Meditating on History

Fairfield High School

Meditating on History

U.S. History IV

Matt Patton

Robert Chapin

In January of 1993, the Institute of Science, Technology and Public Policy at Maharishi International University and Citizens for a Crime Free D.C. proposed the National Demonstration Project to Reduce Violent Crime and Improve Governmental Effectiveness in Washington, D.C. The purpose of this National Demonstration would be to test the hypothesis, “that group practice of the Transcendental Meditation and TM-Sidhi program can positively influence an entire society in objectively measurable ways,” and, “to alert the leaders of society to the existence of such a technology.” (Hagelin et al., 1994)

The hypothesis for this Project was based on previous studies of social stress. The underlying theory is that social problems arise from collective stress in the society as a whole. By relieving stress in individuals, the level of stress in the whole society can be decreased. Therefore, if a proven stress reduction technique were to be practiced by even a relatively small group of individuals, there should be a dramatic reduction in the collective social stress. Scientist and educator Maharishi Mahesh Yogi has recently uncovered such a technique called Transcendental Meditation (TM). TM has been found to promote, “regulation of cortisol and other hormones in the opposite direction to chronic stress.” A number of studies have found that “chronic stress produces abnormally low levels of serotonin, as well as abnormally high levels of cortisol, changes which have been directly correlated with aggression, hostility, and impulsive, violent behavior.” (Hagelin et al., 1997)

$4.2 million were sought from the Washington, D.C. crime budget to finance the Project, but the funds were eventually raised through private donations. The Demonstration was organized over the course of only one month. During this time accommodations were made for 5,000 experts in the practice of TM from 80 different countries. (“Fighting Crime,” 1996) The size of the group increased as participants arrived in three ‘tiers’ over the eight week period of the Demonstration. From June 7 to July 30 the experts, “meditated in the morning until lunch. Then there were conferences in the afternoon, and we meditated until dinner.” (Weiselberg, L., personal communication, June 2, 1998) For experimental purposes, “There was virtually no behavioral interaction between the members of the coherence-creating group and the population deemed to be affected.” (Hagelin et al., 1997) This was done to support a Field Theory of Consciousness, which suggests that the effects of this group's crime ‘intervention’ could be experienced at a distance.

An Independent Project Review Board was established and met with the Institute on the second day of the Demonstration. This Board was organized, “to (1) advise the Institute on the research design, including specification of the dependent and control variables, (2) oversee the research process, (3) independently evaluate the data if they chose to, and (4) give critical feedback on and ultimately approve the research reports on the project.” (Hagelin et al., 1994) The Board consisted of 27 sociologists and criminologists from across the country. It approved the Institute's research protocol on June 17.

The research hypotheses stated that the Demonstration would “create coherence in collective consciousness as indicated by: (1) Decreased violent crime [defined as homicides, aggravated assaults, rapes, and robberies], (2) Improved quality of life [an index comprising 911 calls, suicides, fatal accidents, reported child abuse, etc.], (3) Increased governmental cooperation and efficiency, indicated by increased success of and support for President Clinton.” (Hagelin et al., 1994) The hardest part of testing these hypotheses would be proving whether or not there was a statistically significant reduction in violent crime. The quality-of-life index could be measured once that information was available. Support for the President could be found based on the results of 86 public opinion polls that were conducted before, during, and after the Demonstration.

In September, 1994 the Independent Project Review Board approved the Institute's preliminary report, Results of the National Demonstration Project To Reduce Violent Crime And Improve Governmental Effectiveness In Washington, D.C. The report explained the results from the analysis of unofficial crime data, the public opinion polls, and the quality-of-life data that was available at the time. The report also included background information on the Project, and discussions of some of the technical decisions made in analyzing the effects of the Demonstration.

The preliminary report was easy to interpret despite its technical nature. To estimate the effect (if any) that the Demonstration had on crime rates, four statistical models were designed. These models accounted for all of the variables that were found to have a statistical significance in relation to crime. The first model was based on unofficial daily crime data for homicides, rapes and assaults (HRA). The available preliminary data for robberies was not limited to Part 1, or violent crimes, and therefore was not included in the report. The available data was further limited by the finding of a “discontinuity of model structure between [1992 and 1993].” This discontinuity could not be accounted for, so the data collected for 1988 to 1992 was disregarded. This should have been thoroughly discussed, investigated, and corrected, however the preliminary nature of this report makes the issue negligible.

Significant variables in the first model included temperature, precipitation, day of week, size of the group, and a decay parameter that accounted for long-term effects of the Demonstration. The second model included an additional variable that increased the precision fit of the model. The third model used the square root of the crime rates in an attempt to compensate for ‘skewed’ data.

