Training Needs Assessment: Where We Are and Where We Should Go Introduction

The aim of this article is to systematically review Training Needs Assessment (TNA) scientific literature. Based on two research questions (where are we? where should we go?), we hoped to evaluate the current state of scientific production on TNA and to point out some possible developments. The following databases were Scielo. Fifty-One articles were analyzed. The results show that: (a) there is little agreement on how to measure training needs; (b) most of the current TNA models and methods are reactive and do not consider contextual factors and multiple levels of analysis in a proactive way; (c) there are gaps in TNA and a need for theoretical definitions; (d) there is little concern with building theories and concepts related to TNA. Based on these findings, we point out that TNA practice and research should: (a) be based exclusively on measurable human competences gaps, in multiple possible levels of analysis; (b) not focus only on individual professional roles, but also on internal and external contextual factors that can be important in the future; (c) discuss and criticize in depth what work needs, training needs and training needs assessment mean; (d) elaborate and test TNA theories, concepts, models and methods.


Introduction
New workplace demands and requirements are causing major changes in formal education as well as in professional training.Some factors seem to introduce a new scenario for organizations: the rapid pace of technological change in the information society, the increasing content knowledge required for production, the reduction in the product life cycle, and rapidly changing production processes.The need for workers' continuous learning is one of the various effects of these pressures.
In this context, Training Needs Assessment (TNA) processes have a strategic role because they provide clear guidelines as to which professional skill deficiencies must be remedied and what the profile of future trainees should be.For McGehee and Thayer (1961), training needs come from underdeveloped skills, insufficient knowledge or inappropriate worker attitudes.Mager and Pipe (1979) define training needs as identified differences between the employees' current performance and the performance that the organization expects of them.
Training Needs Assessment refers to the organizational process of collecting and analyzing data that supports decision making about when training is the best option (or not) to improve individuals' performances, define who should be trained, and exactly what content should be taught (Clarke, 2003).For Wright and Geroy (1992) TNA should be a systematic process of collection, analysis and interpretation of data on individual, group and/or organizational skill gaps.They should have seven key characteristics: (a) be based mainly on culture and organizational philosophy; (b) be proactive instead of reactive; (c) have a method that permits the distinction between situations that can be addressed through training and those that cannot; (d) allow various organizational actors who are directly or indirectly interested and involved in training to participate; (e) be based on observable skills rather than leaders', managers' and professionals'' perceptions; (f) consider the varied use of sampling techniques and data analysis; and (g) in the end, have a cost/benefit analysis.
However, despite its importance, research shows that training needs diagnoses have been done in an unsystematic manner in organizational settings (Clarke, 2003;Ferreira, Abbad, Pagotto, & Meneses, 2009;Ford & Noe, 1987;McGehee & Thayer, 1961;Moore & Dutton, 1978;Ostroff & Ford, 1989;Taylor, O'Driscoll, & Binning, 1998;Wexley, 1984).There is still relatively little theoretical and empirical research on TNA (Kraiger, 2003).Literature review devoted to the subject is rare.In Management, studies lack systematic theoretical and methodological approaches which may provide consistency to TNA research and practices.We can say that the theoretical and methodological characteristics of TNA scientific knowledge are, somehow, unknown.It seems that much of what was recommended by seminal authors (Mahler & Monroe, 1952;McGehee & Thayer, 1961;Moore & Dutton, 1978, among others) is still not completely incorporated into TNA research and practice.
For over 50 years, Training, Development and Education (TD&E) literature has been concerned with the importance of systematic procedures for TNA and the investigation of internal and external variables that influence or originate needs for training in work contexts (McGehee & Thayer, 1961).However, the scientific production in the area has yet to provide plausible answers to this and other important questions surrounding the topic.
It is precisely in such a theoretical and empirical context that this article is justified.In order to help find possible ways to fill these gaps, it is of great importance to describe the current state of scientific literature on TNA, bringing to light and evaluating the methods and theories employed until today and drawing some possible scenarios to the future.Thus, this article is based on two research questions.Where are we when it comes to the current state of TNA scientific production?Based on the current state of TNA production, where should (or could) research and practice go?
The primary criterion established for article selection was that it had to be published in a scientific peer reviewed journal.There were 90 articles, which, after reading the summaries, and assessing the adequacy of the subject, were reduced to 61, of which 51 will be analyzed in this study.Our decision to analyze 51 studies is based on operational questions, like the relationship between the complexity of data analysis, time and workforce.The articles examined in this study are a sample of convenience and do not overstrain the knowledge on TNA.

