Microeconometrics - Science topic

Wendy, thank you for writing. I noticed that comment you do not mention the need overlooked even by the WCED 1987 to transition from pollution production economies to pollution-less economies such as the transition from environmental pollution production economies to environmentally clean economies, for which you need first to set up POLLUTION REDUCTION MARKETS LIKE GREEN MARKETS, which then can be driven towards environmental pollution-less markets by closing the renewable energy technology gap to make environmental pollution reduction A PROFIT MAKING OPPORTUNITY...

I appreciate the comment.

Respectfully yours;

Lucio

Osama, since 1987 WCED the world knew that there was a need to transition systematically, locally and globally, to clean economies and leave pollution production economies like the coal based economy and the oil based economy behind. No plan has ever been made, even the paris agreement avoids going beyond managing externalities a la sustainable development.... The paper is above a step by step way to do it if one day the world has to do it..... The sustainability crisis under management sooner or later will backfire and force a more painful and faster local and global transition from pollution production economies to pollution reduction economies and then to clean economies....You can see the trend of the crisis from bad to worse by looking at data WCED 1987 and 2024....while the problem has been under management instead of being fixed....

Thank you for taking the time to comment

Dear Sadam Jamaldin and

Thank you very much for your answers and tips.

They are not green markets because they are not cleared by a green market price, they are environmental externality management based markets.

My apologies for not noticing sooner that this thread continued with a further question. Consider running a random effects panel regression, with each person who is tracked over time as a panel, person-level demographic and other explanators, and dummy variables denoting the group/village each person was in (with no dummy for the comparison group). Run a Hausmann test to assure random effects made sense. To look at the effects of the intervention, initially do not code demographic/geographic differences between the villages as explanators. But in a subsequent run, code those to probe whether differences observed (or lack thereof) resulted from differences between the villages rather than the intervention. Often, the significance of intervention only becomes clear when those village-level controls are added. I'm unsure if ANCOVA also allows a panel formulation.

sounds interesting. I would be happy to receive more details.

1. You need to run a panel regression. Each family is a panel. Since you mentioned HH fixed effects, I think you are aware of that.

2. Poisson is an arrival rate distribution (yes-no), does not make sense in this context, although you could try a mlogit or hazard rate regression on number of years of school attended. Since expenses have a long-tailed range, you probably will want a log-linear model. Rather than OLS, most likely run a generalized linear model, possibly experimenting with a gamma link. Especially if your severely disabled often have no educational expenses, consider a 2-stage model based on Manning, W. G., Basu, A., & Mullahy, J. (2005). Generalized modeling

approaches to risk adjustment of skewed outcomes data. Journal

of Health Economics, 24, 465–488. doi:10.1016/j.jhealeco.

2004.09.011

3. You need a variable to control for birth order (i.e., first child, second child, etc). In that count, exclude children who died before they were school age. So e.g., first child surviving to age 5.

4. I would probably exclude families with a multiple birth.

5. Sex of child obviously is a critical variable and I am assuming all the data come from one country.

6. How will you handle government vs private expenditure, compulsory education rules. (obviously not an issue in some countries) [Extreme example: in the US, severely disabled all are sent to school as the school essentially provides custodial day care or a pupil personnel worker who comes to the home.]

7. If you are collecting data prospectively, use the WHOqol-BREF with the parents. Get a parental assessment of the child's functional capacity using the Health Utility Index-3, the PEDSqol ,or one of the new generation of similar instruments.

8. If you wish to chat further, I'm in the DC suburbs, miller@pire.org

Dear Alexander Shepherd ,

On theoretical grounds, you can find a good and accesible introduction to the topic in th 8th chpater of Mostly Harmless Econometrics (Angrist & Pischke, 2008).

Intuitively, clustered standard errors allow researchers to deal with two issues:

(1) Correlation of observation in the same group (e.g., students in the same class, which are more likely to be similar or share unobserved characteristics than others in the other classes).

(2) Correlation over time of the same units (e.g., students or classes over time)

The use of this approach allows for any sort of correlation within the clusters and over time, so the researcher has not to worry about serial correlation.

