MACHINE LEARNING IN JOB LEAVING DECISION

I. Introduction

The recent decades have witnessed drastic changes in every organizational life; almost all human institutions have modified their production of value, service providing, and upgraded the skills of their human resources and undergoing widespread restructuring. Human resources and human expertise proved to be a significant issue and underlying reason for competition among organization (Richard A. Swanson, Elwood F. Holton III, 2009). Human’s role has become the focus of the attention to the researcher and business sectors managers that the organizations need to develop their human resources to sustain their competitive edge (Drucker, 1994) (J. Quinn, P. Anderson, S. Finkelstein, 1996). As a result, they offer the training courses to their employee. The employee who receives the necessary training can perform in their job and are more likely to feel valued because they are invested in. Therefore, less likely to change employers. Recruitment costs go down due to staff retention.

However, training expenditures are incredibly high. The average training budget for large companies was 19.7 million USD, while midsize companies allocated an average of 2.1 million, and small companies dedicated an average of 355,721 USD (Freifeld, 2018). Training courses’ optimization has become an urgent priority for companies. If companies know which of these candidates want to work for the company after training or looking for new employment, it will help increase the training course’s efficiency (reduce the cost and time) and effectiveness (the quality of training or planning the courses and categorization of candidates).

Thanks to the advances in Machine Learning application, it helps in answering this question. In doing so, the paper aims to present the step by step application of machine learning to classify and predict the probability of the candidates to look for a new job or will work for the company, as well as interpreting affected factors on employee decision based on the available information which is relevant to a candidate such as demographics, education, experience etc.

II. Data Description

The data set is collected from Kaggle, one of the world’s largest data science communities with powerful tools and resources. The whole data divided into train and test. Target is not included in the test but is contained in a separate file named “Answers. npy”

The dataset contains 14 features related to candidates, namely

Enrollee_id: Unique ID for the candidate
City: City code
City_development_index: Development index of the city which measure the level of development in cities developed for the Second United Nations Conference on Human Settlements (Habitat II) in 1996
Gender: Gender of the candidate
Relevent_experience: Relevant experience of the candidate
Enrolled_univerisity: Type of University course enrolled, if any in the current moment
Education_level: Education level of candidate
Major_discipline: Education major discipline of the candidate
Experience: Candidate total experience in years
Company_size: Number of employees in the current employer’s company
Company_type: Type of current employer
Last_new_job: Difference in years between previous job and current job
Training_hours: training hours completed
Target: 0 – Not looking for a job change, 1 – Looking for a job change

The training dataset size includes 19158 observations, while the test dataset size is 2129 observations. In our training dataset, most of the candidate is male (90%), are not enroll university at the moment (73.6%), has relevant experience (72%), has Bachelor or higher degree (87%), and 88% of them has STEM as their primary discipline.

There are some particular highlighted points in our dataset. First, most features in our dataset are categorical (Nominal, Ordinary, Binary)., some with high cardinality. Categorical features are any feature type that can be classified into two major types: Nominal variables have two or more categories that do not have any kind of order associated with them. For example, if gender is classified into two groups, i.e. male and female, it can be considered a nominal variable. On the other hand, ordinal variables have “levels” or categories with a particular order associated with them. For example, an ordinal categorical variable can be a feature with three different levels: low, medium and high. Order is important.

Second, our dataset has missing values. These missing values fall into seven categories. Missing data presents various problems. The absence of data reduces statistical power, and the lost data can cause bias in estimating parameters. Moreover, it can reduce the representativeness of the samples. (Kang, 2013)

Last, our training dataset is imbalanced. The target, which has the value 0 (when the candidate stays with the company), is more than 14.000, occupied 70% of the dataset while the 1-value is more than 4.000 observations. This imbalanced dataset classification problem causes misclassify examples from the minority class, the leaver.

Counting Observations Between Target 1 and 0

III. Pre-processing data

For the convenience of pre-processing data, the training and test dataset are combined and separated after the pre-processing steps finish. The fake target feature in the test dataset is created and has a value of -1. This step is required to facilitate the combination of training and test dataset.

a. Handling with Ordinality Features

The categorical variables should be converted into the machine-readable form (the numeric form) so that a machine learning algorithm understands it. Most machine learning algorithms can not handle this kind of categorical text data unless we convert them to numerical values. Many algorithm’s performances vary based on how categorical variables are encoded. (Claudia Silvestre, Margarida Cardoso, Mario Figueiredo, 2013)

Label Encoding is a popular encoding technique for handling categorical variables. In this technique, each label is assigned a unique integer based on alphabetical ordering. However, one of the challenges of using Label Encoding is that it does not capture our features’ ordinality. As mentioned above, our data have lots of ordinal features whose ordinality might affect the outcome, such as education level, company size, and experience. For example, according to previous research papers, there is an indirect linkage between education level and organizational commitment through job satisfaction in the information technology environment (Norval D. Glenn, Charles N. Weaver, 1982), (EJ Lumley, Melinde Coetzee, Rebecca Tladinyane, Nadia Ferreira, 2011). In this paper, all the ordinal features are converted into numerical features using manually mapping; for the other feature that does not have ordinality character such as city code, Label Encoding would be applied.

b. Handling with Missing Data

There are various approaches to handle the missing value problem in our data set. The feature with substantial amounts of null values could be dropped, but we will lose many precious data, so this method is not recommended. Another way is filling the mean, mode or median value in place of null values. In this paper, K – Nearest Neighbors Algorithm (kNN Imputation) is applied to find missing values and impute them in data. Configuration of kNN imputation often involves selecting the distance measure (e.g. Euclidean) and the number of contributing neighbours for each prediction, the k hyperparameter of the kNN algorithm then the missing value can be replaced with the nearest neighbour estimated values. This method appears to provide a more robust and sensitive method for missing value estimation and surpass the commonly used row average method (as well as filling missing values with zeros) (Olga Troyanskaya, Micheal Cantor, Gavin Sherlock, Pat Brown, Trevor Hastie, Robert Tibshirani, David Botstein, Russ B. Altman, 2001). We want to use kNN to fill the missing value in the features which have the missing value, so first, we separate these features that contain the missing value and the use kNN to address this problem.

c. Handling with Imbalanced Dataset

Finally, as mentioned above, our data set is suffering from imbalanced classification, which involves developing predictive models on classification datasets with a severe class imbalance. The challenge of working with imbalanced datasets is that most machine learning techniques will ignore, and in turn, have poor performance, the minority class, in this case, the leavers that have value 1 in the target feature. The most straightforward approach involves duplicating examples in the minority class, although it does not add any new information to the model. Instead, new examples can be synthesized from the existing examples. This is referred to as the Synthetic Minority Oversampling Technique, or SMOTE (N. V. Chawla, K. W. Bowyer, L. O. Hall, W. P. Kegelmeyer, 2002). “SMOTE first selects a minority class instance an at random and finds its k nearest minority class, neighbours. The synthetic instance is then created by choosing one of the k nearest neighbours b at random and connecting a and b to form a line segment in the feature space. The synthetic instances are generated as a convex combination of the two chosen instances a and b” (Haibo He, Yunqian Ma, 2013). The approach is effective because new synthetic examples from the minority class are plausible; they are relatively close in feature space to existing examples from the minority class. In this paper, SMOTE method is applied to address the imbalanced classification problem. After using SMOTE, we can see our data set is balanced between target 1 and 0, with a size of 28.762 observations.

IV. Model Evaluation

a. Evaluation Metrics

It can be more flexible to predict probabilities of an observation belonging to each class in a classification problem rather than predicting classes directly. This flexibility comes from the way probabilities may be interpreted using different thresholds that allow the operator of the model to trade-off concerns in the model’s errors, such as the number of false positives compared to the number of false negatives. One diagnostic tool that helps interpret probabilistic forecasts for binary classification predictive modelling problems is the Receiver Operating Characteristic Curve or ROC Curve. The ROC curve is a valuable tool for a few reasons. First, the curves of different models can be compared directly in general or for different thresholds. Second, the area under the curve (AUC) represents the degree or measure of separability, which tells how much the model can distinguish between classes can be used as a summary of the model skill (Bradley, 1997). For all these reasons, the AUC – ROC curve is the best metric for our binary classification problem.

b. Modelling

With the increase in the machine learning application, many models have been created and improved, proving their power, efficiency, and effectiveness. In this paper, we will try to apply one of the most loved machine learning algorithms, which is Extreme Gradient Boosting (XGBoost)

XGBoost provides an efficient and effective implementation of the Gradient boosting (GBBoost) algorithm. Both XGBoost and GBBoost are ensemble learner. They both create a final model based on a collection of individual models. These individual models’ predictive power is weak and prone to overfitting, but combining many suck weak models in an ensemble will lead to an overall much-improved result. However, XGBoost and GBBoost’s difference is that XGBoost computes second-order gradients (second partial derivatives of the loss function), which provide more information about the direction of gradients and how to get to the minimum of our loss function, while GBBoost uses the first partial derivatives. Moreover, the XGBoost uses advanced regularization (Lasso and Ridge), improving model generalization (Tianqi Chen, Carlos Guestrin, 2016).

To choose which model is the best fit for our classification problem, we compare the ROC curve of several popular machine learning methods such as Decision Tree, Random Forest, Gradient Boosting, the more advanced method which is Neutral Network and the last which is Extreme Gradient Boosting (XGBoost). The Single Decision Tree and the Neutral Network ROC curves lie below the two others and are inferior. The Random Forest, the Gradient Boosting, and XGBoost ROC curves lie very close to each other. However, the XGBoost ROC curve is slightly better in term of improving its sensitivity without sacrificing specificity. The AUC score of XGBoost (0.921) is also a little bit higher when compared to the Random Forest (0.919) and the Gradient Boosting (0.917). For these reasons, the XGBoost is applied to our classification data set.

c. Hyperparameter Optimization

The last thing we could do to improve our model’s prediction power is hyperparameter optimization. Machine learning models have hyperparameters set to customize the model to our dataset, but how to best set a hyperparameter and combinations of interacting hyperparameters for a given dataset is challenging. A better approach is to objectively search different values for model hyperparameters and choose a subset that results in a model that achieves the best performance on a given dataset. There are some efficient methods for hyperparameter optimization, such as Random Search or Grid Search. This paper will use the more efficient and effective method to choose the best hyperparameters, which is Bayesian Optimization. It provides a principled technique based on Bayes Theorem to direct a search for a global optimization problem. It works by building a probabilistic model of the objective function, called the surrogate function, that is then searched efficiently with an acquisition function before candidate samples are chosen to evaluate the actual objective function. (Jia Wu, Xiu-Yun Chen, Hao Zhang, Li-Dong Xiong, Hang Lei, Si-Hao Deng, 2019)

First, the Bayesian optimization algorithm needs a function they can optimize. In this case, we want to minimize the loss function. We create our custom function name optimize will take the hyperparameter values as arguments, which can be provided to the model directly to configure it. We can define these arguments generically in python using the **params arguments to the function, then pass them to the model via the XGBClassifier(**) function. Now that the model is configured, we can evaluate it. In this case, we will use 5-fold cross-validation on our dataset and evaluate each fold’s accuracy. We can then report the performance of the model as one minus the mean accuracy across these folds. This means that a perfect model with an accuracy of 1.0 will return a value of 0.0 ( 1.0 – mean accuracy)

Next, we can perform the optimization. We want to find the best parameters for the best accuracy, and obviously, the more the accuracy, is better. We cannot minimize the accuracy, but we can minimize it when we multiply it by -1. This way, we are minimizing the negative of accuracy, but in fact, we are maximizing accuracy. Using Bayesian optimization with the Gaussian process can be accomplished using the gp_minimize() function with the name of the objective function and the defined search space. In this model, we try to optimize the model with four hyperparameters named ‘n_estimators’, ‘beta’, ‘gamma’, ‘uniform’, and then make a list of param names; this has to be the same order as the search space inside the main function, last we use functions partial, creating a new function which has same parameters as the optimize function except for the fact that only one param, i.e. the “params” parameter is required. This is how gp_minimize expects the optimization function to be.

Then we can show the best parameters and apply these parameters in our model; we can see our AUC score increases to 0.924. Applying our model to Kaggle competition and using their test set and the target of the test set, which is from the separate file named “Answer.npy” to check our model’s prediction power, we receive the relatively high AUC score (0.754)

V. Features Importance

In term of feature importance, we can see that “relevant_experience” and “enrolled_university” are two features which affect the most on the leaving decision of our candidate (Figure 20). It is quite understandable because the candidate with the more relevant experience can find more attractive offers from other company and organization than the others. It tempts the highly relevant experience candidate to leave the company to find a better place after finishing the training course. The “enrolled university” status is another factor that could affect the staying decision. Our mapping aligned 0 as a stage of no enrolling university, 1 – full-time enrollment and 2 – part-time enrollment at the current moment. The candidates who are attaining full time or part-time enrollment in a university have a motivation to finish their study program first rather than sticking to the company and work as full-time employees after finishing the training course.

VI. Conclusion

With the rapid development of machine learning application, every process of running an organization business has been benefited; one of them is Human Resource Management. This paper tries to apply one of the machine learning techniques, Extreme Gradient Boosting, to find the answer for the question of who will stay after the company’s training course in the data science sector. The model performs exceptionally well, with the AUC score reaching 0.754 with the test set (setting up by Kaggle competition), 0.923 with the validation set and 0.945 with the training set. The paper also points out two significant factors that affect a candidate’s decision to leave. They are the relevant experience of the candidate and the enrolled university status of the candidate. In summary, by building the model to predict the candidate decision problem, the paper aims to give some hints to company organizers to pre-plans their training course more efficiently and effectively

References

Bradley, A. P. (1997). The Use of The Area Under The ROC Curve In The Evaluation Of Machine Learning Algorithms. Pattern Recognition, 30(7), 1145-1159.

Claudia Silvestre, Margarida Cardoso, Mario Figueiredo. (2013). Clustering and Selecting Categorical Features. Progress in Artificial Intelligence, 16, 331-342.

Drucker, P. F. (1994). The Age of Social Transformation. Atlantic Monthly, 274(5), 53-80.

EJ Lumley, Melinde Coetzee, Rebecca Tladinyane, Nadia Ferreira. (2011). Exploring The Job Satisfaction And Organizational Commitment of Employees In The Information Technology Environment. Southern African Business Review, 15(1).

Freifeld, L. (2018). 2018 Training Industry Report. trainingmag.com.

Haibo He, Yunqian Ma. (2013). Imbalanced Learning: Foundations, Algorithms, And Applications. Wiley-IEEE Press; 1st edition.

J. Quinn, P. Anderson, S. Finkelstein. (1996). Managing Professional Intellect: Making The Most Of The Best. Harvard Business Review, 74(2), 71-80.

Jia Wu, Xiu-Yun Chen, Hao Zhang, Li-Dong Xiong, Hang Lei, Si-Hao Deng. (2019). Hyperparameter Optimization for Machine Learning Models Based on Bayesian Optimization. Journal of Electronic Science and Technology, 17(1), 26-40.

Kang, H. (2013). The Prevention And Handling Of The Missing Data. Korean Journal Of Anesthesiology, 64(5), 402-406.

Max Kuhn, Kjell Johnson. (2013). Applied Predictive Modelling. New York: Springer, New York, NY.

N. V. Chawla, K. W. Bowyer, L. O. Hall, W. P. Kegelmeyer. (2002). SMOTE: Synthetic Minority Over-Sampling Technique. Journal Of Artificial Intelligence Research, 16, 321-357.

Norval D. Glenn, Charles N. Weaver. (1982). Further Evidence on Education and Job Satisfaction. Social Forces, 61(1), 46-55.

Olga Troyanskaya, Micheal Cantor, Gavin Sherlock, Pat Brown, Trevor Hastie, Robert Tibshirani, David Botstein, Russ B. Altman. (2001). Missing Value Estimation Methods For DNA Microarrays. Bioinformatics, 17(6), 520-525.

Richard A. Swanson, Elwood F. Holton III. (2009). Foundations of Human Resource Development . Berrett-Koehler Publishers.

Tianqi Chen, Carlos Guestrin. (2016). XGBoost: A Scalable Tree Boosting System. Cornell University.

Artificial Intelligence (AI) And Labor Productivity: The Importance of Complementary Intangible Investment In AI

HUY NGUYEN

“Technology is a gift of God. After the gift of life, it is perhaps the greatest of God’s gifts. It is the mother of civilization, of arts and of sciences.” Freeman John Dyson – The mathematical physicist

ABSTRACT:

Artificial Intelligence (AI) promises to improve existing goods and services and, by enabling automation of many tasks, to increase significantly the efficiency with which they are produced. However, measured productivity growth has declined by half over the past decade, and real income has stagnated since the late 1990s for OCED countries and the United States. This paper considers the importance of complementary intangible investment in preparation for fully gaining the Artificial Intelligence benefit in term of increasing labor productivity growth. It does so by reviewing the main arguments from previous literature and by assessing them accordingly to the economic views and shows that the most impressive capabilities of AI, particularly those based on machine learning, have not yet diffused widely. More importantly, AI full effects will not be realized until waves of complementary innovations are developed and implemented. The required adjustment costs, organizational changes, and new skills can be modelled as a kind of intangible capital. A portion of the value of this intangible capital is already reflected in the market value of firms.

