Charlie Nelson has over 40 years experience in forecasting, statistical analysis and modelling, and market insights. As an employee or a consultant, he has worked in a wide range of industries including telecommunications, transport, automotive, retail, packaged groceries, media, and weather.
He has used a range of forecasting techniques such as regression, Box-Jenkins univariate and multivariate time series models, and multinomial logit models for discrete choice.
While models are a very important component of forecasting, Charlie also uses indicators and judgement in developing forecasts. Indicators can be inputs to both models and judgement and a broad knowledge base is essential in exercising judgement. He has developed unique and powerful indicators and is researching ways of capturing the opinions of people with good judgement.
Charlie has a B.Sc. degree majoring in statistics. He founded Foreseechange in 2000 and he was employed by Nielsen before then.
Development of leading indicator forecasting models
Over recent years, he has developed unique, powerful, leading indicator models for consumer spending. Consumer spending growth is influenced by economic factors, such as income and wealth. But these factors on their own have limited explanatory power and limited predictability.
Existing consumer survey-based indicators, such as the consumer sentiment index (sometimes referred to as consumer confidence), have low predictive power.
Charlie has identified the missing psychological factor in such indicators and developed a proprietary measure for it. It is willingness to spend and is part of a suite of measures collected since 2003.
Consumer survey measures of willingness to spend, combined with survey measures of expectations about income and asset values has proven to be a powerful indicator of spending over the subsequent six to 12 months. In some categories of spending, economic factors which lead spending are also incorporated into the leading indicator models.
Leading indicator models of business spending have also been developed for advertising spending. Such models may be capable of adaptation to other categories of business spending.
Predicting Australia’s economy in 2008 and 2009
In 2007, Charlie developed a scenario for an economic slowdown in Australia in 2008. He identified three factors which could combine to cause a serious slowdown.
One of these was interest rates, which had been rising since 2003 and there were warning signs that consumers interest payment burden was becoming too heavy. Charlie thought it likely that the Reserve Bank of Australia would continue to push up interest rates and that the negative impact on consumer spending would occur from mid-2008. The Reserve Bank did, indeed, raise interest rates – in August and November 2007 then in February and March 2008.
Another factor was the sub-prime home lending issue in the USA. It was considered likely that this would have a negative impact on the US economy during 2008. Both directly and through China, this was expected to negatively impact Australia’s economy. And so it did, with the collapse of Lehman Brothers in September 2008 signalling the arrival of that impact.
The third factor was the Olympic Games in China in August 2008. Charlie had visited Beijing in April 2007 and witnessed the reconstruction of the whole city – housing, offices, retail, and transport. This would all come to a halt by August 2008. As a major supplier of resources to China, this was expected to impact our exports.
These three factors did have a simultaneous negative impact in late 2008. That the vast majority of economic forecasters did not predict this has been well documented.
Australia responded with both fiscal and monetary stimulus. By April 2009, Charlie’s consumer survey measure of willingness to spend had risen from a record low to a record high and he predicted that Australia would avoid recession. For several months after that, both official and private economic forecasters continued to predict a recession which never happened.
This was documented by one of Charlie’s clients, Harold Mitchell, in the Age and the Sydney Morning Herald on 30 April 2009.
Predicting an end to south eastern Australia’s long drought
Curious about the reasons for the 1997 to 2009 drought, Charlie began researching drivers of rainfall in south eastern Australia. Climate scientists had concluded that the cause was climate change and that it was a permanent change. While accepting that climate change may well be one factor, the abrupt nature of the changes suggested that there may be other causes.
Analysis of the historical rainfall data suggested some cyclical influences. Charlie’s knowledge of astronomy led him to conclude that two separate decadal astronomical cycles were having a negative impact on rainfall and that the drought would break.
In early 2010, he wrote to climate scientist Professor David Karoly, and several water authorities with a prediction that the end of the long drought was imminent. It did indeed end later in 2010.
Climate scientists do not accept that astronomical cycles can influence weather so Charlie is currently researching the chain of causality.
Predicting New Vehicle Sales
On Australia Day 1998, Business Review Weekly magazine published Charlie’s forecast for new vehicle sales in 1998. It was a number that “staggered” industry executives and one CEO asked what Charlie had been smoking – which the writer Bill Tuckey translated to a question about the brand of tea was used.
The industry consensus was that sales would be close to the 1997 total and possibly slightly lower – around 720,000.
Charlie’s forecast, based on a model he had developed, was for between 770,000 and 820,000 depending on employment growth.
The outcome was 808,000, towards the top of Charlie’s predicted range and 12.2% above the industry consensus.
The model used was a Box-Jenkins transfer function model. It incorporated economic factors and an adaptive trend component which captured unspecified trend drivers. The model is still in use and providing accurate forecasts in 2016, although some additional economic factors have since been identified and included.
Charlie is currently evaluating using consumer survey data to predict macroeconomic outcomes. The data, which goes back to 2005, asks respondents to estimate the likelihood of particular events occurring over the next 12 months. It has been found that there are biases in these responses (consistent with many studies in behavioural economics).
Not all respondents display bias, however, and with carefully chosen control questions those respondents with the least bias can be identified.
These “wisdom of the masses” surveys are showing good potential.