The fourth model was based on weekly instead of daily data. The day of the week variable was therefore not needed. The precipitation variable was also removed from this model, but no explanation was given for doing that. The report simply states that, “In the weekly analysis, temperature was the only significant control variable.” However, this is contradictory to the fact that, “Precipitation was also found to influence HRA crimes, with a significant decrease in violent crime on rainy days.” This is quickly followed by a footnote which only explains how the weekly data was obtained. This would have been cause for suspicion, but it was the second model that was statistically selected and used as ‘the best model.’ The four models predicted crime reductions attributable to the Demonstration of 17.9%, 17.7%, 18.7%, and 17.0%, respectively.

Quality of life data was only available for two of the variables needed when the report was written. Examination of psychiatric emergency calls found that, “the number of daily calls was increasing significantly ... and this trend reversed to one of decreasing calls since the Demonstration.” It was also found that, “there were 40% fewer complaints against the police in June and July than in April and May.”

“Prior to the beginning of the Demonstration Project, opinion polls showed that President Clinton's popularity had declined steadily since his inauguration and was at the lowest point recorded for any president so early in an administration.” A model similar, yet far simpler than the ones used to analyze crime data showed, “a decline in approval rating of almost 1% of the population (about 1.9 million people) per week.” However, during and after the Demonstration, “There was a new trend of increasing approval for the President, at a rate of an additional .2% of the adult population, approximately 377 thousand people per week.”

Before the Demonstration, “News reports constantly underscored the Clinton presidency’s deteriorating prestige.” During the Demonstration the media frequently mentioned a “marked turnaround in the atmosphere in Washington and the fortunes of the Clinton administration.” A headline of the New York Times on June 21 read, “Bill Clinton's brightening sky. . . .What a difference a week makes.” One satisfying discussion printed in the July 18 edition of the Washington Post reasoned that, “Washington has relaxed. But such a swift reversal of political fortune is not easy to account for. Clinton really turned things around or ... something else is going on . . . almost mysteriously and almost overnight, in the face of government distress, the press seemed to be transformed.”

Aside from a mysterious ‘structural break’ which caused the Institute to throw out most of the collected data, the report flowed smoothly and seemed professionally drafted. Many helpful graphs and charts were included to visually represent the reported 17.7% decrease in crime. I found the report to be believable, knowing that it had been approved by an Independent Project Review Board.

The final report (Hagelin et al., 1997), in contrast, was inconsistent, hard to understand, and seemed to explain all of the wrong things. This was disappointing, as I was looking forward to a clear-cut explanation of the entire Project. The preliminary report (Hagelin et al., 1994) had gone out of its way to describe the Independent Review Board and to indicate that “The Institute sent a draft of this Report to the Project Review Board for review in June 1994, and this final report has been revised in response to members’ comments.” The final report, on the other hand, said nothing more than, “The functions of the Project Review Board were to advise the Institute on the research design and monitor the research process.” I believe that this is a sneaky way of saying that the Review Board had been dissolved before this report was written, and that it was not reviewed independently.

The final report prominently displays on the first page a 23.3% decrease in HRA crimes at the end of the Demonstration. This is followed by a 24.6% decrease in HRA crimes when the previously thrown out data (1988-1992) is used. If you will remember, violent crime has previously been defined as consisting of homicides, forcible rapes, aggravated assaults and robberies. HRA crimes do not include robberies. Two sentences later the report states that “Robberies did not decrease significantly,” and that a model which included HRA crimes as well as robberies showed only a 15.6% reduction in crime overall. These confused numbers permeate the rest of the report.

In addition to these first three models mentioned, the report later discusses several other models that were created. In the first report, a variable called AIC was used to select the ‘best’ model based on having the fewest number of variables (parsimony) and showing the greatest precision of fit. The same was done in the final report, however the AIC was mentioned four pages after it was decided that, “All the variables in the time series model, including dependent, intervention and control variables, consist of data series formed by weekly aggregation; it is well known that aggregation over time reduces the level of noise as a proportion of the mean level of the series and often leads to more parsimonious models.” It was unclear whether the daily crime data was then considered by its AIC. Even though the weekly data may “often lead to more parsimonious models,” the precision of model fit still needs to be considered. Unlike the preliminary report, there was no chart or even a table that could be used to compare the AIC of all of the models which were mentioned.

All of the models that followed were based on the first model, which used the 1993 weekly crime data and average weekly temperature to predict crime rates. The next model that was mentioned included a variation on the method for finding the decrease in crime. It showed a 22.2% decrease in crime, but the first model was, “likely to yield a more accurate estimate.” A fifth model accounted for the fact that another mysterious ‘structural break’ had been discovered. This model showed only a 20.6% decrease.

A sixth model was created to analyze the daily crime data. Unlike the previous five models which differed only by the change of one variable, this model was particularly unique. A footnote informs that, “Daily precipitation and the average rate of HRA crimes for each day of the week during the preceding 5 years were also significant predictors and were also controlled for in the analysis.” Not only did this model use daily data, this obscure note alerts that the data from 1988-1992 has been included, the day of the week was significant in predicting crime, and precipitation also had a significant effect on the crime rate in Washington, D.C.

