Monthly Archives: July 2017

Dr Sylvain Barde

An empirical validation protocol for large-scale agent-based models

by Sylvain Barde and Sander van der Hoog, discussion paper KDPE 1712, July 2017.

Non-technical summary

Despite recent advances in bringing agent-based models (ABMs) to the data, the estimation or calibration of model parameters remains a challenge, especially when it comes to large-scale agent-based macroeconomic models. Most methods, such as the method of simulated moments (MSM), require in-the-loop simulation of new data, which may not be feasible for models that are computationally heavy to simulate. Nevertheless, because ABMs are becoming an important tool for policy making it is a relevant issue to be able to validate them properly, so that they can be compared to other policy-related models.

Our work proposes a new 3-stage protocol for validating computationally demanding simulation models, based on previous work by Salle & Yildizoglu (2014). The major advantage of this protocol is that it relies on a set of metrologies that are all available as ‘off-the-shelf’ software, requiring only a coordination of their implementation:

1. Efficient sampling & Data generation: The protocol start by generating an efficient sampling of the parameter space of the ABM, using the Near-Orthogonal Latin Hypercube (NOLH) design of Cioppa and Lucas (2007). This is followed by the generation of simulated data using the ABM on the sample points provided by the NOLH. This is the computationally heavy step, which only needs to be performed once.

2. Probabilistic modelling using MIC scoring & Empirical data: The simulated data is first used to train the Markov Information Criterion (MIC) algorithm of Barde (2016a,2016b) on each sample point, which is then scored on the empirical data. This is provides a ‘response surface’ by mapping from the pre-determined set of NOLH parameter calibration points into a fitness landscape using the MIC score as the fitness metric.

3. Surrogate modelling of MIC Response Surface using kriging: In the final stage we use kriging (Krige, 1951) to generate a surrogate model of the set of MIC responses and obtain an interpolated ‘MIC response surface’. By optimizing over this simpler model it is relatively quick to find new candidate sample points with possibly better performance.

The 3-stage protocol is applied to the Eurace@Unibi model of Dawid et al. (2016,2017). This macroeconomic ABM displays strong emergent behaviour and is able to generate a wide variety of nonlinear economic dynamics, including endogenous business and financial cycles. In addition, it is a computationally heavy simulation model, so it fits our targeted use-case. The empirical data used for the analysis consists of monthly data for 3 macroeconomic data (Industrial production, CPI and unemployment rate) from 30 OECD countries and the Eurozone.

Following the first step of the protocol 513 distinct sample points are generated using an NOLH design for 8 core parameters of the model. The simulated data generated in the first stage consists of 1000 simulated series of 1000 time periods for each sample point. Because a single series requires 13 minutes and 20 seconds to run, generating the full set requires 114,000 CPU hours. While using a High Performance Computing (HPC) cluster speeds up the process, Eurace@unibi remains a heavy model to simulate. By contrast, the scoring of the data using the MIC in the second stage only requires a modest 513 CPU hours.

The results obtained by applying the protocol to the Eurace@unibi model are promising, as tight estimated are obtained for several parameters of the Eurace@unibi. The quality of the kriging model is tested by re-running the protocol on a new sample of points obtained through the optimistion of the response surface, and by verifying that the predicted MIC values provided by the kriging model match the realised MIC score obtained through the validation exercise. The very high correlation (0.926) between the MIC scores predicted by the kriging model and the ones obtained by re-running the entire protocol confirms that kriging provides an accurate interpolation of the MIC response surface.

While this exercise provides a successful proof-of-concept for the protocol, particularly the kriging stage, several weaknesses are identified. A first is the fact that the NOLH design cannot be extended easily if further samples are required. This can be remedied by using another design, such as a sobol sequence, which can be extended more easily. Another weakness is the fact that the MIC measurement relies on univariate conditioning, which probably introduces measurement errors when used in a multivariate setting and reduces the accuracy of the response surface. Solving this problem requires using a multivariate version of the MIC, which is the focus of ongoing research.

You can download the complete paper here.

Graduation at Canterbury Cathedral

Graduation and prizewinners 2017

It was a pleasure to celebrate with our new graduates and their families at a reception held on 10 July at the Cathedral Lodge in Canterbury.

During the event, the School awarded a number of prizes for outstanding achievement, and you can see all our prizewinners for 2017 on our website under Celebrating Success.

The reception was a fantastic way to mark the achievement of our students, and we would like to wish all our graduates the very best of luck for the future.

You can take a look at the photos from the event on the School’s Facebook page and on the @UniKentAlumni page, plus you can view the whole graduation ceremony on YouTube.

Microeconomic approaches to economic development

The School of Economics hosted a research workshop on ‘Microeconomic Approaches to Economic Development’ on 21-22 June 2017 organised by Anirban Mitra and Zaki Wahhaj.

Professor Debraj Ray (New York University) and Professor Kalle Moene (University of Oslo) gave keynote presentations on, respectively, the relationship between group size and conflict, and the role of merit in economic distribution, in a session chaired by Professor Tony Thirlwall.

The programme also included invited talks from researchers in development economics at the Graduate Institute in Geneva, Royal Holloway College, and the Universities of Bristol, Essex, Oxford, Sussex, Warwick and Kent.

The full programme for the workshop is available here.

Professor Iain Fraser

UK consumers spend more on British meat after horse meat scandal

Consumers are now far more conscious of the origin of fresh and processed meat they buy and willing to pay more for British labelled meat, according to new research by Professor Iain Fraser in the School of Economics.

The shift has, in part, come as a result of the 2013 horse meat scandal when it was discovered that numerous beef products were found to contain horse meat, some of which had been declared unfit for human consumption.

