by János Gáspár Buildecon, Hungary
Nowcasting, short forecasting the project based approach
Credit to The National Bank of England for the fan chart and Cosgrove Hall Films for Jamie and the magic torch.
When it comes to construction market size expressed in value, the typical questions Statistical offices ask when they gather the information are:
- Who invested into what (investment statistics)?
- Who built what (construction statistics)?
In both cases you have preliminary data and final data. Because of the tradeoff between timeliness (freshness) and coverage
(fullness) you have to live with limited information for a while. Final annual investment figures in Hungary become available
t+10 months, while final annual construction figures are out t+9 months, where t is latest full year. Furthermore, the scope (depth) is also subject to change, segment level breakdown (office, hotel, health,
road, railway etc.) is only available as part of the final data.
For all these reasons, Buildecon has been aiming to find a third leg which helps understand the present and to some extent the future performance of the segments better. This research we carry out together with ibuild.info and ELTINGA is about
- systematically processing Hungarys leading construction project information database.
The ibuild.info construction project database about Hungary is something similar to the service provided by Docugroup for Austria, Germany and Switzerland, by Dodge for North-America or by Cordell for Australia and New Zealand. All of these are market based services, providing information on project level to each (potential) stakeholder. Typically, they provide detailed information about past, ongoing and planned projects, covering not only the parameters of the projects, but also contact information. From our viewpoint, their most important feature is that
- project databases are updated daily with incoming new information, so they are very-very timely.
Out of the ibuild.info database, our analysis focuses only on those projects which are currently (at any point in time) under construction or were under construction (completed) from 2000 on. And the parameters of each project we use are
- Start date
- (Planned) Finish date
- Value (net construction cost)
- Function (residential, hotel, office, retail, road, railway, etc.)
- Type (new vs renovation)
- The algorithms we have created aggregate the data of individual projects to segment level indicators.
Visually, the indicators look like this:
The observed period can be a quarter, a half-year or a year. Since we create time series for all indicator types, they become comparable in time. And here it starts to become really interesting. To see it in practice, let me show some parts of the latest EBI Construction Activity Report, which is the second one, and is about H2 2016.
1. Activity-Start indicator e.g. shows that in H2 2016 construction works started in the case of 159 office projects. Besides,
we show that the total expected construction value of these works was recorded to be more than HUF 60 billion in the second
half, which is around 40% larger than the corresponding value in H2 2015.
2. So this Activity-Start indicator is about the projects where construction works began in a given period, in this example in H2 2016. The new and renovation split shows that 25 new and 134 renovation office projects were started and the corresponding value of these works were registered to be around HUF 40 billion and HUF 20 billion. New construction experienced a growth of almost 3 times in value, while there is a drop in started renovation works compared to H2 2015.
3. Beside Activity-Start indicator, which is a sort of forward looking indicator, we calculate Output, which is a coincident
type indicator. It is about the amount spent on construction in a given period. In the current report we included the yearly
figures for Output and we show that according to our records almost HUF 100 billion were invested into office construction
(as total, but new and renovation are calculated separately as well) and that this represented a 9% growth in 2016 over the
4. For Output, we included time series as well into the report, splitting them to new and renovation. From nowcasting viewpoint, these time series are the most important, telling us how strong the connections are between historical figures of the statistical office and segment Outputs coming from project aggregation. (Top 5 ongoing projects from the database were also included for each segment).
So here we stand now with this research. We have segment level information within days after the latest observed period. This
helps nowcast and foresight. But being as honest as The Bank of England, I must add that project databases are not exempt
from revision either. New projects become known to have been started and project parameters become more accurate with time.
The extent of this revision was discovered thanks to this research. And it also has very important consequences. On the one hand, we are able to give feedback to the data owner where to improve project research. On the other, by understanding the nature and the probable directions of the revision, we learn more on how to incorporate the findings into forecast. Parallel to this, with continuously monitoring how revision affects the aggregates in time, we will be able to set an optimal compilation date, after which the aggregates only change insignificantly.
This optimal compilation date is currently under examination, but this new torch we have by aggregating construction projects is in any way a great help in a relatively dark room.
Additional thoughts: In the methodological section of the report more information is available on the classifications applied, on the coverage and on the possible queries available, among others. And here I would like to say thanks to Gergely Répássy (ibuild.info) and Áron Horváth (ELTINGA) for the common work that has brought us to this point.