Where did precipitation suddenly come from again, and where did it go in the other models? Early in the report precipitation is mentioned in a list of ‘potential control variables,’ and then it is not seen anywhere else in the report, except in this one footnote. It can only be assumed that precipitation must have been eliminated by the great AIC. Directly after that first instance of ‘precipitation,’ there is a reference, “(For further rationale on the choice of these control variables, see Hagelin et al., 1994.)” The bibliography entry for this reference point to the preliminary Project report. That would be completely legitimate, however, the Institute, which readily supplied a copy of the final report, would not release a copy of the preliminary report. A copy of the preliminary report was found at the University’s library.

A final model accounted for changes in police activity. Ironically, “the analysis showed that police staffing changes did not have a significant effect on reducing HRA crime.” The model showed that a 22.4% reduction in crime could be attributed to the Demonstration when the police were taken into consideration.

What do all of these confusing variables have in common? Every one of them causes the predicted decrease in crime attributable to the Demonstration to fall. The preliminary report (Hagelin et al., 1994) mentions that “although a time series model can be fitted to the 1993 data without including any control variables, such a model has less predictive accuracy.” Why didn't they create a model without including any control variables? Because, the goal was to account for all variables that could have possibly affected the crime rate.

If a model that can predict crime based on temperature is just as good as a model that can predict crime without any control variables, then why not simplify this jumble of variables and put them all into one model? The beginning of the report's discussion summarizes (Hagelin et al., 1997) that the, “Results were significant when different noise models were used; when the independent variable was specified in different ways; when the base model was differenced to exclude the possibility of contemporaneous trends; whether or not there was a structural break; when all control variables were omitted; when different baseline lengths were used; and when daily data was analyzed. Regardless of these alternative models, the influence of the coherence-creating group size proved to be highly robust.” My point exactly. If so many variables were taken into consideration all at the same time, it is possible that they could account for the decreased crime rates.

I don't see why parsimony plays a role in the model selection. Would maximizing the precision of model fit not satisfy the needs of this analysis? If nothing else, the final report should have explained the model selection better. The only discussion in defense of this argument is that the AIC “satisfactorily balances both underfitting and overfitting risks.” What is overfitting, and why would it be caused by having too many variables? How can so many variables be found significant, and then not be included in the final model? Precipitation, for example, was found to be a significant predictor of crime rates in the daily crime data. I fail to see how rain could change crime on a day-to-day basis and not have an effect on average weekly crime.

The second hypothesis, an improved quality of life index, was all but completely left out of the updated report. It is referred to on the last page of the report. “In a separate study using time series analysis, Goodman found that all seven of the following sociological variables showed significant positive impact of the coherence-creating group after the start of the Demonstration Project...”

A question that I have been asking myself is, was the Demonstration successful? In some ways it was. The final results showed that by the end of the Demonstration there was a 58% decrease in rapes. Of all the numbers floating around, this was by far the most impressive. History was ‘intervented.’ One of the demonstrators recounted (Weiselberg, L., personal communication, June 2, 1998), “We noticed that by the end there were very few sirens. Everything had settled down.” But did the demonstration, “alert the leaders of society to the existence of such a technology?” No. Somewhere the Project had lost steam. There was a lot of excitement around this Project in Fairfield, and then nothing happened.

Unfortunately, the political motivation of the Project suggests that the final report might not have been written objectively. The Project Director, John S. Hagelin, Ph.D., ran twice as the presidential candidate for the Natural Law Party. A successful Demonstration would have brought him a lot of publicity, and his election as president would have been a giant leap towards the, “establishment of permanent coherence-creating groups in major cities across the United States.” (Hagelin et al., 1994)

What if the Project had been very successful? To put it very bluntly, if such a large decrease in crime had been affected, there would be a lot of people still alive today. Perhaps there would be well-established groups of people meditating in every city. America would have taken a step toward enlightenment, and the world would surely follow. This Project had certainly set its aims high.

BIBLIOGRAPHY

Fighting crime with meditation. (1996). Fairfield, IA: Thaddeus Computing, Inc. http://www.thaddeus.com/p13/p1300022.htm

Hagelin, J. S., Orme-Johnson, D., Rainforth, M., Cavanaugh, K., & Alexander, C. N. (1994). Results of the National demonstration project to reduce violent crime and improve governmental effectiveness in Washington, D.C. Fairfield, IA: Institute of Science, Technology and Public Policy.

Hagelin, J. S., Rainforth, M. V., Orme-Johnson, D. W., Cavanaugh, K. L., & Alexander, C. N. (1997). Effects of group practice of the Transcendental meditation program on preventing violent crime in Washington, DC: Results of the National demonstration project, June-July 1993. Fairfield, IA: Institute of Science, Technology and Public Policy.

Use of coherence-creating groups to reduce warfare and violent crime.