The punchline is that a factor database has the extraordinary ability to diversify the risk of a backdoor data breach across network locations. This allows you to multiply the probabilities of a breach at each location for the simple reason that both locations must be breached in order to retrieve any intelligible data. So a 5% chance of breach at location one and a 5% chance of breach at location two means the overall chance of breach is now 0.25%. That's not an insignificant reduction.

Data security efforts are usually devoted to fortification. A single location houses the treasure, and we build armies and forts to protect around it. A factor database takes a different approach.

A factor database contains no intelligible data when at rest. During data insertion, each piece of data is split into factors - a primary factor and a meta factor. The primary factor is the value in a single cell - the attribute value of a record. The meta factor contains information that connects all the values into a relational structure, but contains no value. The primary factor and meta factor can be stored at separate locations.

Location one holds an encrypted database with a single column of all primary factors with no indication of how they are related. For example - a single column of disparate patient first names, last names, and biometric information. The meta factor contains information that relates the values in the primary factor. These factors are stored in separate locations. Information from both locations is necessary to retrieve even a single piece of data.

This is a summary of my economics thesis at Carleton College. Profound gratitude to my advisor Nathan Grawe. Here is the original paper.

In early 2013, India experienced a deep currency crisis. Foreign capital fled the country, and the Indian rupee lost a third of its value relative to the US dollar. Why did foreign capital leave so quickly? There are two broad answers. Perhaps investment conditions in India were no longer favorable. Or perhaps conditions abroad improved considerably. The available evidence is mixed. In early 2013, the US Federal reserve announced that it is considering the tapering of asset purchase programs that have kept interest rates near zero since the 2008 recession and subsequent rebuilding efforts. Investors would be keen take advantage of rising interest rates. At the same time however, Indian media outlets were reeling with reports of rampant government corruption and delayed, underfunded public projects. Was economic policy in the developed world ‘pulling’ investors in, or were economic conditions in developing nations ‘pushing’ investors out?

The literature as a whole suggests that both push and pull effects are present, but the effects have changed over the past two decades. Studies on capital flows in the early to mid 1990s mostly affirm the push hypothesis. Calvo et. al. (1993, 1996), and Fernandez-Arias (1996) highlight the importance of US economic variables including economic policy in predicting capital flows to Latin American countries. When investment conditions deteriorate, investors diversify into riskier assets abroad. On the other hand, later studies like Fratzscher (2012) and Mackowiak (2007) find that country-specific ‘pull’ factors tend to matter more than ‘push’ factors.

This paper uses the econometric methods of Calvo et. al. (1993). That paper affirmed the 'push' hypothesis. I test those results using updated data and a broader set of countries. I use interest rates from developed nations to capture monetary policy, and foreign direct investment inflow into developing nations to capture investment. I follow the principal component analysis methodology outlined in Calvo et. al. (1993) to construct indices for both sets of variables from the many countries in the dataset. These indices are then incorporated into several structural vector autoregressions (VAR) that impose temporal exogeneity on the endogenous variables by specifying the condition that investment variables in developing nations do not affect monetary policy variables in developed nations.

Impulse response functions after structural VAR analysis on world interest rates and FDI inflows to three major regions (South-east Asia, Eastern Europe and Latin America) mostly revealed no significant effects of world interest rates on FDI. This suggests that investors are not influenced by short-term interest rate movements in making foreign direct investment decisions. Substitution effects from those interest rate fluctuations may happen elsewhere. Foreign portfolio investments are one possibility. Although there is high co-movement of interest rates among wealthy nations, some countries move independently. Sustained low Japanese interest rates in the 90s are one example. This suggests the possibility that investors may find small interest rate differences in wealthy nations large enough to not warrant riskier investments in emerging economies.