I. INTRODUCTION

About 70,000 years ago, organisms belonging to the species Homo sapiens started to form even more elaborate structures called cultures. The subsequent development of these human cultures is called history (Harari, 2014). However, despite many millennia of evolution, none of the events discussed so far has mattered very much, at least in comparison to something else – something that bent the curve of human history like nothing before or since (Morris, 2011) (Figure 1). Three important revolutions shaped the course of history: The Cognitive Revolution kick-started history about 70,000 years ago. The Agricultural Revolution sped it up about 12,000 years ago. Then, the Industrial Revolution, which was just over two hundred years ago, made a sudden change in our social development (Figure 2). It was the sum of several nearly simultaneous development in mechanical engineering, chemistry, metallurgy, and other disciplines. Now comes to the second machine age. Computers and other digital advances are doing for mental power – the ability to use our brains to understand and shape our environments – what the steam engine and its descendants did for muscle power.

Nowadays, the rapid advance in the field of “Artificial Intelligence”, which could be called as “the second wave of IT-based technology” (Erik Brynjolfsson, Daniel Rock, Chad Syverson, 2018), has profound implications for the economy as well as society at large. Artificial intelligence has the potential to directly influence products and services and the tasks required to create these goods, with important implications for productivity, employment, and competition. The discussion around the recent patterns in aggregate productivity growth highlights a seeming contradiction. On the one hand, there are astonishing examples of potentially transformative new technologies that could significantly increase productivity and economic welfare. On the other hand, measured productivity growth over the past decade has slowed significantly. This deceleration is substantial, cutting productivity growth by half or more if its level in the decade preceding the slowdown. It is also widespread, having occurred throughout the OECD and, more recently, among many large emerging economies as well. (Syverson, 2017).

This paper will give evidence and explanations for the need for complementary intangible investment to gain AI’s benefits entirely in term of increasing labor productivity. Firstly, the numerous potential applications of AI will be presented, which is the reason why people have a forward-looking technological optimism about the AI future. Second, the paper provides a disappointing recent reality of labor productivity growth in some major countries. Finally, this “paradox” could be explained and addressed by the required complementary intangible capital investment and discusses a linkage between the complementary investment in AI and the labor productivity growth

II. HOW COULD AI INCREASE LABOR PRODUCTIVITY?

1. Investment trends in AI sector from 2009 to 2019

AI has existed for decades: processing voice to text or language translation; real-time traffic navigation; dynamically serving targeted advertisements based on personal data and browsing history and so on. We are supposed to see more widespread, scaled adoption of AI across sectors. Nowadays, more and more companies are investing in AI and managing the complicated process of adopting this new technology. For example, the four leaders in terms of the number of AI startups funded (the United Kingdom, Israel and Germany) attract 80% of the total amount of fund raised in this sector over the 2009-2019 period, representing $8.6 billion out of a total of approximately $10.8 billion in funds raised by AI startups (Figure 3). It is a very positive picture of European dynamism in term of AI investment. However, in comparison with the US market, Europe is not growing as fast as it could, the US remains the indisputable leader of AI startup dynamism. In 2018, the United States counted 70 exits for an overall investment of $4.5 billion and 510 transactions with average fundraising of around $10 million. Among the sector-specific applications of AI, healthcare and biotech witnessed a surge in European AI startups which are representing 13% of AI startups while entertainment, media, culture accounts for 9% followed by financial services (8%) and defense, security (4%) (Figure 4). An analysis of the International Journal of Computer Vision, the most cited European AI journal – between 2015 and 2019 highlights both the strength of AI investment in terms of R&D. As we could see from the Figure 5 And Figure 6, R&D in AI is monopolized at the global level under the tripolar structure composed of the US, China and the UK, which represent more than half of the institutions featured in the journal. At the European level, the UK, France and Germany based institutions represent two-thirds of the institutions featured in the journal.

A considerable investment flows into the AI sector with a witness of an AI development race among major countries. This phenomenon could raise a question about whether there is a hype of AI. If it is not the case, so what, why and how AI could be a potential role in firm-level and country-level development?

2. Two extraordinary AI’s skills

One way of looking at the last 150 years of economic progress is that it is driven by automation. The industrial revolution used steam and then electricity to automate many production processes. Relays, transistors, and semiconductors continued this trend. “Perhaps artificial intelligence is the next phase of this process rather than a discrete break” (Philippe Aghion, Benjamin F. Jones, Charles I. Jones , 2018). Indeed, historically, most computer programs were created by meticulously codifying human knowledge, step by step, mapping inputs to outputs as prescribed by the programmers. Computers have been replacing humans in carrying out a widening range of tasks – filing, bookkeeping, mortgage underwriting, installing windshields on automobile bodies and so on – the list becomes longer each year. However, there is still some type of work which could not be entirely computerized. “Computers have an advantage over humans in carrying out tasks that involve some kinds of information processing. Nevertheless, humans retain an advantage over computers in tasks requiring other kinds of information processing” (Frank Levy, Richard J. Murnane , 2005). “Computers are good at following rules but lousy at pattern recognition” (Brynjolfsson, Erik and Andrew McAfee, 2014). When expressed in computer code, some rule-related works could be replaced by computer thanks to adding algorithms. For example, the mortgage underwriter who decides whether a mortgage application should be approved, the full list of rules might include tests on the applicant’s liquid assets, the number of years with the current employer, and so on. An application that passed every test was approved. Each rule leads to a clean “yes/no” answer and by setting “yes” = 1 and “no” = 0, it is easy to imagine how a computer could be programmed to process mortgage application in this way. However, lie information processing tasks that cannot be boiled down to rules or algorithms. The basic daily example is driving. The driving task requires the human capacity for pattern recognition. The driver has to recognize what he or she is confronting. However, articulating this knowledge and embedding it in software for all but highly structured situations are enormously tricky tasks. Computers cannot easily substitute for humans in jobs like driving. Now, thanks to modern AI, our digital machines have escaped their narrow confines and started to demonstrate broad abilities in pattern recognition, perception, complex communication, and other domains that used to be exclusively human. “Machines that can complete cognitive tasks are even more important than machines that can accomplish physical ones” (Brynjolfsson, Erik and Andrew McAfee, 2014). We are going to see AI do more and more, and as this happens, costs and employment will go down while the outcomes will stay the same or even improve, and our lives will get better. Soon countless pieces of AI will be working on our behalf, often in the background. They will help us in areas ranging from trivial to substantive to life-changing. Trivial uses of AI include recognizing our friend’s faces in photos and recommending products. More substantive ones include automatically driving cars on the road, guiding robots in warehouses, and better matching jobs and job seekers. However, these remarkable advances pale against the life-changing potential of artificial intelligence.

3. Economically and Efficiently

AI is a wonder of modern science that has made a lot of things possible that were unthinkable before. Now thanks to AI, many things can be done more quickly and more effectively. AI has increased the efficiency and productivity of many things in the industry. For instance, the saline wastewater, which is widely generated by industry, can be used for a variety of purposes such as food processing, textile, leather tanning and petroleum industries. However, the composition of saline wastewater depends mainly on the product, supplies, number of units used in the process and the water sources. Thus, saline wastewater may contain high organic loads, oil, grease, suspended solids, phosphorus and nitrogen. These old systems, which are using biological treatment processes, have inadequate organic load removal. Applying AI model, based on the combination of artificial neural networks and genetic algorithms, can increase the organic load removal efficiency above 70% then improve wastewater treatment performance of complex saline industrial wastewaters (Alain R. Picos-Benítez; Juan D. López-Hincapié; Abraham U. Chávez-Ramírez; Adrián Rodríguez-García, 2017).

Another example is that a team from Google DeepMind recently trained an ensemble of neural networks to optimize power consumption in a data centre. By carefully tracking the data already collected from thousands of sensors tracking temperatures, electricity usage, and pump speeds, the system learned how to make adjustments in the operating parameters. As a result, the AI was able to reduce the amount of energy used for cooling by 40% compared to the levels achieved by human experts. Overall, data centre electricity costs in the US are about $6 billion per year, including about $2 billion just for cooling (Erik Brynjolfsson, Daniel Rock, Chad Syverson, 2018). There are many reasons for this. First, instead of replacing jobs, AI’s automation is far more likely to target specific tasks within a role – particularly repetitive ones we would consider to be”low-value” (David H. Autor, Anna M. Salomons , 2017). By fundamentally changing the types of jobs that are being done, AI allows humans to focus on more meaningful works, which could improve efficiency and productivity (David Autor, Anna Salomons, 2018). Seconds, AI could be embraced for the productive savings by using complex calculations, routine tasks and pattern recognition. With these extraordinary abilities, AI can minimize the number of errors and mistake during the production process, then, reduce the costs and improve the efficiency of the manufacture.

4. AI Improvement

AI or machine learning systems are also designed to improve over time. Indeed, what sets them apart from earlier technologies is that they are designed to improve themselves over time. Instead of requiring an investor or developer to codify, or code, each step of a process to be automated, a machine learning algorithm can discover on its own a function that connects a set of inputs X to a set of outputs Y as long as it is given a sufficiently large set of labelled examples mapping some of the inputs to outputs (Brynjolfsson, Erik, Andrew McAfee, 2017). The improvements reflect on only the discovery of new algorithms and techniques, particularly for deep neural networks, but also their complementarities with vastly more powerful computer hardware and the availability of much larger digital datasets that can be used to train the systems (Erik Brynjolfsson, Daniel Rock, Chad Syverson, 2018). More and more digital data is collected as a byproduct of digitizing operations, customer interactions, communications and other aspects of our lives, providing fodder for more and better machine learning application.

5. The spillover effect and Innovation

A similar application of AI could be implemented in a variety of commercial and industrial activities. For instance, manufacturing accounts for about $2.2 trillion of value-added each year. Manufacturing companies like GE are already using AI to forecast product demand, future customer maintenance needs, and analyze performance data coming from sensors on their capital equipment. Recent work on training deep neural network models to perceive objects and achieve sensorimotor control at the same time have yielded robots that can perform a variety of hand-eye coordination tasks. (Levine, Finn, Darrell, and Abbeel, 2016). (Liu, Gupta, Abbeel and Levine, 2017) trained robots to perform several household chores, like sweeping and pouring almonds into a pan, using a technique called imitation learning. In this approach, the robot learns to perform a task using a raw video demonstration of what it needs to do. These techniques will surely be essential for automating manufacturing processes in the future. The results suggest that AI may soon improve productivity in household production tasks as well (Erik Brynjolfsson, Daniel Rock, Chad Syverson, 2018), which in 2010 were worth as much as $2.5 trillion in nonmarket value-added (Bridgman, Dugan, Lal, Osborne, and Villones, 2012).

Moreover, if we think of AI as a type of capital, precisely a type of intangible capital (Erik Brynjolfsson, Daniel Rock, Chad Syverson, 2018), (M. O’Mahony, M. Vecchi, 2009) find that the spillover effect existence is belonging to an intangible-intensive industry by using data on five large OECD economies between 1988 and 1997. To be more specific, the paper finds that the firms operating in most R&D and skill-intensive sectors have from 2-5% higher productivity growth. Similarly, (A. Elnasri, K.J. Fox, 2017) study the case of intangible investments in Australia between 1993-2013. The authors also find that private intangible investments have a general positive TFP effect in Australia, interpreted as a spillover effect.

Last but not least, AI can spur a variety of complementary innovations. For instance, machine learning of AI has transformed the abilities of machines to perform many primary types of perception that enable a broader set of application (Erik Brynjolfsson, Daniel Rock, Chad Syverson, 2018). The significant advances in AI have not been in the form of the “general problem solver” approaches; instead, recent advances in AI are by, and large innovations that require a significant level of human planning and that apply to a relatively narrow domain of problem-solving. Therefore, AI is an area where we might focus on the impact of innovation (improved performance) and diffusion (more widespread application) in terms of job displacement versus job enhancement. Consider machine vision – the ability to see and recognize objects, to label them in photos, and to interpret video streams. As error rates in identifying pedestrians improve from one per 30 frames to about one per 30 million frames, self-driving cars become increasingly feasible (Brynjolfsson, Erik, Andrew McAfee, 2017). (Iain M. Cockburn, Rebecca Henderson, Scott Stern, 2018) gives some quantitative empirical evidence on AI effect on innovation by estimating the evolution of different areas AI in terms of scientific and technical outputs of AI researchers as measured by the publication of papers and patens from 1990 through 2015. Together, these preliminary findings provide that the innovation indicators are rapidly developing while AI application is being applied in many sectors.

III. THE DISAPPOINTING LABOUR PRODUCTIVITY PERFORMANCE

Although the giant AI’s benefits discussed above hold great potential, there is little sign that they have yet affected aggregate productivity statistic. “Labor productivity growth rate in a board swath of developed economies fell in the mid-2000s and have stayed low since then.” (Erik Brynjolfsson, Daniel Rock, Chad Syverson, 2018). For example, labor productivity growth in the OCED area remains weak and well below the pre-crisis rate. Since 2010, annual growth in labor productivity has slowed to 0.9% about half the rate recorded in the 2000-2005 pre-crisis period (Figure 7). The post-crisis slowdown in productivity growth affects all major sectors but mainly manufacturing, where productivity growth rates remain well below last-decade’s rates in most countries (Figure 8). Indeed, in Australia, Israel, and the United Kingdom productivity gains in manufacturing have been negligible since 2010. In the services sector, the picture has been more varied (Figure 9). In Central and Eastern European OCED economies, for example, the catch-up process has helped sustain relatively robust growth, picking up actively in Poland and Slovenia in the most recent years. However, productivity growth remains weak in most other economies, indeed, sclerotic in some, such as Italy and Greece. Even in influential countries, such as Germany, Denmark and France, it remains weak. Wage growth has recovered in many countries but remains below pre-crisis rates in most countries (Figure 10). Growth in real wages, adjusted for inflation (using the consumer price index), has improved almost across the board in recent years compared with the early recovery period but remains below pre-crisis rates in two-thirds of OECD countries. The United States is experiencing the same scenario of a slowdown in measured labor productivity growth. “From 2005 through 2015(Q3), labor productivity growth has averaged 1.3% per year. This is down from a trajectory of 2.8% average annual growth sustained over 1995-2004” (Syverson, 2017). However, these slowdowns do not appear to reflect the effects of the Great Recession only. “In major advanced economies, productivity growth was slowing prior to the Great Recession.” (Gilbert Cette, John G. Fernald, Benoit Mojon, 2016). Both capital deepening and total factor productivity (TFP) growth lead to labor productivity growth, and both seem to be playing a role in the slowdown (Andrews, Criscuolo, Gal , 2016). Disappointing technological progress can be tied to each of these components. TFP directly reflects such progress. Capital deepening is indirectly influenced by technological change because firms’ investment decisions respond to improvements in the capital’s current or expected marginal product.

IV. THE NEED OF COMPLEMENTARY INTANGIBLE INVESTMENT IN AI

1. AI is a General Purpose Technology (GPT)

The inconsistency between forward-looking technological optimism and backwards-looking disappointment gives us a hint that the full AI benefits have not been fully reaped. AI, which has the vast potential to be pervasive, to be improved upon over time, and to spawn complementary innovations, is one of the prominent candidates that embody the characteristics of general-purpose technologies (GPTs). However, “A GPT does not deliver productivity gains immediately upon arrival” (Jovanovic, Boyan, Peter L.Rousseau, 2005)The technology can be present and developed enough to allow some notion of its transformative effects even though it is not affecting current productivity levels in any discernible way. “AI will bring a positive productivity shock to most economic sectors. However, AI capital will need complementary investments in intangible capital, such as complementary investment in firm-specific human capital and organizational structures” (Erik Brynjolfsson, Daniel Rock, Chad Syverson, 2018). This is precisely the state that the economy may be in now. GPTs can at one moment both be present and yet not affect current productivity growth if there is a need to build a sufficiently large stock of the new capital, or if complementary types of capital, both tangible and intangible, need to be identified, produced, and put in place to harness the GPT’s productivity benefits fully. (David, 1989) notes a similar phenomenon in the diffusion of electrification. At least half of US manufacturing establishments remained unelectrified until 1919, about 30 years after the shift to polyphase alternating current began. Initially, adoption was driven by simple cost savings in providing motive power. The most significant benefits came later when complementary innovations were made. Managers began to fundamentally re-organize work by replacing factories’ centralized power source and giving every individual machine its own electric motor. This enabled much more flexibility in the location of equipment and made possible active assembly lines of material flow.

2. The benefit of complementary intangible investment in AI

Consider the benefit of complementary investment in intangible capital when applying GPTs such as IT or AI in firms, (Brynjolfsson, Erik, Lorin Hitt, 2003) found that while small productivity benefits were associated with firms’ IT investment when one-year differences were considered, the benefits grew substantially as longer differences were examined, peaking after about seven years. They attributed his pattern to the need for complementary changes in business processes. If the firm applies the new technologies to its manufactory process, but there is no considerable adjustment to match the firm’s human capital to the new structure of production, the GPTs will not have any noticeable effect on firm’s productivity. “As computers become cheaper and more powerful, the business value of computers is limited less by computational capability and more by the ability of managers to invent new processes, procedures and organizational structures that leverage this capability” (E. Brynjolfsson, L.M. Hitt, 2000) . For instance, when implementing large enterprise planning systems, firms almost always spend several times more money on business process redesign and training than on the direct costs of hardware and software. In fact, (Brynjolfsson, Erik, Lorin Hitt, 2000) also highlighted how investment in Information and Communication Technology (ICT), which is one of the GPTs candidates, needs even higher commitments to modern forms of firms’ organizational structure and to firm-specific human capital to be effective. The authors estimate that the ratio between ICT and complementary intangible investments is 1:9.

