Unlike statistical software, key lessons learned from learning process are never going to be obsolete and will never expire.

 - Aidy Halimanjaya -  

Money for helping the poor is short when it is needed, and quite the opposite, fund rushes from rich countries flowing to developing countries at times when no one is in urgently having a need for it. With Prof. Dr. Geske Dijkstra based in Erasmus Graduate School of Social Sciences and the Humanities, I explored this global financial phenomenon in my early research career. Now after a few years, I take sometime to reflect and to write this experience and to give some ideas and motivation

for those who just warm up the engine to do quantitative research. Perhaps this short blog can be useful for master students who are working on their thesis, either for those students who choose a similar subject on business cycle or some who do time series analysis more generally.  

1. When data is available, a very tight time scale in quantitative research is possible and allows for a researcher to excel beyond what he/she could imagine – so be prepare to stretch your self.

Prof. Dr. Geske Dijkstra’s guidance is priceless in setting up research milestones. I am lucky to have a research supervisor like her. Many students do not have this privilege. Some research supervisors, especially the senior ones can be very busy. Well if they are not available, we can set our own research milestones. I must say that sticking to research milestones and target was helping me to unlock my potential. I was able to stretch my capacity beyond what I could imagine. In less than 3 months during my lovely stay during a spring time in Rotterdam, I learned and applied some new econometric techniques to demonstrate whether the low and high tide of money for the poor (known as overseas development assistance or ODA in short) timely responds to the need of poor countries.  

2. Dealing with data for quantitative analysis will never be an easy start, so allocate time wisely.

Trust the economists when they say data is always far from perfect. It happened to me that two third of my entire research period was dedicated to only cleaning a few variables of time-series data. There are only five variables in total but I totally underestimated how much time needed to organise them. These five variables include ODA inflows from developed countries to developing countries, developing countries’ foreign investment flow, earnings from export, remittances and government revenue. I extracted ODA data from the Organisation for Economic Co-operation and Development (OECD) Development Assistance Committee (DAC) Query Wizard for International Development Statistics and the others from World Development Indicators (WDI). Five do not sound many, but apparently each variable has many dimensions.  

3. Unclear with which measurements of a variable to use will waste your time.

For ODA data, there are two types: disbursement and commitment data. Disbursement is when donors actually spend the money and commitment is when the make commitment to provide the money. Each type of ODA is measured in current US dollar (or the nominal value) and in constant US dollar (or the real value, which is nominal value adjusted for inflation). I was not clear with what to choose and I end up collecting and managing all four sets of data for one variable. Such as waste of time! If I could repeat the process, I would consult with my supervisor to know which type is suitable for my research and aligns with my research objective and which measurement aligns or correspond with other indicators. Again, be clear with each measurement. You can find the explanation of each measurement in glossary or methods paper that backs up the data.  

4. In quantitative research, cleaning data is not an administrative process but it requires the applications of some complex methods, so, be ready to study hard and to try out.

I was thinking that cleaning data only involves organising rows, columns and cells in excel sheet. Apparently, cleaning data is not as easy as I expected. To have time series data ready for econometric analysis requires the applications of some complex methods such as missing-data imputation or estimation of missing data, detrending, which is a statistical technique to take out the slow and gradual change in a time series data so that only short term fluctuations left to be analysed. Missing-data imputation is not recommended when there are too many missing observations.  

5. Many decisions in applying some methods are based on the rule of thumbs and guided by intuition.

Some questions like should I or should I not use missing-data imputation? To answer this question, your call as a researcher is truly important. That is why the more research you do, the shaper you are to decide the ‘sweet spot’, like to use or not to use a certain method.  There are pros and cons for each decision. So be clear about that. To learn more about the paper I wrote and some related resources, you can read the summary of the paper in my website under research and project titled ‘Floods and draughts of aid: its cyclical inflows into developing countries'.

If you are a keen reader or a master student who is interested in doing research on a similar topic, you can access the thesis, which is stored in Erasmus University thesis repository. You may want to harness some great ideas from journal articles by Bulir and Hamann (2008) Volatility of Development Aid, From the Frying Pan into the Fire, by Chauvet and Guillaumont (2009) Aid, Volatility, and Growth Again: When Aid Volatility Matters and When It Does Not, and by Pallage and Robe (2002) Foreign Aid and the Business Cycle