Criteria and procedures for article analysis
Twelve criteria for analysis of selected articles were determined, as shown in Table 1.The articles were analyzed by the authors and two members of a research group.

Results
We analyzed articles ranging from 1978 to 2010.The scientific literature on TNA experienced considerable quantitative growth between 1990 and 2010.Research in the area remained practically nonexistent in the period 1970 to 1989.
As to the country(ies) in which research data was collected, according to Table 2, there is a clear predominance which took place in England (15).There is also a considerable number of studies on TNA in the United States (11).The need to map leaders' and medical students' perceptions of their training needs.

Holton, Bates and Naquin (2000)
United States To develop and implement a TNA method in the public sector.
Ad-hoc approaches to TNA.Employee participation in the TNA process is crucial.

Discussion
Keeping in mind our research questions and objectives and the recommendations by Baumeister and Leary (1997), we present the discussion as follows.
TNA: where are we?
This section is dedicated to show where our analysis suggests TNA scientific knowledge currently is, in methodological and theoretical terms.Our recommendations about what to do (where to go) given such results are presented in the next section.In sum, one can say that TNA approaches (in practice and research) had a considerable methodological advancement in past decades, shifting from ad-hoc frameworks (Clarke, 2003;Ferreira et al., 2009;Ford & Noe, 1987;McGehee & Thayer, 1961;Moore & Dutton, 1978;Ostroff & Ford, 1989;Taylor et al., 1998;Wexley, 1984) to a more professional and scientific basis.But there are still several methodological weaknesses and a very long path to move forward in theoretical terms.It is also important to say that TNA research experienced a great growth in publications in the last two decades, especially in the 2000's (Kraiger, 2003).
Regarding the main research questions, one can say that the analyzed studies aimed, primarily, to respond: . How can one respond to workers' qualification needs?
. How can one systematize and operationalize TNA processes and practices?
. How can one identify and measure training needs?
. What are the possibilities and limitations of practice, research and current TNA models?
As to the theoretical and empirical issues investigated, it can be said that most studies aimed at addressing one or more of the following: . Diagnose training needs for professionals; . Describe challenges for TNA practice; . Describe weaknesses in current TNA approaches; . Describe/propose TNA procedures; . Construct TNA instruments.
It seems that TNA practice and research still have an almost exclusively diagnostic/procedural and reactive focus, concerning how to do it in the present.Apparently, prospective TNA approaches, based on literature on competence and competences management (Boyatzis, 1982;Cockerill, 1989;McClelland, 1973;Prahalad & Hamel, 1990;Sparrow & Bognanno, 1994), are still rare.Research are mainly applied, investigating methodological or practical problems and solutions related to TNA systems.This is of great value, but there is still a critical lack of theory development and/or evaluation.Still, there is no apparent concern with developing organizational policies on TNA.Some important theoretical issues that are almost absent in the studies are: the relationships between the TNA concepts, work needs, and competence or competences management based on future scenarios (Sparrow & Bognanno, 1994); the missed conceptual link between individual and organizational needs; and to propose new kinds of needs, as learning needs, educational needs, development needs, avoiding practices and research to be dependent on only one kind of possible instructional solution to meet competence gaps (training).Methodologically, it is possible to note the prevalence of survey-type studies (34.63%).Eleven (11) theoretical essays, three (3) case studies and three (3) action research studies were also obtained.There is a relative predominance of quantitative studies (20).Mixed studies (qualitative/quantitative) and qualitative data showed moderate frequency (11 and 10, respectively).There is also a relative dominance of questionnaires as data collection instruments (26.49%).Some research (9) used questionnaires and interviews, which suggests consonance with qualitative/quantitative studies.Four (4) studies reported using only interviews.Thirteen (13) research theoretical reports were not subject to this analysis criterion.The methodological diversity of scientific knowledge on TNA must be prized (Baumeister & Leary, 1997).The use of multiple data collection methods (e.g.questionnaire, interviews, focus groups) and analysis (e.g.content analysis, descriptive and inferential statistics) is highly desirable to investigate complex phenomenon, as in the social and behavioral sciences.On the other hand, we can note that this methodological diversity is accompanied by a high diversity in results and conclusions as well, even regarding the same object (training needs at work).There is no convergence of results: some authors define training needs as a occupational competence gap (e.g.Borges-Andrade & Lima, 1983); others understand it as a performance gap at multiple levels (e.g.Asku, 2005); and others suggest it being the number of vacancies in an organizational sector (e.g.Castley, 1996).Therefore, we do not know if the methods are flawed, the object is too complex, or both (or even none of these).We risk saying that this area of knowledge is still seeking its object (consequently, the way(s) to theorize and measure it).
The research design most commonly employed in the analyzed studies also deserves attention.Survey-type studies, descriptive or correlational, and with purposive samples imply a series of limitations regarding external validity, generalization, inference robustness, and conclusion validity (among others).In practical terms (to managers' decision-making), this may be a minor problem, but, in scientific terms, we should look at this more carefully.As soon as we do not have robust research designs in the area (e.g.experimental or quasi-experimental), it is hard to separate what is the phenomenon per si (training needs at work), its antecedents and consequents, and what are methodological flaws.
As for the levels of analysis, in most studies (19) the question of levels does not apply, because the author did not argue or discuss such a question.There is a relative predominance of studies that investigated the micro level of TNA ( 16).In relation to the meso and macro levels, there are a balanced number of studies (8).It seems that in Management research focus is on the macro level of analysis, while in Psychology there is a shift to the micro level.These results contradict the findings of Chiu, Thompson, Mak, and Lo (1999), who said that the most studied level of analysis was macro (organizational), followed by the meso (groups, tasks and processes), with the individual being the least studied.Otherwise, it is clear that multilevel analysis and modeling is still a neglected technique when it comes to TNA practice and research, even with clear indications that theory and data in this area can have a hierarchical structure (Koslowski, Brown, Weissbein, Cannon-Bowers, & Salas, 2000;McGehee & Thayer, 1961;Mossholder & Bedeian, 1983;Ostroff & Ford, 1989), which is recommended by multilevel literature as a premise to use such methods (Hox, 2010;Hox & Roberts, 2011;Kreft & De Leeuw, 1998;Raubendush & Bryk, 1986;Snijders & Bosker, 1994).
Regarding the area, there is a predominance of studies in Management (25), followed by studies applied to Medicine (11) and Psychology (7).Others areas also include TNA techniques, Education, Public Management, Marketing and Information Technology.This is quite interesting and shows that needs at work is a multi/inter/trans-disciplinary research object, being of interest to multiple knowledge fields.Perhaps this justifies the field's theoretical and methodological diversity.
Regarding the independent and dependent variables, few studies (6 out of 51) aimed to correlate variables, which is a measure adopted for testing models and hypotheses.Some independent variables were: clarity of mission in the area of personnel development, presence of total quality programs, level of investment in personnel development.Some dependent variables were: TNA, instructional design, employee satisfaction, productivity, communication, TD&E expenditures.Methodologically, these were quantitative and survey-type studies, with purposive samples, using questionnaires.All studies report significant relationships between the variables of interest.An  (Hansson, 2007).But is important to affirm that these results can have several alternative explanations, since the methods employed do not permit causality inferences (Baumeister & Leary, 1997).
TNA: where should (or could) we go?
According to the TNA strengths and weaknesses in practice and research presented earlier, we have a rich research agenda that could be structured.
