Undoubtedly, everyone would be benefited from knowing the future. Everyday, we plan our activities by referring to the weather forecasts. It reflects how important a prediction is. If you invest, no matter in stocks, real estate, bonds or commodities, price prediction is the way to make profit.
However, the difficulties of making accurate predictions are commonly recognized. In financial theory, unpredictable prices is considered an evidence of an efficient market — Efficient Market Hypothesis (EMH).
But this article is not about predictions, but post-dictions. A new term invented here to discuss about a kind of forecast of something happened. It sounds illogical because if something had happened, its information is ‘history’.
Yet, nowadays as the statistics are so complicated that they normally lag behind a few days to a few months before the statistics can be compiled and announced. For example, GDP figures are commonly released in a quarterly basis in at least one-month lag. Housing price index is another good example that it normally lags by at least 2 weeks. Internationally, it lags by about 2 months.
Since the market is strongly affected by the announcements of these important statistics. A post-diction of the statistics can still help make a profit in the market. Take Hong Kong Housing Price Index as an example, the official one released by the Rating and Valuation Department of the government is usually announced in the first week of each month. But the announced index figure is the one of 2-month ago.
Can we post-dict accurately? Let’s try. For example, Figure 1 shows the house price index (HPI) of Hong Kong from July 2019 to December 2020. The latest released figure of All Class HPI in Feb 2021 for the index of Dec 2020 is 379.3. It is subject to further adjustment next month as indicated by the asterisk. It continues the downward trend from mid 2020, but a year-on-year comparison shows almost no change in this year.
How would you post-dict the figure for January 2021?
In fact, we can use typical prediction tools to post-dict, or gather a lot of information from the market to post-dict, such as the CCL index which reports weekly, or the trading volume figures (i.e. number of housing transactions), or the stock price changes of housing related sectors. Also, we can carry out sentiment analyses of the housing markets in the past few weeks.
- Excel Forecast Sheet
Let’s start by using the simplest forecast sheet in Excel to post-dict the trend. Figure 2 shows the Exponential Smoothing (ETS) forecast using Excel forecast sheet. I have incorporated 6-month seasonality function. If you forget how to use the Excel forecasting function, please refer to my article Yiu (2019). The results show a continuous downward trend in the first few month of 2021. For example, the index is post-dicted to be 379.0 (-0.09% mom) and 374.0 (-1.08% mom) in January and February 2021. We can check their accuracies next month.
Besides forecasting by level terms (HPI), it is sometimes more plausible to forecast by the year-on-year change (HPI-yoy), especially if you are studying the housing price cycles. Figure 3 shows the HPI-yoy and indicates a -0.5% yoy fall in January 2021. This is the ETS forecast with a 6-month seasonality function. The result can be quite different for different parameters. For example, the automatic choice of 9-month results in a bigger fall. Yet, the patterns of the forecasts in level terms (Fig 2) and in yoy (Fig 3) seem to be quite similar.
2. CCL index
Figure 4 shows the weekly Centa-City Index which is about 2-week lags. In other words, the index on 2021/01/18–2021/01/24 probably reflects the market price in the first week of 2021. It records 175.04 which is lower than the two previous weeks 176.16 and 176.52. Yet, the index seems to be rebounding since the 2nd week of 2021. If we take the average of the indices in the first 3 weeks of 2021, it is 175.93. In comparison with the average of the previous 4 weeks, that is 176.76, it shows a fall of about 0.5% month-on-month in January 2021. Putting back into the RVD index, i.e. it post-dicts the January index at 377.4, which is even lower than the ETS forecast.
3. Trading Volume
In the past, trading volume changes shows a strong and positive correlation with housing price changes. But the association is weakened in recent years due to the special stamp duties and mortgage policy interventions, etc. Anyway, as the trading volume data is only lagging one month, it provides January data as shown in Figure 5. It shows a sharp reduction in the number of 1st hand sale in January 2021 (from 2,111 to 586), whereas the number of transactions in the 2nd hand market is about the same. Since the fall happened in the 1st hand market only, it is more likely to be due to the lack of supply in the 1st hand market rather than a poor market sentiment. In other words, the information from trading volume implies a horizontal trend of housing prices similar to the trading volume in the 2nd hand market.
Both the ETS forecasting tool and the weekly index post-dict a moderate downward trend, and the trading volume model post-dict a steady trend. A reasonable guess is a minor downward adjustment of about 0–0.9%.
You may further try using the sentiment analysis or referring to stock market to post-dict the January 2021 HPI of Hong Kong. Let’s discuss the answer early next month when the official index is announced.
Yiu, C.Y. (2019) 2 forecasting methods by excel, Medium, Apr 15. https://ecyy.medium.com/2-forecasting-methods-by-excel-1991b5b7d111