The following image shows one cell in a correlation matrix at the intersection of GDP B and Citizen Education spending.  Note that when GDP B rises, Spending on high school and masters degrees increases, while spending on associate and bachelors’ degrees declines. 

Note in the following correlation cell of Producer Price Index against Education Spending there a similar relationship, but shifted to the left, or more negative for each level of education.

This third correlation cell shows Agriculture Employment against Education Spending with exactly the opposite relationships at each level of Education.

The ReNews team is in the process of examining a list of such patterns across society in order to :

1. 1. Evaluate whether each pattern is beneficial or harmful across American communities.
   2. Identify whether there is opportunity to magnify positive forces, or counteract harmful ones. 
   3. Commission original Context Journalism to cover the causes and human impacts of the most damaging patterns.
   4. Document and monitor policy proposals in each domain, and the vested interests supporting or opposing each proposal.
   5. Record patterns of Citizen support and opposition across their deep but anonymous ReNews belief profiles, and help Citizens determine whether their beliefs align with their best interests.
   6. Map media campaigns attempting to influence Citizens in one direction or another, and the vested interests and budgets behind the campaigns.

This is only one example of the methods ReNews Context Journalism applies to identify and cover patterns within societal performance which might indicate inequity or policy opportunity.

Increases in some of these primary measures are often accompanied by increases in secondary variables.  For instance, when Employment increases beyond a level near full employment, the competition between companies to hire employees results in competition in wages.  This in turn leads to increases in average wages across higher growth industries, and sometimes the entire economy.  In other cases, industry demand exceeds supply of qualified Citizens coming out of the education system motivating policies to support immigration and foreign workers.

The ReNews Context Engine and non-linear experience of news allows Citizens to explore the complexity of these relationships at their own pace, and according to their own interests.  There is no algorithm because there are no ads and no subscriptions to motivate an algorithm.

Levels of Correlation

There are three ranges of correlation visible in these heat maps:

1. 1. Highly positive correlations – relationships with correlations at or near +1.0.  In these cases both variables consistently move in the same direction, up or down, together.
   2. Highly negative correlations – relationships with correlations at or near -1.0.  In these cases both variables consistently move in opposite directions, up or down, from each other.
   3. Middle range correlations – relationships with correlations between +0.67 and -0.67.  In these cases the variables move less consistently, whether in the same direction or opposing directions or not at all.
      1. From 0 to +0.25 and from 0 to -0.25 there is very little consistency in movement.
      2. From +0.25 to +0.67 and from -0.25 to -0.67 there is moderate consistency in movement which Renews will explore in more detail.

Cross-correlations Expose the Obvious

Cross-correlations of economic variables against each other reveals which variables are consistently aligned, opposed, or influenced by other factors.

Some of the strong positive and negative cross-correlations which appear logical include:

• Correlation of GDP growth with employment growth.
• Reduction in manufacturing production correlates with increased importation of durable goods.
• Increased inflation negatively correlates with discretionary purchasing but is not significantly correlated with essentials.

While these are not new insights, ReNews believes that statistical validation of core truths can go a long way toward anchoring all Citizens in a common set of agreed-upon societal information.  That, in turn, can lead to better collaboration in the policy development process.

Unexpected Correlations

Sometimes Renews’ multi-variable correlation analysis or ‘correlation mining’ turns up unexpected relationships which are not intuitive.  In these cases, we let our news and data analysts apply higher level math to dig deeper into the meanings of these relationships in the ReNews hypergraph.

For example, the apparel and miscellaneous categories of personal expenditures are less directly correlated to increases in positive economic indicators than other categories and in some cases are negatively correlated.

We select these observations for further investigation because they differ significantly from the correlations in other categories.  We plan to consider likely inflection points we should see in the data, like:

1. the emergence of “Fast Fashion” and
2. changes in overall apparel average price due to increased secondhand sales and rentals via the internet. 

Citizen Profile Correlations

When ReNews earns the trust of a sufficient number of Citizens, we will be able to analyze the population of anonymized Citizen belief profiles for:

• correlations of their profile data with their policy positions. 
• the demographic and belief mix within clusters of Citizens aligned with or against a position.
•  correlation of cluster membership with societal and economic data. 
• correlation of policies with Citizen outcomes to reveal patterns of Citizen best interests to guide future policy development.  

When ReNews reaches this threshold of membership it will suddenly become more valuable than consumer surveys, which have been rapidly declining in reliability since the advent of cell phones.

It would be difficult to architect a better means of engaging reliably with Citizens about their beliefs than in a

1. completely transparent news and contextual information platform,
2. in which each Citizen feels protected by perfect anonymity.

Regression Measures a Mix of Correlations

Once we determine which separate variables are correlated with a Citizen outcome, ReNews is able to look one level deeper to see which of the correlated variables explain more or less of the change in Citizen income.  To do this we apply regression analysis.  Instead of looking for correlation between two variables, regression analysis looks for a mixture of correlations between several input variables and a single output variable.  The result shows the percentage that each input variable contributes to determining a change in the output variable, Citizen income in this case.