In the onset of a new strong increase in the Blue Chip Swap premium – it grew from 40% to 65% in two weeks – the elusive question on whether the non-commercial FX has an impact on inflation rises again.

To aid in this search we estimate a Vector Autoregressive model with our high-frequency inflation dataset – 360 weekly datapoints since 2013 -, trying to capture the short term relationship between inflation (core, regulated and seasonal), fx (official and BCS) and monetary policy (aggregates, interest rates).  We tested dozens of variables and model specs (see methodological appendix) and found that:Core inflation is highly persistent. As much as 80% of core inflation variation can be explained by autocorrelation.Regulated price hikes do not seem to have a persistent effect on core inflation. The coefficients are systematically non-statistically significant once inflation dynamics are controlled by FX.Besides inertia, official FX is the most significant variable. Cumulative impulse-response on core inflation from an “official FX” shock is 15% and it explains roughly 20% of total core inflation variation. Blue Chip Swap has no significant incidence on short term core inflation! The coefficients are, again, systematically nonsignificant in all the tested specificationsWe also weren’t able to find any short term significance on money and interest rates.We tested on other variables (i.e. wages) and for many “ad-hoc” specs (nonlinearities, dummies such as price or capital controls, non-symmetric responses) and results look robust. Results were also consistent with other longer-term and lower frequency (monthly) datasetsOur results do not imply, and we do not believe, that other non-significant macro or policy variables are irrelevant, but on the other hand that official FX seems to be the most relevant transmission channel to explain short term inflation. Also, the nature of the model and the data seek to capture short term relationships (week. months) and not mid/long term (quarters, years).

What does that imply? Reserves dynamics and official FX seem to be the relevant variables to follow in the near future. As long as the government is able to hold the grip on official FX, inflation should remain under control. Of course, this does not rule out non-FX, money related inflationary shocks, but this kind of shocks has not been present in the recent past, even in high money printing periods such as 2011-2015.

Also, results should be analyzed in the context of variables moving within in-sample ranges and should not be extrapolated to “out of sample” ranges (i.e BCS premium above 100%). We don’t yet foresee variables moving out of “in-sample” ranges, although odds are not low: BCS has already been moving much faster than expected. 
 Historical Blue Chip Swap premium
Methodological annex. Vector Autogresive inflation Model
We estimate a Vector Autoregressive model with our high-frequency inflation dataset trying to capture the short term relationship between inflation, fx and monetary policy.
SEIDO High Frequency CPI vs INDEC

The dataset has 360 weekly datapoints since 2013 and includes core, regulated and seasonal price indexes, official and BCS FX, interest rates (monetary policy, Baldar), money aggregates (seasonally adjusted m0, m2, m3, cash holdings), wages (monthly) and monetary regime dummies (FX controls, inflation targeting).

We tested dozens of models specs (in levels, differences, lags, inclusion and exclusion of variables, etc) and opted to report the results of a “simple” spec where most of the findings prevail: Core and regulated inflation and official and BCS fx, all expressed in weekly variations.  Most of the results were robust to different specifications.

Standard specification tests were also made (Stationarity tests, AIC and Schwarz for lags selection, errors autocorrelation, etc)
Summary of coefficient results

Cumulative Impulse responses

Forecast Error Variance Decomposition