First, TNA initiatives should have a broader and more proactive focus, shifting from exclusively reactive and diagnostic to a theory development and review framework.The issue on whether a competence gap should be faced by training or others types of instructional events (such as development, instruction, education or even informal learning at work) has not yet been discussed.It seems inappropriate to define a priori that a competence gap necessarily signifies a training need.We suggest that another types of needs should be discussed, such as learning needs, educational needs, development needs, among others.A deep theoretical and epistemological refinement of needs at work concepts and methods could achieve such an agenda.
It is also important that TNA practice and research do not focus only on present competences related to professional roles (Borges-Andrade & Lima, 1983), but also on emerging competences that can be important to the organization in the future (Felstead & Ashton, 2000).Besides promoting training events with higher probabilities of positive impact at work, this would also allow the development of long-term training and TNA policies, which, by the way, is neglected in the studies analyzed.Still, concerning competences, we noted that there is no consensus about it as the only construct that permits investigating training needs.This is of great concern since training investment decisions are being made based on a diversity of indicators that depend more on the context and less on the individual (Asku, 2005;Castley, 1996), constituting, in fact, other types of needs than training.Thus, we suggest that the literature on competence and competences management (Boyatzis, 1982;Cockerill, 1989;McClelland, 1973;Prahalad & Hamel, 1990;Sparrow & Bognanno, 1994) should be used as a primary resource for TNA practice and research, assuming that (gap of) competences are the only way to investigate human training needs at work.Methodologically, we can say that TNA practice and research has advanced in the past decades, employing scientific techniques to develop instruments (Hennessy & Hicks, 1998), proposing TNA models (Al-Khayyat & Elagamal, 1997) and being marked by methodological diversity, for example.However, TNA current methods still present flaws, as we showed earlier.First of all, we should have more mixed techniques, based on qualitative and quantitative data simultaneously (While, Ullman, & Forbes, 2007).Research and practice should also utilize a more heterogeneous and probabilistic sample, avoiding questioning only top managers, and including randomly chosen employees at different levels.This would allow triangulation of evidence, highly important to scientific and professional decision-making.Secondly, studies, especially TNA research, should employ more robust designs, such as quasi-experimental or experimental studies.This would allow for a more reliable set of conclusions about whether or not to invest in training and would help to improve TNA theories and concepts.
Regarding level of analysis, the adoption of multilevel modeling in TNA research is urgent.Since 1950, several studies have suggested that TNA theory and data can have a hierarchical arrangement (Koslowski et al., 2000;McGehee & Thayer, 1961;Mossholder & Bedeian, 1983;Ostroff & Ford, 1989).Nevertheless, current practice and research neglects to discuss this issue.Research should be based in at least three levels and their respective variables, followed with statistical multilevel regression analysis: internal and external organizational contexts (as laws, technology, politics, structure); organizational area or unit characteristics (number of employees, training budget, organizational level); and individual characteristics (training needs, competences BAR, Rio de Janeiro, v. 10, n. 1, art. 5, pp.77-99, Jan./Mar.2013 www.anpad.org.br/bardomains, age, education, learning style).In this scenario, we could more deeply investigate needs antecedents and consequents and needs definitions between levels (including groups of individuals).
Regarding areas of application, TNA must continue to expand.We can suggest that TNA is applicable to several areas of knowledge, as our results indicated.Wherever one is interested in professional education planning and executing, TNA approaches can be employed (technology, medicine, management, marketing, mental health, education, psychology).Perhaps, this is why TNA is such an interesting and exciting theme.

Table 1 Selected Articles' Analysis Criteria and Their Definitions
7Research field and participants Location(s) in which data collection occurred and description of participants.Continues

Table 3
shows the authors, countries, aims, and key research questions.

Table 4
shows the overall results according to design, nature and instruments.

Table 6
presents the independent and dependent variables used in research sample.