Furthermore, (Bresnahan, Timothy, Erik Brynjolfsson, and Lorin Hitt, 2002) find evidence of three-way complementarities between IT, human capital, and organizational changes in the investment decisions and productivity levels. (Brynjolfsson, Erik, Lorin Hitt, Shinkyu Yang, 2002) show each dollar of IT capital stock is correlated with about $10 of market value. They interpret this as evidence of substantial IT-related intangible assets and show that firms that combine IT investment with a specific set of organizational practices are not just more productive: they also have disproportionately higher market values than firms that invest in only one or the other. “It is plausible that AI-associated intangibles could be a comparable or greater magnitude” (Erik Brynjolfsson, Daniel Rock, Chad Syverson, 2018). This pattern in the data is consistent with a long stream of research on the importance of organizational and even cultural change when making GPTs investment such as IT, ICT or AI and technology investments more generally (Henderson, Rebecca, 2006) (Orlikowski, 1996)

However, such changes take substantial time and resources, contributing to organizational inertia. Firms are complex systems that require an extensive web of complementary assets to allow the GPT to transform the system entirely. Firms that are attempting transformation often must reevaluate and reconfigure not only their internal processes but often their supply and distribution chains as well (Erik Brynjolfsson, Daniel Rock, Chad Syverson, 2018). For example, considering the retail sector, (Micheal D. Smith, Joseph Bailey, Erik Brynjolfsson, 1999) show that the difficulties incumbent retailers had in adapting their business processes to take full advantage of the internet and electronic commerce. Many complementary were required. The sector as a whole required the build-out of the entire distribution infrastructure and employee’s training. Customers had to be “retrained”. None of this could happen quickly.

Another example is the case of self-driving cars. Consider what happens to the current pools of vehicle production and vehicle operation workers when autonomous vehicles are introduced. Employment on the production side will initially increase to handle R&D, AI development, and new vehicle engineering. Moreover, learning curve issues could well imply lower productivity in manufacturing these vehicles during the early years (Steven D. Levitt, John A. List and Chad Syverson, 2013). Thus, labor input in the short run can actually increase, rather than decrease, for the same amount of vehicle production. These changes can take time, but managers and entrepreneurs will direct invention in ways that economize on the most expensive inputs (Acemoglu D. , Restrepo P., 2017). According to LeChatelier’s principle, elasticities will, therefore, tend to be greater in the long run than in the short run as quasi-fixed factors adjust (P. Milgrom, J. Roberts , 1996) .

3. The Predicted Labor Productivity Indicator

While implementing AI initially depresses labor productivity due to increasing the employment in R&D, the slow productivity growth today does not rule out faster productivity growth in the future. (Erik Brynjolfsson, Daniel Rock, Chad Syverson, 2018) used the data of US, productivity indices from 1948 to 2016 and ran the regression to test whether the past productivity growth rates are the good predictors of future productivity growth. As it turns out, the data shows that it would have been hard to predict the decrease in productivity growth in the early 1970s or foresee the beneficial impact of information technology (IT) in the 1990s. The regressions in Table 1 allow for autocorrelation in error terms across years (1 lag). Table 2 shows the results which cluster the standard errors by decade. In both cases, the R² of these regressions is low, and the previous decade’s productivity growth does not have statistically discernable predictive power over the next decade’s growth. Although the intercept in the regression is significantly different from zero, the coefficient on the previous period’s growth is not statistically significant. The lack of explanatory power of past productivity growth is also apparent in the scatterplots (Figure 11)

Instead of relying only on past productivity statistics to predict productivity growth, we should consider the technological and innovation environment we expect to see shortly (Erik Brynjolfsson, Daniel Rock, Chad Syverson, 2018). Brynjolfsson et al. (2018) used data comparing the US labor productivity between a period after portable power technologies had been invented and were starting to be placed into production (1890 – 1940) and a period which IT technologies were implemented (1970 – now). The authors see that labor productivity during the portable power era shared remarkably similar patterns with the current series (Figure 10). In both eras, there was an initial period of roughly a quarter-century of relatively slow productivity growth. The productivity growth slowdown we have experienced after 2004 also has a parallel in the historical data, a slowdown from 1924 to 1932. As can be seen in the figure, and instructive to the point of whether a new wave of AI and associated technologies could re-accelerate productivity growth at the end of the portable power ear rose again, averaging 2.7 per cent per year between 1933 and 1940.

V. CONCLUSION

There is a bounty of AI benefits which can be implemented in a variety of our life activities. Thanks to two essential and extraordinary skills gaining in AI, which are perception and cognition, AI now can replace more and more human jobs which can help the production input cost such as employment wage or production material go down while the outcome is improved significantly, thus, increasing the labor productivity. Moreover, the spillover effect is another factor that brings AI plays an important role the country-level development. The AI application is not only be implemented in one sector but also a wide range of sectors in the national economy, which enables complementary innovations that could multiply their impact. However, the productivity growth has slowed down recently and what gains there have been are unevenly distributed, leaving many people with stagnating incomes, declining metrics of health and well-being. This gloomy scenario could suggest that the breakthroughs of AI technologies already demonstrated are not yet affecting much of the economy.

By surveying the literature at the country, industry and firm level, this paper found evidence of the increasing importance of business intangibles in explaining labor productivity growth dynamics. Moreover, according to the results in the surveyed papers, to fully reap benefits of investment in Artificial Intelligence (AI), complementary investments in business intangibles are also essential. It points to organizational complements such as new business processes, new employee’s skills and new organizational and industry structures as a significant driver of the contribution of AI. These complementary investments may be as much as an order of magnitude larger than then investments in the AI itself. However, both the AI investments and the complementary changes are costly, hard to measure. They take time to implement, and this can, at least initially, depress productivity as it is currently measured.

Realizing the benefits of AI is far from automatic. It will require effort and entrepreneurship to develop the needed complements, and adaptability at the individual, organizational, and societal levels to undertake the associated restructuring. Theory predicts that the winners will be those with the lowest adjustment costs and that put as many of the right complements in place as possible. This is partly a matter of good fortune, but with the right roadmap, it is also something for which they, and all of us, can prepare.

3. Economically and Efficiently

4. AI Improvement

5. The spillover effect and Innovation

III. THE DISAPPOINTING LABOUR PRODUCTIVITY PERFORMANCE

IV. THE NEED OF COMPLEMENTARY INTANGIBLE INVESTMENT IN AI

1. AI is a General Purpose Technology (GPT)

2. The benefit of complementary intangible investment in AI

3. The Predicted Labor Productivity Indicator

V. CONCLUSION

3. Economically and Efficiently

4. AI Improvement

5. The spillover effect and Innovation

III. THE DISAPPOINTING LABOUR PRODUCTIVITY PERFORMANCE

IV. THE NEED OF COMPLEMENTARY INTANGIBLE INVESTMENT IN AI

1. AI is a General Purpose Technology (GPT)

2. The benefit of complementary intangible investment in AI

3. The Predicted Labor Productivity Indicator

V. CONCLUSION

References

A. Elnasri, K.J. Fox. (2017). The Contribution of Research and Innovation to Productivity. Journal of Productivity Analysis, 47, 291-308.

Acemoglu D., Restrepo P. (2017). The race between machine and man: Implications of technology for growth, factor shares and employment. (N. B. Research, Ed.) 22252.

Alain R. Picos-Benítez; Juan D. López-Hincapié; Abraham U. Chávez-Ramírez; Adrián Rodríguez-García. (2017). Artificial intelligence based model for optimization of COD removal efficiency of an up-flow anaerobic sludge blanket reactor in the saline wastewater treatment. Water Science & Technology, 75(6), 1351-1361.

Andrews, Criscuolo, Gal . (2016). The Best versus the Rest: The Global Productivity Slowdown, Divergence across Firms and the Role of Public Policy. OECD Productivity Working Papers.

Bresnahan, Timothy, Erik Brynjolfsson, and Lorin Hitt. (2002). Information Technology, Workplace Organization, and the Demand for Skilled Labor: Firm-Level Evidence. Quarterly Journal of Economics, 117(1), 339-376.

Bridgman, Dugan, Lal, Osborne, and Villones. (2012). Accounting for Household Production in the National Accounts 1965 – 2010. Survey of Current Business, 92(5), 23-36.

Brynjolfsson, Erik and Andrew McAfee. (2011). Race Against the Machine. Digital Frontier.

Brynjolfsson, Erik and Andrew McAfee. (2014). The Second Machine Age: Work, Progress, and Prosperity in a Time of Brilliant Technologies. WW Norton & Company.

Brynjolfsson, Erik, Lorin Hitt. (2003). Computing Productivity: Firm-level Evidence. Review of Economics and Statistics, 85(4), 793-808.

Brynjolfsson, Erik, Andrew McAfee. (2017). What’s Driving the Machine Learning Explosion? Harvard Business Review, 18, 3-11.

Brynjolfsson, Erik, Lorin Hitt. (2000). Beyond Computation: Information Technology Organizational Transformation and Business Performance. Journal of Economic Perspectives, 14(4), 23-48.

Brynjolfsson, Erik, Lorin Hitt, Shinkyu Yang. (2002). Intangible Assets: Computer and Organization Capital. Brookings Papers on Economic Activity, 2002(1).

David Autor, Anna Salomons. (2018). Is Automation Labor-Displacing? Productivity Growth, Employment, and The Labor Share. NBER Working Paper Series No.24871.

David H. Autor, Anna M. Salomons. (2017). Robocalypse Now – Does Productivity Growth Threaten Employment? European Central Bank Sintra Forum Conference Paper.

David, P. (1989). Computer and Dynamo: The Modern Productivity Paradox in A Not-Too Distant Mirror. The Warwick Economics Research Paper Series.

E. Brynjolfsson, L.M. Hitt. (2000). Beyond Computation: Information Technology, Organizational Transformation and Business Performance. Journal of Economic Perspectives, 14, 23-48.

Erik Brynjolfsson, Daniel Rock, Chad Syverson. (2018). Artificial Intelligence and The Modern Productivity Paradox: A Clash of Expectations and Statistics. NBER Chapters, in The Economics of Artificial Intelligence: An Agenda, 23-57.

Frank Levy, Richard J. Murnane . (2005). The New Division of Labor: How Computers Are Creating the Next Job Market. Princeton University Press.

Gilbert Cette, John G. Fernald, Benoit Mojon. (2016). The Pre-Great Recession Slowdown in Productivity. European Economic Review, 88, 3-20.

Harari, Y. N. (2014). Sapiens: A Brief History of Humankind. Harper.

Henderson, Rebecca. (2006). The Innovator’s Dilemma as a Problem of Organizational Competence. Journal of Product Innovation Management, 23, 5-11.

Iain M. Cockburn, Rebecca Henderson, Scott Stern. (2018). The Impact of Artificial Intelligence on Innovation: An Exploratory Analysis. NBER Chapter, in: The Economics of Artificial Intelligence: An Agenda, 115-146.

Jovanovic, Boyan, Peter L.Rousseau. (2005). General Purpose Technologies. Handbook of Economic Growth, 1B, 1181-1224.

Levine, Finn, Darrell, and Abbeel. (2016). End-to-end Traning of Deep Visuomotor Policies. Journal of Machine Learning Research, 17(39), 1-40.

Liu, Gupta, Abbeel and Levine. (2017). Imitation from Observation: Learning to Imitate Behaviors from Raw Video via Context Translation.

M. O’Mahony, M. Vecchi. (2009). R&D, Knowledge Spillovers and Company Productivity Performance. Research Policy, 38, 35-44.

Micheal D. Smith, Joseph Bailey, Erik Brynjolfsson. (1999). Understanding Digital Markets: Review and Assessment . MIT Press.

Morris, I. (2011). Why The West Rules — For Now: The Patterns of History, and What They Reveal About the Future . Picador .

Orlikowski, W. J. (1996). Improvising Organizational Transformation Over Time: A Situated Change Perspective. Information Systems Research, 7(1), 63-92.

P. Milgrom, J. Roberts . (1996). The LeChatelier Principle . American Economic Review , 173-179.

Philippe Aghion, Benjamin F. Jones, Charles I. Jones . (2018). Artificial Intelligence and Economic Growth. National Bureau of Economic Research in The Economics of Artificial Intelligence, 237-282.

Roth, F. (2019). Intangible Capital and Labour Productivity Growth: A Review of the Literature. Hamburg Discussion Papers in International Economics, No.4.

Steven D. Levitt, John A. List and Chad Syverson. (2013). Toward an Understanding of Learning by Doing: Evidence from an Automobile Plant. Journal of Political Economy, 121(4), 643-681.

Syverson, C. (2017). Challenges to Mismeasurement Explanations for the US Productivity Slowdown. Journal of Economic Perspectives, 31(2), 165-186.

Tables and Figures

Figure 1: Human Social Development Index

Source: Brynjolfsson, Erik and Andrew McAfee. (2014). The Second Machine Age: Work, Progress, and Prosperity in a Time of Brilliant Technologies. WW Norton & Company.

Figure 2: Human Social Development Index After Steam Engine Was Introduced

Source: Brynjolfsson, Erik and Andrew McAfee. (2014). The Second Machine Age: Work, Progress, and Prosperity in a Time of Brilliant Technologies. WW Norton & Company.

Figure 3: The AI startup funds raised among major countries from 2014 to 2019 [USD million]

Source: Roland Berger

Figure 4: The AI application categorizing in different sectors and AI investment between European and US

Figure 5: The number of publishing research papers in the International Journal of Computer Vision (2015-2019) among countries

Figure 6: Distribution of patent registrations among European countries, the US and China (2015-2019)

Figure 7: Labor productivity growth in the OCED and European Area from 1995 – 2018

Figure 8: The labor productivity change in the manufacturing sector among some major countries from 2000-2018

Figure 9: Labor Productivity in business services excluding real estate

Figure 10: Growth in real wages before and after the crisis

Figure 11: 10 Year Average Labor Productivity Growth Scatter Plot

Table 1

Table 2

THE URBAN CONCENTRATION AND ECONOMIC GROWTH IN ASIAN COUNTRIES

HUY NGUYEN

Abstract:

Development in Asia is tied to the growth of sustainable cities. Economically dynamic cities are central to future economic growth and continuing reduction in poverty. Economic growth and the rapid growth of cities have brought enormous change to most Asian countries, raising living standards but at a considerable social and environmental cost. The paper investigates whether there is a significant relationship between productivity growth and the degree of urban concentration, as measured by primacy, or the share of the most extensive metro area in a national urban population in Asian countries. Is there reason to believe many Asian countries have excessive primacy and how costly is excessive (or insufficient) primacy? Based on the Henderson (2000) paper, the paper estimates growth effects, using a panel of 48 Asian countries from 1996-2017. The paper finds that urban concentration does have a positive effect on productivity growth in the Asian area. However, instead of the best degree of urban concentration, the paper finds that there is a “worst degree of primacy” in Asian countries. It is not mean that the paper’s result is not inconsistent with Henderson (2000) paper but rather the root of this difference just lies at the selection bias in Asian countries.

I. INTRODUCTION

It has been argued that strong urban economies are the backbone and motor of the wealth of nations (Jacobs, 1985). As countries become more reliant on manufacturing and services and less on agriculture, urban areas are more likely to become essential for fostering externalities, nourishing innovation, providing a hub for trade, and encouraging human capital accumulation. Such economies should be particularly important for Asian countries, which are almost developing countries (United Nation (2012)), since trends in urbanization show that the share of the urban population has increased substantially in these since the 1950s. For example, between 1950 and 2000 urbanization, defined as the share of urban to the total population, increase by 124 per cent in developing countries compare to 38 per cent in the industrialized world (United Nation (2002)), so that the gap in the relative size of the urban population between developing and developed countries has narrowed substantially. Urbanization and economic growth in developing countries also go hand in hand. The simple correlation coefficient across countries between the per cent urbanized in a country and GDP per capita (in logs) is about 0.85 (Henderson, 2000). Economic development involves the transformation of a country from an agricultural-based economy to an industrial-service based economy (Henderson J. , 1988). Production of manufacturing and services is much more efficient when concentrated in dense business-industrial districts in cities. Close spatial proximity, or high density, promotes information spillovers amongst producers, more efficiently functioning labor markets, and savings in the transport costs of parts and components exchange among producers and of sales to residents. (P. Ciccone, R. Hall, 1995) (Glaeser, 1992) (Fujita, 1999) (Kolko, 1999).

However, recently economists have tended to focus on the issue of urban concentration, rather than urbanization per se (Henderson V. , 2003). Countries and international policy officials worry about whether key cities are too big or too small (Renaud, 1981) (UN, 1993) and international agencies presume that many of the world’s mega-cities are over-populated, at considerable cost to those economies. In the economic development literature, there is the (Williamson, 1965) hypothesis, as adapted to an urban context (Hansen, 1990), which states that a high degree of spatial or urban concentration in the early stages of economic development is helpful. The concentration enhances information spillovers and knowledge accumulation at the time when the economy is “information deficient” (Henderson V. , 2003). However, as development proceeds, the de-concentration occurs for two reasons. The economy can afford to spread economic infrastructure and knowledge resources to hinterland areas. Second, the cities of initial high concentration become a high cost, congested locations that are less efficient locations for producers and consumers. Indeed several authors find the pattern of first increasing and then decreasing urban concentration across countries as income rises (El-Shakhs, 1972) (Alonso, 1980) (J. Davis, 2002). Growth rates of the very largest cities tend to slow, while those of medium and large size cities continue to increase.

This paper tries to replicate all these results of previous literature to estimates the effects of urban concentration on productivity growth in the Asian region. Using the cross-country panel context and basing on the model in (Henderson V. , 2003), the paper formulates the cross-country productivity growth regressions controlled by urban concentration and some economic indicators. The results find the support for the idea that urban concentration drives growth in Asian countries. However, there is no best degree of urban concentration in the case of Asia. On the contrary, the results point out the “worst primacy level” which means that the country will gain the lowest productivity growth rate if it stays at this critical primacy point. The paper argues that most of the Asian countries are experiencing a very early stage of development; thus, the data does not include the full information about another development stage in the Asian region.

