Making Statistics Talk 2 – Identifying the key message PPT2 CONTRACTOR IS ACTING UNDER A FRAMEWORK CONTRACT CONCLUDED WITH THE COMMISSION
Summary Identifying the data changes to be covered Identifying and communicating the key message Identifying contributions to aggregate changes The identification of trends / turning points in data Identifying a key message in a data set
Aggregate data changes to cover Direction, size of aggregate change over year (or other reasonably long period) Identify change in data since last period Place data in context of at least a year’s observations
Short-term changes Look at observation to observation (month to month) data changes Identify impact on most recent & baseline observations Large, symmetrical changes may be ephemeral
The more that data is disaggregated, the less accurate it becomes
Disaggregated data changes Identify main sector contribution(s) to change i.e. not the largest change in magnitude but, for example, the largest (sector share) x (change) Ranking of Member States is often structural May not be of interest in Main statistical findings Absolute values of outliers of interest Martin Wolf: ‘only three countries have negative inflation and only a fifth of items in the consumer price index have fallen in price.’ Changes in ranking can be of interest
Trends and changes Fitting trend lines & moving averages to data helps understanding of data One observation does not usually mean an end to a trend Even if a new secular trend is established, cycles (seasonal, business cycle) will often overwhelm Basically, very unlikely to identify a new cycle without applying auto-regression methods
Exercise 3 Identify the main aggregate and disaggregated data changes of interest in HCPI data, either the series we have been discussing or at Member State level Data provided (EU, Euro-zone, Member States): m/m-12, m/m-1 aggregate and m/m-12, m/m-1 at COICOP top level (12 groups) 20 minutes, 20 minutes discussion
Identifying the main story Two possibilities: Biggest effect Versus Biggest surprise HICP example: ‘Headline’ (all items) inflation rate unchanged Versus Increase in ‘core inflation’ (ex. energy, food, alcohol, tobacco)
My choice: ‘unchanged inflation’ I am more certain about the behaviour of the aggregate data Based on the aggregate figure, I can develop a narrative that includes the (apparently conflicting) disaggregated data effects: ‘Core’ inflation increase; fall in month to month inflation; clothing price fall (potentially aberrant data); number of member states with falling price levels; price stability of these countries Result: a cohesive narrative that incorporates conflicting data changes
Summary of main story selection Main effect v. main surprise Focus on aggregate data Incorporate disaggregated data effects within this narrative Ensure uncertainty about disaggregated effects is appropriately identified Avoid reference to external causes / analysis Can use analysts concerns to identify areas of interest
Exercise 4 Identify the main story for a Member State’s HICP data Note how you would incorporate other data developments within the narrative 10 minutes 10 minutes discussion
Thank you CONTRACTOR IS ACTING UNDER A FRAMEWORK CONTRACT CONCLUDED WITH THE COMMISSION