In your case, the motivation is only the 2nd, since you have countries (with no smaller units inside you are considering) and years. Therefore, your clusters should be, in principle, countries. This can be easily implemented in Stata (command xtreg):

xtreg depvar [indepvars] [if] [in] [, re RE_options] vce(cluster group)

where group would be country in your case.

The problem is that you only have 19 countries here. It is not clear which is the threshold for this (most of the literature suggests at least around 50).

A remedy for this is proposed by Cameron and other authors, and you can implement it thourgh the package clustse in Stata (you have to download it).

Another possibility is to model the serial correlation. Depending on the topic, one must have some ideas of how the serial correlation might be. For instance, fertility usually follows and AR(1) process. A possibility could be to test for the robustness of differents forms of serial correlation (e.g., )

Related to this, it is useful to comment at which level should you cluster. You only have to take into account this issue when you introduce variables other than dummies at a certain level (if they are just dummies and you are not interested in the coefficient of them, no problem with that level). Sometimes you have two levels that do no overlap, e.g., firms and workers with workers moving from one to another firm over time. In that case, you must use two-way clustering (in Stata, you have to use the package reghdfe). Your case is not this one as far as I know. When you have two levels that overlap (e.g., regions and countries, with countries comprising regions and region- and country-level variables), you have just to cluster at the highest level (e.g., country).

I hope it helps.

All the best,

Nacho

You can Search it in Bloomberg Database . But it's a very costly database and may not available in all places.so contacts any IIM and request them to access bloomberg and you can get that data on every country in this world.

Thank you Valentina. I later came to understand that cost efficiency indices from Frontier 4.1c software are greater than one due to the nature of error term (v+u)

Dear Sebastian, As you only have two waves just pool your data and add a time dummy. You can add some time*Xs where you think appropriate. Cheers, Marc

Seems to me that the initial problem is that many items have no variance, or very little (if everyone answered 4 or 5, then they are essentially endorsing all the items.

In a case like this, you are up against the very limits of science, which attempts to relate variation in one property to variation in another. With no variation to model, the system just isn't possible. The distinction between a 4 and a 5 on a Likert scale is driven by factors like response recency, education etc, so much of the variance in the data is going to be non-construct-related but correlated between items.

I am doubtful if anything can be done in this case, I'm sorry to say.

thank you sir.

Radka,

If your ultimate goals is to be able to assess the overall level of those systemically stable elements of a business's activities which have enduring 'local impact'and at the same time recognise that business activities are not events occurring at a point in time but are processes of which time is an attribute of each of the effects of each of its actions, then you have immediately placed your work outside of the CGE orthodoxy of mainstream economic analysis,  Not an uncommon place to be for most development economists when they become focussed on delivering real results for real communities rather than marketable policy recommendations for PR and political rhetoric purposes of central governments.

In the former case your route should take you into the methods, tools and concepts of the evolutionary and institutional approaches within economics.  It should encompass all the behavioural, organisational and legislative institutions whose operation flows through time and create an institutional landscape characteristic of the context of each target community, and of those more remote communities whose landscapes' characteristics are themselves determinant factors in the conduct of the local operations of those firms whose ownership and control is influenced by the norms and values embedded in them and which differ from those of the local target's institutional context (as described by its local institutional landscape).

It sounds complex but, when you have visualised it, the idea is quite simple.  With that idea you have a framework inside of which you able to more easily conceive of those factors which you seek and to design the measurement methodologies which are appropriate to their nature.  Overall assessment of what will be a profile of disparate metric types might best be approached by another 'landscaping' approach. This time based upon a 'radar diagram' representation may be rather than the multi-dimensioanl 'surface analogue' by which institutional landscapes might be compared and assessed.

This part of a such a methodology might equip you in seeing how business size might be correlated with the systemic exportation through transfer payments and profit attribution of labour value to the detriment of local goals for long-term development, stability and the social cohesion which changes in income disparities tend to damage.