II. THE EFFECTS OF URBAN CONCENTRATION ON GROWTH DEVELOPMENT

Urbanization in Asia is proceeding at a scale that is unprecedented in human history. In Asia in 1950, some 232 million people, or 17% of the population lived in urban areas (Table 1). Over the following 55 years to 2005, the urban population grew nearly sevenfold to an estimated 1,562 million, 40% of the population in the Asia region, will be urban representing an increase of over 70% or 1,100 million in the next 25 years. Over this same period, the rural population is expected to decline by 6% or 133 million. Almost all future population growth in Asia will be in towns and cities. While the urbanization process is occurring in virtually all developing countries, it is now centred on Asia. Aggregate population data are dominated by the impact of the two largest countries, the People’s Republic of China (PRC) and India. These countries account for 19% and 15%, respectively, of the projected total world urban population growth during 2006-2030 and 31% and 25%, respectively, in Asia (Table 1).

Why urban concentration affects productivity growth? According to (Henderson V. , 2003) paper, losses from excessive or deficient primacy in static urban models come from GDP losses from resource misallocation, where, for example, under excessive primacy where urban development is concentrated in just one or two primate cities, these cities are subject to exhausted scale economies, excessive congestion, and excessive per capita infrastructure costs, while smaller cities have unexploited scale economies and often deficient capital investment (G. Tolley, 1979), (Fujita M. , 1989) (J.V. Henderson, 2000). Moreover, city size affects positively the degree of local information spillovers, which interactively affects local knowledge accumulation, promoting productivity growth. (D. Black, 1999). However, cities of excessive size draw resources away from investment and innovation in productive activity to try to maintain quality of life in a congested local environment.

A key issue concerns how to measure urban concentration. The paper measures the urban concentration by using urban primacy (Mutlu, 1989) (R.E. Hall, 1999) (A.F. Ades, 1995). Primacy is measured typically by the share of the largest metro area in the national urban population. Is urban primacy as measured by the share of the largest city in the national urban population a reasonable measure? Because such shares are typically massive, primacy measures tend to be closely correlated with Hirschman-Herfindahl indices. Since Hirschman-Herfindahl indices contain squared shares, they are dominated by the most significant share if that is a high number (e.g., over 0.25). Average primacy in our sample, over Asian countries and years, is .35 (In (Henderson V. , 2003) paper, the average primacy over countries is .31). This idea of close correlation is also supported by evidence on Zipf’s Law (Gabiax, 1999). Within countries when we rank cities from largest (rank 1) to smallest, rank times population size is approximately the same constant for all cities. Thus the size of the largest city in the country defines all other city sizes and is sufficient information to calculate any comparative index of national urban concentration.

1. Urban Concentration and Economic Growth

To incorporate considerations of urban concentration into an economic growth framework, the paper uses the model which was presented by (Henderson J. V., 2000) and (Henderson V. , 2003). First, output in an economy is usually specified as produced according to an aggregate Cobb-Douglas function of the form

Y_i = (K_i(t))^α (A_i(t)N_i(t))^1-α (1)

Where K(t) is capital, N(t) is labor (proportional to population, with the factor of proportionality normalized to 1), and A(t) the level of technology. Rearranging (1), taking logarithms, and differencing we have:

ln(Y_i(t)/N_i(t) – ln(Y_i(t-1)/N_i(t-1)) = a[ln(K_i(t)/N_i(t)) – ln(K_i(t-1)/N_i(t-1))] + (1 – a)[ln A_i(t) – ln A_i(t-1)] (2)

Equation (2) states that labor productivity growth is a function of changes in capital per worker and technology. Given the Cobb-Douglas form in (1), labor productivity growth due to changes in technology is also equivalent to total factor productivity growth. Changes in technology, ln A_i(t) – ln A_i(t-1), are modelled in equation (3) below as a function of a base period characteristic such as human capital, which affects the ability to adopt new technologies (R. Nelson, 1966) (G. Grossman, 1991) (S. Durlauf, 1998) and internal country considerations affecting growth in term of efficiency, such as openness or urbanization or primacy. (Glaeser, 1992) (Henderson J. V., 2000).

The productivity growth in equation (2) could be formed as a function of primacy, human capital, and base period output per worker, representing the level of development (Henderson V. , 2003)

ln A_i(t) – ln A_i(t-1) = f (primacy_i(t-1), national scale_i(t-1), ln(Y_i(t-1)/N_i(t-1), human capital_i(t-1)) + d_t + m_i + e_it (3)

The error structure discussed below consists of d_t common shocks across all nations, m_i a country fixed effect where geography, culture, and institutions affect productivity growth, and e_it time-varying error term. We will look at whether these variables affect the level of technology ln A_i(t), as opposed to its growth. Of course, if there are growth effects, in some sense there must be level effects since levels are an accumulation growth. However, econometrically, growth effects are much easier to quantify than level effects

The key is to establish the primacy affects productivity growth, and that, in any context, there is the best degree of primacy and deviations from that degree significantly reduce productivity growth. For the growth formulation in equation (3), the “best degree” of primacy will be defined as one that maximizes productivity growth, other things being equal. To have the best degree of primacy, apart from the human capital, the f(.) function in (3) will be specified as a quadratic in primacy. For a₀ primacy + b₀primacy² where a₀ > 0 and b₀ < 0, the best degree of primacy is the peak point: – a₀/(2b₀). In summary we give the f(.) function in (3) the form

(a₀+ a₁ln(national scale) + a₂ ln(Y/N))primacy + b₀primacy² + b₁(human capital) (4)

The working hypotheses would be that the collection of terms multiplying primacy is positive, while b₀ < 0, and a₁ < 0 and a₂ < 0 so that the best degree of primacy declines as output per worker or national scale increase, where best primacy is given by

Primacy = -( a₀+ a₁ln(national scale) + a₂ ln(Y/N))/(2b₀) (5)

The national scale is measured by the urban population, but it will turn out that scale effects are not always statistically significant. The paper will focus on the specification in equation (4), first without national scale variables (just primacy and ln (Y/N) variables) and then with scale variables.

2. Data and Error Structure

The data cover 1996-2017 in three-year intervals (i.e., τ = 3). Data on output per labor, capital per labor, and human capital index are from the Penn World Tables Mark 9.1 (the newest version). All data are converted to international dollars using purchasing power parity rate (PPP) and at constant 2011 price. Population data on total population, urban population and primacy (population of the largest metro area/national urban population) are from UN World Urbanization Prospects.

Means and standard deviations of all variables are given in Table 2. Given the three-year intervals, for any year (e.g.,1996), for output per labor, capital per labor, human capital index and so on, the annual average rates over t₂ – 1 to t₁ (e.g., 1996-1998 for 1999). The human capital index, which is introduced by Peen World Table 9.1, is measured based on the average years of schooling from (Robert J. Barro, 2013) and an assumed rate of return to education, based on Mincer equation estimates around the world (Psacharopoulos, 1994). The primacy level is measured by a share of the population of the largest city of a country in the total urban population of the country, which is discussed above. Table 3 presents the largest cities in all Asian countries.

First, it is noticeable that the average of output per labor or productivity growth in Asian countries is quite high (10.28%) in three-year intervals (the average of output per labor in (Henderson V. , 2003) paper is 8.89%)) . However, this growth is depended mainly on the growth in investment capital per labor, which we can see in the regression results below. Many previous works of literature also point out the critical factor of investment capital in driving productivity growth in Asian countries. The domestic investment and foreign direct investment (FDI) flowing to Asian countries are integral elements in the growth process in Asian economies (X. Liu, 2009). Moreover, the average of primacy level in the Asian region (.388) is also slightly higher than (Henderson V. , 2003) paper (.305). This number is not surprising when we have been experiencing a rapid urbanization process in Asian countries with remarkable examples are China, Thailand, Vietnam and so on.

Furthermore, in the estimation of equation (2) and (3), the error structure is critical. While variables such as primacy, output worker, and urbanization are correlated, we want to identify the “causal” effect of primacy on productivity growth. In equation (3), as noted earlier, the d_t is time shock trends across countries. The m_i is country fixed effects representing unobserved country time-invariant factors such as geography and culture. These will affect both growth and covariates, a fundamental problem in identifying primacy effects. To deal with the fixed effect m_i, the paper uses the fixed effects and the random effects estimations to eliminate it.

Another problem is that the contemporaneous e _it shocks could affect growth from (t-1) to t, such as internal country innovations and changes in political or legal regimes, the paper assumes it would be exogenous to predetermined values of covariates in (t-2). Nevertheless, e_it does effect covariates in t and even potentially (t-1), such as primacy was determined by migration to the dominant metro area in the country; that is, covariates are not strictly exogenous, which becomes relevant trying to account for fixed effects. Thus, we will use the Generalized method of moments (GMM) Arellano-Bond estimator to get deal with this problem and later the Sargan test will be used to prove the assumption of exogenous e_it shocks.

One issue concerns the viability of instruments: why should past levels of variables be suitable instruments for current changes? Part of the answer in the underlying national economic growth process, where for example, past GDP per worker is a predictor of future output per worker changes through the growth process. Another part lies in frictions in domestic capital and labor market (Rappaport, 2000). Migration frictions relate current primacy changes to past primacy, and capital market frictions and accumulation processes relate current changes in capital stock or investment rates so past levels or rates.

3. Results and Discussion

The paper starts by presenting results on the basic productivity model and then turn to the results from incorporating primacy.

a. Basic Productivity Model

The key dependent variable in this model is the change in labor productivity growth (𝜟Ln[Y(t)/N(t)] = Ln[Y(t)/N(t) – Ln[Y(t-1)/N(t-1)), controlling by the change in investment capital per labour, human capital (Table 4). After running the Breusch-Pagan test for unobserved heterogeneity in the data and the Hausmann test for the correlation between fixed effect (m_i) and explanatory variables, the random effect estimation’s results are more valid because of its consistency and efficiency. Therefore, the paper will use the results of random effect estimation and GMM to present.

First, as we can see in Table 4, the coefficients of the change in capital per labour variable are all highly positively statistically significant at one per cent level in four estimations, which means that the increase in the investment capital per labour growth point will lead to the increase in the productivity growth point in Asian countries. To be more specific, one growth point in investment capital per labour leads .590 growth point in productivity growth point, other factors fixed. The coefficient of the change in capital per labor is quite higher in GMM estimator (.913) although it is still highly significant. These results are consistent with (X. Liu, 2009) (Ching-Cheng Chang, 1999) and (Marcel P. Timmer, 2000) paper which shows that investment capital is one of the sources of productivity growth in Asian economies.

Another factor, which plays a role in driving productivity growth in the Asian region, is human capital. The paper finds that 1-point increase in the human capital index will help increase .047 productivity growth point in Asian countries. This result is also consistent with (Ahmed, 2010) which reveals that the contribution of human capital intensities to labor productivity growth of the ASEAN5 countries through the contribution of TFP per unit of capital growth. Unfortunately, we do not get a significant coefficient of the human capital variable in GMM estimator.

Also, the adjusted R² and t-values do not indicate multicollinearity in the model. At the same time, the year and countries effect is controlled to eliminate the bias which comes from the exogenous events.

b. Basic Primacy Results

At first glance, we can see there is a sign of a positive linear relationship between primacy and productivity growth rate (Figure 1). However, considering the fitted value, in Asian case, the function of productivity depended on primacy may not a concave function but instead a convex function. Therefore, we might not get a maximization point of the quadratic function of the best degree of primacy as (Henderson V. , 2003) paper did. However, we will have a “worst primacy degree” which is a minimization point of the quadratic function.

The econometric regression results are in Table 5, which contain the critical results of the paper along with results in Table 7, columns (1) – (3) where there is a quadratic form to primacy, and it is interacted with output per worker to allow “worst primacy” to vary with output per worker. All these coefficients of explanatory variables are highly statistically significant at one per cent level except from primacy square variable, which is significant at five per cent level. The explanatory variable results under different estimation methods do differ, although pooled OLS and random effects estimations results are very similar. Columns (4) – (5) of Table 5, the paper reports on a simple quadratic without the interaction term between primacy and output per worker, to make the point that there is a “worst degree of primacy”, although almost all the independent variables have no statistically significant in both random effect and GMM estimators.

The urban concentration measured by primacy has a positive effect on driving labor productivity growth in Asian countries. One additional percentage point in primacy could increase 0.009 productivity growth point in the Asian country. The result is also consistent with (Henderson J. V., 2000) paper. Furthermore, the coefficient of primacy square turns out to be positive. Hence, as we discussed above, we will get the minimization point of primacy which the paper coins the term “worst primacy degree”. It means that the country will get the lowest labor productivity growth point if it stays at this critical primacy level. The negative coefficient of the interaction term between primacy and output per labor suggests that the worst primacy increase linearly with output per labor. It means that the high output per labor countries will have a higher worst primacy degree compared to the low output labor countries (Table 6). Thus, if they do not want to stay at the point where they reach the lowest labor productivity growth, they need to gain a higher primacy compared to their worst primacy level. In other words, the high concentration seems very important to growth at developing progress in the Asian region.

Moreover, any deviations from the worst primacy are profitable. For example, in the case of Thailand, 5-point per cent deviation increase (decrease) in primacy above (below) its worst value will increase 0.0011 growth points over three years or about 0.036% year. This growth increase changes with output per worker. In terms of whether countries are above or below their worst primacy levels, in 2017, 40% are above, 44% are below, and the rest next to the worst level (Table 7).

The next issue concerns how to assess whether particular countries have too little or too much primacy. Since country sizes vary and worst primacy should vary substantially by country size for any output per worker level, we need to control for country scale, as in equation (4), the country scale is measured by the urban population. After running the Breusch-Pagan test for unobserved heterogeneity in the data and the Hausmann test for the correlation between fixed effect (m_i) and explanatory variables, the fixed effect estimation’s results, which presented in Table 8, are more valid because of its consistency and efficiency. The interacting variables of the primacy variable with output per worker and the measure of national scale – national urban population, are added. This gives the worst primacy level as in equation (5), one that is a linear function of output per labor and national scale. For the fixed effect results, worst primacy increase with output per labor and with national scale. Although as national urban population rise, worst primacy changes very little.

c. Discussion

However, the critical discussion here is why we cannot get the best primacy level with the Asian countries data? Is the Asian case not consistent with the (Henderson V. , 2003) paper?. The answer to these questions lies in the development stage, which most of the Asian countries are staying.

Considering the neoclassical economics, Solow-Swan model coined the term conditional convergence which predicts that the income levels of developing countries will tend to catch up with or converge towards the income levels of rich countries if the developing countries have similar savings rates for both physical capital and human capital as a share of output (Rao, 2010) (Figure 2). As an example, output per labor in Japan, a country which was once relatively poor, has converged to the level of the rich countries. Japan experienced high growth rates after it raised its saving rates in the 1950s and 1960s, and it has experienced slowing growth of output per labor since its savings rate stabilized around 1970, as predicted by the model (Jorgenson, 1988). It means that most of the Asian countries have been experiencing the first stage of their development (Gong, 2016), where their economic growth is drove mainly by the investment capital flow and labor. Moreover, the urban concentration plays a vital role in attracting the investment flow, also the spillover effect of technology, information and knowledge, which also affects profoundly in Asian economics growth. The benefit of urban concentration suppresses urban over-concentration consequences. The more density of the city is the more rapid increase in its economic growth. Besides, there are a few Asian countries which step across the “turning point” from the first stage development to the second stage development (maybe just Japan) even China, at the moment, is just currently at the intersection between the first and second stage (Gong, 2016). Therefore, the Asian countries data, which included 48 Asian countries are not fully captured all the big picture of the primacy effect in economic growth in the Asian region. We are now just seeing the “take-off” stage of economic development of Asian economic where the primacy is a profound importance factor in boosting economic through attracting investment flows. However, the backside of the excessive primacy in Asia has not complied in our data when the increase of economic growth continue unabated in Asian economics. The paper, thus, argues that our results are not inconsistent with (Henderson V. , 2003) paper. The difference here, instead, is just a selection bias in our data which just focus on the Asian region. Nevertheless, it does not mean that the paper results are meaningless in term of interpretation. It does give another evidence supporting the idea about the positive effect of primacy in economic growth, and it is also consistent with another piece of literature discussing the miracle transformation in term of economic growth in Asian countries.

4. Robustness and Other Specifications

To eliminate the heteroskedasticity problem in our data, all the regression results presents the heteroskedasticity-robust standard error. Another problem is that some single-city countries, for example, Singapore, HongKong or Macao could raise a bias in our result because such that these countries have a high level of urban concentration and a rapid economic growth rate. It could affect our results when estimating the primacy effect in output per labor growth rate. Therefore, Table 9 presents the results, excluding all the single-city countries. All the coefficient of our essential explanatory such as primacy, primacy square and the interaction terms still show the same sign and magnitude to the dependent variable, which is the output per labor growth rate. Overall, the direct productivity formulation and the paper results are compelling and robust.

III. CONCLUSION

This paper argues that urban concentration does have a positive effect on economic growth in Asian countries. This paper explores this statement econometrically based on replicating the (Henderson V. , 2003) paper, using a panel of 48 Asian countries every three years from 1996 to 2017. The results do not show the best primacy level property because most of the Asian countries are staying at their first development. The benefit of primacy level suppresses any urban over-concentration consequences. Therefore, instead of best primacy level, the paper points out the term of “worst primacy level” in Asian case.

