In this series of posts, we explore using the analysis of news to help determine medium term trends in the economy, with a focus on the timing of economic cycles, investment allocations, and recessions. Our general conclusion, perhaps unsurprisingly, is that news can be used to help market professionals better understand broad economic and financial conditions and trends. Further, news indices calculated for a variety of countries are shown to compare favorably to leading economic indicators in providing insights into future economic activity.
Our study is unique in a number of important ways.
- Focusing on medium term trends and the economic cycle is important as it allows news analysis to be used for things such as economic forecasts and investment allocation decisions.
- Widening the scope of news to many different countries creates a more global perspective, and allows for confirmation of significant trend shifts, by comparing the timing of inversions.
- Looking at many decades of news encompasses a wide variety of market and economic conditions, and gets around the potential problem of fitting the analysis to dominant themes of one particular cycle during a confined time period.
In basic terms, we take lots of news, from a variety of sources in a variety of countries, some in English and some not, over an extended time frame to test the idea that news tends to trend in such a way so as to reveal economic and financial conditions which in a broader perspective form cycles.
Like every form of analysis, it has advantages and disadvantages. And, we should strongly highlight the fact that, to date, no panacea-form of analysis has been created and that professionals should incorporate a variety of information into any analysis or decision.
Advantages:
- Commonality – news brings an extremely disparate, diverse, and global group of investment market professionals together in the sense that they all stay up-to-date with the news and presumably incorporate it into their outlook. Such commonality makes news an important factor in determining any perceived trend.
- Novel Yet Recognized Data – a systematic approach to incorporate news into a broader economic and financial analysis adds novel data and insights and helps to improve overall analysis. Interestingly, the approach and output are novel, yet the input (news) is an obviously recognized data source.
- Depth of Data – news in general provides a very deep source of information. Our study looks at news starting from the 1960s in the US. Additionally, we look at news from a variety of Developed and Emerging Markets in a variety of languages.
- Identifying Important Inversions – news tends to go to extremes around important inversions of the economic cycle. Such extremes tend to make the timing of important inversions easier.
- Large Divergences – at times, such as near the end of a significant bull or bear market, the news indicator can create large divergences which can be used as an alert of an impending inversion (which might resemble monthly RSI divergence with the country’s equity index which often appear around similar times).
- Time Series Confirmation – news time series, as we will see, can be created around different themes as well as using different news sources from different countries. Around important inversions, these time series tend to confirm one another (for instance many turned up from a very low level at the beginning of 2009 as economies and markets were passing their worst).
- Highly Cyclical Economies and Markets – the methodology used tends to correctly identify key inversions during highly cyclical periods or conditions, and can be useful when looking at cyclical economies, industries and companies.
- End of Recession Signals – compared to popular recession probability indicators, news indices does an exceptional job at determining the end of recessions.
- Works Well with Other Indicators – news trends seem to work well in conjunction with other market and economic indicators in that it helps to put conditions in context and also helps to confirm larger trends.
- Daily Updates – news is updated in real-time, and end-of-day analysis is fairly standard. Daily data offers a rather important addition to broad economy-wide analysis – for the most part, investors go to monthly data for similar insights which are also often delayed, making daily news insights an extremely attractive alternative.
Disadvantages:
- Volatility – news, as it updates in real-time, tends to produce a fairly volatile time series, which can be difficult to analyze. This is a classic problem with real-time data.
- Difficulty with Mid-Cycle – somewhere around the middle of a longer term bull market or economic expansion, oscillators can lose some of their accuracy. The way we have constructed the main news time series makes them work more along the lines of oscillators, making them somewhat weaker signals around this mid-cycle period.
- Divergence between Economy and Market – normally, economic and market trends tend to move in unison, which creates a supportive environment for leveraging the news for both economic and financial trends. In cases when there is a divergence between economic trends and market trends, a consolidated news time series can lose some of its accuracy – think of a ‘goldilocks’ economy that has somewhat lower growth rates but with a steady outlook coinciding with a booming stock market. Another example, is from 2001/02 when many economic indicators signaled marginally improving conditions while most equity markets hit their respective lows only in late 2002 or early 2003. These divergences between economic and equity trends can at times produce unusual signals from news trends.