This approach might also provide the framework within which might be studied the effect of re-purposing and redirecting of local labour into directly creating and evolving local institutions that are conducive to the sustained development, support and maintenance of local social, physical and organisational infrastructure that are conducive to raising the efficiency and effectiveness of the local community in producing and increasing value which feed directly into the 'well-being value' experienced as a lifetime flow for each individual and their households.

The same methodology, applied to the current institutional landscape, and the firms operating within it would identify those behavioural and policy institutions, operating upon and within the local community, whose main function are, in practice, to facilitate the 'export'  to more central, and hence remote, communities value produced by local labour through their local economic activities.

Given the characterisation and detailed understanding of these two states it would create a locally relevant contextual framework of the present and future institutional landscapes and suggest the policy paths, whose evolution and implementation would lead to the evolutionary diminution and, perhaps, abandonment of those institutions of the present which are antagonist to the desired development goal.  Similarly, those policy paths would include elements which lead to the creation and enhancement of those institutions which are then being seen as promoting the local development goals.

Moral and ethical integrity, of global international force and validity, can be embedded within the target developmental institutional landscape for each locality.  This can be done by integrating into the desired institutional landscape those elements which are consistent with that set of institutional elements, that as a set are regarded as necessary and sufficient, for that institutional proto-landscape which is deemed to describe and define the conditions essential for any society or community capable of durably, resiliently and sustainably conducting its life in strict accordance with the aggregate agreed wisdom of humanity as expressed through the normative declarations of those statutory international bodies we have established since 1948 and which are now evolving out of their infantile stage into some form of co-responsibility and credibility.

The same methodology of landscape building applied to the current body of pronouncements as a whole would perhaps be very instructive and salutary for us all to consider.  It would be expected to identify those elements in what has already been said and agreed which are in fact antagonistic to the overall goals and perhaps serve various local interests and have no place in the setting of a holistically relevant, universally applicable vision of a humanitarian future by which humanity may not only survive and evolve but in which every lifetime lived can also take pleasure in doing so.

Radko,  The approach you choose to take now cannot employ a methodology that is not yet in existence.  However, the birth of that methodology can be facilitated by researchers such as yourself developing, in the course of their own work small elements and tools, and employing specific conceptual frameworks, in argument, which are consistent with its ideas.

Human behaviour has and continues to evolve.  Societal development, at the social and the economic levels, is one of continual evolution and these elements are completely intertwined and are characterised by flows, and flows within flows.  Some order is imposed upon that river flowing through time by degradable and modifiable conduits of different strength, diameter and length... they are our institutions.  The river has many tributaries.. the isolated communities in our history. But, we can now see that it will always be characterised as being a flow passing through a web of channels around many obstacles... these are our geographic and environmentally determined regions.  They lie in ever changing patterns as they present themselves and we pass them by... but always as one river.

This river of institutions and regions is itself comprised of molecules of water... each one of which is you and I, and every other person.  The river is us.  No molecule is any different from any other, and through the twists and turns as the river runs along its bed of time any molecule may end up in any channel, in any conduit and each remains a part of the whole, entitled to be in the liquid which is the water of the river.  End of metaphor...before I stretch it too far.

The point being that you cannot model the behaviour of any part of a river using a time series of statically calculated equilibria at each moment along the chosen axis of time and extrapolate and predict future states of equilibria beyond the present moment.  Using our metaphor, you know that there are conduits, ending, starting, flexing and cracking.  But you do not know which, where or when.  Just that they will.  The same goes for the obstacles up ahead, the channels opening up and those which are closing ahead of you, even dead-ends forming.  You know that they will occur and again do not know when,where and which.  

Any physicist would die laughing at you if tried to build such a model and would throw a tome or two describing basic fluid mechanics at you as he took his last gasping, hilarious breath.  You may be still alive.  But you are wrong and he was right!  At least, demonstrably more so than yourself.  Fluid mechanics may only hold a few useful tools, or even just concepts to help our thinking about ourselves.  But we can be quite sure that no living creature has or ever will be able to rationally arrive at any optimal solutions to most of life's problems in any useful way at the individual level, and so when aggregated their behaviour cannot be expected to be done in any accurately representational manner.