There are three main sets of findings. First, at any level of development, there is the worst degree of national urban concentration. The worst degree increases sharply as income rises. The worst degree of concentration also increases with the country scale measured by the national urban population. The benefits of deviating the worst primacy level are also substantial. The benefits tend to rise with the level of development. The results are very robust. Second, in a group of 48 Asian countries in 2017, approximately 19 countries have been above their worst primacy level, 21 of them are below, and the rest is next to their primacy level. Last but not least, in term of the national scale, the worst primacy level increase with national scale. However, when national urban population rise, worst primacy changes very little. The paper also tries to check the result’s robustness by using heteroskedasticity-robust standard error and run the regression excluding all these single-city countries such as HongKong, Singapore, Macao and so on to eliminate the bias which comes from the selection bias. However, the results are still statistically significant in both sign and magnitude, which means that our results are robust.

Although the paper results do not find the best primacy level in Asian countries because of the unique Asian economic development characterization, it does find another evidence which advocates the positive effect of primacy in economic growth and our results are consistent with previous literature which try to discuss and explain the factors driving the rapid growth in Asian economies.

Bibliography

A.F. Ades, E. G. (1995). Trade and Circuses: Explaining Urban Giants. Quarterly Journal of Economics, 110, 195-227.

Ahmed, E. (2010). Human Capital and ICT per Capital Contribution To East Asian Productivity Growth. International Social Science Review, 85, 40-55.

Alonso, W. (1980). Five Bell Shapes in Development. Papers of the Regional Science Association, 45, 5-16.

Bank, W. (2000). Entering the 21st Century World Development Report 1999/2000. Oxford University Press.

Ching-Cheng Chang, Y.-H. L. (1999). Efficiency Change and Growth in Productivity the Asian Growth Experience. Journal of Asian Economics, 10(4), 551-570.

D. Black, J. H. (1999). A Theory of Urban Growth. Journal of Political Economy, 107, 252-284.

El-Shakhs. (1972). Development, Primacy, and Systems of Cities. Journal of Developing Areas, 7, 11-36.

Fujita, M. (1989). Urban Economic Theory. Cambridge University Press.

Fujita, P. K. (1999). The Spatial Economy. Cambridge: MIT Press.

G. Grossman, E. H. (1991). Innovation and Growth in the Global Economy. MIT Press.

G. Tolley, J. G. (1979). Urban Growth Policy in a Market Economy. New York: Academic Press.

Gabiax, X. (1999). Zipf’s Law for Cities: An Explanation. Quarterly Journal of Economics, 114(3), 739-767.

Glaeser, H. K. (1992). Growth in Cities. Journal of Political Economy, 100, 1126 – 1152.

Gong, G. (2016). Two Stages of Economic Development. ADBI Working Paper 628.

Hansen, N. (1990). Impacts of Small and Intermediate-Sized Cities on Population Distribution: Issues and Responses. Regional Development Dialogue, 11, 60-76.

Henderson, J. (1988). Urban Development: Theory, Fact and Illusion. Oxford University Press.

Henderson, J. V. (2000). The Effect of Urban Concentration on Economic Growth. NBER Working Papers 7501.

Henderson, V. (2003). The Urbanization Process and Economic Growth: The So-What Question. Journal of Economic Growth, 8, 47-71.

J. Davis, J. H. (2002). Evidence on the Political Economy of the Urbanization Process. Journal of Urban Economics.

J.V. Henderson, R. B. (2000). Political Economy of City Sizes and Formation. Journal of Urban Economics, 48, 453-484.

Jacobs, J. (1985). Cities And The Wealth of Nations: Principles of Economic Life. New York: Random House.

Jorgenson, D. W. (1988). Productivity and Economic Growth in Japan and The United States. The American Economic Review, 78(2), 217-222.

Kolko, J. (1999). Can I Get Some Service Here: Transport Costs, Cities, and the Geography of Service Industries. Harvard University.

Marcel P. Timmer, A. S. (2000). Productivity Growth in Asian Manufacturing: the Structural Bonus Hypothesis Examined. Structural Change and Economic Dynamics, 11(4), 371-392.

Mutlu, S. (1989). Urban Concentration and Primacy Revisited: An Analysis and Some Policy Conclusions. Economic Development and Cultural Change, 37, 611-639.

Nation, U. (2012). Developing Countries. United Nation.

O, J. (1993). Reform and Urban Bias in China. Journal of Development Economics, 29(4), 129-148.

P. Ciccone, R. Hall. (1995). Productivity and Density of Economic Activity. American Economic Review, 86, 54-70.

Psacharopoulos, G. (1994). Returns to Investment in Education: A Global Update”. World Development, 22(9), 1325-1343.

R. Nelson, E. P. (1966). Investment in Human, Technological Diffusion, and Economic Growth. American Economic Review, 56, 69-75.

R.E. Hall, C. I. (1999). Why Do Some Countries Produce So Much More Output Per Worker than Others? Quarterly Journal of Economics, 83-116.

Rao, B. B. (2010). Estimates of The Steady-State Growth Rates For Selected Asian Countries With An Extended Solow Model. Economic Modelling, 27(1), 46-53.

Rappaport, J. (2000). Why are Population Flows so Persistent? . Federal Reserve Bank of Kansas City.

Renaud, B. (1981). National Urbanization Policy in Developing Countries. Oxford University Press.

Robert J. Barro, J.-W. L. (2013). A New Data Set of Education Attainment in The World 1950-2010. Journal of Development Economics, 104, 184-198.

S. Durlauf, D. Q. (1998). The New Empirics of Economic Growth. NBER Working Paper No. 6422.

UN. (1993). World Urbanization Prospects: The 1992 Revision. New York: United Nations.

Williamson, J. (1965). Regional Inequality and the Process of National Development. Economic Development and Cultural Change, 3-45.

X. Liu, C. S. (2009). Trade, Foreign Direct Investment and Economic Growth in Asian economies. Applied Economics, 41(13), 1603-1612.

APPENDIX

Table 1: Urbanization Trends in Asia, 1950 – 2030

Table 2: Descriptive Statistics

	Mean	Standard deviation
Ln(output per labor)	10.286	1.106
Ln(capital per labor)	11.394	1.286
Human capital index	2.446	.586
Primacy	.388	.251
ln(urban population)	8.742	1.825
Urban share	.592	.259

Table 3: The most populous cities among Asian countries

Countries	The most populous metropolitan area
Armenia	Yerevan
Azerbaijan	Baku
Bahrain	Manama
Bangladesh	Dhaka
Bhutan	Thimphu
Brunei Darussalam	Bandar Seri Begawan
Cambodia	Phnom Penh
China	Shanghai
Hong Kong	Hongkong
Macao	Macao
Cyprus	Nicosia
Georgia	Tbilisi
India	Mumbai
Indonesia	Jarkata
Iran	Tehran
Iraq	Baghdad
Israel	Jerusalem
Japan	Tokyo
Jordan	Amman
Kazakhstan	Almaty
Kuwait	Al Ahmadi
Kyrgyzstan	Bishkek
Lao People’s DR	Vientiane
Lebanon	Beirut
Malaysia	Kuala Lumpur
Maldives	Malé
Mongolia	Ulaanbaatar
Myanmar	Yangon
Nepal	Kathmandu
Oman	Muscat
Pakistan	Karachi
Philippines	Quezon
Qatar	Doha
Republic of Korea	Seoul
Saudi Arabia	Riyadh
Singapore	Singapore
Sri Lanka	Colombo
State of Palestine	Gaza
The Syrian Arab Republic	Aleppo
Taiwan	Taipei
Tajikistan	Dushanbe
Thailand	Bangkok
Turkey	Istanbul
Turkmenistan	Ashgabat
United Arab Emirates	Dubai
Uzbekistan	Tashkent
Vietnam	Ho Chi Minh City
Yemen	Sanaa

Table 4: A Basic Productivity Growth Equation

(dependent variable: ln(Y_i(t)/N_i(t) – ln(Y_i(t-1)/N_i(t-1)))

The results are reported in heteroskedasticity-robust standard error

The random effects estimator is more efficient and consistent

	Pooled OLS (1)	Fixed effects (2)	Random effects (3)	GMM (4)
𝜟ln(capital/labour)	.596*** (.111)	.577*** (.178)	.590*** (.127)	.913*** (.113)
Human capital(t-1)	.048** (.022)	.025 (.067)	.047** (.023)	-.034 (.157)
Year effects	Yes	Yes	Yes	Yes
Countries effects	Yes	Yes	Yes	Yes
Adj R²	.272	.262	.272
Sargan p-value				.246

*** Significant at 1% level

** Significant at 5% level

* Significant at 10% level

Table 5: Productivity Growth With Primacy

(dependent variable: ln(Y_i(t)/N_i(t) – ln(Y_i(t-1)/N_i(t-1)))

The results are reported in heteroskedasticity-robust standard error

The random effects estimator is more efficent and consistent

	Basic model	Primacy Alone
	Pooled OLS (1)	Fixed effects (2)	Random effects (3)	Random effects (5)	GMM (6)
𝜟ln(capital/labour)	.495*** (.107)	.539*** (.172)	.506*** (.121)	.578*** (.130)	.705*** (.164)
Human capital(t-1)	.054*** (.019)	.063 (.055)	.054*** (.019)	.045* (.023)	-.161* (.086)
Primacy(t-1)	.009*** (.002)	.037*** (.007)	.009*** (.002)	.001 (.001)	-.016 (.086)
Primacy sq(t-1)	210^-5* (9.55*10^-6)	5*10^-5 (.00007)	.00002** (.00001)	-9.8910^-6 (8.810^-6)	.0002 (.0002)
Primacy(t-1)*Ln[Y(t-1)/N(t-1)]	-.001*** (.0002)	-.003*** (.0007)	-.001*** (.0002)
Year effects	Yes	Yes	Yes	Yes	Yes
Countries effects	Yes	Yes	Yes	Yes	Yes
Adj R²	.355	.237	.355	.265
Sargan p-value					.344

*** Significant at 1% level

** Significant at 5% level

* Significant at 10% level

Table 6: Worst primacy level varies with output per labor growth rate in some countries in 2017

Countries	Output per labor growth rate (ln(Y_i(t)/N_i(t))	Worst Primacy Level (%) in 2017
Myanmar	9.16	4.23
Bangladesh	9.18	4.70
Vietnam	9.34	8.51
Kyrgyzstan	9.37	9.32
Lao People’s DR	9.50	12.65
India	9.69	17.48
Pakistan	9.73	18.27
Bhutan	9.89	22.38
Philippines	9.94	23.61
China	10.08	27.12
Thailand	10.25	31.29
Israel	11.09	52.41
Republic of Korea	11.18	54.51
Japan	11.20	55.13
Turkey	11.24	56.10
Taiwan	11.37	59.37
Singapore	11.62	65.47
Saudi Arabia	11.96	73.99
Qatar	12.39	84.94

Table 7: Asian countries are above, below or next to their worst primacy level in 2017

Countries are below their worst primacy level	Countries are above their worst primacy level	Countries are next to their worst primacy level
Armenia	Bahrain	Azerbaijan
Bangladesh	China	Brunei Darussalam
Bhutan	India	Lebanon
Cambodia	Indonesia	Maldives
HongKong	Iran	Pakistan
Macao	Iraq	The Syrian Arab Republic
Cyprus	Israel	Thailand
Georgia	Japan	Uzbekistan
Kuwait	Jordan
Kyrgyzstan	Kazakhstan
Mongolia	Lao
Myanmar	Malaysia
Nepal	Oman
Philippines	Qatar
Palestine	South Korean
Taiwan	Saudi Arabia
Tajikistan	Singapore
Vietnam	Sri Lanka
Yemen	Turkey
	Turkmenistan
	United Arab Emirates

Table 8: Productivity Growth: Primacy Effects With Country Scale

(dependent variable: ln(Y_i(t)/N_i(t) – ln(Y_i(t-1)/N_i(t-1)))

The results are reported in heteroskedasticity-robust standard error

The fixed effects estimator is more efficient and consistent

	Pooled OLS (1)	Fixed effects (2)	Random effects (3)	GMM (4)
𝜟ln(capital/labour)	.496*** (.107)	.559*** (.174)	.508*** (.122)	.616*** (.115)
Human capital(t-1)	.055*** (.019)	.074 (.055)	.055*** (.019)	.026 (.073)
Primacy(t-1)	.010*** (.002)	.053*** (.009)	.010*** (.002)	.108*** (.013)
Primacy sq(t-1)	.00002*** (9.05*10^-6)	.00003 (.00007)	.00002** (9.49*10^-6)	.0003* (.0001)
Primacy(t-1)*Ln[Y(t-1)/N(t-1)]	-.001*** (.0002)	-.003*** (.0007)	-.001*** (.0002)	-.008*** (.0009)
Primacy(t-1)*Ln[urban population]	-.00007 (.0001)	-.002** (.0008)	-.00009 (.0001)	-.004*** (.001)
Year effects	Yes	Yes	Yes	Yes
Countries effects	Yes	Yes	Yes	Yes
Adj R²	.356	.250	.357
Sargan p-value				.422

*** Significant at 1% level

** Significant at 5% level

* Significant at 10% level

Table 9: Productivity Growth: Primacy Effect With Country Scale Excluding All single-city countries

	Pooled OLS (1)	Fixed effects (2)	Random effects (3)	GMM (4)
𝜟ln(capital/labour)	.510*** (.114)	.648*** (.193)	.548*** (.139)	.768*** (.138)
Human capital(t-1)	.053** (.020)	-.0333 (.098)	.0521** (.021)	-.032 (.093)
Primacy(t-1)	.006 (.004)	.0788*** (.012)	.0073* (.004)	.115*** (.033)
Primacy sq(t-1)	.00005** (.00001)	.0001 (.00009)	.00004** (.00002)	.0004 (.0004)
Primacy(t-1)*Ln[Y(t-1)/N(t-1)]	-.0009*** (.0002)	-.006*** (.001)	-.0009*** (.0002)	-.0103*** (.0026)
Primacy(t-1)*Ln[urban population]	-.00001 (.0002)	-.002** (.0009)	-.00001 (.0002)	-.003 (.003)
Year effects	Yes	Yes	Yes	Yes
Countries effects	Yes	Yes	Yes	Yes
Adj R²	.381	.322	.380
Sargan p-value				.625

*** Significant at 1% level

** Significant at 5% level

* Significant at 10% level

Figure 1: The relationship between productivity growth and primacy

Figure 2: Conditional Convergence in Solow-Swan models

Đức có thật sự chủ quan trước đại dịch Covid-19?

Những tuần vừa qua là những tuần không mấy yên bình đối với nước Đức nói riêng và cả thể giới nói chung. Đức tình tới thời điểm hiện tại, số người dương tính với Covid-19 đã chạm mốc hơn 4100 ca, 8 ca trong số đó đã tử vong (Coronavirus-Monitor, n.d.). Tuy nhiên, điều dễ nhận ra là người Đức ở đây khá là “bình thản” trước con Covid-19 này mặc dù WHO đã thông báo đây là đại dịch toàn cầu vào chiều ngày 11.03 (Ducharme, n.d.). Hơn nữa, khi so sánh với Việt Nam, đã và đang kiểm soát dịch khá tốt cho tới thời điểm hiện tại cả về phòng chống lẫn chữa trị thì Đức, một đất nước có hệ thống y tế được WHO xếp hạng 25 trên thế giới (Ajay Tandon, Christopher JL Murray, Jeremy A Lauer, David B Evans), lại tỏ ra “yếu kém” trong việc kiểm soát bệnh dịch khi số ca hàng ngày liên tục tăng mạnh. Đồng thời, có nhiều ý kiến cho rằng châu Âu đang áp dụng biện pháp “chấp nhận lây lan” để đạt được “miễn dịch cộng đồng” (herd immunity), mà theo ý kiến cá nhân của mình thì rất là sai và không có cơ sở cũng như nghiên cứu khoa học nào ủng hộ cái biện pháp đấy cả (mình sẽ giải thích bên dưới). Vậy câu hỏi đặt ra là có thật là Đức đã “thất bại” trong việc kiểm soát dịch bệnh hay không? Tại sao Đức lại tỏ ra “bình thản” và “chấp nhận” cái con số lây lan nhanh đến như thế? Trong thời gian nằm nhà trách dịch, mình sẽ cũng thảo luận về các câu hỏi trên.

