in line with those of the San Francisco Fed and the Cleveland Fed.

Based on the above, what is your US forecast for 2024?

US Headline CPI around 2% in December 2024, US core CPI around 2½ percent. We produce two types of forecasts. The first one is a model-based medium-term forecast. This is our main focus because it is the priority of the central banks. The second type of forecast is a bottom-up model-based (judgmentally adjusted). In this moment, both approaches suggest the same conclusion: headline CPI around 2%, core CPI around 2½ percent in December 2024. An Excel file with the profile of our bottom-up CPI MoM sectoral forecast can be downloaded here.

(Note: we remind the reader that the details/data of our forecast are always available upon request)

Putting everything together

In this second part of the note, we have estimated the impact of renewed supply disruptions on core inflation. The results of our research confirms the estimates of the existing literature (some of which are  in part I here). The main result is that the current shock is expected (if any) to put only limited upper pressure on consumer prices. As usual, it is difficult to anticipate the extent (and the length) of the disruption. However, conditional on what can be seen today in the three indexes, it seems likely that the shock can put only a few tenths of upper pressure on core inflation (in the US, in the EA, and in Japan).

Contact

Want something more tailored?

We provide tailored consulting on ad-hoc projects.