- Interpretation Suggested – sometimes a peaking news time series is a red flag while at other times it could mean the economy and/or market could begin to settle in for a less volatile up trend. A mid-cycle environment is such an example. As is the case with the vast majority of indicators and analysis methods, news trends should be put into context and interpreted along side other indicators.
- Global Influence on National News – the impact of global trends on a nation’s news can be substantial, which can distort trends at times if their respective cycles are not in unison.
- Uniqueness of Cycles – each economic cycle is unique, which could produce distortions in the analysis of news.
- Changing Language – language is not static, with new terms emerging and meanings shifting over time, all of which can produce issues with longer term analysis.
- No Panacea – news trends is an exceptional tool, but alas not perfect as we will see.
- Uncommon – not many market professionals use news analysis and as such there is a general barrier to entry. In your morning meeting, if you say the ‘200 day moving average turned up’ or ‘consumer confidence came out better than expected’, people will immediately understand your point, while commenting on the trend of news is still uncommon.
On balance, many of these advantages and disadvantages are similar in scope to other economic and market indicators and analysis techniques. From an investment professional’s viewpoint, perhaps the largest advantage is simply that analyzing news trends offers an additional metric with which to better understand the economic and market environment. Additionally, the fact that its input is text makes the form of analysis rather unique in the broader spectrum of tools available to investors, economists and analysts.
In this series of posts, we focus on three broad datasets. First, we look at exclusively US news, starting from the 1960s. Having a multi-decade news dataset has allowed us to analyze many different economic and market conditions, helping to show that news analysis can add value under a variety of circumstances. Second, we analyze Brazilian news, starting from the early 2000s. This is of interest as we wanted to test these concepts not only in a different country, but one with a different language and different economic and market dynamics (Brazil being a good example of an Emerging Market that is more commodity driven while the US being a good example of a Developed Market that is more technology driven). Third, we look at a combination of news from many different countries, starting from the early 1990s, which we refer to as Multi-Country. Finally, we pull together the various time series and create an estimate for Global News.
In order to test news time series, we compare them to a number of generally recognized economic and financial time series. For macro-economic trends, we use leading indicators, which over many different economic cycles have done an exceptional job of providing investors insights into important economic inversions ahead of actual turning points. The main problem with leading indicators, like almost every macro-level indicator, is the time delay between the creation of the underlying data and the leading indicator’s release to the public. News has the potential to offset this issue as daily time series do not create such delays and can offer insights into the most recent underlying conditions and therefore reduce problems with reliance on delayed monthly macro data.
For financial markets, we use three equity market indices – the Dow Jones Index (DJI), the Hang Seng Index (HSI), and the EWZ (the Brazilian USD ETF as a proxy for the Brazilian equity market). The DJI works well for analyzing long term trends. The HSI is a long standing Emerging Market index, that tends to be more interest rate sensitive. And, Brazil, via EWZ, is a long standing Emerging Market that tends to be more commodity driven. All of these instruments are sensitive to the global economic cycle.
We also look at the news cycle in terms of the main positive and negative economic / financial themes that had occurred during that particular cycle. Unsurprisingly, we find that very positive themes tend to develop during up-cycles in global news and negative themes during down-cycles. Improving news indices are generally associated with bull markets and economic expansion while deteriorating indices with bear markets, recessions, and crises. Obviously, better understanding news trends would therefore significantly benefits investors.
Lastly, we test how news indices did in comparison to popular forecasting tools in terms of timing the end of US recessions. Though many methods exist, we focus on the US Leading Indicator (source: OECD), Smoothed U.S. Recession Probabilities (source: Federal Reserve Bank of St. Louis), Real-time Sahm Rule Recession Indicator (source: Federal Reserve Bank of St. Louis), and Probability of US Recession Predicted by Treasury Spread (source: Federal Reserve Bank of New York). US News Indices performed extremely well against these well-known tools, and on average tend to time the end of US recessions as well or better, which is even more impressive once you consider news updates daily in comparison to monthly delayed macro data on which most of the popular recession probability tools are based.