11.03.2020, Berlin: Bundeskanzlerin Angela Merkel spricht neben Jens Spahn (CDU, l), Bundesminister für Gesundheit, vor einer Pressekonferenz der zur Entwicklung beim Coronavirus. Foto: Michael Kappeler/dpa +++ dpa-Bildfunk +++

Trước tiên, để xem Đức có thật sự “chủ quan” đối với Covid-19 không thì hãy cũng quay ngược thời gian vào ngày 13.01 ca dương tính Corona virus đầu tiên ngoài Trung Quốc xảy ra tại Thái Lan. Lúc này, Đức đã chuẩn bị kĩ càng cho việc phát hiện Covid-19, bằng chứng là sau đó hơn 1 tuần, trường hợp nghi nhiễm đầu tiên đến từ một người phụ nữ có chuyến bay từ Trung Quốc hạ cánh ở thủ đô Berlin vào ngày 25.01. Tuy kết quả xét nghiệm là âm tính, nhưng cũng cho chúng ta thấy được, nước Đức không chủ quan. Bộ trưởng Y Tế Đức, Jens Spahn (CDU) phát biểu trong ngày hôm đấy: “Nhìn chung, nguy cơ ảnh hưởng của Covid-19 đến nhân dân nước Đức khá là thấp, tuy nhiên chúng tôi đã có kế hoạch rõ ràng cho dịch bệnh này”. (Vitzthum, 2020)

Đến ngày 28.01, ca xác nhận dương tính đầu tiên xuất hiện ở bang Bavaria, một nhân viên của nhà máy sản xuất mái ngói ở Stockdort bang Bavaria, có tiếp xúc với đồng nghiệp người Trung Quốc của mình. Tất cả các mối liên hệ từ bệnh nhân đầu tiên này đều được cách li và theo dõi cẩn thận. Gần một tháng, Đức kiểm soát số ca nhiễm bệnh dưới 25 ca chủ yếu đến từ công ty mà bệnh nhân đầu tiên bị nhiễm nhờ phương pháp theo dõi và cách li các mối quan hệ tiếp xúc với người bệnh, tương tự như Việt Nam ta đang thực hiện ngay lúc này. Tuy nhiên, cho đến ngày 26.02, dịch bùng phát ở phía bắc nước Ý rất mạnh, Áo đã tạm thời chặn các chuyến tàu từ Áo tới Ý trong khoảng thời gian này, một cặp vợ chồng người Đức bị nghi ngờ lây nhiễm ở Milan, dương tính với Covid-19 và đã tham dự lễ hội Kölner Karneval và tiếp xúc hàng trăm nghìn người tại lễ hội. Một lần nữa, bộ trưởng bộ Y tế Đức Jens Spahn phát biểu trước họp báo: “Đức không còn có thể truy tìm chuỗi lây nhiễm được nữa và chuẩn bị đương đầu với đại dịch bùng phát ở nước này” (Berkeley Lovelace Jr., William Feuer, Dawn Kopecki, 2020). Đồng nghĩa với việc, phương pháp theo dõi và cách li đã không còn có thể sử dụng được nữa vì số người tiếp xúc với người bệnh là quá đông. Đây có thể coi là bước ngoặt làm cho Đức không thể áp dụng phương pháp các nước như Việt Nam, Đài Loan hay Singapore đang làm và mà phải có phương án B cho tình hình diễn biến sắp tới.

Đức có “chấp nhận” lây nhiễm diện rộng để đạt “miễn dịch cộng đồng”?

Thủ tướng Đức, bà Angela Merkel phát biểu trong họp báo ngày 11.03: “Mức độ của cuộc khủng hoảng Corona vẫn chưa thể lường trước được. Hiện vẫn chưa rõ về hệ miễn dịch đang được hình thành trong cộng đồng”. Mục tiêu của Đức hiện nay là cố gắng làm chậm sự lây lan hết mức có thể. “Wir müssen Zeit gewinnen” để hệ thống y tế không bị quá tải (Herrmann, 2020). Theo đó là hàng loạt các sắc lệnh được ban hành: Hàng loạt các địa điểm tham quan công cộng đóng cửa, nhà trẻ trường học đại học tạm đóng cửa dời lịch học, sử dụng bus cửa sau (bình thường phải lên cửa trước để mua vé, nhưng nhờ Covid, các phương tiện công cộng được miễn phí và đi cửa sau để tránh tiếp xúc vời tài xế), các quán bar club hoạt động về đêm đóng cửa (ở Hamburg là Reeperbahn), các lễ hội tập trung hơn 1000 người bị tạm ngừng (hội sách Leipzig, Musikmesse Frankfurt), Bundesliga tạm ngừng giải cho đến hết tháng 3, thực hiện kiểm tra đo thân nhiệt ở các đường biên giới… Nếu Đức chấp nhận lây nhiễm cộng đồng thì có cần làm những biện pháp cứng rắn làm thiệt hại nền kinh tế hàng tỷ Euro như thế này không? (Look, 2020).

Hơn thế nữa, việc lây nhiễm cộng đồng gây ra một gánh nặng rất lớn cho hệ thống y tế nước Đức. Theo thống kê, nếu có 100.000 người lây nhiễm thì Đức chỉ có thể cung cấp 800 giường bệnh, con số này đã là tốt nhất so với châu Âu nói riêng. Tổng đài hotline ở riêng thành phố Bonn (tại bang North Rhine-Westphalia – đang là tâm dịch của nước Đức, hiện tại đang có hơn 1600 ca dương tính với Covid-19) đang tiếp nhận hơn 100 cuộc gọi đến một ngày. Bang North Rhine – Westphalia cũng thông báo đang thiếu nguồn nhân lực y tá và bác sĩ trầm trọng mặc dù con số nhiễm bệnh đang là hơn 1600 (Kaschel, 2020)

Do đó, chúng ta có thể thấy rõ mục tiêu sống còn của cả nước Đức bây giờ là cố gắng làm chậm tốc độ lây lan của dịch bệnh “flattening a curve“ , thời điểm gần là tới mùa hè, xa hơn là tới khi có vacxin cho Covid-19. Nếu không, hệ thống y tế sẽ quá tải, không đủ trang thiết bị y tế cho người bệnh, chúng ta có thể thấy ví dụ rõ ràng thông qua Trung Quốc và Ý (Godoy, 2020)

Bài viết là ý kiến cá nhân của tác giả, không nhằm việc bênh vực hay ủng hộ cách phòng ngừa của nước Đức, mà đơn giản chỉ muốn cung cấp một góc nhìn khác hay lí do đằng sau các biện pháp chống dịch mà nước Đức đã và đang đề ra. Một lần nữa, tác giả có niềm tin mạnh mẽ rằng Việt Nam, Đức nói riêng và cả thế giới nói chung sẽ vượt qua đại dịch lần này trong tương lai gần, có thể trong vòng 3 đến 4 tháng tới, thậm chí sẽ sớm hơn.

HUY NGUYEN – DU HỌC SINH VIỆT NAM TẠI CHÂU ÂU

Bibliography

Ajay Tandon, Christopher JL Murray, Jeremy A Lauer, David B Evans. (n.d.). Measuring Overall Health System Performance For 191 Countries. GPE Discussion Paper Series: No. 30.

Berkeley Lovelace Jr., William Feuer, Dawn Kopecki. (2020, 2 26). CNBC. Retrieved from Coronavirus: US confirms new case, Germany says it’s at the beginning of an epidemic: https://www.cnbc.com/2020/02/26/coronavirus-latest-updates.html

Coronavirus-Monitor. (n.d.). https://interaktiv.braunschweiger-zeitung.de/corona-virus-karte-infektionen-deutschland-weltweit/?fbclid=IwAR2KL337tJGOue4MaZTgmIP9bHQ_eBunLOvIjYl0hCuPwk5b40UAK78aVTI. Retrieved from https://interaktiv.braunschweiger-zeitung.de/corona-virus-karte-infektionen-deutschland-weltweit/?fbclid=IwAR2KL337tJGOue4MaZTgmIP9bHQ_eBunLOvIjYl0hCuPwk5b40UAK78aVTI.

Ducharme, J. (n.d.). World Health Organization Declares COVID-19 a ‘Pandemic.’ Here’s What That Means. Retrieved from https://time.com/5791661/who-coronavirus-pandemic-declaration/.

Godoy, M. (2020, 03 13). Flattening A Pandemic’s Curve: Why Staying Home Now Can Save Lives. Retrieved from NPR: https://www.npr.org/sections/health-shots/2020/03/13/815502262/flattening-a-pandemics-curve-why-staying-home-now-can-save-lives?t=1584215052803

Herrmann, B. (2020, 03 11). Süddeutsche Zeitung . Retrieved from Merkel: Wir müssen Zeit gewinnen: https://www.sueddeutsche.de/politik/coronavirus-merkel-wir-muessen-zeit-gewinnen-1.4840925

Kaschel, H. (2020, 03 06). Coronavirus in Germany: Health care system under pressure. Retrieved from DW: https://www.dw.com/en/coronavirus-in-germany-health-care-system-under-pressure/a-52663510

Look, C. (2020, 02 21). Bloomberg. Retrieved from German Economy Faces Coronavirus Hit As Export Orders Sink : https://www.bloomberg.com/news/articles/2020-02-21/german-economy-faces-coronavirus-hit-as-export-orders-sink

Vitzthum, T. (2020, 01 26). Welt. Retrieved from Deutschland bereitet sich auf erste Corona-Fälle vor: https://www.welt.de/politik/deutschland/article205359037/Lungenvirus-Deutschland-bereitet-sich-auf-erste-Corona-Faelle-vor.html

What if Italy leaves the European Monetary Union?

I. INTRODUCTION

For most of the big, rich countries of the West, the financial crisis of 2008-2010 and its worrying aftermath came as a genuine shock, the worst such economic shock since the Second World War. For one country, a country that had been a founder member of the European Union (EU) in 1957, whose capital city gave the EU’s founding treaty its name, and which, deep in its history, could claim to have invented or at least developed many of the foundations of western capitalism and civilization. That country is Italy. However, the current situation is not a bright one for Italy. At more than 2.5 trillion euro, Italy has the world’s sixth-largest total debt load. At roughly 135 per cent of gross domestic product (GDP), it’s more than twice what EU rules allow.

Italy’s expansionary budget plans are drawing the ire of the European Union and Italians are even tired of living under austerity. Not surprisingly, the debate about the exit from the Eurozone is particularly hot in Italy, a peripheral country experiencing high levels of public debt, high unemployment and low competitiveness. If Italy takes the unprecedented step of leaving the euro currency and returning to the Lira, it could cause massive losses to investors across the continent, potentially triggering another financial crisis. Using the breakdowns of Argentina or Uruguay a decade ago as a guide rather the relatively mild European Exchange Rate Mechanism (ERM) disruptions (Ansgar Belke, Daniel Gros, 2002), it is reasonable to assume that the external value of the Lira might fall by up to 60 per cent vis-à-vis the “rump euro” bloc. The reasonable benefit of leaving EMU is that re-gaining the monetary independence which Italy gave up on the day they joined EMU. Therefore, they could redenominate into their own currency, for example, Lira, on the same day they announce their willingness to leave the EMU. “The Euro took away two key adjustment mechanisms: control over interest rates and exchange rates. And instead of putting anything in their place, it introduced tight strictures on debts and deficits” (Stiglitz, 2018). However, large-scale devaluation and the reintroduction of monetary sovereignty will also have negative economic and political effects. (Erlanger, 2018)

There has been some discussion of what would happen if Italy and the EU could not reach an agreement on fiscal policies, in other words, an Italy withdrawal from EMU. This paper tries to address all the major issues, economic and political repercussions Italy would face when exiting the currency union.

II. HOW COULD ITALY LEAVE THE EURO?

The predicament of the eurozone is both financial and economic. The financial element centres on debt. Several countries have public debt burdens which are unsustainable. In some cases, private debt is also overwhelming. Meanwhile, excessive debt in the public or private sectors threatens the stability of the banking system. The economic problem is that whereas monetary union was supposed to bring convergence, in several members costs and prices continued to rise rapidly relative to other members of the union, and indeed the outside world, thereby causing a loss of competitiveness. “The way the euro was designed led to divergence: when some country had an adverse “shock”, stronger countries gained at the expense of the weaker” (E.Stiglitz, 2016). This resulted in large current account deficits and the build-up of substantial net international indebtedness. Often it is the same countries that suffer acutely from both financial and economic problems. As a result of poor competitiveness and the burden of excessive debt, several members of the euro-zone suffer from a chronic shortage of aggregate demand, which results in high levels of unemployment. This worsens the debt position of both the private and public sectors, thereby weakening the position of the banks. “The eurozone was flawed at birth. The structure of the eurozone – the rules, regulations, and institutions that govern it – is to blame for the poor performance of the region, including its multiple crises” (E.Stiglitz, 2016). Meanwhile, other countries enjoy current account surpluses, often accompanied by more favourable debt positions in both the public and private sectors. Moreover, the favourable position of the surplus countries is partly the direct result of the weaker members’ loss of competitiveness.

Allow to briefly sum up the logic and the rationale behind a possible Eurozone breakup. In “The Breakup of the Euro Area” (Eichengreen, 2010) and “The Euro – How A Common Currency Threatens The Future of Europe” (E.Stiglitz, 2016) give us a good explanation of which is the real problem of the Eurozone. Interestingly, it guides us step-by-step logic instruction to leave the Eurozone. “A similar analysis would apply to any other country contemplating leaving the Eurozone” (E.Stiglitz, 2016). The purpose is trying to limit the economic also the political damage as much as possible – an amicable divorce.

Cattedrale di Santa Maria del Fiore, Florence, Italy

In theory, keeping a country’s planned exit secret for as long as possible would help that country to minimize the disruptive effects likely to be caused by the disclosure of its plans to leave. Such effects might include: large capital outflows from the country as international investors and domestic residents withdrew their funds; associated falls in asset prices and increases in bond yields; runs on banks, perhaps causing a banking crisis; and negative effects on consumer and business confidence (Eichengreen, 2010). Together, these effects could make it more difficult for a country to leave the currency union in an organized and orderly manner. However, there are also some disadvantages associated with keeping exit plans secret. This would prevent a broader discussion and debate on the best way for a country to leave, which would probably result in a sub-optimal plan for exit. It would also preclude or limit public involvement in the decision, potentially damaging the democratic process and leading to social and political unrest. There could, for example, be no referendum on the question. This might also preclude the possibility of a cross-party political consensus, hence weakening the new policy arrangements and reducing confidence among both the public and international markets that the new monetary framework would succeed. In the case of the euro, keeping a planned exit secret would be much more difficult than in numerous historical examples, not least because so much would be at stake.

Fontana di Trevi, Rome, Italy

In addition, one key part of the transition to a new currency, namely the printing of new notes and the minting of new coins, takes organization and time. (E.Stiglitz, 2016) suggests that, in order to guarantee the smooth transaction function, modern technology could be used to provide the basis of a new financial system. “With an electronic money, leaving the euro can, in principle, be done smoothly, assuming there is cooperation with other European authorities” – (E.Stiglitz, 2016). The main argument is that this electronic currency would instantly introduce without any waiting time to be printed or minted like any tradition note and coin currency. It would be the money inside the national banking system. In effect, this money could be “locked-in”. But anybody could transfer the money in his bank account to that of anyone else. Thus, everybody has, in effect, almost full use of his money. Moreover, the national bank, at this moment, could regulate the credit creation because of its re-gaining the monetary independence from EMU. Then, the bank had to have seed deposits in order to lend which forces the flow of bank credit. “Banks effectively create credit out of thin air, backed by general confidence in government, its ability and willingness to bail out the banks, which includes its power to tax and borrow” (E.Stiglitz, 2016). Thus, by setting up a well-functioning banking system, Italy could restore domestic control over credit creation and step-by-step managing the current account deficit, debt restructuring and gaining the competitiveness in the international market.

However, I find very hard to believe the redenomination into national fiat currency could be done easily and the cost for Italy contemplating withdrawing from EMU is relatively enormous compared to the benefits which it could bring.

II. SHOULD ITALY LEAVE THE EURO?

It is still highly questionable if leaving the Euro is a feasible option for Italy and if there is a legal way to do so. What is most probable is that, even if Italy could find a way out, this would come at huge economic and political costs. Imagining that despite everything, Italy does decide to leave. The process of leaving the common currency itself still poses many problems. First of all, we need to consider the economic costs.

1. THE ECONOMIC COSTS OF LEAVING

a. The massive withdrawal of cross-border capital

Cross-border capital flows will decline significantly as the consequences of losing their credibility as reliable partners if Italy plans to exit the Euro. Assets and liabilities, which is a clear nexus to a country, would be instantaneously rebalanced following the reinstatement of Lira currency. Market participants would be aware of this fact. A key challenge is that the movement of capital in and out of the country for a transition period. Deposit holders would likely begin sending their money to perceived safe havens as soon as they began to suspect Italy might be about to leave the euro. Households and firms anticipating that domestic deposits would be redenominated into the Lira, which would then lose value against the euro, would shift their deposits to other Eurozone banks, German banks for example. “As the euro crisis emerged, money left the banking systems of the weak countries, going to those of the strong countries.” (E.Stiglitz, 2016). A system-wide bank run would follow. Investors anticipating that their claims on the Italian government would be redenominated into Lira would shift into claims on other Eurozone governments, leading to a bond-market crisis and ECB would unlikely provide extensive support. “EU’s willingness to cooperate with a country that has controversially left the monetary union some days before is, at the very least, questionable.” (Santambrogio, 2018). The Italian government was already in a weak fiscal position, it would not be able to borrow to bail out the banks and buy back its debt. This would be a disaster.

