Blog

In September 2024, I published a post in which I observed that it was possible that the Canadian Dairy Commission (CDC) will adjust the price of milk down in February 2025. That observation was based on my estimates of the cost of production and the observed inflation rate based on the Consumer Price Index (CPI), the two components of the National Pricing Formula. In a press release, the CDC calculates from the formula that the price of milk should decrease marginally. My calculations produced a much larger decline. In this post I seek to explain the difference.

The CDC released on October 4, 2024, the 2023 Cost of Production study. After indexation, it finds that the cost of production for one hectolitre of milk dropped by 2.93%. I found in my September post a 6% drop. Regarding inflation, in its press release, the CDC notes that the CPI increased 2.89% for the 12-month period ending in August 2024, while I found that it increased by 2%. Using the National Pricing Formula, the CDC finds the price of milk should decline by 0.02% in February 2025. I had found that it should drop by 2%.

I was surprised of the difference between the CDC’s cost estimates and mine. I find there were a few details of the CDC calculations that I did not understand or assumed different. Accordingly, I corrected the way I estimate milk production costs. For the change in the CPI, I did not think I could get this wrong. It turns out that the CDC calculates the change in the CPI in an unconventional way.

This will be very wonky as I will be going into some technical details. I am critical of some methods used by the CDC but I’m not arguing for the CDC to modify its procedures. Using a formula to adjust prices is always flawed in many ways. However, if the National Pricing Formula and the whole process that surrounds it are accepted by stakeholders, then I do not see the point in modifying them. I think consistency over time is more important to the credibility of the process than getting every little detail right. Nonetheless, improvements to the process should be considered every five years or so to keep up with new realities.

Cost of production estimates

The CDC makes available the results of the annual cost of production surveys at this page, along with a publication that describes the process of estimating the cost of production. The process has become a lot more transparent over the last two years. I have been slowly gaining a deeper understanding of the whole process, and accordingly improving my models.

I go over the details I did not get right in my September post.

Indexation

One detail I did not get right was the reference period the CDC uses for the indexation. I believe the CDC refers to this as the base period in its cost study, although it does not define it. Previously I had used the three months ending in August of the year of a cost of study. But after asking the CDC, who kindly responded very quickly, the reference period is the whole year, reflecting that farms report costs for the whole year.

Correcting this had a small impact on my calculations. I am able to replicate CDC’s indexation. However, I made a small change to the indexation: I index transportation costs using data from Statistics Canada table 18-10-0281 because I do not have data for P5 transportation costs.

Comparing costs across years

When calculating the latest annual change in the production cost, the CDC compared the indexed costs for the three months ending in August 2024 to the indexed costs for the three months ending in August 2023 as calculated in 2023. I previously thought the CDC used the 2023 cost study to index the costs in the three months ending in August for 2023 and 2024.

Although I find the practice of comparing costs from two indexed cost studies questionable, I do not think it’s wrong. This approach has the advantage of consistency because it compares indexed costs. However, it relies on an older study to evaluate costs in a year for which the results of cost study are available. That is, the current practice is to use the 2022 cost study to estimate costs in the three months ending in August 2023. However, the 2023 cost study could be used to index costs in the three months ending in August 2023. I consider this would be a better choice because it uses the most recent and best information available.

Purpose and use of the cost study

The CDC conducts its cost study with the purpose of estimating the annual change in milk production cost. I adapt the cost study to estimate costs month to month. Although I index costs following the CDC’s method, I make several modifications to follow costs through time.

I adapt the CDC methodology by incorporating information from multiple cost studies. The annual study surveys a small, but reasonably large, number of farms. Two hundred and fifty-four farms in the last survey. Incorporating information from previous cost studies is akin to increasing the number of farms surveyed in a given year. That is, previous studies contain information that is useful in estimating costs in the current year and to forecast them. I combine multiple cost studies using regressions on the weights calculated for the indexation of costs.

Figure 1 compares estimated monthly costs for the CDC annual indexation method to my method which combines multiple cost studies. The two methods track quite well until 2020 and begin to differ more significantly afterwards. The pandemic year appears odd with costs jumping up using the CDC method. For that reason, I assigned a lower weight to 2020 in my regression models. Notice also the jumps in the cost estimates across years when following the CDC method. This is most obvious from 2018 to 2019, from 2019 to 2020 and from 2020 to 2021. My method tends to smooth out cost estimates, which is appropriate given my objective to estimate costs through time.

Figure 1: Estimated milk total production cost

The biggest reason for the difference in the estimated costs between my calculations and the CDC’s was that I combined the results from several studies to estimate costs. You can observe in Figure 1 that difference for the three months ending in August 2024. Note also that I did not have access yet to the 2023 when estimating costs in September. The indexation of costs in Figure 1 includes data from the 2023 cost study.

Since my September post, I modified the regression models for the estimation of the indexation weights. If you compare my cost estimates in Figure 1 to those in September, the main difference is what happens once the costs are no longer calculated from actual indexes but forecasts of indexes. I believe the new regression models give a more appropriate index weighting scheme than the previous ones.

Consumer price index

According to Statistics Canada, the CPI rose 2.0% on a year-over-year basis in August 2024. I was quite surprised to find that the CDC considers that it increased by 2.89%.

The way Statistics Canada calculated the annual change in the CPI is by taking the value of the CPI in August 2024, 161.8, and the value of the CPI in August 2023, 158.7, giving a 2% inflation rate (161.8/158.7 - 1 = 0.0195). This is the conventional way of calculating the change in the CPI. I figured out that the CDC took the mean of the CPI between September 2023 and August 2024, 159.9, and the mean of the CPI between September 2022 and August 2023 (155.4) to find a 2.89% increase in the CPI (159.9/155.4 - 1 = 0.0289).

I cannot say that the method used by the CDC is wrong but it is certainly unconventional. I’m not able to find good reasons to proceed that way. Seasonality should not be an issue as it is an annual rate that is calculated. Taking averages smooths out monthly shocks but large sudden shocks rarely ever happen on the CPI, in particular when calculating an annual rate.

Looking back at previous announcements, I found that the CDC has used the same method to calculate the change in the IPC for the last three years. I cannot confirm for 2021 and prior as changes in the CPI are not mentioned in the announcements.

If the CDC sticks to the same calculation method in the long run, I consider that it is acceptable, although the method is unconventional. Given enough time, Statistics Canada and CDC calculation methods should yield the same average annual change in the CPI. What the CDC cannot do, for credibility reason, is being inconsistent in the way it calculates the change in the CPI.