NP-hard problems make up most of life's decisions, we do not calculate solutions to them, nor could we for the most part.  We have evolved, and continue to evolve, other means to handle life's complexity.  Much of that methodology is now institutionalised in our newly acquired technologies and tools.

For more discussion of this I refer you back to what I mentioned many words earlier.   A seminal textbook, bringing together foundational and coherent argument, more suited to handling and studying the realities of life, its societies and its economic activities, in particular, is in the making here on RG.   Prof Shiowaza being one.  One might also delve into http://afee.net/

I expect I haven't been much help to you in producing you next paper, but it can sometimes be useful to take you eyes off the ground at your feet and gaze for a while at the horizon....  if only to check that your path is still set towards the sun and not away from it.  :-))

Robin

It's a bit computationally intensive, but out-of-sample prediction might work. It's a bit tricky with panel data but easily implemented in Stata.

Thank you Amir for your reply.

I've already tried to increase the number of iteration without success, so I think the problem is related to the first point. Even if I don't have convergence, I obtain the value of the parameters and the measure of BIC and AIC. Are these results reliable? At the end it turns out that the lowest value of BIC and AIC are obtained with 6 classes (in which I don't have convergence). Do you suggest to use only the results of the 3 classes  as the results of 6 classes are not reliable? Thank you

May be these papers will be useful

I understand that if T tends to infinity, the estimator of fixed effects is consistent. However, if T is fixed and N tends to infinity (typical in models with reduced-sized), only the estimator β of fixed effects is consistent. While the estimator of individual effects (α+μ_i) won't be consistent because the number of parameters is increased when N also increases.

Maybe, you can find more information in:

I think it is important to take into account market expectations which affect price. It is important to have in mind that that price is that average in the negotiations.

Regards

Francisco

Dear Muhamad,  This is a question that has been troubling economists for years.  never mind the semantics of the English and I focus on a production function or non parametric approach.  You need to be specific and clear about what and how you're measuring things.  Now assuming some kind of neo-classical production function TFP approach a la Jorgenson et al, or alternatively a torquist index approximation of same, then TFP is a residual that could contain (changes in the growth rate index approach): economies of scale, economies of scope, reorganisation of economic activity among the many actors/producers (See Gollop, Ball, Hawke and Swinand) (this may be a deviation from perfect comp), changes in input quality not captured in your current input quantity measurement programme, X-efficiency/inefficiency-non maximising producer behaviour (this is part of technical efficiency), slow adaption of new technologies (technical progress -- this may be optimal based on optimal replacement rates of long lived capital equipment - see the many articles by Jorgenson, etc).  If you're specifying a functional form then there are potential issues there too.  In anycase, if any of the factors such as scale etc, overwhelm the technical progress and technical efficiency, then one could be negative while the others positive.

BTW, I would not call better crop varieties technical efficiency, I would call these technological progress, and adopting best practices as technical efficiency (see Coelli at al textbook).  this is mostly just nomenclature.

To disentangle these, you need a methodology such as those developed by Fare et al.  or a translog distance function if going parametrical estimation route.

All the above answers are correct

If a function u(x_1,x_2,...x_n) represents preferences, any increasing transformation of this function g(u(x_1,x_2,..x_n)) also represents same preferences. The marginal utility, however, changes from u'(x_1,...) to g'(u(...))u'(x_1,..). Therefore, economists care mostly about ratios of marginal utilities.

1. Concentrated Ownership simply refers to the case where majority of shares are held by few owners. Further, if more than 5%, 10%, 20% (different levels) of shares are held by the state, or group of companies, or a family or a foreign investor then you are in a position to identify the owner or block holder of the company. 

2a. To calculate Ownership Concentration, you have to define the levels. For instance you can measure how much ownership is concentrated at Top 1 level. this means what percentage of shares are held by the largest shareholder of the company.

Similarly in literature, Concentration of Ownership at level 3 and 5 (percentage of shares held by 3 largest and 5 largest shareholders) has been calculated.

2b. To calculate family ownership you have to calculate percentage shares held by a family. 