Colosseum, Rome, Italy

Therefore, the Italian government must implement very strict capital controls to prevent a large current account deficit due to the increasing demand for cash withdrawal. It could cost the Italian government in time and effort. Authorities should forbid or authorize on a case by case the purchase of foreign financial and real assets by residents. Resident households and businesses are forbidden from acquiring foreign assets, investing overseas or holding bank accounts outside Italy. However, there is a risk that capital flight might begin sometime before the Italian authorities announced the euro-zone exit. There might be a leak during the planning state or market forces might simply make it clear that exit was inevitable sooner or later. Keeping the withdrawal plan as secret as possible could solve these problems but there are still many obstacles to make the plan in secret. It could be putting a ceiling on daily cash withdrawals in order to prevent a collapse of the banking system, which would no longer have access to liquidity provided by the ECB. In particular, from the announcement of the redenomination until banks were able to distinguish between euro and national currency (Lira) withdrawals, banks and cash machines could be shut down (Eichengreen, 2010). The shutdown of ATMs can be administered relatively simply by the “switching companies” that connect the machines to the various networks. Italy could simply declare a bank holiday in which all banking transactions, including those conducted electronically, were prevented. Withdrawals could be permitted again as soon as arrangements could be made to treat these as withdrawals of foreign currency debited against the national deposits according to the prevailing exchange rate. One example of capital controls, in case of redenomination of the national currency, comes from Czech and Slovak. In 1993, there was a more binary change (as no parallel currency was allowed). During that episode, currency separation was implemented during an extended bank holiday. The border was closed, and the law stipulated mandatory stamping of all currency, to distinguish between Czech and Slovak money (Dědek, 1996). Hence, notes were effectively redenominated into a new currency unit (Eichengreen, 2010). In any event, a system of capital controls is a very complex bureaucratic machine that cannot possibly be set up in one night. Capital controls should be put in place long before the exit decision is made public. Another way to prevent capital flight is by keeping a country’s planned exit secret for as long as possible. It would help Italy to minimize the disruptive effects likely to be caused by the disclosure of its plan to leave.

b. The cost of new currency operation

The costs of re-nationalizing the new Lira currency, the operational costs, are necessary for adjusting the systems to the new currency. In an ideal case, new notes and coins would be available to coincide with the launch of the new currency. In practice, though, there are long lead times associated with printing notes and minting coins. It is unlikely that this can be reduced below a few weeks. Notes would have to be printed and coins to be minted and this would require some time. It is doubtful that such printing and minting arrangements could be kept secret, thereby opening up all the downsides of openness discussed in Section I. The Italian central bank must propose a detailed plan to carry this task and the new currency must be implemented throughout Italy. This policy could take an inexhaustible amount of time and effort to plan and implement. Computers will have to be reprogrammed. Vending machines will have to be modified and so on. It follows that there would be a period of high uncertainty where it would be unclear which is the legal tender.

2000 Italian Lire banknote

Some suggest the possibility of using no cash until notes are ready. Some other argues that “Italian euros” – Euros “stamped” by Italians authority, so they cannot be confused with “normal” Euros – should be printed and used as they were Libras and later recollected when the actual lira is ready to be distributed (Santambrogio, 2018). Another likely option is creating a new electronic currency as Stiglitz has already recommended. Although Italian government could save their energy from printing notes and minting coins, it is still a difficult task for Italian policymakers to set up a cryptocurrency monetary regulation because the new electronic currency’s volatility which is increased by uncertainty regarding cryptocurrency relationship with the centralized authorities (i.e., nations) that currently exist (Scharding, 2019). The huge fluctuation in the value of the electronic currency generates risk, which is a threat to national stability. (DeVries, 2016)

c. Redenomination of contracts

Redenomination of contracts is another great issue to deal with. It could be incredibly painful to privates, especially to the ones, banks or firms, who operate locally but have borrowed abroad and would see their liabilities instantaneously increased. Redenomination of the Italian government debt is another problem. In order to manage the debt restructuring, “First, the government should declare all euro-denominated debts payable in national currency” (E.Stiglitz, 2016). However, it is not easy like that. Not all such contracts are with Italian creditors, nor are all issued under Italian law. A further complication arises since, as (Eichengreen, 2010) notes “contracts are not simply being redenominated from one Italian currency to an Italian currency. Foreign courts might, therefore, take EU law as the law of the currency issuer and invalidate the redenomination of certain contracts”.

Monumento Nazionale a Vittorio Emanuele II, Rome, Italy

For example, Italian bonds (both corporate and sovereign) issued with reference to Italian governing law and under Italian jurisdiction, are highly likely to be redenominated into a new Italian currency, if Italy exits the Eurozone. On the other hand, so-called Euro-bonds issued by an Italian corporate in international markets, typically using English law and under the jurisdiction of English courts, would not be easy to redenominate. This is the case, for example, of loans by German banks to Italian corporations or purchases of parts in Germany by Italian manufacturing firms. Italian courts would presumably rule in favour of the redenomination of all loans to Italian borrowers, including those from German banks, but German courts might rule against redenomination. This opens the door to litigation and to an extended period of uncertainty (Eichengreen, 2010). However, (Proctor, 2011) also states that: “This question [of redenomination] must, in turn, depend upon the original, contractual intention of the parties, and this will be determined by reference to the law applicable to the contract as a whole”. Therefore, such obligations may well stay in Euro’s regardless of the policies of Italy with respect to a new currency. This means that Italy would have to prepay in Euro those contracts, which account for almost 30% of the total, a share we cannot underestimate (Santambrogio, 2018). The former European Central Bank (ECB) Governor, Mario Draghi, also took on the issue. In a letter to two Italian lawmakers in the European Parliament, Draghi basically threatened Italy saying that “If a country were to leave the Euro system, its national central bank’s claims on or liabilities to the ECB would need to be settled in full”. (Fugazzi, 2017) also noted that “Although leaving the monetary union may be legally possible, this is the main challenge would be managing the transition from the euro to the revived domestic currency”. Since according to European Commission’s TARGET2 Balance Report of the time, Italy was owning 358.6 billion of Euro, “Italexit” has potential to wreak a wave of economic disaster into both the Italian economy and its banking system. Devaluation of the currency relative to that of the euro will instantaneously boost the debt value.

d. Devaluation and real wages

Currently, Italian workers have been facing low productivity when compared to other European neighbours such as Germany or France and some argued that the devaluation of the new Italian currency could address this problem. However, in case of devaluation, inflation could follow, substantially reducing real-wage value. (Eduardo Borensztein, José De Gregorio, 199) shows that about 30% of the devaluation is offset by higher inflation after three months, and the offset climbs to about 60% after two years, with a significant real depreciation present for longer periods. In fact, wages could not keep up the pace of increasing prices, and the resulting wage inflation would neutralize any benefits in terms of external competitiveness. “Securing the objective of increased competitiveness depends upon the full inflationary impact of devaluation not being passed through the system. If it is, there will be no improvement in competitiveness and the devaluation will fail” (Bootle, 2012).

Vennice, Italy

In another aspect, alleging that Italy is manipulating its exchange rate to gain an advantage in trade, European Union might establish a compensatory duty on Italian exports that, joint with renewed transaction cost, would strongly reduce any competitiveness gain. They could even impose taxes on investments towards Italy, on the basis that it is unfairly attracting them. “A country that reintroduced its national currency at levels that stepped down its labour costs by 20 per cent might be required to pay a 20 per cent compensatory duty when exporting to other members of the EU, reflecting concerns that it was unfairly manipulating its currency and solving its economic problems at the expense of its neighbours” (Eichengreen, 2010). Whatever the compensatory tariff, collecting it would require the reestablishment of customs posts and border controls, adding to transactions costs. Other EU members might seek to tax foreign investment outflows on the grounds that the defector was using an unfair monetary-cum-exchange-rate policy to attract FDI. In this climate of ill will and recrimination, they might seek to limit the freedom of movement of its citizens (Buiter, Willem , 1999). Considering the geography, Italy is embraced tightly by the EU, it would be a nightmare for them if Italy would be isolated in an economic term, especially in trading. Unless, the Italian government could reach individual trade agreements with the remaining Eurozone members, which is like the Swiss-EU relationship, there would not be any gain in competitiveness by devaluating the currency and the whole thing would have no point.

e. The Euro benefit of credibility

The Euro brought a wide range of benefits to its members, which would all be lost when leaving it. Italy was account for more than 10 billion euros EU’s expenditure in 2018 (Sebastian Hauptmeier, A. Jesus Sanchez-Fuentes, Ludger Schuknecht, 2011). Furthermore, for those whose commitment to price stability was previously weak and whose interest rates were high and greatly variable, the Euro has proved really be helpful and restored their credibility on international markets. (Francesco Paolo Mongelli, Juan Luis Vega, 2006). The advent of the euro has brought credibility benefits to members whose commitment to price stability. Enhanced expectations of price stability have brought down domestic interest rates, biding up bond, stock and housing prices. Foreign capital has flooded in to take advantage of this convergence play. The cost of capital has declined, investment rises in the short run. Households feeling positive wealth effects, consumption rises as well. (Bini-Smaghi, Lorenzo , 1998). Therefore, if Italy’s economy left the eurozone, they would see the rise in its transaction and other costs associated with returning to a national currency. (Tavlas, 2004)

Vennice, Italy

The ones supporting the exit from the Euro underrate the possible economic and political consequences of a default and of the loss of credibility on international markets. The international reputation of Italy would arguably suffer. This could lead to credit rating downgrades and higher sovereign spreads. In turn, this would mean higher debt-servicing cost which could spark an even bigger debt crisis. (Biais, Bruno, Fany Declerc, James Dow, Richard Portes,Ernst-Ludwig von Thadden, 2006)

2. THE POLITICAL COSTS OF LEAVING

Political costs are another problem which the Italian government must address. “Besides economic sanctions, the most painful consequences that Europe could make Italy suffer are, again, of political nature” (Santambrogio, 2018). A country that reneges on its euro commitments will antagonize its partners. It will not be welcomed at the table where other EU-related decisions were made. (Eichengreen, 2010) suggests that “the deflector would be relegated to second-tier status in intra-European discussions of non-monetary issues”. Suspended voting rights in the council, joint with widespread political enmity, might mean a de facto political expulsion for Italy. If Italy attaches any value to its active role within the European integration process this cannot happen. Furthermore, legal issues concerning the redenomination of debt and the establishment of capital controls could exacerbate political hostility towards Italy. “The balance sheets of other Member State’s banks and surely would reduce markets’ trust in the Euro and in Europe as a whole, driving out investments from the continent” (Santambrogio, 2018). Diplomatic tension and political resentment could follow, and cooperation in any sector, also nonmonetary issues, would suffer. The EU has effective means to respond to a hypothetical exit of a Member State from the EMU. If the Commission recognizes that such exit constitutes a breach of the law of the Treaties, it could take a variety of actions against the guilty Member State. There are economic sanctions, fines and European Funds’ cut.

Duomo di Milano, Milan, Italy

In addition, taking domestic political costs into account, “the challenge would be to create a stable demand for a new and potentially weaker new domestic currency” (Nordvig, 2014). In order to encourage investment in this new, seemingly fragile, national currency, the Italian government must implement major institutional reforms such as labour market or law environmental regulations which could address their competitiveness problems. Moreover, not only Italy would suffer the damage of Italexit but also the whole European Union gets hurt. The trust in the Euro and the whole European community would be reduced by other Member States’ banks, driving out investments from the continent. Diplomatic tension and political resentment could follow, and cooperation in any sector, also nonmonetary issues, would suffer. Besides, the Italexit could raise doubts about the future of the monetary union. Residual members would suffer a further loss of competitiveness. This would cause a domino effect as other peripheral countries would see their situation further worsened and might want to follow Italy’s example.

III. CONCLUSION

The possibility that an incumbent member of the euro area might reintroduce its national currency cannot be excluded. The EU is still an entity whose residents identify themselves as citizens of nation-states. Differences in national history and identity imply differences in preferences over monetary policy. A country contemplating exit in order to obtain the kind of real depreciation needed to address problems of chronic slow growth and high unemployment would be deterred if it thought that its efforts to engineer a real depreciation would be frustrated by the inflationary response of domestic wages and prices, or if it thought that leaving the monetary union would significantly raise its debt servicing costs. But if the defector strengthens the independence of its central bank and the efficiency of its fiscal institution, then it is at least conceivable that these negative economic effects would not obtain.

In contrast to some other authors, in this paper, I have tried to address, and later disprove, the main rationale which led many authors to believe that Italy should leave the Euro. “The Euro is irrevocable” – The former president of European Central Bank, Mario Draghi also made a strong statement about the possibility of the breakup of the Euro. Italexit could still happen as the unwilling and messy result of an unbearable deterioration in public finances and economic performance, combined with misguided political will and financial market turmoil. It would be a huge mistake. Much better, and less costly, would be to address the underlying problems, allowing Italy to survive and thrive within the Euro by enhancing potential growth and economic resilience. It would be wrong to conclude that Italexit or exit from the monetary union by any other Member State, is going to be an easy process that can be evaluated with a straight cost-benefit analysis and smoothly managed in an orderly way. In the case of Italy leaving EMU, redenomination and default would become very likely and would cause a negative effect on both the economy and polity of the country. Hence, Italexit would not address the issues its proponents claim it would address while producing significant financial instability. Just mentioning it as a viable solution as part of a political platform would simply risk of making it a self-fulfilling prophecy. The economic, social, and political consequences would be enormous and last for years. In light of everything discussed, I can then conclude that leaving the Euro, especially because of economic and political costs, is not a feasible option for Italy.

Hamburg, 28.02.2020 in the peak of the coronavirus invasion.

HUY NGUYEN – Du học sinh Việt Nam tại CHLB Đức

References

Alberto Bagnai, Brigitte Granville, Christian A. Mongeau Ospina. (2017). Withdrawal of Italy From The Euro Area: Stochastic Simulations of A Structural Macroeconometric Model. Economic Modelling, 64, 524-538.

Ansgar Belke, Daniel Gros. (2002). Monetary integration in the Southern Cone. The North American Journal of Economics and Finance, 13(3), 323-349.

Biais, Bruno, Fany Declerc, James Dow, Richard Portes,Ernst-Ludwig von Thadden. (2006). European Corporate Bond Markets: Transparency, Liquidity, Efficiency. London: Centre for Economic Policy Research. .

Bini-Smaghi, Lorenzo . (1998). The Democratic Accountability of the European Central Bank. Banca Nationale del Lavoro Quarterly Review 205, 119-143.

Bootle, R. (2012). Leaving The Euro: A Practical Guide. Capital Economics.

Buiter, Willem . (1999). Alice in Euroland. Journal of Common Market Studies 37, 181-209.

Dědek, O. (1996). The break-up of Czechoslovakia : an in-depth economic analysis. University of California: Avebury, Print.

DeVries, P. D. (2016). An Analysis of Cryptocurrency, Bitcoin, And The Future. International Journal of Business Management and Commerce, 1(2).

E.Stiglitz, J. (2016). The Euro – How A Common Currency Threatens The Future of Europe. New York, London: Norton & Company.

Eduardo Borensztein, José De Gregorio. (199). Devaluation And Inflation After Currency Crises. International Monetary Fund.

Eichengreen, B. (2010). The Breakup of the Euro Area. The University of Chicago Press. Retrieved from http://www.nber.org/books/ales08-1

Erlanger, S. (2018, 05 29). The New York Times. Retrieved from Negotiations in Italy Push Euro to Fore, The Last Place Europe Want It: https://www.nytimes.com/2018/05/29/world/europe/italy-euro-european-union.html

Francesco Paolo Mongelli, Juan Luis Vega. (2006). What Effects is Emu Having on the Euro Area and its Member Countries? An Overview. ECB Working Paper No. 599.

Fugazzi, S. (2017). Fake News Fake President . United Kingdom : Lulu.com. Retrieved from https://www.amazon.co.uk/Fake-News-President-Stefano-Fugazzi/dp/0244318972

Nordvig, J. (2014). Cost And Benefits of Eurozone Breakup: The Role of Contract Redenomination And Balance Sheet Effects in Policy Analysis. Manuscript.

Proctor, C. (2011). The Euro – Fragmentation And The Financial Market. Capital Markets Law Journal, 6(1), 5-28.

Santambrogio, G. (2018). Leaving The Euro. A Feasible Option For Italy? Working Papers CSE 2018, 18(3).

Scharding, T. (2019). National Currency, World Currency, Cryptocurrency: A Fichtean Approach To The Ethics of Bitcoin. Business And Society Review, 124(2), 181-298.

Sebastian Hauptmeier, A. Jesus Sanchez-Fuentes, Ludger Schuknecht. (2011). Towards Expenditure Rules And Fiscal Sanity In The Euro Area. Journal of Policy Modeling, 33(4), 597-617.

Stiglitz, J. (2018, 04 19). How To Fix The Eurozone. Retrieved from Politico: https://www.politico.eu/article/opinion-italy-germany-how-to-exit-the-eurozone-euro-reform/

Tavlas, G. S. (2004). Benefits And Costs of Entering the Eurozone. Cato Journal, 24(1-2), 89-106. Retrieved from https://heinonline.org/HOL/P?h=hein.journals/catoj24&i=91

2019 – NĂM “VÀNG” CỦA THẾ GIỚI

Viết cho những ngày cuối năm 2019 . . .

Mình vẫn còn nhớ những khoảnh khắc cuối năm 2018 bước sang năm 2019, thời điểm cuộc chiến tranh thương mại Mỹ – Trung vẫn đang đà leo thang, giá dầu (Crude oil) tại thời điểm đó giảm từ 75$/barrel xuống gần 42$/barrel, chỉ số Dow Jones Industry (DOW J) của Mỹ giảm, chạm mốc dưới 22000 điểm, một tổng thống Donal Trump đầy bất ổn, vấn đề môi trường, nội chiến là Yemen, một Venezuela trên bờ sụp đổ, Brexit, khủng hoảng nhập cư ở châu Âu, hay mối nguy về chiến tranh hạt nhân ở Bắc Triều Tiên… tất cả như báo hiệu một năm 2019 đầy u ám và ảm đạm. Tuy nhiên, nếu nhìn lại, người ta có thể lạc quan mà khẳng định rằng 2019 vẫn là một trong những năm tuyệt vời nhất mà thế giới có thể chứng kiến.

Kinh tế – một năm đầy thắng lợi cho thị trường tài chính

Bên cạnh chiến tranh thương mại, xung đột vũ trang và vấn đề nợ công bao trùm nền kinh tế toàn cầu (OECD công bố tốc độ tăng trưởng kinh tế toàn cầu trong năm 2019 là -3.3%), năm 2019 vẫn là một năm tuyệt vời đối với các nhà đầu tư. Những con số trên thị trường tài chính vẫn rất hấp dẫn, tổng số vốn hóa thị trường chứng khoản toàn cầu tăng lên mức 10 nghìn tỷ dollars, thị trường trái phiếu vẫn đang nóng lên từng ngày, giá dầu đã tăng lên 25% so với đầu năm (hiện tại đang là 62$/barrel), các nền kinh tế vừa trải qua khủng hoảng ở châu Âu như Hy Lạp hay Ukraine đang dần được phục hồi, thậm chí giá vàng cũng tăng trưởng ấn tượng.