Research by Professor Fraser, and Dr Mohamud Hussein from Agribusiness Solutions Hub, uncovered that because of this incident UK consumers now place a higher value on meat products with an origin label from Britain and are willing to pay around £2/kg more than for meat with no country of origin information.

Beef products in particular have seen the biggest rise in consumers valuing country of origin labels given it was beef products that were found to have been affected by horse meat contamination.

Responding to this change in buyers’ behaviours, the researchers found that food retailers have voluntary increased the amount of origin labelling they provide on products to keep consumers informed. As such, mandatory food labelling requirements appear unnecessary, at least for the short term.

The findings will be published in a forthcoming issue of the Journal of Agricultural Economics.


Article by Dan Worth, University of Kent Press Office

Press coverage generated by this article:


Dr Anirban Mitra

Cash for Votes: Evidence from India

by Anirban Mitra, Shabana Mitra and Arnab Mukherji, discussion paper KDPE 1711, June 2017.

Non-technical summary

Campaigning in elections is costly. In countries without public funding of election campaigns, such financing necessarily relies on donations by private corporations and individuals. If, furthermore, such donations are subject to restrictive legal limits – which is often the case in several developing nations – then it suggests that election campaigns must be frugal. However, in reality elections in such contexts are rarely low-key: there exists ample anecdotal evidence of voters being bribed with cash or actual consumption goods prior to elections. The media often reports on cash seized during various elections. For example, in India the amounts ranged from 19.5 million INR (about 0.3 million US$) in the eastern state of Assam to 155 million INR (about 2.4 million US$) in the southern state of Tamil Nadu during the 2014 parliamentary election. There is, however, a clear lack of hard evidence of the extent and form of vote-buying. This is unsurprising, largely because neither political parties nor voters have any incentives for revealing any details regarding the cash (and “kind”) that changes hands.

In this paper, we propose a methodology to empirically assess the nature and extent of vote-buying using data on elections from all the major Indian states. Our approach to the problem is novel: we look at the consumption patterns of households and examine how they vary before and after elections. The idea is to capture the actual change (presumably, rise) in expenditure by the voters as a result of any cash transfer they might receive from the campaigning parties.

Our two key sources of data are the National Sample Survey (NSS) rounds on household consumption expenditure, conducted during 2004-2012, and state assembly elections data for that period. Each NSS consumption module contains detailed information on the surveyed households’ monthly consumption expenditure on over 300 different commodities. Each of these survey rounds takes a year to complete and covers all states. For every surveyed household we have information on the date of the survey. Combining this with the data on state assembly elections, we are able to ascertain whether a household is reporting on consumption close to elections. Given that in a particular year only some states have elections, we have a sample with different groups: there are households that reported their consumption just a few days before they voted and those that did so many days before or after voting. In fact, we construct ‘time windows’ of different lengths prior to election dates to see how the consumption pattern changes. We compare these groups with the ‘reference group’, which comprises households in neighbouring (non-election) states that were surveyed on similar dates. In this manner, we tackle the main challenge regarding identification since the timing of surveys is independent of that of state assembly elections.

We find that households tend to spend more on a range of staples, and, to an extent, on ‘intoxicants’. The expenditures on education-related items (books, school uniforms, etc.) increase too. Moreover, the effects are quite substantial. Take the case of pulses: there is an increase in consumption of pulses worth around 50 INR per-capita for households surveyed close to election dates. Given that the average per-capita monthly spending on pulses is around 460 INR, this implies about a 10% increase. These “spikes” disappear with (chronological) distance from elections. Using our estimates, the approximate monetised value of the consumption spikes in a district on average turns out to be 2,900 million INR. This figure, when aggregated over a 5–year period (to allow for all states to have elections) comes to around 9% of India’s GDP. These estimates are too substantial to be explained by legal public spending and indicates the presence of the “black economy” in Indian elections.

You can download the complete paper here.

Dr Alex Klein

The Determinants of International Migration in Early Modern Europe

by Alexander Klein and Jelle van Lottum, discussion paper KDPE 1710, June 2017.

Non-technical summary

Migration was a common feature of pre-industrial societies. Because the vast majority of migrations took place within the confines of a country, a province or even a parish, in most cases such moves occurred over relatively short distances. However, long-distance migrations, involving moves of hundreds of kilometres or more, often of individuals entering foreign territories, were hardly rare (Page Moch 2003; Manning 2005; Van Lottum 2007; Bade et al. 2013). Recent estimates show that international mobility levels increased strongly after the medieval period, peaking in the late seventeenth century. In the latter half of the seventeenth century an estimated 8 percent of European individuals (residents of Russia excluded) could be considered an international migrant (Lucassen and Lucassen 2009). These numbers were surpassed only during the mass migrations to the New World in the nineteenth century (Hatton and Williamson 1998). Traditionally, studies on early modern international migration focus on two groups in particular: refugees and elite migrant groups (or individuals from them). Notwithstanding the substantial cultural and economic importance of these migrant groups, in reality they constituted only a fraction of Europe’s internationally mobile population (Lucassen 2012).

This paper offers the first multivariate regression study of international migration ‘common
men and women’ in pre-industrial Europe. Using unique eighteenth-century data about
maritime workers, we created a data set of migration flows among European countries to
examine the role of factors related to geography, population, language, the market and chain
migration in explaining the migration of these workers across countries. We show that among
all factors considered in our multivariate analysis, the geographical characteristics of the
destination countries, size of port towns, and chain migration are among the most robust and
quantitatively the most important factors influencing cross-country migration flows.

You can download the complete paper here.