Similarly you can calculate, insiders shareholding, state ownership, foreign ownership etc.

3. Yes, the concentrated ownership means high control of a firm.

4. Private firms are unlisted.

The effect of research on teaching is that it helps to teach a specific topic with confidence and with evidence, instead of vaguely teaching.

Dear Vera,

For the outcome equation: Competing risks models can be readily estimated for each destination separately, you do not need a special package for that: just stset your data accordingly and re-estimate the same models for each outcome, marking the spells that experience the respective other events as zero until that point and censored afterward. Your risk set will be the same each time, but the number of events will vary.

As for the selection equation I remember the routine had some rigidities in that respect, but maybe there, too, you can work with separate binary response models?

Sorry I cannot be of more help.

Best, Jonas

It depends on your research goal.

normally, there is relatonship between income and consumption;

Generally, if I do research on household consumption, income are often less than consumption, and then ask the respondents again; it is difficult to get income sometimes, in this situation, I often use consumption rather than income to get some results. for example, demand income elastisity often is replaced by demand expenditure elastisity,

That's a good question. I have never heard anyone raise that objection before, however, and I don't think you need to worry about it -- certainly plenty of authors have ignored the problem! But it is interesting: Suppose a given case had an erroneously high reported number of hours. That would result in (a) an understatement of their hourly wage, and (b) an increase in the weight attached to that observation. That would tend to bias the weighted average hourly wage downwards for the whole sample. But as long as that bias is roughly similar across subgroups of the population and over time it should not have much effect on the Oaxaca decomposition.

Here's another random Oaxaca tip. If you use Stata you have the option of specifying "pooled", which specifies that the coefficients used to evaluate the "explained effects" of differences in X variables between the two periods are the coefficients from a regression that pools the data, and includes a dummy variable to distinguish the two periods. The thing I like about that specification is that it does not rely on the period 1 or the period 2 coefficients, but rather on a kind of average of the two which seems prudent. Also, the total "unexplained effect" will come out exactly the same as the coefficient on the dummy variable in the pooled regression: it is the part that cannot be explained by the differences in X.

Best of luck,

Tom

Hi Vera, I am not aware of the paper by Boehmke et al. (2006), but with single observation per unit data you can control uncorrelated (with X) unobserved heterogeneity(UH), with multiple observation per unit even for correlated UH (see eg Van den Berg, 2001, Handbook). Concerning dynamic selection, you should have a look at the so called "Timing of events" approach by Abbring/Van den Berg (2003). In addition, I do remember having seen something on selection and duration models by Denis Fougère or Jen-Pierre Florens. I haven't found it now on the spot, but it might be worthwile looking at Florens/Fougère/Mouchart (2008) or Fougère/Kamionka (2008). Best Alfred

The covariance matrix for the Hausman test is only positive semi-definite under the null. It also does not necessarily have the obvious degrees of freedom. Take a simple example. Consider adding an additional variable to an OLS regression. The initial coefficients are consistent and efficient if the coefficient on the additional variable is 0. They are consistent if the new variable is orthogonal to the original variable. Under the null of a zero coefficient, the difference between the covariance matrix with the additional variable and the covariance matrix without this variable is positive semi-definite, but if, in fact, the additional variable has a lot of explanatory power, the standard errors on the original coefficients can come down. Moreover, as this example shows, Hausman tests can often have fewer degrees of freedom than the rank of the covariance matrix and therefore end up being positive semi-definite rather than positive definite under the null. If memory serves me, Hausman discusses this for IV in his original article. I think Paul Ruud had a good review article sometime in the early to mid 1980s and the general issue is discussed in Davidson and MacKinnon's text.

I do not think there is a solution within this non-linear setting without specifing a fully parametric 'first stage'. This is pretty unattractive because (as opposed to the linear model) this first stage has most likely literally to be true to avoid inconsistencies. An alternative is to treat durations as outcome variable and set-up a non-parametric LATE analysis along the lines of Imbens & Angrist. If covariates are needed, the Froelich (2007, Journal of Emetrics) paper would be the appropriate reference for this.