Thị trường chứng khoán tăng hơn 10 nghìn tỷ dollars trong năm 2019

Chỉ số của Wall Street và MSCI (Morgan Stanley Capital International – một trong những công ty tài chính lớn nhất ở Mỹ, có trụ sở chính ở New York) xác lập kỷ lục khi lần lượt tăng 30% và 24%. Các chỉ số tài chính ở khu vực Châu Âu, Nhật Bản, Trung Quốc và Brazil cũng tăng thấp nhất là 20% theo giá dollar.

Trước động thái cắt giảm lãi suất và nới lỏng định lượng (Quantitive Easing) của FED (Cục dự trữ liên bang) giúp làm nóng thị trường trái phiếu. Trái phiếu chính phủ Mỹ (U.S Treasuries) tăng 9.4%, trái phiếu chính phủ Đức (German Bunds) tăng gần 5.5% theo giá euro.

Về thị trường hàng hóa, giá dầu tăng gần 25% kể từ đầu năm, giúp cho thị trường chứng khoán ở Nga tăng trưởng cao nhất so với phần còn lại của thế giới trong năm 2019 với mức tăng trưởng 40% và làm đồng Rúp Nga lọt trong top ba ngoại tệ có mức cầu cao nhất.

Các công ty công nghệ lớn tiếp tục chiếm ưu thế trên thị trường chứng khoán. Mặc dù Apple có thể mất ngôi vương – công ty có giá trị lớn nhất trên thế giới – vào tay Saudi Aramco, tuy nhiên hãng “táo khuyết” vẫn có thể tự an ủi bản thân bằng sự tăng trưởng 77% trong năm 2019. Facebook tăng 57%, Microsoft 53%, Google 30%, Netflix 24% và Amazon 19%, Alibaba 53%.

Thị trường Crypto cũng nóng không kém với sự trỗi dậy của Bitcoin, giá trị Bitcoin tăng 260% vào tháng 6/2019.

Về tình hình an sinh xã hội

Thoát khỏi vùng “cực nghèo”

Kể từ năm 1981, 42% dân số trên toàn thế giới nằm ở mức “cực nghèo” (được định nghĩa mức sống dưới 2$/ngày tức khoảng dưới 46k VND/ngày). Con số này giảm xuống chỉ còn dưới 10% trong năm 2019.

Phần trăm tỷ lệ dân số nằm ở mức cực nghèo từ năm 1980 – 2019

Tỷ lệ tử vong ở trẻ em trên toàn thế giới giảm đáng kể

Các bệnh như bại liệt, phong cùi, mù lòa hay phù nề ở trẻ em được giảm đáng kể. Với sự chung tay nỗ lực của các nước trên thế giới, căn bệnh thế kỉ AIDS cũng được đẩy lùi.

Tỷ lệ dân số biết chữ cũng được cải thiện

500 năm về trước, phần lớn dân số trên thế giới mù chữ, tuy nhiên theo thống kê của World Bank, tính cho đến năm 2019, con số này giảm còn dưới 10%, nghĩa là hơn 90% dân số trên toàn thế giới biết chữ, đặc biệt là số phụ nữ được phổ cập giáo dục được nâng cao. Sự nhận thức về quyền bình đẳng giới tính cũng được cải thiện.

Tỷ lệ dân số biết chữ ở các nước trên thế giới

Việt Nam

Chiến tranh thường mại Mỹ – Trung tuy gây thiệt hại hàng tỷ dollar cho nền kinh tế toàn cầu nhưng Việt Nam lại nằm trong số ít quốc gia được lợi từ cuộc chiến tranh này. Việt Nam được coi là phương án thay thế Trung Quốc cho các doanh nghiệp nước ngoài khi lựa chọn địa điểm sản xuất, với lợi thế cạnh tranh là lực lượng nhân công giá rẻ, năm 2019, dự báo GDP Việt Nam sẽ đạt 7.15%, thuộc nhóm các nước tăng trưởng cao nhất trong khu vực và trên thế giới, nâng quy mô GDP lên khoảng 266 tỷ dollars, thu nhập bình quân đầu người đạt gần 2800 dollar, đáng mừng nhất là tỷ lệ nợ công so với GDP giảm mạnh xuống còn 55%. Nguồn vốn đầu tư nước ngoài FDI cũng thể hiện rõ lợi thế này khi liên tục tăng trong 10 tháng đầu năm. Tại hội nghị trực tuyến toàn quốc giữa chính phủ và các địa phương, Tổng bí thư – Chủ tịch nước Nguyễn Phú Trọng có lời phát biểu ấn tưởng về sự tăng trưởng của Việt Nam trong năm 2019: “Mây đen phủ lên toàn cầu, nhưng mặt trời vẫn đang tỏa nắng ở Việt Nam”.

Bên cạnh các chỉ số đẹp về kinh tế, Việt Nam cũng phải đối mặt với những vấn đề môi trường- xã hội như chỉ số ô nhiễm đang trong tình trạng báo động, năng suất làm việc vẫn chưa cao. Tuy thu hút được vốn FDI nhiều trong những năm gần đây, nhưng chủ yếu vẫn dựa vào nguồn nhân công giá rẻ, nếu chúng ta không cải thiện được chất lượng nguồn lao động thì sẽ không thể giữ chân được các nhà đầu tư nước ngoài và Việt Nam sẽ mãi là “sân sau” gia công của thế giới. Nếu muốn phát triển nhanh và bền vững thì theo ý kiên cá nhân chỉ có một cách là đầu tư vào nguồn vốn con người để cho ra nguồn lao động có chất lượng cao, từ đó giảm sự phụ thuộc vào nền công nghiệp khai thác cũng như gia công chế biến các nguyên liệu thô. Đầu tư vào nguồn vốn con người sẽ là lời giải cho hai bài toán đối nghịch môi trường và tăng trưởng kinh tế ở Việt Nam. Ngoài ra, tham nhũng cũng cần phải giải quyết triệt để. Chống tham nhũng không chỉ làm giảm thiệt hại về kinh tế cho đất nước mà còn củng cố lòng tin ở nhân dân đối với bộ máy chính quyền. Theo World Economic Forum, tham nhũng gây tổn hại cho các nước đang phát triển 1.26 nghìn tỷ dollars mỗi năm tuy nhiên gần một nửa dân số ở các nước này cho rằng hối lộ và tham nhũng là có thể chấp nhận nếu nền kinh tế ở trong giai đoạn suy thoái.

Nhìn chung bên cạnh những thách thức vẫn còn đang được đặt ra, năm 2019 đã đạt được nhiều thành tựu nổi bật về kinh tế và xã hội trên toàn thế giới, mở ra nhiều hy vọng cho một thập kỷ mới khởi sắc hơn cho toàn thể công dân toàn cầu nói chung và cho Việt Nam nói riêng. Là một người con của đất Việt, không mong gì hơn khi được chứng kiến sự phát triển từng ngày của đất nước cả về kinh tế, môi trường và quan trọng nhất là con người, nguồn “vốn” mà người viết đặt niềm tin mạnh mẽ rằng, là then chốt quyết định để đưa Việt Nam ra đấu trường khu vực và cả quốc tế, sánh ngang với các nước phát triển khác.

Bài viết cho những ngày cuối năm ở trời Tây. Rất mong được sự góp ý của bạn đọc. Chúc mọi người một năm mới an khang thịnh vượng – Einen guten Rutsch ins neue Jahr (Một câu chúc mừng năm mới của người Đức mà mình thấy rất hài hước, dịch ra là chúc bạn có một cú “trượt” trơn tru trong năm mới, ai bảo người Đức không có khiếu hài hước nhỉ?)

Hamburg, 30.12.2019

HUY NGUYEN – Du học sinh Việt Nam tại CHLB Đức

CHÂU ÂU – KỶ NGUYÊN RỰC RỠ HAY LỤI TÀN

22904723_1643647709020238_5262929551187090749_o

“Liên minh châu Âu đang đứng trên bờ vực của sự sụp đổ, các nhà lãnh đạo châu Âu nên thức tỉnh và hành động, nếu không, sự biến mất của khối EU là không thể tránh khỏi” – Emmanuel Macron, đương kim tổng thống Pháp, phát biểu trước báo The Economist.

Lục địa già châu Âu từ lâu đã được thế giới xem như là cái nôi của nền văn hóa – nghệ thuật và cả khoa học hiện đại của nhân loại, cố tổng thổng Anh, Winston Churchill, trong bài diễn văn của mình tại đại học Zurich, Thụy Sĩ có đề cập đến tầm quan trọng của Hợp chủng quốc Châu Âu (United States of Europe) đối với 28 quốc gia thành viên nói riêng và thế giới nói chung – “ Sẽ là niềm hạnh phúc và sự thịnh vượng vô hạn cho 400 triệu người khi Liên minh châu Âu được hợp nhất và cùng nhau chia sẻ những thành tựu đã và đang đạt được”.

Tuy nhiên, châu Âu có thật sự đang chạm tới cánh cửa thiên đường, sự thịnh vượng và hạnh phúc trường tồn? Tại sao, tổng thống Pháp Macron lại phát biểu đầy tính bi quan về tương lai của châu Âu như vậy?

Năm 2010, các nước EU chấn động bởi cuộc khủng hoảng tài chính bắt nguồn từ nợ công Hy Lạp, mặc dù ngân hàng Trung Ương châu Âu (ECB) đã kịp thời ngăn chặn sự lây lan, tuy nhiên, hậu quả kinh tế mà các nước thành viên phải gánh chịu là không thể tránh khỏi khiến EU bước vào thời kì suy thoái. Hơn thế nữa, ngay lúc này, Ý – “cơn bệnh” của châu Âu – đang manh nha đe dọa EU bằng một cuộc khủng hoảng tài chính khác có sức công phá gấp đôi hoặc gấp ba so với năm 2010. Quỹ tiền tệ thế giới (IMF) cũng đã cảnh báo rằng “vỡ nợ công ở Italy (nước có nền kinh tế đứng thứ 8 thế giới) là không thể cứu chữa- không một tổ chức tài chính nào có thể cung cấp gói hỗ trợ cho quả bóng khổng lồ này”. Italy đang có nợ công ở mức 134% GDP với quy mô kinh tế là 2 nghìn tỷ USD với tốc độ tăng trưởng thấp, chỉ 1.5% (2017), tỷ lệ thất nghiệp ở Italy là 10.5% (2018). Nếu như điều này xảy ra, sự tan rã ở châu Âu là tất yếu, các nước thành viên kể cả Đức, Pháp, hay Hà Lan đều chìm vào kỷ nguyên đen tối và nền kinh tế sẽ mất cả thập kỉ để phục hồi.

Vậy nguyên nhân do đâu đã đẩy châu Âu đến bờ vực của sự tan rã?

Nhà kinh tế đoạt giải Nobel (2001) Joseph E. Stigliz trong cuốn sách The Euro đã chỉ ra rằng việc có đồng tiền chung Euro thực chất là một bài toán khó cho các nước thành viên. Tưởng tượng một nhóm học tập ở lớp đại học, chúng ta có các thành viên nổi bật chịu khó chăm học như Đức, Pháp, Hà Lan, Áo… nhưng cũng có những thành viên “lười” hay trốn họp nhóm như Hy Lạp, Ý … nhưng cả nhóm đều nhận được điểm chung – đồng tiền chung euro – Đức với hiệu suất làm việc cao gấp hai hoặc ba lần Ý hay Hy Lạp, vô tình chung đã làm giảm giá sản xuất và tăng sức mạnh của đồng euro cao hơn nhiều lần gây nên áp lực cho các nước thành viên trong khối EU. Ví dụ, một công nhân Đức một giờ được trả 10€ thì sản xuất đc 10 cái bút thì công nhân Hy Lạp nhận được mức lương tuy thấp hơn 7 € nhưng chỉ sản xuất một giờ được 5 thậm chí 2 cái bút. Do đó, giá thành sản xuất ở Đức rẻ hơn, làm tăng sức cạnh tranh trên thị trường quốc tế. Hy Lạp, Ý không thể bắt kịp hiệu suất làm việc như vậy nhưng các công ty vẫn phải trả mức lương cao ở đồng euro, dẫn đến việc sức hút đầu tư giảm, tỷ lệ thất nghiệp gia tăng kéo theo nền kinh tế đi xuống. Chính phủ muốn thúc đẩy tăng trưởng kinh tế chỉ còn cách thông qua chính sách tài khóa, mở rộng chi tiêu chính phủ, hậu quả là nợ công tăng cao. Tuy nhiên do sức mạnh của đồng Euro cũng như sự bảo trợ của ECB và các nền kinh tế mạnh như Đức, Pháp, Hà Lan. Hy Lạp và Ý, vẫn được các quỹ tài chính quốc tế như IMF thậm chí cả ECB cung cấp các khoản vay thông qua việc mua lại trai phiếu chính phủ. Vấn đề là nếu trong dài hạn, Hy Lạp, Ý vẫn phụ thuộc vào chi tiêu chính phủ để thúc đẩy kinh tế mà không tập trung cải cách nguồn lao động, hiệu suất làm việc hay vốn con người (human capital), thì một cuộc khủng hoảng tài chính tiếp theo là không thể tránh khỏi.

Đức, từ lâu vẫn bị xem như là thủ phạm gián tiếp gây ra “lỗi” ở đồng Euro. Lời cáo buộc cho rằng, Đức lợi dụng nền kinh tế tự do, không có rào cản thuế quan ở châu Âu và bằng hiệu suất làm việc cao của mình, chiếm hầu hết thị phần ở thị trường châu Âu nói riêng và quốc tế nói chung. Các nước thành viên EU có nền kinh tế kém hiệu quả hơn không thể dùng rào cản thuế quan hay giảm giá trị tiền tệ để đối phó với sự bành trướng của Đức, đã vô tình bị đẩy vào thế phải đi vay. Hơn thế nữa, dù quy mô kinh tế tăng cao, nhưng chính phủ Đức vẫn áp dụng chính sách tài khóa thắt chặt so với các nước còn lại, có lẽ vì đã từng có trải nghiệm không mấy tốt đẹp về siêu lạm phát vào những năm 1921 – 1923 mà anh bạn Đức của chúng ta tuy rất giàu nhưng chi tiêu rất dè xẻn khiến đồng Euro trở nên khá “mạnh” so với các ngoại tệ khác. “Germanexit” là một trong những giải pháp để chữa lành cơn sốt cho EU.

Nguy cơ chiến tranh Thương mại Mỹ – Trung đang làm dấy lên mối lo cho ECB về một đợt suy thoái tiếp theo, tháng 9/2019 lần đầu tiên kể từ năm 2016 ECB giảm lãi suất dưới mức 0, chính sách này trong ngắn hạn mong muốn thúc đẩy cầu nền kinh tế thông qua sự cho vay của ngân hàng Thương mại, hạn chế khoản tiền dự trữ của ngân hàng trung ương ở ECB (lãi suất âm là ngân hàng Thương mại sẽ phải mất phí khi gửi tiền dự trữ ở ECB). Mình thấy rõ điều này qua giá cả bất động sản ở Đức, cụ thể là ở Hamburg trong năm gần đây. Ngân hàng cung cấp những gói vay (mortgage) khá hấp dẫn cho người dân, làm cầu bất động sản tăng mạnh, đẩy giá nhà đất lên cao. Tuy nhiên, nhiều nhà kinh tế cho rằng đây là con dao hai lưỡi, điều này làm mất đi lợi nhuận của ngân hàng Thương mại, khiến họ dè chừng và cẩn thận hơn trong việc cho vay và từ đó làm cung nền kinh tế giảm xuống. Khiến cho tình hình suy thoái ngày càng trầm trọng hơn.

Tình hình kinh tế bất ổn ở Ý, khủng hoảng nhập cư, đồng Euro, nguy cơ suy thoái và Donald Trump đang là những vấn đề làm đau đầu các nhà lãnh đạo EU. Có thể thấy tổng thống Macron đã đúng khi nói Liên minh châu Âu chưa bao giờ “mong manh” đến thế. Mong rằng mùa giáng sinh sắp tới sẽ xoa dịu nỗi lo của người dân ở lục địa già này. Thôi lấy mì tôm ăn đã. Cảm ơn đã đọc đến cuối bài.

HUY NGUYEN – Du học sinh Việt Nam tại CHLB Đức

1.https://www.cnbc.com/2019/11/11/france-president-emmanuel-macron-the-eu-is-on-the-edge-of-a-precipice.html

2.https://www.youtube.com/watch?v=giilcPJsYuw

3.https://www.bloomberg.com/opinion/articles/2019-04-30/maybe-europe-can-t-recover-from-the-financial-crisis

4. https://scholar.harvard.edu/files/markesposito/files/eurocrisis.pdf

5. https://www.cfr.org/in-brief/does-italy-threaten-new-european-debt-crisis

6. https://www.dw.com/en/why-is-italy-still-the-sick-man-of-europe/a-48500455

7.https://www.forbes.com/sites/simonconstable/2019/06/20/doomed-how-theres-no-way-out-of-the-debt-crisis-for-italy/

8. https://www.dw.com/en/debt-and-doom-loops-the-eurozones-italian-nightmare/a-50019077

9. https://www.amazon.com/Euro-Common-Currency-Threatens-Future/dp/039325402X

10. https://www.nber.org/chapters/c11654.pdf

11. https://www.nber.org/chapters/c11654.pdf